Proceedings of the Third Pacific Basin Conference on

Adsorption Science
and Technology
This page intentionally left blank
Proceedings of the Third Pacific Basin Conference on

Adsorption Science
and Technology
Kyongju, Korea May 25-29,2003

Editor

Chang-Ha lee
Yonsei University, Korea

r heWorld Scientific .
New Jersey London Singapore Hong Kong
Published by
World Scientific Publishing Co. Re. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: Suite 202, 1060 Main Street, River Edge, NJ 07661
UK oflce: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

ADSORPTION SCIENCE AND TECHNOLOGY

Proceedings of the Third Pacifc Basin Conference
Copyright 0 2003 by World Scientific Publishing Co. Re. Ltd.
All rights reserved. This book, or parts thereof; may not be reproduced in any form or by any means,
electronic or mechanical, includingphotocopying, recording or any informationstorage and retrieval
system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923,USA.In this case permission to
photocopy is not required from the publisher.

ISBN 981-238-349-2

Printed in Singapore
Preface

The Third Pacific Basin Conference on Adsorption Science and Technology was held
fiom May 25 to May 29, 2003 in Kyongju, Korea. The theme for this conference was
thinking about adsorption at a splendid, enjoyable, and sound conference.
It was the first time that an international conference on adsorption was ever held in
Korea. Since the previous conferences organized by Professor K. Kaneko (1997) and
Professor D. D. Do (2000) were very successful, I was very excited as well as very
nervous when I was asked to organize this conference for I wanted to make this one as
successful as the previous ones.
The main purpose of this conference was to encourage the development of new
adsorption science and technology as well as to reflect the growth of this area. The
conference covered a variety of adsorption-related fields from fundamentals to
applications. The conference consisted of plenary and invited sessions, oral sessions and
poster sessions. And the conference areas were as follows: Fundamentals of Adsorption
and lon Exchange, New Materials, Adsorption Characterization, Novel Processes, Energy
and Environmental Processes.
I was very happy to see many contributions from 16 countries, with more than 120
papers. The plenary lectures of Professors D. D. Do (Univ. of Queensland, Australia), K.
E. Gubbins (North Carolina State Univ., USA), K. Kaneko (Chiba Univ., Japan), M.
Morbidelli (ETH Ziirich, Switzerland), A. L. Myers (Univ. of Pennsylvania, USA), D. M.
Ruthven (Univ. of Maine, USA), R. Ryoo (KAIST, Korea), S. Sircar (Leigh Univ. USA),
M. Suzuki (United Nations Univ., Japan), and R. T. Yang (Univ. of Michigan, USA) set
the tone for the theme of the conference. Also, I would like to thank Professors G.Baron,
A. Neimark, and L. Zhou for their contribution as invited speakers. Plenary speakers
presented an in-depth overview of key research areas: Materials, Characterization,
Molecular Simulations, Equilibria, Kinetics, and Processes. Furthermore, many
contributed papers were of high standard. 1 hope that this conference was a worthwhile
and memorable one for all the participants.
I would like to thank all the participants for all the contributions to the conference. I
would like to take this opportunity to thank all the reviewers for their efforts to review
papers within a very short period of time. Thanks should also go to the members of the
organizing committee and secretary, the advisory and scientific committee and session
chairs for their input and assistance. Also, special thanks goes to Professor J. Ritter (Univ.
of South Carolina, USA) for the organization of the US side. And I would like to express
my gratitude to Yeong-Joo Park, my wife, for all of her support and I thank my graduate
students for all of their hard work to make this conference work.
The conference would not have been possible without the generous financial support
from many organizations such as the National Science Foundation, Yonsei University,
Korean Institute of Chemical Engineers, KOSEF, KRF, Daesung Sanso Co., Research
Institute of New Energy and Environmental Systems at Yonsei Univ., Yonsei Center for
Clean Technology, Chonnam National University, Chungnam National University, NRL
for Themophysical Properties, ERC for the Advanced Bioseparation Technology, NRL
for Separation Process, NRL for Environmental Materials & Process.

Professor Chang-Ha Lee

Chairman of the Third Pacific Basin Conference on Adsorption Science and Technology
Department of Chemical Engineering, Yonsei University, Korea

V
Sponsors
Yonsei University
Korean Institute of Chemical Engineers
The Korean Federation of Science and Technology Societies
Korea Science and Engineering Foundation
Korea Research Foundation
Daesung Sanso Co.
National Science Foundation (USA)

Co-Sponsors
Research Institute of New Energy and Environmental Systems
at Yonsei University
Yonsei Center for Clean Technology
Chonnam National University
Chungnam National University
ERC for the Advanced Bioseparation Technology
NRL for Themophysical Properties
NRL for Separation Process
NRL for Environmental Materials & Process

vi
Conference Chair
Chang-Ha Lee (Yonsei Univ., Korea)

Conference Advisory Committee

B. H. Ha (Hanyang Univ.)
H. Lee (Yonsei Univ.)
H. K. Rhee (Seoul National Univ.)

Organizing Committee
S. H. Cho (KIER)
D. K. Choi (KIST)
S. H. Hyun (Yonsei Univ.)
J. W. Jang (SK Eng. & Construction)
H. Kim (Seoul National Univ.)
Y. M. Koo (Inha Univ.)
C. S. Lee (Korea Univ.)
H. Moon (Chonnam Univ.)
D. S. Park (Daesung Sanso Co.)
S. K. Ryu (Chungnam Univ.)
Y. G. Shul (Yonsei Univ.)
J. E. Sohn (Dong-A Univ.)
J. Yi (Seoul National Univ.)

International Advisory Committee

A. S. T. Chiang Wational Central Univ., Taiwan)
D. D. Do (Univ. of Queensland, Australia)
K. E. Gubbins (North Carolina State Univ., USA)
K. Kaneko (Chiba Univ., Japan)
2. Li (South China Univ. of Technology, China)
M. Morbidelli (ETH Zurich, Switzerland)
A. L. Myers (Univ. of Pennsylvania, USA)
J. Ritter (Univ. of South Carolina, USA)
J. L. Riccardo (UNSL, Argentina)
D. M. Ruthven (Univ. of Maine, USA)
H. Tamon (Kyoto Univ., Japan)
R. T. Yang (Univ. of Michigan, USA)
H. Yoshida (Osaka Prefectural Univ., Japan)
L. Zhou (Tianjin Univ., China)

Scientific Advisory committee

T. Bandosz (The City College of New York, USA)
G . Baron (Vrije Universiteit Brussel, Belgium)
M. Bulow (The BOC Group, Inc., USA)
G. Carta (Univ. of Virginia, USA)
K. Chihara (Meiji Univ., Japan)
C. B. Ching (The National Univ. of Singapore, Singapore)
J. Izumi (Mitsubishi Heavy Chem. lnd., Ltd., Japan)
M. Jaroniec (Kent State Univ., USA)

Vii
J. U. Keller (Univ. of Siegen, Germany)
J. M. Lee (KRICT, Korea)
K. H. Lee (Pohang Univ., Korea)
N. Lemcoff (The BOC Group, Inc., USA)
M. D. LeVan (Vanderbilt Univ., USA)
M. Mazzotti (ETH ZUrich, Switzerland)
G. McKay (Hong Kong Univ. of Sci. and Tech., Hong Kong)
F. Meunier (Laboratoire du Froid CNAM, France)
J. K. Moon (KAERI, Korea)
A. Neimark (TRI/Princeton, USA)
Y. D. Park (FUST, Korea)
A. E. Rodrigues (Univ. of d Porto, Portugal)
W. Rudzinski (Maria Curie-Sklodowsk Univ., Poland)
R. Ryoo (KAIST, Korea)
A. Sakoda (Univ. of Tokyo, Japan)
N. Seaton (Univ, of Edinburgh, UK)
S. Sircar (USA)
W. A. Steele (Pennsylvania State Univ., USA)
M. Suzuki (United Nations Univ., Japan)
0. Talu (Cleveland State Univ., USA)
Y. Teraoka (Kyushu Univ., Japan)
Y. Xie (Peking Univ., China)

Organizing Secretary
C. H. Cho (Youngdong Univ.)
S. S. Han (KIER)
K. T. Lee (Yonsei Univ.)
J. Y. Yang (SK Eng. & Construction)

...
Vlll
Contents

Preface V

Plenary Papers
Adsorption Equilibria of Sub-critical and Super-critical Fluids in
Carbonaceous Materials
D. D. Do and H. D. Do 1
FreezingMelting in Porous Carbons
E R. Hung, R. Radhakrishnan, E Beguin, M. Sliwinska-Bartkowiak
and K. E. Gubbins 9
Measurement of Diffusion in Microporous Solids
D. M. Ruthven 17
Ordered Mesoporous Carbons with New Opportunities for Adsorption
Studies
R. Ryoo and S. H. Joo 27
Quantum Micropore Filling and its Application Possibility
T Tanaka, I: Hattori, K. Murata, T. Kodaira, M. Yudasaka,
S. I@ma and K. Kaneko 35
Adsorption in Microporous Materials: Analytical Equations for TYPE I
Isotherms at High Pressure
A. L. Myers 44
New Sorbents for Desulfurization of Transportation Fuels
R. T. Yang, A. Hernandez-Maldonado, A. Takahashi and E H. Yang 51
Optimization of Continuous Chromatography Separations
2. Y. Zhang, M. Mauottiand M. Morbidelli 64
Adsorption Technology for Gas Separation
S. Sircar 72
Carbon Composite Membranes
M. Suzuki, A. Sakoda, S.-D. Bae, T Nomura and Y.-Y. Li 79
Invited Papers
On the Dominant Role of Adsorption Effects in Heterogeneous
Catalysis
J. E Denayer, G. V Baron, D. Devos, J. A. Martens
and R A. Jacobs 87

iX
Supercritical Adsorption: Paradox, Problems, and Insights
L. Zhou 91
Contributed Papers
Microwave Drying for Preparation of Mesoporous Carbon
H. Tamon, T. Yamamoto, T. Suzuki and S. R. Mukai 99
Computer Simulation of Transport in Cylindrical Mesopores
S. K. Bhatia and D.Nicholson 104
Multicomponent Mass Transfer Diffusion Model for the
Adsorption of Acid Dyes on Activated Carbon
K. K. H. Choy, J. E Porter and G. McKay 109
Sorption Thermodynamics of Nitrous OxideLSX Zeolite Systems
M. Biilow, D. Shen and S. R. Jale 114
Activated Carbon Membrane with Carbon Whisker
S.-D. Bae and A. Sakoda 121
Mesoporous Silica with Local MFI Structure
S. P. Naik, A. S. T. Chiang, R. W Thompson, E C. Huang
and H.-M.Kao 126
Infinite Dilution Selectivity Measurements by Gas Chromatography
S. Gumma and 0. Talu 131
Adsorption Properties of Colloid-Imprinted Carbons
M. Jaroniec and 2.-J. Li 136
On the Role of Water in the Process of Methyl Mercaptan
Adsorption on Activated Carbons
S. Bashkova, A. Bagreev and T. J. Bandosz 141
Studies on the Adsorption Properties of Ion-Exchanged Low Silica
X Zeolite
H. Jiang, W Tang, J. P. Zhang, B. I! Zhao and I! C. Xie 147
Carbonization of Organic Wastes Using Super-Heated Water Vapor
and Their Adsorption Properties
H. Yoshida, N. Miyagami and M. Terashima 152
Further Successful Applications of the New Theoretical
Description of Adsorptioflesorption Kinetics Based on the
Statistical Rate Theory
#? Rudzinski and T. Panczyk 157
Characterization and Ethylene Adsorption Properties of
Silver-Loaded FER Zeolite Potentially Used as Trap Material
of Cold-Start Hydrocarbon Emission from Vehicles
!I Teraoka, H. Onoue, H. Furukawa, I. Moriguchi, H. Ogawa
and M. Nakuno 162

X
Pressure-Dependent Models for Adsorption Kinetics on a CMS
Z-S.Bae, I!-K. Ryu and C.-H. Lee 167
Preparative Enantioseparation of Fluoxetine by Simulated
Moving Bed
H.-W Yu and C. B. Ching 172
Optimization Based Adaptive Control of Simulated Moving Beds
G. Erdem, S. Abel, M. Mauotti, M. Morari and M. Morbidelli 177
Mono-Methyl Paraffin Adsorptive Separation Process
S. Kulprathipanja, J. Rekoske, M. Gutter and S. Sohn 182
Chromium (VI) and (111) Species Adsorption from Aqueous Solutions
by Activated Carbon Fibers
0. Astachkina, A. Lyssenko and 0. Muhina 189
Treatment of Complex Wastewaters by Biosorption and Activated
Carbon: Batch Studies
C. Gerente, 2. Reddad, Z: Andres, C. Faur-Brasquet
and I? le Cloirec 194
Adsorption Characteristics of Protein-Based Ligand for Heavy Metals
M. Terashima, N. O h ,T. Sei, K. Shibata and H. Yoshida 199
Preparative Chromatography at Supercritical Conditions
A. Rajendran, M. Mauotti and M. Morbidelli 204
Adsorptive Separation of Oligosaccharides: Influence of
Crosslinking of Cation Exchange Resins
J. A. Vente, H. Bosch, A. B. de Haan and P. J. T. Bussmann 209
Identification and Predictive Control of a Simulated Moving
Bed Process
I.-H. Song, H.-K. Rhee and M. Mazzotti 214
Quick and Compact Ozonation Using Siliceous Zeolite
H. Fujita, T. Fujii, A. Sakoda and J. Izumi 219
Time Resolved Multicomponent Sorption of Linear and Branched
Alkane Isomers on Zeolites, Using NIR Spectroscopy
A. F: P. Ferreira, M. Mittelmeijer, M. Schenk, A. Bliek
and B. Smit 224
Pore Size Effects in the Liquid Phase Adsorption of Alkanes
in Zeolites
J. F: M. Denayer, K. de Meyer, J. A. Martens and G. K Baron 229
Detection of Freezing Point Elevation in Slit Nanospace by
Atomic Force Microscopy
M. Miyahara, M. Sakamoto, H. Kanda and K. Higashitani 234

xi
Modeling of High-pressure Equilibrium Adsorption of Supercritical
Gases on Activated Carbons. Determination of Pore Size
Distribution Using a Combined DFT and EOS
E. A. Ustinov and D. D. Do 239
On the Peculiarity of the Minimum of N-Hexane Permeability in
Activated Carbon
J.-S. Bae and D. D. Do 244
Simplified Experimental Method to Analyse Intra-Activated
Carbon Particle Diffusion Based on Parallel Diffusion Model
I: Miura, I: Otake, H. T. Chang, N. Khalili, S. Iwasawa
and E. G. Furuya 249
In-Situ Characterization of Ion Adsorption at Biomimetic
Airwater Interfaces
T. E Kim, G. S. Lee and D. J. Ahn 254
Single and Multi Component Adsorption of Volatile Organic
Compounds onto High Silica Zeolites - Discussion of
Adsorbed Solution Theory
I? Monneyron, M. -H. Manero and J. -N. Foussard 259
Influence of VOCs Molecular Characteristics on Exothermicity
of Adsorption onto Activated Carbon
l? Pre, C. Faur-Brasquet and I? le Cloirec 264
The Influence of Ar and He on the Rate of Adsorption and on the
Adsorption Equilibrium of Alkanes in Zeolites
M. C,Mittelmeijer-Hazeleger,A. E l? Ferreira and A. Bliek 270
Modeling the Discharge Behavior of Metal Hydride Hydrogen
Storage Systems
S. A. Gadre, A. D. Ebner, S. A. Al-Muhtaseb and J. A. Ritter 276
The Advanced Modeling Technology for Periodic Adsorption
Process: Direct Determination of Cyclic Steady State
J.-H. Yun, A. C. Stawarz and E 0. Jegede 28 1
Adsorption and Desorption Characteristics of Zeolite Impregnated
Ceramic Honeycomb for VOC Abatement
H.-S. Kim, l!-J. Yoo, E-S.Ahn, M.-K. Park, K.-T. Chue
and M.-H. Han 286
Reverse Flow Adsorption Technology for the Recycling of
Homogeneous Catalysts: Selection of Suitable Adsorbents
J. Dunnewijk, H. Bosch and A. B. de Haan 29 1
Molecular Simulation of Gas Separation by Adsorption Processes
J. I? B. Mota 296

xii
Metal-Doped Sodium Aluminium Hydride as a Reversible
Hydrogen Storage Material
J. Wang, A. D. Ebner, K. R. Edison, J. A. Ritter and R. Zidan 30 1
Synthesis and Dehumidification Behaviors of Monodisperse Spherical
Silica Gels with Different Pore and Chemical Structures
C,H. Cho, Y. J. Yoo, J. S. Kim, H. S. Kim, Y. S. Ahn
and M. H. Han 306
Production of Hard Carbons for Lithium Ion Storage by the
Co-Carbonization of Phenolic Resin Precursors
S. R. Mukai, T. Tanigawa, T Harada, T. Masuda and H. Tamon 313
Novel Bioactivite Carbomineral Sorbents, Including Cluster
and Carbon Nanotubes for Superselective Purification of Biodiesel
Fuel - Liquid Hydrocarbons and Carbonhydrate from
Sulfur Containing Impurities
D. I. Shvets 318
Titanosilicate ETS-10: Synthesis, Characterization and Adsorption for
Heavy Metal Ions
G. X. S. Zhao, J. L. Lee and f? A. Chia 324
Ordered Macroporous Materials Structurally Templated by Colloidal
Microspheres
Z. Zhou, ?C.
-I ? and G. X. S. Zhao
Ong 329
Adsorption of Nitrogen, Oxygen and Argon in Transition and Rare
Earth Ion Exchanged Zeolites A and X
R. K Jasra, J. Sebastian and C. D. Chudasama 334
Adsorption of Methylene Blue from Water onto Activated Carbon
Prepared from Coir Pith, an Agricultural Solid Waste
C. Namasivayam and D. Kavitha 339
Separation of Oxygen-Argon Mixture by Pressure Swing Adsorption
X . Jin and S. Farooq 344
Dual Reflux Pressure Swing Adsorption Cycle for Gas Separation
and Purification
A. D. Ebner and J. A. Ritter 349
Simulation of a Coupled MembranePSA Process for Gas Separation
I. A. A. C. Esteves and J. I! B. Mota 354
I3CO and l2C0 Separation on Na-LSX using Pressure-Swing Adsorption
at Low Temperatures
J. Izumi, N. Fukuda, N. Tomonaga, H. Tsutaya, A. Yasutake,
A. Kinugasa and H. Saiki 359

xiii
High Purity Oxygen Generation PSA Process by Using Carbon
Molecular Sieve
J.-G. Jee, T-H. Kwon and C.-H. Lee 365
A Study on the Preparation of Deodorizing Fibers by Coating TiOz
S. W Oh, H. J. Kim and S. M. Park 370
Composite Adsorbents for the Removal of Cs and Sr Ions in
Acidic Solutions
J. K. Moon, C. H. Jung, S. H. Lee, E. H. Lee, H. T. Kim
and I! G. Shul 375
Dehumidification Behavior of Metal(Ti, Al, Mg) Silicates
Impregnated Ceramic Fiber Sheets
I! S. Ahn, C. H. Cho, I! J. Yoo, J. S. Kim, H. S. Kim
and M. H. Han 38 1
Synthesis of Zirconia Colloids from Aqueous Salt Solutions
and Their Applications
K. Lee, P. W Carr and A. V McCormick 387
Comparison of Nano-Sized Amphiphilic Polyurethane (APU)
Particles with SDS, an Anionic Surfactant for the Soil Sorption
and the Extraction of Phenanthrene from Soil
I . 3 . Ahn, H.-S. Choi and J.-Z Kim 392
Synthesis of Mesoporous Activated Carbon with Iron Ion-Aided
Activation
Z Seida, K. Watanabe and Y: Nakano 398
Separation of Peptides from Human Blood by RP-HPLC
S.-K. Lee, I! Polyakova and K.-H. Row 403
Separation of Acanthoside-D in Acanthopanax Senticosus by
Preparative Recycle Chromatography
S . 2 Hong, D.-X. Wang and K.-H. Row 408
Use of Various Forms of &aft Lignin for Toxic Metal Uptake
D. R. Crist, R. H. Crist and J. R. Martin 413
Removal of Uranium Ions in Sludge Waste by Electrosorption
Process
C.-H. Jung, J.-K. Moon, S.-H. Lee, Z-G.Shul and W-Z. Oh 417
Ion Exchange Characteristics of Palladium from a Simulated
Radioactive Liquid Waste
S.-H.Lee, C.-H. Jung, J.-K. Moon, J. H. Kim and H. Chung 422
Application of Characterization Procedure for Complex Mixture
Adsorption in Water and Wastewater Treatment
S.-H.Kim, T.-W Kim, D.-L. Cho, D.-H.Lee and H. Moon 427

xiv
Surface Characteristics of MCM-41 on Cr(1II) and Cr(V1)
Adsorption Behaviors
S. J. Park, B. R. Jun and M. Han 432
Influence of Anodic Oxidation of Activated Carbon Fibers
on the Removal of Heavy Metal in Aqueous Solution
S. J. Park, !I M. Kim and J. R. Lee 437
Kinetics and Diffusion Processes for Reactive Dye Adsorption
by Dolomite
S. J. Allen, G. M. Walker, L. Hansen and J.-A. Hanna 442
Permeate Flux Behavior During Microfiltration of Protein-Adsorbed
Microspheres in Stirred Cell
I: Chang, S.-W Choi, T-G. Lee, S. Haam and W-S. Kim 447
Surface Fractional Dimensions of the Adsorbents from Industrial
Sludge
J. H. You, H. M. Wu and 2. X. Fang 452
Adsorption of Acidic Peptide on Crosslinked Chitosan Fiber:
Equilibria
N. Kishimoto and H. Yoshida 458
Removal of Salt and Organic Acids from Solution used to Season
Salted Japanese Apricots (Ume) by Combining Electrodialysis
and Adsorption
W Takatsujiand H. Yoshida 463
Studies on the One-Column Analogue of a Four-Zone SMB
Y. S. Kim, C. H. Lee, Y. M. Koo and I? C. Wankat 468
Advanced' Flue Gas Treatment by Novel de-SOX Technology over
Active Carbon Fibers
M.-A. Yoshikawa, A. Yasutake and I. Mochida 474
Adsorption of Natural Gas Components on Activated Carbon for
Gas Storage Applications
I. A. A. C. Esteves, M. S. S. Lopes, I? M. C. Nunes,
M. E J. Eusibio, A. Paiva and J. I? B. Mota 479
Prediction of Breakthrough Curves for Toluene and
Trichloroethylene onto Activated Carbon Fiber
J.-M? Park, S.-S. Lee, X-W Lee and D.-K. Choi 484
Catalytic Reduction Mechanism of Nitric Oxide over
ACFslCopper Catalyst
S. J. Park, B. J. Kim and X S. Jang 489

xv
NO Removal of Activated Carbon Fibers Treated by Cu
Electroplating
S. J. Park, J. S. Shin and J. R. Lee 494
The Appliance Study of 02-PSA in the Oxygen Activated
Sludge Process
S. H. Lee, P. S. Yong, H. M. Moon and D. S. Park 499
Application of Solid Adsorbent for VOC Monitoring Sensor
0. J. Joung and E H. Kim 504
PSA for Solvent Recovery with USY-Type Zeolite; an
Experimental and a Simulation Study
K. Chihara, T. Kaneko, T. Aikou and S. O h 509
Azeotropic Adsorption of Organic Solvent Vapor Mixture on
High Silica Zeolite, Experimental & Simulation
K. Chihara, K. Hijikata, H. Yamaguchi, H. Suzuki
and E Takeuchi 514
VOC Enrichment by a VSA Process with Carbon Beds
J. Yang, M. Park, J.-W Chang and C.-H. Lee 519
Isobutane Purification by Pressure Swing Adsorption
S.4. Han, J.-H. Park, J.-N. Kim and S.-H. Cho 524
Characteristics of Gas Separation by Using Organic Ternplating
Silica Membrane
J. Moon, S. Hyun and C.-H. Lee 529
Adsorber Dynamics of Binary and Ternary Hydrogen Mixture
in Activated Carbon and Zeolite 5A Beds
M.-B. Kim, J.-S. Kim, C.-H. Cho and C.-H. Lee 534
Temperature Programmed Adsorption ( P A ) of Various
Hydrocarbons on Adsorbers of Honeycomb Type
D. J. Kim, J. E. fie, E S. Oh and J. M. Kim 539
Non-Isothermal Dynamic Adsorption and Reaction in
Hydrocarbon Adsorber System
D. J. Kim, W. G. Shim, J. E. Yie and H. Moon 544
Sorption of U(V1) onto Granite: Kinetics and Reversibility
M. H. Baik and P. S. Hahn 549
Adsorption of Uranium (VI) on Kaolinite: Speciation
and Mechanism
M. J. Kang, B. E. Han and P. S. Hahn 554
High-Temperature Adsorption of Hazardous Metal Chlorides
Using Activated Kaolinite
H. C. Yang, J. S. Yun, E J. Cho and J. H. Kim 559

xvi
Probing the Cut-Off for Intracrystalline Adsorption on Zeolites:
Pore Mouth Adsorption
R. Ocakoglu, J. E M. Denayer, J. A. Martens, G. B. Marin
and G. I! Baron 564
The Low-Temperature Sorption Behaviour of Cryosorbent Materials
C. Day and I! Hauer 569
Thermal and Surface Mechanism Studies on
Adsorption-Temperature Programmed Desorption of Nitrogen
Oxides over Chemically Activated Carbon Fiber
H.-J. Kim, X-W Lee, E. Lee and D.-K. Choi 574
Surface Properties of Activated Carbons Containing Basic
Hydroxide Ions and NOx Adsorption-Desorption Process
X W Lee, H. J. Kim, D. K. Choi, J. W Park C. H. Lee
and B. K. Na 579
Adsorption Characteristics of Nitrogen Compounds on Silica Surface
H. J. Kim, C.-H. Lee, X G. Shul and W S. Min 584
Adsorption Characteristics of VOCs on Mesoporous Sorbents
W G. Shim, M. S. Yang, J. W Lee, S. H. Suh and H. Moon 589
Molecular Simulation for Adsorption of Halocarbons in Zeolites
K. Chihara, T. Sasaki, S. Miyamoto, M. Watanabe,
C. E Mellot-Draznieks and A. K. Cheetham 595
Adsorption of BTX on MSC in Supercritical C02,a Chromatographic
Study
K. Chihara, N. Omi, E Znoue, T Yoshida and T Kaneko 600
Porous Alumina with Bimodal Pore Size Distribution as an
Organic Adsorbent
I: Kim, C.Kim, P Kim, .IC. Park and J. yi
. 605
Storage and Selectivity of Methane and Ethane into
Single-Walled Carbon Nanotubes
X-G. Seo, B. H. Kim and N. A. Seaton 610
Hydrotalcites for Carbon Dioxide Adsorbents at High Temperature
J. I. Yang, M. H. Jung, S.-H. Cho and J.-N. Kim 615
Effect of Polarity of Polymeric Adsorbents on Desorption of
VOCs under Microwave Field
X. Li, 2. Li, H. X. Xi and H. Wang 620
Mixed-Gas Adsorption on Heterogeneous Substrates in the Presence
of Lateral AD-AD Interactions
A. J. Ramirez-Pastor, E M. Bulnes and J. L. Riccardo 625

xvii
Adsorption on Correlated Disordered Substrates
R. H. Lopez, E M. Bulnes, E Rojas, J. L. Riccardo
and G. Zgrablich 630
Temperature Effects on the Scaling Properties of Adsorption
on Bivariate Heterogeneous Surfaces
E Rom*, E Bulnes, A. J. Ramirez-Pastor and G. Zgrablich 635
Adsorption of Polyatomic Species: An Approach from
Quantum Fractional Statistics
J. L. Riccardo, A. J. Ramirez-Pastor and E Rom 640
Multilayer Adsorption with Multisite-Occupancy
E Rom, A. J. Ramirez-Pastor and J. L.. Riccardo 645

xviii
Plenary Papers
This page intentionally left blank
 ADSORPTION EQUILIBRIA OF SUB-CRITICAL AND SUPER-CRITICAL
FLUIDS IN CARBONACEOUS MATERIALS

D.D.DO AND H. D.DO

Department of Chemical Engineering, University of Queensland, St. Lucia, Qld 4072, Australia
E-mail: [email protected]

In this paper, we present an overview of a number of techniques used to characterize the adsorption
equilibria of sub and super-critical fluids in nonporous carbon black and porous activated carbon.
Tools such as the grand canonical Monte Carlo simulation (GCMC), Density Functional Theory
(DFT), Molecular Layer Structure Theory (MLST) of Ustinov and Do, and enhanced molecular
layering and pore filling proposed by Do and his co-workers will be discussed. Although the GCMC
provides the brute force calculation of molecular interactions and its results have been used as a
benchmark for other techniques to compare with, its extremely time consuming computation makes
the other techniques more appealing to engineers and experimentalists. Despite the ever-increasing
computing power of today personal computer, the advantage of simpler methods is warranted for
their role in solving practical problems. Furthermore, added to this advantage is the good
perfarmance in terms of the prediction power of the simpler methods. AN these tools will be
discussed in this paper regarding their applications to adsorption of super and sub-critical fluids in
carbonaceous materials, such as graphitid thermal carbon black and activated carbon.

1 Introduction

Adsorption equilibria and kinetics are important for the proper design of adsorption
processes. The equilibria information of adsorption isotherm. is clearly the first hand
information that one needs to approximately size the adsorber. Since the adsorption
affinity can vary by many orders of magnitude. Ranging from very low for weakly
adsorbing gases to very high for strongly-adsorbing hydrocarbon vapours, it is very
important that we know the value of this adsorption affinity. Experimentally this
information can be obtained from careful experimentation of adsorption isotherm
measured from very low pressure (where adsorption affinity can be calculated) to very
high pressure where saturation capacity can be determined. Alternatively, the adsorption
affinity can be determined from some appropriate theories or computer simulation. With
the advances of high speed computer and the development of modem tools to deal with
inhomogeneous fluids in confined space such as micropores, the second approach is
gaining ground and new theories are constantly developed allowing engineers and
scientists to calculate adsorption isotherm from minimum amount of information.
Nevertheless, that does not mean to say that we can make do without experimental data.
We still have some ground to cover before that stage can be reached. At the present time,
careful and reliable experimental data are still required for validation of theories and even
confirmation of molecular simulations.
In this paper we will discuss some modem tools for studying adsorption equilibria of
super and sub-critical fluids on non-porous surface and in porous carbons. In particular,
the tools of grand canonical Monte Carlo (GCMC) simulation [l, 21, Density Functional
Theory [3], Molecular Layer Structyre Theory (MLST) [4, 51, and the enhanced
molecular layering and pore filling of Do, first developed in 1998 [6]and later applied in
a number of practical systems [7-101, will be discussed. Their applications to
experimental systems are illustrated to highlight the advantages and disadvantages of
these modem tools of equilibria characterization.

1
2 Tools of characterisation

2. I Grand canonical Monte Carlo simulation

In the molecular simulation of adsorption in confined space such as pores of adsorbent,
the most widely used and successful ensemble is the grand canonical Monte Carlo
simulation. In this ensemble, we specify the chemical potential of the fluid,p, that the
candidate pore is immersed in, the size of the pore, and the temperature.
In the GCMC, the simulation can be carried out in the same procedure suggested by
Metropolis at al. [ 111, and the density distribution (hence average pore density) can be
obtained as a direct result of the simulation. The GCMC simulation is usually carried out
by starting with a very low value of chemical potential. Once the simulation
corresponding to this chemical potential has been completed, the chemical potential is
increased incrementally and the particle configuration of the last run is used as the initial
configuration and the new simulation is then carried out. This process is repeated until
the final chemical potential is performed. In the simulation, the simulation box is
constraint by the width of the pore and the L, and L, in the x and y directions. The box
lengths L, and L, are chosen large enough, and usually chosen as twice the cut-off
distance. In our work we choose r, = 5ofi and periodic boundary conditions are applied
in the x and y directions. The procedure of the GCMC involves three basic moves:
1. DisDlacement of Darticle: A particle is selected in random and is given a new
conformation (translation and rotation). The move is accepted with a probability
p = min[l, exp(-AU / kT)]
where AU is the difference between the new and old configuration energies. In this
move the displacement step and the rotation angle are chosen in such a manner that
the acceptance rate is between 25 and 50%.
2. Insertion of Darticle: The second move is the particle insertion move. A particle is
generated at a random position. Its acceptance must satisfy the probability:
- -
p =min 1,
1 A~(N+~)
exp& - U(N + 1)+ U(N)]/ kT)

where V is the volume of the simulation box and A is the thermal de Broglie
wavelength. Usually in the simulation we supply the activity, z = A- exp(p/ kT) ,
instead of the chemical potential. For bulk gas phase which behaves as an ideal gas,
this activity is the density of the gas phase.
3. Removal of Darticle: The third move in the GCMC simulation is the removal of a
particle. A particle is chosen in random and removed. The probability of acceptance
this removal is

The GCMC is successfully applied to many adsorption systems. For temperatures

greater than the pore critical temperature, the adsorption isotherms exhibit a smooth
behaviour. However, for temperatures less than the pore critical temperature, there is
a possibility of transition. When there is transition, the equilibrium point may be
obtained by applying the thermodynamic integration method [ 121.

2
2.2 Density Functional Theory (DFT)
The DFT method was popularized in the sixty and was increasingly modified and applied
to many problems involving inhomogeneousfluids in the 80s. Among the many versions
of DFT, the one proposed by Tarazona and co-workers [3] remains the popular one in
solving adsorption of confined fluid in pores. Their method is the non-local DFT and is
applied on a grand canonical ensemble. The starting point of the method is the grand
potential of the system
= F(p(l)) + jP@)ucxtcr)dx - jP@>P& (la)
where p(rJ is the singlet particle density. The first term on the RHS of the above equation
is the Helmholtz free energy of the system. The Helmholtz free energy is expressed as a
sum of two terms. The first term is that obtained from a reference system (hard sphere
fluid is chosen as one) and the other is the perturbed component (which is due to the
attractive component of the intermolecular fluid-fluid interaction), that is
F(PCr)) = FH(P@N+ FA-( P O
The Helmholtz free energy of the reference hard sphere system, in turn,can be expressed
as two terms. One is due to the ideal gas contribution (accounting for the momenta of all
particles) while the other term is the repulsive interaction among the hard spheres. It is
FH(P(E)) = jP@
[In(A3P(IN - 11dr + j P ( 0 f m G(INdI
where the smoothed density is given by
-
P(I) = I P W w(x-il;p(l))dzI
Assuming a mean field approximation, the perturbed component of the Helmholtz free
energy due to the attractive force is
1
FA-(P(I)) =?j j P 0 P W %(I-d)dd dx (1b)
The equilibrium density profile can then be obtained by finding minimum of the grand
potential as defined in eq.(la). Tarazona used the Carnahan-Starling equation to derive
the excess function f(p(I)) as (here, d is the hard sphere diameter)
f(&)) = (4y - 3 ~ )/(I - y) ; y = (nd 16) &)

2.3 Molecular Layer Structure Layer Theoty (MLST)

The Molecular Layer Structure Theory (MLST) was first developed [4, 51 to study the
vapour liquid equilibrium and surface tension of many substances over a wide range of
temperature. In this method, the fluid is considered as parallel molecular layers, whose
surface densities are different for inhomogeneous
- fluids. Similar to the DFT method, we
define the following grand potential:
R= cp;k(p;)+cpj
i
+urf - jl]

where py is the surface density, Ask)is the intrinsic Helmholtz free energy of the layer
j, pj is the interaction energy between one molecule on the layerj and all surrounding
layers, and u; is the external potential exerted on the layerj. In this method of MLST,
the intrinsic Helmholtz free energy is calculated from the equation of state of
homogeneous fluid, and pi is calculated from the integration of 12-6 Lennard-Jones

3
potential energy. The equilibrium density is calculated by minimizing the grand potential
with respect to density as well as distances between layers [4,5].

2.4 The Do-Method

Recently, Do and co-workers [6-101 have proposed a very simple method but it does
reveal the mechanistic pictures of what are occurring in pores of different size. The
process of adsorption in pore is viewed as follows. Molecules in pore are constantly
in motion but statistically there is a spatial distribution of these molecules due to the
interactive forces between them and the surface atoms. We treat this spatial distribution
as a step function; uniformly high density near the surface and uniformly low density in
the inner core of the pore. Due to the long range interaction of the surface, the pressure
of the fluid in the inner core is not the same as that in the bulk phase. Assuming a
Boltzmann distribution, the pressure of the inner core is related to the bulk fluid as
Pp = P.exp(- a E/ kT) (2)
where E is the average potential of the inner core, which is a function of pore width and
the layer thickness. The adsorption process is basically treated as a molecular layering
process, and it can be described by a layering equation, for example an equation taking
the form of the BET equation. The affinity parameter C, of this equation the same
as the C-parameter for a flat surface. Rather they are related through the following
relation
c,(H) = c.~xP[(Q,(H)-Q)/RT~ (3)
where Q is the heat of adsorption of the flat surface and Qp is the correspondingvalue for
a pore. These values may be taken as the minimum value of the potential energy between
a molecule and a flat surface and that between a molecule and a pore. The layering
equation can be written as
t / t, = f(P, (HI, c, (HI) (4)
where ,t is the thickness of a monolayer. This layering process is followed by a pore
filling process. Here the term pore filling is used in its most general sense, that is pore
is filled with molecules by either two-dimensional condensation in small pores or three
dimensional condensation in large pores. We argue that this general pore filling process
is governed by the equation

H / 2 - t - ~ , =)*
where ~0 is the position at which the solid-fluid potential energy is zero. Although this
form is similar to that of the Kelvin equation, the significant difference rests on the use of
the pore pressure P, in the above equation. If we substitute the pore pressure of eq.(2)
into the above equation, we get
P M
(H/2 - t - 2,)
For small pores, E(H) dominates the RHS of the above equation and hence the pore
filling pressure in small pores is dominated by the strength of the potential field created
by the overlapping of the fields of the two walls. On the other hand, for larger pores
where &H) = 0 , the RHS is dominated by surface tension and this equation reduces to
the well known modified Kelvin equation.

4
 Using eqs. (2) to (5) we can readily
obtain the pressure at which'a pore is
completely filled with adsorbate
molecules. This pressure is called the
pore filling pressure. Figure 1 shows
the plot of the reduced pore filling
pressure (P/Po) versus pore width for
nitrogen adsorption in slit pores at 77.3
K and for argon at 87.3 K. The results
of DFT and GCMC are also shown as
symbols, and it is seen that the
0 10 20 30 40 50 agreement between the DFT., GCMC
Pore Width (A) and the Do method is very good, even
Figure 1: Pore tilling p r e ~ s ~versus
n pore width the minimum position of this curve.
for argon at 87.3 K and nitrogen at 77.3 K

3 Results and Discussion

Having presented briefly the working procedures of the various methods, we now would
like to illustrate their applications to adsorption of super and sub-critical fluids on non-
porous carbon surface and in porous carbonaceous solids having slit pores. But first it is
worthwhile to compare the time scales of computation of these methods:
Do-method < MLST <4 DFT < GCMC
(seconds) (minutes) (hours) (day)

3.1.1 Non-porous surfaces

We show in Figures 2 the adsorption isotherm of argon on graphitized thermal carbon
black at a number of tempemtures. The symbols in the figure are experimental data [131.
10

0 5 10
Resslm,(MPa)
Figure 2n: MLST prediction of argon isotherms
15
0
0
-
5 10
WPa)
15

Figare Z b Comparison between GCMC & MLST

The continuous lines in these plots are fiom the results of MLST theory and GCMC. We
see that the agreement between the MLST theory and the data [131 is excellent (Figure
2a). In patticular, the temperature dependence of the isotherms is well described by the

5
MLST model. The results obtained from GCMC are shown in Figure 2b, and it is
surprising that the GCMC results are not as good as the MLST results. It describes the
data well at low pressure but it consistently over-predicts the amount adsorbed at higher
pressures. One probable explanation for this might be the possibility of temperature and
density dependence of the molecular parameters in GCMC simulation for high pressures.
This reason needs some justification because the MLST also uses molecular parameters,
and they are kept constant in the calculations. This illustrates that the molecular
simulation, despite its great potential, is less effective than the MLST for this problem of
high pressure adsorption of super-critical fluids on non-porous surfaces. The MLST is
very effective in the study of the effects of parameters and operating conditions, such as
surface carbon density and solid-fluid interaction energy, on the behaviour of surface
excess at extremely high pressures. This will be discussed in details at the conference.
What we observed for argon are also repeated with Kr, methane, ethylene, propane
and sulphur hexafluoride. Figures 3 show the comparison between GCMC and MLST
for ethylene at 283 JC, which is just greater than the critical temperature.

We see that at this temperature, the surface

excess versus pressure shows a distinct
spike at a pressure very close to the critical
pressure of 5.036 MPa Beyond this
pressure, the surface excess decreases very
sharply and this is due to the large change
in density in the bulk fluid for a small
change in the pressure. It is remarkable
that the MLST describes the data [14] very
0.1 1 10 100
well. The GCMC, on the other hand,
Pmssurs (MPa)
Figure 3: Comparison between GCMC & MI.ST for describes well at low pressures but badly
ethylene adsorption at 283 K predicts at higher pressures.

3.1.2 Slit pores

For slit pores, we compare the results of DFT and GCMC for slit pores having width 8
and 14 A in Figures 4 for argon adsorption at a super-critical temperature of 298K.
0.8 1.o

0.8
0.8

-&
0.4

B
0.2
0.2

0.0 0.0
104 104 10= 10' lo-' 100 10' 10' lW 103 101 10' 16 10' 102
Activity. z d AdMty, zd
Figure 4n: GCMC versus DFT for slit pore H = 8 A Figure 4b:GCMC versus DFT for slit pore. H = 14 A

We fmt note the gradual behaviour of the adsorption isotherms, and this is typical for
super-critical fluid adsorption in pores of all sizes. For sub-critical adsorption which we
will consider later, there is a possibility of phase transition for certain pore widths.

6
 We see that the agreement between the DFT and GCMC results is very good for
large micropores of 14 A, while there is a distinct deviation in small micropore of 8 A.
This is due to the mean field approximation of the DFT method (eq. lb). In terms of
computation time, the DFT method is modestly faster than the GCMC (depending on the
error criteria as well as the grid points used in the determination of singlet density
distribution) but not substantially fast enough to warrant it in its use in routine process
applications.

3.2 Sub-criticalfluid

3.2.1 Non-porous surface

Sub-critical fluids are widely applied for surface and pore characterization. Figure 5
shows the surface excess adsorption of argon at 87.3K on graphitized thermal carbon
black. The results are obtained fiom the GCMC and DFT. The subtle difference in the
formation of second and higher layers is observed and this is due to the mean field
approximation. As an illustration of the MLST method, we show in Figure 6 the results
in the prediction of argon adsorption isotherm on graphitised carbon black at 87.3 K [IS].
The theory exhibits a kink formation of the second layer and this deviation from the data
could probably be due to the heterogeneity of the graphene surface. It is interesting to
note that that kink exists in the DFT as well as GCMC simulation (see arrow in Figure 5).

10 1b 18 1@ 101
Gas Density. po -M (Pa)
pigpre 5: GCMC v m DFT for open surfacc for Ar Figure 6: Paformance of MLST for Ar at 87.3 K

3.2.2 Slit pores:

Here we would like to show the difference between the three methods used in pore
characterization: GCMC, DFT and Dos method. Figures 7 show the average pore
density versus pressure for argon adsorption in slit pores having width of 8 and 30A.
Here we see that the difference between the DFT and GCMC for 8 A pore is quite
significant, and not only in terms of the pore filling pressure but also on the behaviour of
the adsorption isotherm curves. The local isotherms obtained from the Do-method for
these pores, although not completely in agreement with the GCMC results, qualitatively
match these results rather well over the complete range of pressure. The fast computation
of the Do-method can be the greatest advantage among the various methods, and it can be
effectively used in process optimizationor in routine pore characterisation.

7
 *r-K!f,Pb Qaa Den*, Fa?
Figure 7a: GCMC versus DFT for slit pore H = 8 A Figwe m: G M C vermu, DFC for slit pore H-30 A

4 Conclusions

We have presented in this paper an overall but brief summary of various tools commonly
used in the adsorption equilibria calculations in carbon. These tools rage fmm simple
methods such as the Do-metho& the MLST of Ustinov and Do to advanced methods such
as the DFT and the GCMC simulation. Although the DFT is modestly faster than the
GCMC, they are both computer time-hungry and unsuitable in any applications that
require routine computation of isotherms, especially when parameters are varied. The
simple methods,in particular the Do-method,might provide an answer to this.

5 Acknowledgements

This work is supported by the Australian Research Council. Support b m PBAST3 is

also gratefully acknowledged.

References

1. A&, D. J., Mol. P h y ~29

. (1975) pp. 307-3 11.
F~nkel,D. and Sinit, B., Understanding Molecular Simulation (Acad. Press, 2002)
2.
Tarazona, P., Marconi, U. and Evans, R., Mol. Phys. 60 (1987) pp. 573-595.
3.
Ustinov, E. and Do, D. D., J. Colloid I n f e ~ a cScience
4. e (in press, 2002)
5. Do, D. D., Ustinov, E. and Do, H.D., FluidPhase Equilibria (bpress, 2002)
6. Do, D. D.,presented at The International Symposium on New Trends in Colloid and
Interface Science (14-26 September, 1998, Chiba, Japan).
7. Nguyen, C. andDo, D. D.,Lungmuir 15 (1999) pp. 3608-3615.
8. Do, D. D.,Nguyen, C. and Do, H.D.,Colloids Sur$acces 187-188 (2001)- _pp. _ 51-71.
9. Do, D.D. and Do,H.D.,Lungmuir 18 (2002) pp. 93-99.
10. Do, D.D. and Do, H.D., AppliedSu@ace Science (in pms, 2002)
11. Metropolis, N., Rosenblutb, A,, Rosenbhth, M., Teller, A. and Teller, E., J. Chem.
. (1953) pp. 1087-1092.
P h y ~21
12. Peterson, B. and GubbinS, K.E., Mol. Phys. 62 (1987) pp. 215-226.
13. Specovius, J. and Findenegg, G., Ber. Bun. Phys. Chem. 82 (1983) pp. 174180.
14. Findenegg, G., Funhentaki of A&otption, May 6-1 1 (1983) pp. 207-218.
15. Olivier, J., J. Porous Materials 2 (1997) pp. 9-17.

8
 FREEZINGMELTING IN POROUS CARBONS

F. R. HUNG AND K. E. GUBBINS

Department of Chemical Engineering, North Carolina State University,
Raleigh, NC 2 7695, USA

R. RADHAKRISHNAN
Department of Chemistry and Courant Institute of Mathematical Sciences,
New YorS NY 10003, USA

F. BEGUIN
Centre de Recherche sur la Matiire Divisee, UMR-6619, CNRS-Orleans University,
Orlt!ans945071 cedex 02, France

M. SLIWINSKA-BARTKOWIAK
Institute of Physics, Adam Mickiewicz University, Umultowska 8.5, 61-614 Poznan, Poland

We report molecular simulation and experimental results for the freezing/melting behavior of simple
fluids adsorbed in porous carbons having slit-shaped or cylindrical geometry. In the simulations we use
simple models for activated carbon fibers of pore width 1.5 nm and multi-walled carbon nanotubes of
diameter 5 nm. Experimental studies employ dielectric relaxation spectroscopy and nonlinear dielectric
effect measurements. We show that the solid-fluid phase behavior depends on two main parameters,
the pore width H (or diameter D? and a parameter measuring the ratio of the fluid-wall to fluid-fluid
attractive interaction (a).For slit-shaped pores, values of a larger than unity lead to an increase in the
freezing temperature of the confined fluid with respect to that of the bulk, and values of a smaller than
unity produce a decrease in the freezing temperature. Contact layer and hexatic phases were also
observed for some systems. The hexatic phase was found in slit-shaped pores with larger a values, and
is stable over extended temperature ranges. For cylindrical pores, our results for a D = 9 . 7 q multi-
walled carbon nanotube show no formation of regular 3D crystalline structures. Our results also suggest
that the outer layers experience a slight increase in the freezing temperature, while the inner layers
experience a depression in the freezing temperature with respect to the bulk freezing point. These
simulation results are in good agreement with the experimental results.

1. Introduction
An understanding of fieezing phenomena for fluids confined within nano-scale pores is
important in the fabrication of nano-structured materials, and in nanotribology, adhesion,
and characterization of porous materials. However, studies of these phenomena are
plagued by several difficulties. Experimental investigations must address the difficulties
of the lack of well-characterized porous materials with appropriate pore size, the
unambigous determination of the nature of the confined phase, and the prevalence of
long-lived metastable states. Molecular simulation studies do not suffer fiom any of these
difficulties; moreover, it is possible to determine free energies of the confined phases, and
so determine true thermodynamic equilibrium and the order of the phase transition.
Simulations experience other difficulties, however, in particular uncertainties concerning
intermolecular potentials and pore characterization, as well as limitations due to the speed
of current supercomputers. The difficulties found in experiments and simulations make
the two approaches complementary, so that combined experimental-simulation studies
can be rewarding.
Recent studies for pores of simple geometry have shown a rich phase behavior associated
with fieezing in confined systems -I3]. The freezing temperature may be lowered or
raised relative to the bulk fieezing temperature, depending on the nature of the adsorbate
and the porous material. In addition, new surface-driven phases may intervene between

9
the liquid and solid phases in the pore. "Contact layer" phases of various kinds often
occur, in which the layer of adsorbed molecules adjacent to the pore wall has a different
structure fi-om that of the adsorbate molecules in the interior of the pore. These contact
layer phases have been predicted theoretically, and confirmed experimentally for several
systems [4*91. In addition, for some systems in which strong layering of the adsorbate
occurs (e.g. activated carbon fibers), hexatic phases can occur; such phases have quasi-
long-ranged orientational order, but positional disorder, and for quasi-two-dimensional
systems occur over a temperature range between those for the crystal and liquid phases.
These are clearly seen in molecular simulations, and recent experiments provide
convincing evidence for these phases [14]. It has been shown that this apparently complex
phase behavior results ftom a competition between the fluid-wall and fluid-fluid
intermolecular interactions. For a given pore geometry and width, the phase diagrams for
a wide range of adsorbates and porous solids can be classified in terms of a parameter a
that is the ratio of the fluid-wall to fluid-fluid attractive interaction t1,97131.
In this paper we report molecular simulation studies of fteezing in slit-shaped pores and
in multi-walled carbon nanotubes. Simulations make use of the Landau-Ginzburg
formalism for the free energy[I3] and parallel tempering"']. Experimental results from
dielectric relaxation spectroscopy are reported for several fluids in these materials.

2. Methods
2.1. Simulation

We performed grand canonical Monte Carlo (GCMC) simulations of Lennard-Jones (LJ)

fluids adsorbed in pores of slit-shaped and cylindrical geometry. The slit-shaped pore of
width H was modeled using the 10-4-3 Steele potential [I6], whereas the cylindrical pore
of diameter D was modeled using the potential due to Peterson et al."71. Systems of up to
64000 molecules were considered. We calculated the Landau free energy as a function of
an effective bond orientational order parameter @, which is sensitive to the degree of
order in the system. The Landau fi-ee energy approach used in earlier s t ~ d i e s [ ~was
*'~~
extended to incorporate spatial inhomogeneity in the order parameter, and a generalized
Landau-Ginzburg approach was developed to determine the Landau free energy surface
of inhomogeneous fluids"31,with the latter given by:

A[@(r)] = -k,T ln(P[@(r)]+ constant

where k~ is the Boltzmann constant, T is the temperature and P[@(r)]is the probability of
observing a system having an order parameter value between @ and @+&D at the position
given by r. The probability distribution function P[@(r)] was determined during the
simulations using umbrella sampling in pores of slit-like geomerry [I3], and parallel
tempering "'I for pores of cylindrical geometry. We found significant ordering into
distinct molecular layers in the systems considered; therefore it is possible to define the
order parameter @(r)using a two dimensional bond orientational order parameter, defined
as tw.

10
where the fmt sum is over the adsorbed molecular layers and ij is the coordinate of the
plane in which molecules in layerj are most likely to lie on. We expect @(r)=l for two-
dimensional hexagonal crystals and @(r)=O for two-dimensional liquids. YGJ(p)measures
the hexagonal bond order at position p = x ex + y ey within each two-dimensional layer j.
Nb is the number of nearest neighbors of a molecule at position and Bk is the orientation
of each nearest neighbor bond with respect to an arbitrary axis E31. For slit-shaped pores,
these two-dimensional layers lie in the xy plane. For cylindrical pores, a quasi-two
dimensional configuration of each molecular layer can be obtained by cutting each one of
the concentric layers along the axial direction and unrolling it flat. The grand fiee energy
of a particular phase is then related to the Landau fkee energy by [13]:

The probability distribution fhction P[@(r)] was determined during the simulations
using umbrella sampling in pores of slit-like geometry [13], and parallel tempering [I5] for
pores of cylindrical geometry I19]. The method of parallel tempering is a Monte Carlo
scheme that has been derived to achieve good sampling of systems that have a fiee energy
landscape with many local minima [*'I. Yan and de Pablo [''I implemented this scheme in
the grand canonical ensemble, performing an MC simulation in n-systems which differ in
both temperature and chemical potential. In addition to the standard MC trial moves, they
proposed configuration swaps in this method. A swap attempt between configurations i
a n d j is accepted with probability given by

where A P l l k J , - l/kBT,, and AU and are the difference in potential energies and
number of particles, respectively, between configurations i and j. We have used as many
as 50 different configurations in our calculations with the parallel tempering Monte Carlo
scheme, to cover all the phase space of interest and to guarantee frequent swaps between
replicas. We have checked the evolution of the parallel tempering simulation as a function
of Monte Carlo steps and after the equilibration, we verified that each configuration
visited many sets of (T,p) along a single simulation run.The nature of the phases in each
molecular layer was determined by measuring the two-dimensional bond orientational
order parameter and by monitoring the two-dimensional, in-plane positional and
orientational pair correlation b c t i o n s within each layer. The positional pair correlation
function in layerj is given by the familiar radial distribution function g@), measured
within the two-dimensional plane formed by each layer j. The orientational pair
correlation function is given by G,,J( p ) = (Y; (0)VJ( p ) ) .

2.2. Experiments: Dielectric Relaration Spectroscopy

The relative permittivity of a medium, K'=Kr-iK,, is in general a complex quantity whose
real part K~ (also known as the dielectric constant) is associated with the increase in
capacitance due to the introduction of a dielectric[211.The imaginary component K, is
associated with mechanisms that contribute to the energy dissipation in the system, due to
viscous damping of the rotational motion of the molecules in alternating fields; this effect
is frequency dependent. The experimental setup consisted of a parallel plate capacitor of
empty capacitance Co.The capacitance C and the tangent loss, tan(b) of the capacitor
filled with the sample were measured at different frequencies and temperatures, using a
Solartron 1260 Gain Analyzer. For the case of CC4 in multi-walled carbon nanotubes, the
sample was introduced between the capacitor plates as a suspension of the adsorbent
(multi-walled nanotubes with average internal diameter of 5 nm, average external
diameter of 10 nm)in the pure adsorbate (CC14). The relative permittivity is related to the
measured quantities by K,=C/C~,~ = t a n ( S y Kwhere
, 6 is the angle by which current leads
the voltage. Further details of the experimental methods are described elsewhere

3. Results
3. I . Activated Carbon Fibers (Slit-Shaped Pores)
In Figure 1 we show the Landau free energies for LJ CC14 confined in activated carbon
fibers of pore width H*=3, at P 3 3 5 K and T=290 K [I3]; pores of this width can just
accommodate two adsorbed layers of CC4. We also show in this figure the grand free
energy of the phases as a h c t i o n of temperature. The free energies show three local
minima. Examination of the spatial and orientational correlation functions show that these
correspond to liquid-like (L), hexatic (H) and crystalline (C) phases. Thus for the liquid-
like phase g(r) and G6,,(r)both showed short range order, whereas for the crystalline
phase they both showed long-range order; for the hexatic phase g(r) showed short range
order, but G6J(r)showed algebraic decay, indicating quasi-long range bond orientational
order. For E-347K and TG90 K, the thermodynamically stable phases are the liquid and
hexagonal crystalline phases, respectively. Between these two temperatures, the hexatic
phase is the thermodynamically stable phase. The existence of the hexatic phases was
verified using the Landau-Ginzburg formalism and system size scaling analysis on very
large system sizes[14]. Experimental measurements of the nonlinear dielectric effect on
carbon tetrachloride (CCL,) and aniline confined within activated carbon fibers showed
divergences at temperatures close to those found in the simulation^"^^, and the
divergences obeyed the scaling laws predicted by the KTHNY theory, thus providing
strong evidence for the hexatic phase.
, 1 6 .

TJK

@)

Figure 1. (a) Landau free energy in the two molecular layers of LJ CCIJ confined in a graphitic, slit-like pore of
width H=3ojj(1.5 nm),at T=335 K and P290 K. (b) Grand free energy as a function of temperature for liquid,
hexatic and crystalline phases, for the same system. From
A corresponding states a n a l y ~ i s ~ ~
shows
" ~ ] that the shift in the freezing temperature, as
well as the temperature of any other surface-driven phase transitions, should depend on
three reduced variables, namely H*=H/c~fi a = ~ , E ~ O & A /and E ~o-j&fi here f and w
refer to confined fluid and wall, respectively, avis the atom density of the pore wall, and
A is the spacing between the atomic layers that make up the pore wall. For relatively

12
small fluid molecules the size ratio parameter is found to have little effect on the freezing
behavior, except in the molecular sieving regime. Thus it is possible to plot global
freezing diagrams in terms of a, the ratio of fluid-wall to fluid-fluid attractive forces.
Such a diagram is shown in Figure 2 pores of width H*=3, obtained from both simulations
and experiments [I3. The diagrams from simulations and experiments are in qualitative
agreement, and demonstrate that a parameter determines the qualitative nature of the
fieezing behavior. Large values of a lead to an increase of the freezing temperature of the
confined fluid and to the formation of hexatic phases. As the fluid-wall interactions
become weaker, a decrease on the freezing temperature is observed and the temperature
range where the hexatic phase is stable decreases. Similar diagrams for larger pores, e.g.
H*=7.5, using molecular simulations [I show the presence of contact layer phases, in
which the layer of adsorbed molecules adjacent to the pore wall has a different structure
from that of those molecules in the interior of the pore. These simulations show that for
simple fluids the role of a is to determine the nature of the fkeezing behavior, while that of
H* is primarily to determine the magnitude of shifts in the transition temperatures.

.(a) (b)

Figure 2. Global phase diagram of a fluid in a slit pore of width H=303 (I .5 nm) from (a) simulations, and (b)
experiments. The experiments are for various adsorbates confined within activated carbon fibers: 1=H20,
2=C&N& 3=c&NH2, kCH@H, 5=CCL, 6 e . From and references therein.

3.2. Carbon Nanotubes

The freezing behavior of LJ CCl, adsorbed in model multi-walled carbon nanotubes of
internal diameter D=5 nm (D*=9.7)was studied using parallel tempering. The formation
of five concentric layers of adsorbate was observed when the confined fluid solidified. In
Figure 3(a) we show the average values of the two-dimensional, bond orientational order
parameter in the individual layers, as a function of temperature. We observe a
discontinuity in the average order parameters of the contact and second layers around 260
K, suggesting a phase transition in these two layers. The change in the order parameter for
the third layer is less abrupt than in the first two layers and starts around 255 K. In the
fourth layer, the order parameter value increases continuously until it jumps around 215
K, reaching a maximum value of a-0.7 at lower temperatures. Our results suggest that
for this specific pore diameter, the contact and first inner layers freeze at temperatures at
or slightly higher than the bulk freezing point, whereas the inner layers experience a
depression in the freezing temperature when compared to that of the bulk. These
observations are corroborated by our results for the positional and orientational pair
correlation functions in the unwrapped layers. Some of our results for the contact layer
are shown in Figure 4(b). At T=262 K, the isotropic positional pair correlation hnction
and the exponential decay in the orientational pair correlation function are signatures of

13
an isotropic liquid. At T=252 K, the features observed in both pair correlation functions
are characteristic of a two-dimensional hexagonal crystal "". Similar features were found
for the other layers.

0.8

0n
0.6

0.4
e
02

(a) (b)
Figure 3. (a) Two-dimensional,bond orientational order parameter average values in the molecular fluid layers
of IJ CCL confined in a multi-walled carbon nanotube of diameter D = 9 . 7 9 (5 nm). Triangles, squares,
diamonds and circles represent the order parameter values for the contact, second, third and fourth layers,
respectively. The dotted line represents the bulk solid-fluid transition temperature. (b) Positional and
orientational pair correlation functions in the unwrapped contact layer of LJ CC4 confined in a multi-walled
carbon nanotube of diameter D = 9 . 7 0 ~ ( run)
5 showing liquid phase at 7'=262 K and crystal phase at T=252 K.

Our results for the D=9.7ufcarbon nanotube did not show any sign of ordering of the
confined LJ CCl, into regular, three-dimensional crystal structures. This finding is in
agreement with previous simulation [lo] and experimental 110*22-241 studies, where it was
shown that for pore diameters below 200- only partial crystallization occurs. The same
studies found that for silica materials, the lower limit below which there are no 3-D
crystal domains in the system was around D=120- 110*22-241. Our simulations for carbon
nanotubes of smaller diameters show that the freezing temperatures increase as the pore
diameter decreases. We are currently performing free energy calculations to determine the
thermodynamic stability of each of the phases, as well as to determine the exact phase
transition temperatures for the confined layers [''I.

In Figure 4 we show the behavior of the experimentally determined capacity C as a

function of temperature for CCl, confined in multi-walled carbon nanotubes with average
internal pore diameter of 5 nm (average external diameter equal to 10 nm). The sample
was introduced between the capacitor plates as a suspension of CC14-filled carbon
nanotubes; however, the volume of bulk C C 4 was very small compared to that of the
confined CCl4, so that the results represent the confined adsorbate. The results indicate
that melting in the amorphous inner layers starts at about 205 K, and proceeds
continuously up to 234 K, where there is a sharp increase in C associated with the melting,
in good agreement with the simulations. The bulk melting point is 250 K. The decrease
in C for temperatures above 259 K indicates the final melting of remaining solid CCl, in
the pore at 259 K, and may correspond to melting of the contact layer(s). The features
found at T=175 K may be associated with a solid-solid transition, since such a transition
(from monoclinic to rhombohedric crystal) occurs in bulk CCl, at 210 K. The melting

14
temperatures found in the experiments are in qualitative agreement with those found in
our simulations.

Figure 4. Dependence of capacity C on T for CCL in multi-walledcahon nanotubes with average pore diameter
of 5 nm (10 nm average external diameter), from dielectric relaxation spectroscopy. Symbols represent results
obtained at different frequencies: circles, 30 kHz; squares, 100 kHz; and triangles, 6OOkHz. The signals are for
both bulk and confined CCL.

4. Conclusions
Our results in slit pores show that the solid-fluid phase behavior for this case depends on
two main parameters, the reduced pore width H and a parameter measuring the ratio of
the fluid-wall to the fluid-fluid (a)attractive potential. An increase of the freezing
temperature of the confined fluid relative to the bulk is observed for large values of a,
whereas a decrease is observed for small values of a. We found evidence of new surface-
driven confined phases, such as contact layer and hexatic phases. The hexatic phase
was found in slit-shaped pores with larger a values, and it is stable over extended
temperature ranges. Phase diagrams of the transition temperature vs a were depicted for
several LJ fluids confined in activated carbons H*=3. These phase diagrams are in
qualitative agreement with experimental results. For cylindrical pores, our results for a
D=9.7oS multi-walled carbon nanotube show no formation of regular 3D crystalline
structures, in agreement with experimental results. Our results also suggest that the outer
layers experience an increase in the freezing temperature, while the inner layers
experience a depression in the freezing temperature with respect to the bulk freezing
point. Simulations for carbon nanotubes of smaller diameters show that the freezing
temperatures increase as the pore diameter decreases. Preliminary experimental results
show solid-fluid transition temperatures for CCl, in carbon nanotubes that are in
qualitative agreement with those determined in our simulations.
This work was supported by a grant from the National Science Foundation (Grant No.
CTS-9908535). Supercomputer time was provided under a NSF/NRAC grant
(MCA93S011).

15
[I1 For a review to mid-1999 see: L.D. Gelb, K.E. Gubbins, R. Radhakrishnan and M. Sliwinska-
Bartkowiak, Phase Separation in Confined Systems, Rep. Prog. Phys. 62 (1999) pp. 1573-1659
I[' M. Miyahara and K.E. Gubbins, FreezingMelting Phenomena for Lennard-Jones Methane in
Slit Pores: a Monte Carlo Study, J. Chem. Phys. 106 (1997) pp. 2865-2880
r31 R. Radhakrishnan and K.E. Gubbins, Free Energy Studies of Freezing in Slit Pores: An Order-
Parameter Approach Using Monte Carlo Simulation, Mol. Phys. % (1999) pp. 1249-1267
r41 M. Sliwinska-Bartkowiak, J. Gras, R. Sikorski, R. Radhakrishnan, L.D. Gelb and K.E. Gubbins,

Phase Transitions in Pores: Experimental and Simulation Studies of Melting and Freezing,
Langmuir 15 (1999) pp. 6060-6069
['I H. Dominguez, M.P. Allen and R. Evans, Monte Carlo Studies of the Freezing and Condensation
Transitions of Confined Fluids, Mol. Phys. 96 ( 1998) pp. 209-229
[61 K. Kaneko, A. Watanabe, T. Iiyama, R. Radhakrishnan and K.E. Gubbins, A Remarkable Elevation
of Freezing Temperatureof CCL, in Graphitic Micropores, J. Phys. Chem. B, 103 (1999) pp. 7061-
7063
[71 A. Watanabe and K. Kaneko, Melting Temperature Elevation of Benzene Confined in Graphitic
Micropore", Chem. Phys. Lett. 305 (1999) pp. 71-74
[*I R. Radhakrishnan, K.E. Gubbins, A. Watanabe and K. Kaneko, Freezing of Simple Fluids in
MicroporousActivated Carbon Fibers: Comparison of Simulation and Experiment,J. Chem. Phys.
111 ( 1999) pp. 9058-9067
l9] R. Radhakrishnan, K.E. Gubbins and M. Sliwinska-Bartkowiak, Effect of the Fluid-Wall
Interaction on Freezing of Confined Fluids: Towards the Development of a Global Phase Diagram,
J. Chem. Phys. 112 (2000) pp. 11048-1 1057
['I M. Sliwinska-Bartkowiak,G. Dudziak, R Sikorski, R. Gras,R. Radhakrishnan and K.E. Gubbins,
MeltingRreezing Behavior of a Fluid Confined in Porous Glasses and MCM-41: Dielectric
Spectroscopyand Molecular Simulation, J. Chem.Phys. 114 (2001) pp. 950-962
[ I i 1 M. Sliwinska-Bartkowiak,G. Dudziak, R. Sikorski, R. Gras,K.E. Gubbins and R. Radhakrishnan,
Dielectric Studies of Freezing Behavior in Porous Materials: Water and Methanol in Activated
Carbon Fibers, Phys. Chem. Chem. Phys. 3 (2001) pp. 1 179-1184
[*'] M. Sliwinska-Bartkowiak,R. Radhakrishnan and K.E. Gubbins, Effect of Confinement on Melting
in Slit-Shaped Pores: Experimental and Simulation Study of Aniline in Activated Carbon Fibers,
Mol. Sim. 27 (2001) pp. 323-3337
["I R .Radhakrishnan, K.E. Gubbins and M. Sliwinska-Bartkowiak, Global Phase Diagrams for
Freezing in Porous Media, J. Chem. Phys. 116 (2002) pp. 1147-1155
[14] R. Radhakrishnan, K.E. Gubbins and M. Sliwinska-Bartkowiak, On the Existence of a Hexatic
Phase in Confined Systems, Phys. Rev. Lett. 89 (2002) art. 076101
["I Q. Yan and J.J. de Pablo, Hyper-Parallel Tempering Monte Carlo: Application to the Lennard-
Jones Fluid and the Restricted Primitive Model, J. Chem. Phys. 111 (1999) pp. 9509-95 16
[I6] W.A. Steele, Physical Interaction of Gases with Crystalline Solids: 1. Gas-Solid Energies and
Properties of Isolated Adsorbed Atoms, Sut$ Sci. 36 (1973) pp. 3 17-352
[I7] B.K. Peterson, J.P.R.B. Walton and K.E. Gubbins, Fluid Behavior in Narrow Pores, J. Chem.
Soc., Faraday Trans. 2 82 (1986) pp. 1789-1800
[lS1 R.M. Lynden-Bell, J.S. van Duijneveldt and D. Frenkel, Free-Energy Changes on Freezing and
Melting in Ductile Metals, Mol. Phys. 80 (1993) pp. 801-814
[I9] F.R. Hung, M. Sliwinska-Bartkowiakand K.E. Gubbins, in preparation (2002).
I'[ D. Frenkel and B. Smit, Understanding Molecular Simulation: From Algorithms to Applications
2"* Edition (Academic Press, London, 2002)
["I A. Chelkowski, Dielectric Physics (Elsevier, North-Holland, New York 1980)
I'[ D. Morineau, G. Dosseh, C. Alba-Simionesco and P. Llewellyn, Glass Transition, Freezing and
Melting of Liquids Confined in the Mesoporous Silicate MCM-41,
Philos. Mag. B 79 (1999) pp. 1847-1855
["I G. Dosseh, D. Morineau and C. Alba-Simionesco, Benzene Confined in MCM-41 Below its
Melting Point: A Proton NMR Study, J. Phys. IV 10 (2000)pp. 99-102
[241 D. Morineau, F. Casas, C. Alba-Simionesco, A. Grosman, M.-C. Bellisent-Funel and N.
RatovClomanana, A Neutron Scattering Investigation of the Structural Properties of
Glassforming M-Toluidine Confined in MCM-41, J. Phys. IV 10 (2000) pp. 95-98

16
 MEASUREMENT OF DIFFUSION IN MICROPOROUS SOLIDS

DOUGLAS M. RUTHVEN
Department of Chemical and Biological Engineering, University of Maine, Orono, ME 04469
USA
E-mail: druthven!dumche.maine.edu

The experimental methods for measurement of transport and self-diffision in zeolite crystals (and in
other microporous materials) are reviewed. Large discrepancies between different techniques are
commonly observed and appear to be related to the scale of the measurements, suggesting that
structural defects may be more important than is generally believed.

1 Introduction

An understanding of diffision in microporous solids is essential for the rational design

and optimization of catalytic processes, adsorptive separation processes and inorganic
membranes yet, despite intensive study, our understanding is still far from complete. A
wide range of different experimental techniques have been developed but, for many
systems, the difisivities determined by different methods show discrepancies of more
than an order of magnitude, making it difficult to establish the true value. This paper
presents a short overview of the major experimental approaches with a more detailed
discussion of one particular technique (the zero length column or ZLC method) which
was developed several years ago in our laboratory [ l , 21 and has been widely applied to
study many different systems.

2 Definition of Diffusivities

Molecular transport and tracer diffision in porous solids are conveniently correlated in
terms of difisivities defined in accordance with Ficks first equation:

J = -D(q). aq- .
9 J* = - D * ( q ) . Y l
az az 4

The self-difisivity defined on the basis of the random walk representation of a diffusive
process:
~

2n t 2n r

is identical to the tracer diffbsivity. The relationship between the transport and tracer
diffusivity is more complex.
Considering the chemical potential gradient rather than the concentration gradient as
the fundamental driving force leads to:

17
where Do = BRT is the corrected or intrinsic transport diffusivity. Do and D
coincide only in the Henrys Law region where d lnp/d Inq = 1.O. For a type 1 isotherm
Eq. 3 implies a strongly increasing trend of di hi vity with loading. Eq. 3 is often
attributed to Darken [3] but this formulation actually goes back to the much earlier work
of Maxwell and Stefan [4, 51. Krishna and coworkers have applied the Maxwell-Stefan
approach to micropore diffusion by considering the unoccupied sites (vacancies) as the
n+l component in the system [6,7]:

where D:. measures the rate at which components i and j can exchange directly and
0; measures the rate of migration to vacant sites, the velocity of which (u,+,) is zero.
For single component diffusion (e,=O, N,=O) this reduces to:

J=--.-.-
DV q dp
(1-8) RT dz
1 = D 1A which yields [8]:
and for self diffision we have J A = - J g and DAB

This formulation suggests that at low loadings ( b 0 ) the corrected and tracer diffisivities
should coincide but at higher loadings D < Do except when D is very large.
These formulations are applicable only to normal systems in which diffusing
molecules can pass each other in the channels. When molecules cannot pass we have
single file or string of pearls diffusion which has entirely different features. This topic
has been well reviewed by Kiirger [9] and will not be considered here.

3 Experimental Methods

A summary of the major experimental techniques that have been applied to study
diffusion in microporous solids is given in Table 1. These can be classified according to
the nature of the measurement (transient or steady state), the scale of the measurement
(micro, meso or macro) and the type of process (transport or self-diffision). Additional
details with complete references are given in recent reviews [lo-121.
Most of the microscopic techniques are restricted to relatively small and rapidly
diffising species since a measurable mean molecular displacement must be achieved
within a short time. In general the microscopic techniques show reasonable consistency
with other micro techniques but, as the length scale of the measurement increases, the
apparent diffisivities decrease and show increasing discrepancies between different
techniques.

18
 TABLE 1

Classification of Experimental Techniques for Diffusion Measurements

Transport Diffusion Self-Diffusion

Transient Steady State

Macroscopic Sorption Rate (12) Membrane Tracer methods (12)

Methods IR spectroscopy (16, 17) permeation (14, 15)
Frequency response (20) Effectiveness factor
Chromatography (12) (18919)
ZLC (1 3)
Differential adsorption
bed (21)
TAP reactor (22)
Imaging (23)

Mesoscopic IR microscopy (16, 17) Single crystal Single crystal

Methoa3 membrane (14, 15) membraneltracer

Microscopic Interference microscopy PFG NMR (12)

Methods (24325) NMR-lifetime (12)
Coherent QENS (26) Incoherent QENS

4 The ZLC Technique

The ZLC technique [ 131 depends on measuring the desorption curve for a small sample of
adsorbent pre-equilibrated with sorbate under well defined conditions. The essentials of
the experimental system are shown in figure 1. The column consists of a very small
sample of adsorbent held between two suiter discs in a Swagelock fitting. The
desorption curve is determined by both the adsorption equilibrium and the kinetics but
these two effects can be easily separated by making measurements over a range of flow
rates. In the low flow rate regime intraparticle diffusion is relatively rapid. The
gradient of sorbate concentration through an adsorbent particle is very small and the
desorption rate is determined by convection (the rate of removal of sorbate fiom the
adsorbent surface). In contrast, at high flow rates, the concentration at the external
surface of the particle is always very small and the desorption rate is controlled by
diffusion out of the particle. Thus, by making measurements over a sufficiently wide
range of flow rates both the equilibrium and kinetic parameters may be determined.
This approach is illustrated in fig. 2 which shows a family of ZLC desorption curves
for COz fiom a sample of a carbon monolith, measured over a wide range of flow rates.
The isotherm is linear so at low flow rates (equilibrium control) the response is given by:

19
 r 1

when plotted against the volume of purge (Ft) the response should be invariant with flow
rate and linear in coordinates of In (c/co) vs Ft. The slope yields the sum V,+KV,. For
strongly adsorbed species the dead volume (V3 is usually negligible but for weakly held
species (as in this example) the correction may be important. V, is conveniently
determined from a blank experiment with no adsorbent present.
The general response (treating the adsorbent as a parallel sided disk of thickness 21)
is given by: ,
-c- - a 2 L e x p - P 2I
c
co n=l L + p2(1+ y ) + (L - y#)2

where Pn is given by the roots of

2
L - YPn - Pn tan Pn = 0 (9)

and y = VdKV,. The simplification for y + 0 is obvious. Varying the flow rate
changes L and thus provides a simple experimental test for consistency with the
isothermal d i m i o n model. The corresponding expression for spherical particles is also
available [131.
The ZLC method offers advantages of speed and simplicity and requires only a very
small adsorbent sample thus making it useful for characterization of new materials. The
basic experiment using an inert carrier (usually He) measures the limiting transport
diffusivity (Do)at low concentration. A variant of the technique using isotopically
labeled tracers (TZLC) yields the tracer diffisivity and counter diffision in a binary
system may also be studied by this method. To obtain reliable results a number of
preliminary experiments are needed, e.g. varying sample quality, nature of the purge gas,
the flow rate and, if possible, particle size to c o n f m intracrystalline diffiision control.

5 Comparison of ZLCPFG N M R Diffusivities

PFG NMR measurements can be applied over a wide range of diffusivities and the scale
of the measurement can be varied from a few unit cells to perhaps half the crystal or
particle diameter. For a number of systems good agreement between ZLC and PFG
NMR measurements has been demonstrated; examples are shown in fig. 3. The data for
methanol-NaX are particularly interesting as the unusual concentration trend is revealed
by both sets of data and at low concentrations the tracer diffusivities converge to the
limiting value of Do as is to be expected. Unfortunately such agreement is far 6om
universal. Figure 4 shows data for some systems for which large discrepancies are
found, both in absolute values of the diffusivities and in the concentration trends.

6 Comparison of ZLCWrequency Response Diffusivities

Diffusion of aromatic hydrocarbons in silicalite has been widely studied by several

different methods although these systems are not amenable to NMR measurements

20
because of their short relaxation times and relatively low diffisivities. ZLC and
frequency response (FR) diffusivities for benzene are compared in Fig. 5. The data
show reasonable agreement as to the magnitude of the diffisivities but when trends with
loading are examined some striking discrepancies emerge. For both benzene and p-
xylene the TZLC data show the self-diffisivity to be almost independent of loading
whereas the FR data suggest that Dodecreases quite strongly. For benzene the ZLC data
show that tracer and corrected transport diffisivities are almost the same but for p-xylene
the transport diffisivity is smaller. For both species diffusion in ZSMS (A1 rich) is
slower than in silicalite, presumably as a result of blocking by the strong A1 sites.
The diffusional behavior of p-xylene is complicated. The FR measurements reveal
two different diffusivities corresponding to movement through the straight and sinusoidal
channels. The ZLC method increases only the average diffisivity which is similar to the
value for benzene but it is possible that the difference between the self and transport
diffisivity results from the two channel behavior revealed by the FR data [33].

7 Variation of Diffusivity with Chain Length of Linear Alkanes

The extensive diffusivity data for linear alkanes in silicalite have been recently reviewed
by Talu et a]. [151. The values obtained by several different techniques are shown in fig.
6. It is clear that for these species the discrepancies between different measurements
amount to several orders of magnitude. We have measured diffisivities for a series of
linear alkanes in several different zeolites by the ZLC technique. In all cases we see a
continuous but slow decline at higher carbon numbers with no evidence of local maxima
as shown by the membrane data. For offi-etite-erioniteour results agree well with those
of Maghalaes et al. [34] and show no evidence of the window effect originally
suggested by Gorring [35]. It should be noted that the membrane technique measures
diffusion along the long axis of the crystal whereas other methods measure an average
diffisivity. However the long axis diffisivity difisivity is expected to be lower so the
difference between membrane, FR and ZLC data is puzzling.

8 Concluding Remarks

The diffisivities predicted from molecular simulation are generally too high. For small
rapidly diffusing species the values are comparable with the highest experimental values
but for slower (larger) species the simulation values are higher than even the largest of the
experimental values. For example, it is clear from fig. 6 that the simulation results do
not capture the decreasing trend of difiivity with carbon number which is observed by
all experimental techniques.
In general the shorter range measurements yield higher diffisivity values than the
longer range measurements and this difference becomes greater for larger molecules.
The different microscopic methods commonly show fair agreement but differences
between different macroscopic measurements are often very large.
In a recent experimental study of the adsorption of methanol in a large crystal of
CrAPO by interference microscopy, Lehmann et al. [36] observed that, even at
equilibrium, the distribution of sorbate through the crystal is far from uniform. It seems
clear that access is controlled largely by the defect structure and the growth planes of the
crystal. This observation may provide a plausible explanation for the discrepancies
observed between different diffusion measurements. The impact of the defect structure

21
is minimal over short distances (a few unit cells) but becomes important as the length
scale of the measurement is increased. Furthermore, one might expect that the effect of
defects will be more severe for larger sorbates and the defect structure may differ
substantially between different samples, thus explaining the discrepancies between
different microscopic measurements.

Notation

B intrinsic mobility D' see Eq. 7

c concentration (gas or fluid phase) F purge flow rate
c, initial value of c J flux (relative to adsorbent framework)
D difisivity (transport) K Henry's Law constant (dimensionless)
Do corrected diffusivity L F12/KV,D - dimensionless parameter
D tracer or self-diffusivity 1 half thickness of adsorbed slab
n dimensionality of pore system u velocity in Maxwell Stefan
expression (Eq. 4)
p partial pressure of sorbate V,,V, Volume of gas (dead volume),
volume of solid
q adsorbed phase concentration z distance
qs
- saturation limit P parameter in Eq. 8
r mean square displacement Y V&V,
R gasconstant 8 q/qs hctional loading
t time h step length (random walk)
T temperature (K) T dimensionless time M/l2 Eq. 8;
time between molecular jumps Eq. 2

References

1. M. Eic and D.M. Ruthven, Zeolites 8,4045 (1988).

2. D.M. Ruthven and M. Eic, Amer. Chem. SOC.Symp. Ser. 368,361-375 (1988).
3. L.S.Darken, Trans. A.I.M.E. 175,184(1948).
4. J.C. Maxwell, Phil. Trans. Roy SOC.m , 4 9 - 7 9 (1866).
5. J. Stefan, Bet. Akad. Wissen. Wien 65,323-363 (1872).
6. R. Krishna, Chem. Eng. Sci. a,
1779 (1990).
7. R. Krishna, Chem. Eng. Sci. 48,845 (1993).
8. D. Paschek and R. Krishna, Chem. Phys. Letters, 333,278-284.
9. J. Ktirger, Single File Diffusion, Ch. 7 in Molecular Sieves: Science and
Technology, H.G. Karge and J. Weitkamp eds., Wiley-VCH (2002).
10. D.M. Ruthven and M. Post, Diffusion in Zeolite Molecular Sieves, Ch 12 in
Introduction to Zeolite Science and Practice, H. van Bekkum, E.M. Flanigen, P.A.
Jacobs and J.C. Jansen eds. Elsevier, Amsterdam (2001).
11. J. Ktirger and D.M. Ruthven, Diffusion and Adsorption in Porous Solids, Ch. 5 in
Molecular Sieves: Science and Technology, H.G. Karge and J. Weitkamp eds.,
Wiley-VCH (2002).

22
12. J. Kager and D.M. Ruthven, Di#&on in Zeolites and other Microporous Solids,
John Wiley, New York (1992).
13. D.M. Ruthven and S. Brandani in Recent A&. In Gas Sep. by Microporous Ceramic
Membranes,N.K. Kanellapoulos ed., p. 187-212 Elsevier, Amsterdam (2000).
14. D.T. Hayhurst and A. Paravar, Proc. Sixth Internat. Zeolite Con$, p. 217-224,Reno
(1 983)D. Olson and A. Bisio eds. Butterworth, Guildford (1 984).
15. O.Talu,M.S.SunandD.B.Shah,A.I.ChEJl.~,681-694(1998).
16. W. Niessen and H.G. Karge, Microporous Materials L, 1-8(1993).
17. R. Schurnacher and H.G. Karge, Microporous Materials 1,1-8(1993).
18. W.O. Haag, R.M. Lago and P.B.Weisz, Faraday Discussions Chem. SOC. 72, 317-
330 (1982).
19. M.F.M. Post, J. Amstel and H.W. Kouwenhoeven, Proc. Sixrh Internat. Zeolite Con$
p. 5 17-527,Reno, (1983)D. Olsson and A. Bisio eds. Butterworth, Guildford (1 984).
20. L.V.C. Rees and L. Song in Recent Advances in Gas Separationby Microporous
Ceramic Membranes p. 139-186, N.K. Kanellopoulos ed., Elsevier, Amsterdam
(2000).
21. D.D. Do, X. Xu and P.L.J. Mayfield, Gas. Sep. PuriJ 5,35-48(1991).
22. J.R. Ebner and J.T. Gleaves, U.S. Patent 4,626,412(1986).
23. J.T. Gleaves, J. R. Ebner and T.C. Kuechler, Catol. Revs. Sci. Engg. 30, 49-116
(1988).
24. U. Schemmert, J. Kiirger, C. Krause, R.A. Rakoczy and J. Weitkarnp, Europhys.
Letters 46,204-210 (1 999).
25. U. Schernrnert, J. K&ger, J. Weitkarnp, Microporous and Mesoporous Materials 32,
101-110 (1999).
26. H. Jobic in Recent Advances in Gas Separation by Microporous CeramicMembranes
p. 109-137 N.K. Kanellopoulos ed. Elsevier, Amsterdam (2000).
27. D.M. Ruthven, Zeolites 13,594(1993).
28. S. Brandani, D.M. Ruthven and J. Karger, Zeolites l5,494-496(1995).
29. S.Brandani, J.R. Hufion and D.M.Ruthven, Zeolites l5,624-631 (1995).
30. S. Brandani, Z. Xu and D.M. Ruthven, Microporous Materials 3, 2791-2798
(1 998).
3 1. S.Brandani, M. Jama and D.M. Ruthven, Microporous and Mesoporous Mats. 35-36
283-300(2000).
32. D.M. Ruthven, M. Eic and E. Richard, Zeolites ll,647-653(1991).
33. L. Song and L.V.C. Rees, Microporous and Mesoporous Materials 35-36 301-314
(2000).
34. F.D. Magalhaes, R.L. Lawrence and C.W. Conner, AIChE Jl. a,68-86(1996).
35. R.L. Gorring, J. C a t a l s , 13 (1973).
36. E. Lehmann, C. Chmelik, H. Scheidt, S. Vasenkov, B. Staudte, J. Kgrger, F. Kuemer,
G. Zadvozna and J. Kornatowski, J. Am. Chem. SOC.- in press.

23
 c

Fig. 1. Schematic diagram of system for vapor phase ZLC or tracer ZLC
measurements.

0.1

0
0
\
0

0.01

0 2 4 6 8 10 12

t (9
Fig. 2. ZLC response curves for C G h m a piece of carbon monolith at 303K. The
four curves obtained at flow rates h m 11.4 Mmin &=l) to I 0 4 d m i n
(L=9.5) are fitted with the m e parameters: D/12 = 0.92 sec-. KV,= 0.24 ml,
Vf-0.14 ml.

24
:eIr

2-
I.,
1i

Fig. 3 Comparison of ZLC and PFGHMR diffosiviticsfor (a) C& and k

(Do and al (271 and (b) CHjOH-NaX (61[ZS].

n
I

m
Y.

Fig. 4 (a) Comparisonof PFG hMR data (325K)and mcer ZLC &fa
(358IQ for propane and propene inNaX oeoiite crystals [29].
(b) Cmpakon of PFG W R , FR and mer ZLC difbivities for
bepane-wax at 468K (201.

25
 1 E-07

Benzene

u) l.E-08 .
*-
E x +- F R a . 1 1331
0- D = S . 2 6 ~lo- c--~
b
e 1.E-09 .

0 NZLCTrmsD.
0 TZLCTracera
A NZLC Richud De(1991)
1.E-10
1.9 2.1 2.3 25 2.7 2.9 3.1 33
4ooofr
Fig. 5 krhenius plot showing comparison between ZLC and FR
diffusivitits for bmzene in silicalite. Data from 116 and 31-331

D or o0 i m s-

10%

109
1o-O
lo-
10-

10-3

0 2 4 6 8 10 12 14 16 18
Number of carbon atoms

Fig. 6 Variation of diffisivity with carbon number for n-paraffins in

silicalite at 300K showing comparison between data obtained by
different techniques: (0)M.D.simulations, (e) hierarchical
simulation, (+)
QENS, (V)skgie crystal membrane. (A)PFG
NMR,(A)ZLC [15,26J.

26
 ORDERED MESOPOROUS CARBONS WITH NEW OPPORTUNITIES
FOR ADSORPTION STUDIES

R.RYOO AND S . H.JOO

Centerfor Functional Nanomaterials, Department of Chemistry (School of Molecular Science-
BK21),Korea Advanced Institute of Science and Technology,Daejeon. 305 -701, Republic of
Korea.
E-mail:[email protected]

Mesoporous carbons with various structures were available in recent years, due to the development
of the synthesis route using mesoporous silica templates. The synthesis of these mesoporous carbons,
designated as CMK-1-5, was performed by carbonization of sucrose, furfury1 alcohol or other
suitable carbon sources inside the template pore systems. The mesoporous materials were obtained
in a high-purity carbon state after the silica or aluminosilicate templates were removed with NaOH
or HF solution. The structures of these carbons, e.g., hexagonal p6m. cubic Pm3n and cubic 141/u,
correspond to faithful replication of the silica mesopores to carbon frameworks. The ordered
arrangement of the nano-structured carbon frameworks imparts the carbon materials with uniform
mesoporosity. The X-ray diffraction patterns of these carbon materials exhibit several Bragg
diffractions at small angles, similar to those for the MCM-41-type silica templates. The pore
diameters are uniform and tailorable, typically, in the range between 2 and 10 nm. These carbons
show many new possibilities for applications to adsorbents, catalysts and electrode materials.
Particularly for adsorption studies, it is expected that the well-defined mesoporous structures would
be also useful as a reference pore system for the development of characterization methods and
theoretical modeling for the adsorption on carbon surfaces.

1 INTRODUCTION

The microporous (pore diameters less than 2 nm) and mesoporous (2-50 nm) carbons are
widely used in many areas of modem science and technology, including water and air
purification, gas separation, catalysts preparation, chromatography, fuel storage, and
manufacturing of electrochemical devices [ 1,2]. The wide-spread use results from their
high specific surface areas, large pore volumes, ability for hosting catalytic components
(e.g., metal particles), chemical inertness, good mechanical stability, high electrical
conductivity, and excellent affinity with organic species. The technological importance
has led to many scientific researches on the adsorption of guest species, their diffusion
into the pore systems, and the interaction with the carbon frameworks. Many works on
experimental measurements and theoretical modeling are still being actively performed in
order to characterize accurately the adsorption phenomena as a function of the pore
diameters, shapes and pore-wall structures.
For the accurate characterization of the adsorption phenomena, it is necessary to
obtain accurate information on pore structures. However, most of ordinary microporous
carbons and mesoporous carbons are obtained with amorphous structures that are
characterized by irregular arrangements of non-uniform pores. X-ray (or electron)
diffraction (XRD) techniques are not useful for such carbons because there are no well-
defined structural factors to correlate with the adsorption behavior. Moreover, porous
carbons exhibit wide varieties of the surface functional groups and the thickness of the
pore walls, depending on the details of the synthesis conditions. The lack of distinct XRD
lines makes it difficult to distinguish structural differences between samples which causes
many works to depend empiricaIly on specific samples.

27
 Recently, we have discovered a synthesis route to highly ordered mesoporous carbon
molecular sieves using mesoporous silica templates with various structures such as MCM-
48 (bicontinuous cubic Zu34, SBA-1 (cage-type cubic Pm3n) and SBA-15 (channels
arranged in a hexagonal p6m structure) [3-71.The mesoporous templates are synthesized
via the synthesis route using surfactants [8-lo]. The carbon synthesis procedure consists
of the impregnation of an organic carbon precursor such as sucrose, furfury1 alcohol or
acetylene gas into template pores, subsequent pyrolysis of the precursor and removal of
the template frameworks. The carbon materials designated as CMK-n are released after
the removal of the silica template with NaOH or HF. CMK stands for carbon
mesostructured by KAIST. Several XRD lines appear below 10 due to the mesoscale
structural regularity corresponding to that of the silica templates.
The synthesis of the CMK-n carbons is controlled to various pore shapes,
-
connectivity, diameters (typically, 1 10 nm in diameter) and pore wall thickness. These
carbons exhibit high specific surface areas (typically, the BET specific surface areas up to
2000 m2g-),uniform pore diameters, large adsorption capacities, and high thermal, acid-
base and mechanical stabilities. The CMK-type carbons are also suitable for the formation
of well-defined nanocomposite with organic polymers, so that the nanopore walls can be
modified with various functional groups. These carbons show new possibilities for various
applications in adsorption, catalysis and electrochemistry.
Particularly for adsorption studies, it is expected that the well-defined mesoporous
structures would be useful as a reference pore system for the development of
characterization methods and theoretical modeling for the adsorption in carbon pores. In
this regard, we briefly review on the synthesis strategy, structure characterization, and
their perspectives.

2 SYNTHESIS METHOD

The principle of the carbon synthesis is shown in Fig. 1. Suitable carbon sources such as
sucrose, furfuryl alcohol, phenol-resin monomers and acetylene gas are converted to
carbon frameworks inside mesoporous silica template by pyrolysis. An effective method
for the restriction of carbonization to inside the template is to incorporate a suitable
catalyst such as Al, Sn and Fe onto the silica pore walls prior to the use as template. The
template after the carbonization is removed using ethanol-water solution of HF or NaOH.
As shown in Fig. 1, rod- or tube-type carbons are obtained depending on the synthesis
conditions. Rod-type carbons are prepared if cross-linkable carbon precursors such as
sucrose and phenol resin are carbonized after the template pores are filled with the carbon
sources [3-51. Carbon deposition outside the template can be prevented under the present
carbonization conditions using catalyst. Moreover, the carbon formation can be controlled
to occur uniformly throughout the entire volume of the template pore system. Normally,
the conversion of the organic compounds leads to a significant decrease in volume, and
this is accommodated by the generation of micropores in the carbon nano-frameworks.
The micropore volume depends on the source of carbon and the details of carbonization
conditions such as vacuum, nitrogen flow and heating rate.

28
 Figure 1. Schematic representation of the templated synthesis route using mesoprous silicas.

Tube-type carbons are obtained when organic compounds are carbonized in a thin-
film state on the template pore walls 161. The tube-type carbon can be obtained even after
the entire volume of pores is filled with carbon source, if the excess carbon source is
removed before the carbonization is completed. For example, cylindrical pores are
generated along the center of the carbon frameworks due to the systematic volume
decrease when furfuryl alcohol is pyrolyzed under vacuum after the initial polymerization.
Alternatively, the tube-type carbons can be synthesized as follows: carbonization can be
controlled to occur partially by catalyst at the pore walls at moderate temperatures. The
remaining carbon source is removed by evacuation, and the carbonization is completed by
pyrolysis at high temperature. Chemical vapor deposition on the pore walls can also be
used to produce the tube-type carbon [ 111 as well as the aforementioned rod-type carbons
[12]. The structure of the resultant carbon depends on the thickness of the carbon
deposition.

3 STRUCTURES AND NITROGEN ADSORPTION PROPERTIES

3.1 CMK-I
CMK-1carbon was the first carbon material reported to exhibit well-resolved XRD lines
characteristic of ordered arrays of carbon mesopores [3]. The synthesis of the carbon was
achieved by carbonization of sucrose inside the MCM-48 mesoporous silica. As shown in
Fig. 2, the XRD pattern exhibits a new diffraction line around 1.4. compared with its
MCM-48 template. This change can be explained by the formation of two separate carbon
networks in the bicontinuously mesoporous MCM-48 template. After the separating silica
frameworks are removed, the two carbon networks join together. The joining of the two
carbon networks attributes to the symmetry change from cubic Zu3d to either 14,h or
lower [12]. The new ordered mesoporous structure is indicated by the XRD pattern and
transmission electron microscopic image shown in Fig. 2.

29
 2 4 6 8 1 0
XI (deg-1
Figure 2. (a) XRD patterns for MCM-48 silica template and the CMK-1 carbon synthesized using the MCM-48
template. (b) Transmission elechon microscope image of CMK-I . (c) Scanning electron microscope image of
CMK-I .

The synthesis of carbon with MCM-48 was also reported by another research group
using phenol-formaldehyde resin, following the report on CMK-1 [ 131. However, this
carbon exhibited the same XRD pattern as CMK- 1.
As shown in Fig. 3, nitrogen adsorption isotherms of CMK-1 feature well-pronounced
capillary condensation steps similar to those of ordered mesoporous silicas and indicative
of high degree of mesopore size uniformity. The isotherms reveal that the CMK-1 carbon
has high nitrogen BET specific surface area (1500-1800 m2 g-I), and large total pore
volume (0.9-1.2 cm3 g-') [14]. The adsorption capacity is comparable or larger than that of
MCM-48 template. The pore-size analysis (calibrated BJH analysis) shows that typical
CMK-1 has uniform mesopores about 3 nm in size, which is accompanied by a certain
amount of micropores when sucrose is used as the carbon source.

30
 0.0 0.2 44 0.6 48 1.0
Relativepressure

Figure 3. Nitrogen adsorption isotherm for CMK-1. (inset) Correspondingpore size distribution obtained from
adsorption branch by calibrated BJH method [14].

3.2 Other Rod-Type CMK Carbons

Other rod-type carbons with various structures have been synthesized with the SBA- 1,
SBA-15 and SBA-16 (large cage-type, cubic Im3m) mesoporous silica templates. These
carbons (designated as CMK-2 [4], CMK-3 [5] and CMK-6 [15, 161, respectively) exhibit
very similar XRD patterns to those for their silica templates, as shown in Fig. 4. This
result indicates that the structures are maintained in the same space groups during the
synthesis of the carbons from silicas, unlike the case of the CMK-1 synthesis. It is
therefore reasonable that these silica templates are composed of 3-dimensional (3-D)
mesoporous networks of the same continuity. The carbon synthesis within the mesoporous
networks of the same continuity gives the CMK-2, CMK-3 and CMK-6 mesoporous
carbons corresponding to faithful replication of the template pore systems.
It is also noteworthy that, despite the apparently same 2-D hexagonal structures, the
SBA-15 and MCM-41 silicas are distinguished by their markedly different pore
connectivity in addition to the difference in the pore diameters. The MCM-41 silica has 1-
D channels that are not interconnected. However, the large 1-D mesoporous channels
(typically, 9 nm in diameter) of the SBA-15 silica are interconnected through the so-called
complementary pores, which are around or less than 3.5 nm in diameters, and randomly
located perpendicular to the 1-D channels. Because of the 3-D channel structure, the
structure of the SBA-15 silica can be converted to the negative carbon replica exhibiting
the same kind of structural symmetry. Accordingly, the structure of the CMK-3 carbon is
composed of a hexagonal arrangement of 1-D carbon rods as shown by the structural
model in Fig. 4(c). On the other hand, replication of the MCM-41 silica with carbon
results in the formation of carbon fibers that do not retain the 2-D hexagonal arrangement
r141.

31
Figure 4. XRD patterns for rod-type carbons: (a) CMK-2, (b) CMK-3. (c) Structural model for CMK-3.

3.3 Tube-Type Carbons

CMK-5 is the first example of the ordered tube-type mesoporous carbons that can be
characterized with well-defined Bragg diffractions by ordinary XRD instrument [6]. The
XRD pattern of the CMK-5 carbon is distinguished from that of CMK-3 by the much
lower intensity of the (100) diffraction. The structure of CMK-5 may be described by the
substitution of the carbon nanorods in CMK-3 with nanopipes. The CMK-5 carbon is
synthesized using SBA-15, similar to CMK-3, but the carbon source and synthesis
condition are somewhat different from those for CMK-3. The synthesis method for the
tube-type carbon can be extended to the SBA-16 mesoporous template. The resultant
CMK-7 carbon has a bicontinuous mesoporous structure [ 151.
It is noteworthy that the pore-size distribution curve obtained by the N2 adsorption
has exhibited two sharp peaks with the maxima corresponding to the inside diameter of
the carbon nanopipes (typically, 5.5 nm) and the pores formed between the adjacent pipes
(4.2 nm), respectively [6]. It is reported that the outside diameter of the nanopipes is
tailored by the pore diameter of the template SBA-15, while the wall thickness of the
carbon nanopipes are also controllable to a certain degree. The specific BET surface area
of the CMK-5 varies from 1500 to 2200 m2g-'depending on wall thickness [ 171.

4 PERSPECTIVES

The methods developed for the synthesis of ordered mesoporous carbons are simple and
cost-efficient, and the pore size can be tailored. The synthesis process can be scaled up for
production in bulk quantities. Recent works on the synthesis of mesoporous silicas
brought about much improvement in the cost-efficient and custom-tailored synthesis of the
templates [18]. The discovery of new mesoporous silicas is also expected to provide
additional promising templates for the synthesis of new mesoporous carbons. The
resulting high-surface-area materials with uniform pores promise to be suitable as

32
adsorbents, catalyst supports, sensors, and materials for other advanced applications.
The presence of the distinct XRD patterns provides us new opportunities for precisely
monitoring various physico-chemical phenomena that take place inside the well-defined
carbon pores or at the pore walls such as adsorption, impregnation, framework changes,
formation of metal clusters and grafted functional groups. The materials constructed with
such well defined and controllable pore diameters are suitable as standards or references
for the characterization of porosity of the carbons, similar to the already well-known case
of the MCM-41silica.
In addition to this significance, the carbon frameworks can be grafted with various
organic and organometallic functional groups on the carbon-pore walls [19]. The carbon
can be used to compose systematic nanostructures between organic polymers and carbons
[20]. In addition, there was a recent report on the synthesis of an ordered microporous
carbon using NaY zeolite [21]. This result shows possibilities to extend the templating
route to other zeolite-type carbons with various structures. Furthermore, the template
synthesis methods have been being fully advanced for the synthesis of ordered
mesoporous carbons exhibiting graphite-like atomic orders [22]. These new CMK-nG
carbons will give many new possibilities in the adsorption science and technology, in
addition to the CMK-type mesoporous carbons composed of disordered atomic
arrangement.

References

1. Bansal C. R., Donnet J.-B. and F. Stoeckli, Active carbon, (Marcel Dekker, New
York, 1988).
2. Foley H. C., Microporous Muter. 4 (1995) pp. 407-433.
3. Ryoo R., Joo S. H. and S. Jun, J. Phys. Chem. B 103 (1999) pp. 7743-7746.
4. Ryoo R., Joo S. H., Kruk M. and Jaroniec M., Adv. Muter. 13 (2001) pp. 677-671.
5. Jun S., loo S. H., Ryoo R., Kruk M., Jaroniec M., Liu Z., Ohsuna T. and Terasaki 0..
J. Am. Chem. Soc. 122 (2000) pp. 10712-10713.
6. Joo S. H., Choi S. J., Oh I., Kwak J., Liu Z., Terasaki 0. and Ryoo R., Nature 412
(2001) pp. 169-172.
7. Ryoo R., Joo S. H., Jun S., Tsubakiyama T. and Terasaki O., Stud. Surf: Sci. Cutul.
135 (2001) p. 150.
8. Kresge C. T., Leonowicz M. E., Roth W. J., Vartuli J. C. and Beck J. S., Nature 359
(1992) p. 710-712.
9. Huo Q., Margolese D. I., Ciesla U., Feng P., Gier T. E., Sieger P., Leon R., Petroff P.
M., Schuth F. and Stucky G. D., Nature 368 (1994) pp. 317-321.
10. Zhao D., Huo Q., Feng J., Chmelka B. F. and Stucky G. D., J. Am. Chem. Soc. 120
(1998) pp. 6024-6036.
11. Bang W.-H., Liang C., Sun H., Shen Z., Guan Y., Ying P. and Li C., Adv. Muter. 14
(2002) pp. 1776- 1779.
12. Kaneda M., Tsubakiyama T., Carlsson A., Sakamoto Y., Ohsuna T., Terasaki O., Joo
S. H. and Ryoo R., J. Phys. Chem. B 106 (2002) pp. 1256-1266.
13. Lee J., Yoon S., Hyeon T., Oh S. M. and Kim K. B., Chem. Commun. (1999) pp.
2177-2178.
14. Kruk M., Jaroniec M., Ryoo R. and Joo S. H., J. Phys. Chem. B 104 (2000) pp. 7960-
7967.
15. Ryoo R. et al., manuscript in preparation.

33
16. Yu C., Stucky G.D. and Zhao D. In Abstracts of 3rdInternational Mesostructured
Materials Symposium, July 8- 11,2002, P A 4 P. 54
17. Ryoo R. et al., unpublished results
18. For example: Kim S. S.,Pauly T. R. and Pinnavaia T. J., Chem. Commun., (2000) pp.
1661- 1662.
19. Jun S., Choi M., Ryu S.,Lee H.-Y. and Ryoo R., Stud. Sulf: Sci. Card. submitted
20. Choi M. and Ryoo R., manuscript in preparation.
21. (a) Ma Z., Kyotani T. and Tomita A., Chem. Commun. (2000) pp. 2365-2366. (b) Ma
Z., Kyotani T., Liu Z., Terasaki 0. and Tomita A., Chem. Muter. 13 (2001) pp. 4413-
4415.
22. Kim T.-W., Park 1 . 4 . and Ryoo R.,manuscript in preparation.

34
 QUANTUM MICROPORE FILLING AND ITS APPLICATION POSSIBILITY

T. TANAKA, Y. HA'ITORI', K. MURATAS, T. KODAIRA**, M. YUDASAKA*,

S.IIJIMA*+, AND K. KANEKO
Department of Chemistry,Faculty of Science and Centerfor Frontier Electronics and Photonics,
Chiba University I-33 Yayoi, inage, Chiba 263-8522,Japan.

#)Instituteof Research and Innovation, 1201 Takada, Kashiwa, Chiba 277-0861, Japan

*I Japan Science and Technologv Corporation, c/o NEC Corporation, 34 Miyukigaoka, Tsukuba
305-850 I , Japan.

+) Department of Physics, Meijo Universiv, I501 Shiogamaguchi, Tenpaku7

Nagoya 468-8502, Japan

**) Nanoarchitectonics Research Center, National tnstitute of Advanced Industrial Science and
Technology,AiST Tsukuba Central 5, 1-1-1 Higashi
Tsukuba, Ibaraki 305-8565, Japan
Adsorption isothenns of Ne on AIPO4-5 and of H2 and D2 on single wall carbon nanohom (SWNH)
were measured at 27 K and 20 K, respectively. The comparison of the experimental Ne adsorption
isotherm at 27 K with that calculated from classical density functional theory (DFT)showed the
presence of quantum effect. The comparison of the experimental isotherm of HZon SWNH at 20 K
with the isotherm simulated by classical grand canonical Monte Carlo simulation suggested the
presence of more remarkable quantum effect than Ne on AIP04-5. Adsorption isothenns of HZand
Dz on SWNH at 20 K were compared with each other. Although both isotherms are very similar in
shape, the adsorption amount of 9is larger than that of H2 over the whole PPo range. This result
was ascribed to a marked quantum effect of H2 molecules, indicating the possibility of quantum
molecular sieving by nanocarbons.

1. Introduction

Recent research activities on nanoporous materials have stimulated hdamental studies

on adsorption mechanism in micropores [ 1 - 51. Both of the precise measurement of
high resolution adsorption isotherms fiom the low P/Po region and molecular simulation
showed the presence of monolayer adsorption on the micropore walls and further filling
in the residual spaces after monolayer completion for supermicropores (0.7 nm < pore
width w < 2 nm) ; the contribution by the monolayer to the filling in the residual spaces is
comparable to that by the pore walls [6- 101. Systematic researches on activated carbon
fiber (ACF) having slit-shaped micropores[ 1 1,121 have contributed to elucidation of the
mechanism of micropore filling to develop better adsorbents in adsorption and separation
engineering.
In case of micropore filling in ultramicropores (w < 0.7 nm), the monolayer adsorption
itself is highly enhanced. As this micropore filling in ultramicropores begins from an
extremely low pressure and is sensitive to the molecular size, the ultramicroporous
materials have been widely applied to separation engineering as molecular sieves.
Although it has been believed that the molecular sieve characterisitics arises from the
geometrical fitting of the pore to the molecule, the molecular sieving mechanism is not
necessarily well unveiled[l3 - 151. Accordingly we need to elucidate the molecular

35
sieving effect more to establish a more efficient separation technology. At the same time,
a new possibility of the molecular sieving should be challenged.
Recently nanocarbons such as single wall carbon nanotube (SWNT)[16,171 or single
wall carbon nanohorn (SWNH) have offered an important possibility for hydrogen and
methane storage [19-221. Although reliable adsorption data are not available for
previous SWNT samples[23,24], highly pure samples of SWNH is going to provide a
precise understanding of their storage potential [22]. In particular, nano-scale windows
can be donated on the wall of SWNH and thereby micropore filling in internal and
interstitial nanospaces can be clearly separated experimentally. Hence, SWNH leads to a
better understanding of adsorption properties of nanocarbons. Theoretical researches on
hydrogen storage by SWNT predicted the important contribution of quantum effect due to
the small mass of the hydrogen molecule [23-251. At the same time, these studies
suggested that the quantum effect should be taken into account in general nanoporous
systems and some experimental results on zeolites were discussed from the basis of this
quantum effect [26]. According to these theoretical studies, the zero-point vibration
cannot be negligible for a light molecule and the quantum molecule behaves as if it had a
larger size than the classical molecule. Some papers on quantum effect in nanoporous
systems were published before the above studies. Kaneko et a1 applied a simple
rectangular box model for He adsorption in slit-shaped graphitic nanospaces of ACF,
suggesting the presence of quantum excitation of translational motion of He atoms in the
motion perpendicular to the pore-walls [27]. Beenakker et a1 [28] proposed the possibility
of quantum sieving using hard spheres in a square-well cylindrical tube.
As D2 is a quite important gas in the present industry, an efficient separation of D2 fiom
H2 has been requested. If the quantum effect can be applied to the separation, a new
separation technology should be introduced. Tanaka et a1 examined the temperature
dependence of Ne adsorption on well-crystalline A1P04-5, showing the presence of an
explicit quantum effect [29]. Murata et a1 and Tanaka et al studied hydrogen adsorption on
SWNH over the wide temperature range from the boiling temperature to 303 K [30]. This
paper will review the quantum effect in micropore filling of Ne, H2, and D2 on A1P04-5,
nanocarbons, and ACF.

2. Classical DFT calculation, classical GCMC simulation and effective potential for a
quantum molecule

The molecular potential of a quantum molecule in a model SWNT is described here. For
simplicity, we assumed a homogeneous cylindrical pore for a model of open-ended
SWNT. Thus classical solid-fluid interactions can be calculated using the Lennard-Jones
(LJ) potential integrated over an infmitely long cylinder [3 I]:

&J4=&,E,fQsf
2

[
-
32
63 F(-4.5,-4.5,1.0;P2)
[R*(l-P2)]'o

-3

36
 where F( a,p, y ; x ) is a hypergeometric function, R and pr are the radius of the pore
and the density of solid atoms in the pore wall (38.21m-~).The interaction parameters for
H2, ad and Gp/R are 0.3 18nm and 32.1 K, respectively.
The quantum correction to the classical potential can be calculated by Feynmans
procedure, which employs an effective potential [32,33].

where r is the vector between two particle, 1 = h2/(6mk7)and m is the molecular mass of
hydrogen isotopes. Fig. I shows the effective potentials of hydrogen H2 and deuterium D2
inside the (3,6) nanotube (d = 0.62 nm) at 20 K fiom eq. (3). There is a large difference
in effective potentials for H2 and D2.This suggests that hydrogen is easily excluded from
hydrogen isotopes mixture with the (3,6) nanotube at 20 K. Therefore, such a nanotube
with the small diameter is predicted to exhibit very large seectivity and the other porous
materials that have pores corresponding to the size of the (3,6) nanotube should be
suitable for the hydrogen isotope separation.
The Ne adsorption isotherms on model A1P04-5 micropores were calculated fkom thc
Tarazonas version of the nonlocal density functional theory [34,35] which has beer
actually applied to the study on micropore filling [36,37]. The necessary parameters werc
obtained from the adsorption isotherms of Ne on A1P04-5 at 27K and 30K in a low
pressure range.

Fig. 1 Effective potentials of hydrogen isotopes inside SWNT at 20 K for (3,6)SWNT.

The solid line and (+) symbols denotes hydrogen and deuterium, respectively.
The adsorption isotherms of N2at 77 K and classical H2 at 20 K were calculated wi
classical GCMC simulation for the SWNH aggregate. Here we approximated SWNH 1

37
SWNT. GCMC simulation was carried out using the established procedures. We used the
12-6 Lennard-Jones(LJ) potential for the N2-N2 interaction. The used LJ parameters for N2
are E~ I kB= 95.2 K and = 0.375 nm. As to H2, -1 kB= 36.7 K and aff= 0.2959 nm
were used. The classical solid-fluid interactions given by eq. I were used for the
simulation. We used an established technique of the repeated cell determined by the
triangular and square arrays of SWNHs [7,9,38 - 401. The thickness of the graphene wall
was assumed to be 0.34 nm.

3. Adsorption measurement below 30 K

The cryogenic adsorption system was specially developed to measure adsorption

isotherms of H2 and D2. This system is equipped with a closed helium cycle two-stage
Gifford McMahon refiigerator to operate under cryogenic conditions. The adsorption
temperature can be kept constant within f 0.03 K at 20 K. Adsorption isotherms are
obtained by gas adsorption manometry. This method is based on the measurement of the
gas pressure in a calibrated, constant volume, at a known temperature. The dead space
volume was calculated tiom a helium calibration measurement at the temperature of
interest. Thermal transpiration effect was calibrated according to the work by Takaishi
and Sensui [4 11.

4. Quantum effect in Ne on AIP04-5

A1Po4-5 has one- dimensional pores of wider and narrower parts whose widths are
1.137 and 1.002 nm, respectively. Also well-crystalline samples are available to provide
reliable experimental data. The framework of A1Po4-5 consists of alternate tetrahedral
aluminum and phosphorous atoms bridged by oxygen atoms and thus is electricallyneutral.
The pores of AIP04-5 are not interconnected and form one-dimensional channels parallel
to the crystallographic c -axis [42].
The adsorption isotherms were calculated using DFT based on the classical potential
for narrower pore (radius R of structure = 0.501nm: AP-50) and the average size pore
models (R = 0.5347nm: AP-53). The adsorption increased steeply below P/Po = lo4,
being almost saturated above P/Po = The higher the measuring temperature, the
larger the rising PIPo. The DFT isotherms for both models are completely different from
each other. AP-53 model gives the two-step isotherm, whereas AP-50 model leads to the
single step isotherm which coincides with the experimental one, as shown in Fig.2.
Although the average pore model of AP-53 is expected to be fit for description of the
experimental result, the narrower pore model of AP-50 is better after the correction of the
pore volume by Ar adsorption.
The most probable cause for this marked discrepancy from the prediction is the
quantum effect. The quantum effects are included by replacing parameters E and 0 in V,,
with temperature and A-dependent parameters E and d (where A = h/a(m&).)).The
parameters E and d can be estimated by Feynmans effective potential method. For
quantum neon pairs, the location of the minimum of the effective potential is shifted to
26d and the value at the minimum to 4,and the quantum contributions are dla=
1.014 and &I& = 0.947 at 27 K, respectively. Therefore, the quantum effect decreases the
fluid-fluid interaction. That is, quantum Ne can behave as if they had a bigger size and the

38
moelecule-wall interaction becomes weak. Then, quantum Ne can be adsorbed
monolayerly on the pore-wall, as if the classical Ne molecules do in the narrower
micropores.

Fig.2 Experimental adsorption isotherm of neon on AIP04-5 at 27 K ( 0 ) and theoretical

adsorption isotherms: AP-50 model (+ ); AP-53 model (a). The dashed line denotes the
corrected experimental adsorption isotherm.

5. Micropore structures and quantum effectof SWNH

Nitrogen adsorption isotherm on as-grown SWNH was measured at 77 K to

characterize the SWNH assembly structure (Fig. 3). Then, theoretical nitrogen adsorption
isotherms on several SWNT models at 77 K were calculated by GCMC simulations
treated the classical Lennard-Jones systems to understand the experimental data. The
model adsorption spaces are interstices of triangular-packed and square-packed arrays of
SWNTs (the pore width at the nuclear position d = 3 nm) and external surface of an
isolated SWNT (d = 3 nm), respectively. Simulated isotherms on the three models are also
shown in Fig. 3. While the triangular array model cannot describe the experimental
isotherm in the low PDo range, the square array model well coincides with the
experimental data over a wide P/Po range. This suggests that SWNHs are not perfectly
aligned as the triangular (close-packed) model, but they are roughly assembled in square
arrays. However the square array model cannot reproduce the experimental isotherm in a
high P/Po range. This remarkable upward deviation of the experimental data from the
square model calculation should come from the multi-layer adsorption on the external
surfaces of nanohorns.

39
 10-10 lo* 106 lo4 10-2 lo0
PRO

Fig. 3. Experimental adsorption isotherm of N2 on SWNH at 77 K (0)and simulated isotherms

of N2on three models at 77 K: triangular array (+ symbols); square array (0);
adsorption on
the external surface of an isolated model ( A ) .

04

03

c 02
.-
c
a.
b
v1
0
U 01

n- -
10-17 10-6 lo-* 10-11 10-9 10-7 10 10-3 10-1

P/ Po

Fig. 4. Experimental adsorption isotherm of H2 on SWNH at 20 K (0)and simulated

isotherms of H2 on two models at 20 K using the classical GCMC simulations: square
array (0);adsorption on the external surface of an isolated SWNH (A).
 Fig. 4 shows comparison between the experimental isotherm of hydrogen on the
as-grown SWNH at 20 K and simulated isotherms for the square SWNT array and the
isolated SWNT models from the classical GCMC simulations. Although the simulated
nitrogen adsorption isotherm for the square. orientation model well agrees with the
experimental one in the low P/Po range, the GCMC-simulated adsorption isotherm of the
classical hydrogen does not coincide with the experimental one. This discrepancy should
stem fiom the predominant quantum effect at 20 K.

6. Quantum molecular sieving effect of SWNH

Experimental adsorption isotherm for D2on the as-grown SWNH at 20 K is compared

with that of hydrogen in Fig. 5. While two isotherms are very similar in shape, the
adsorption amount of D2is larger than that of H2 over the whole P/Po range. As the mass
of D2 is a half of H2,the quantum effect of D2should be less marked than that of H2,as
shown in Fig. 1. Even if the extent of the effective expansion of the H2molecule is only
0.01 nm, such effect is quite sensitive to the filling of molecules in the nanospaces.
Accordingly, H2 molecules cannot be adsorbed in narrower nanospaces to which D2
molecules can be accessible. This adsorption difference between H2 and D2 is a
representative of the quantum molecular sieving. It is expected that SWNH having
nanowindows can show a more remarkable quantum effect for H2 and DZ,providing a
hopeful applicant for hydrogen isotope separation.

25
?
n
20

15
W
E
.-0 10

E
2
71 5
4
rl
0 0.2 0.4 0.6 0.8 I

Fig. 5 Experimental adsorption isotherms of H2 and D2on SWNH at 20 K. H2 ( A ) , D2(0)

7. Acknowledgement

This work was funded by Nanocarbon project from the New Energy and Industrial
Technology Development Organization of Japan.

41
References

1. J. S. Beck, J. C. Vartuli, W. J. Roth, M. E. Leonowicz, C. T. Kresge, K. D. Schmitt,

C. T-W. Chu, D. H. Olson, E. W. Sheppared, S. B. McCullen, J. B. Higgins, J. L.
Schlenker, J. Am. Chem. SOC.,114, 10834 (1992).
2. S. Inagaki, Y. Fukushima, K. Kuroda, J. Chem. Soc., Chem. Commun., 680 (1993).
3. M. Kondo, T. Okubo, A. Asami, S. Noro, T. Yoshitomi, S. Kitagawa, T. Ishii, H.
Matsuzaka, K. Seki, Angew. Chem. Inr. Ed. 38, 140 (1999).
4. R. Ryoo, S.H Joo, S. Jun J. Phys. Chem. B 103,7743.(1999).
5 . E. Bekyarova and K. Kaneko, A h . Muter., 12, 1625 (2000).
6. N.A.Seaton, J.P.R.B.Walton, andN.Quirk, Carbon, 27,853 (1991).
7. C. Lastoskie, K. E Gubbins, N. Quirke, J. Phys. Chem., 97,4786 (1993).
8. K. Kaneko, R. F. Cracknell, D. Nicholson, Langmuir, 10,4606 (1994).
9. R. F. Cracknell, K. E. Gubbins, M. Maddox, D. Nicholson, Acc. Chem. Res. 28 281
(1995).
10. T.Ohba, T.Suzuki, and K. Kaneko, Chem. Phys. Lett., 326, 158(2000).
1 1 . K. Kaneko, Carbon, 38,287 (2000).
12. T. Iiyama, T. Ohkubo, K. Kaneko, In Recent Advances in Gas Separation by
Microporous Ceramic Membranes, Ed. N.K. Kanellopouos, Elsevier, pp.35-66
(2000).
13. Y. D. Chen, R. T. Yang, and P. Uawithya, AIChE Journal, 40,577 (1994).
14. C. Nguyen and D. D. Do, Carbon, 33, 1717 (1995).
15. T. Suzuki, R. Kobori and K. Kaneko, Carbon, 38,630 (2000).
16. S. Iijima,Tkhihashi, Nature, 363,603 (1993).
17. D. S. Bethune, C. H. Kiang, M. S. de Vries, G. Gorman, R. Savoy, J. Vazquez, R.
Beyers, Nature, 363,605 (1 993).
18. S. Iijima, M. Yudasaka, R. Yamada, S. Bandow, K. Suenaga, F. Kokai, F.,
K.Takahashi, Chem. Phys. Lett. 309, 165. (1999).
19. K. Murata, K. Kaneko, F. Kokai, K. Takahashi, M. Yudasaka, S. Iijima, Chem. Phys.
Lett., 331, 14 (2000).
20. K. Murata, K. Kaneko, W.A. Steele, F. Kokai, K. Takahashi, K. Kasuya, K. Hirahara,
M. Yudasaka, S. Iijima, J. Phys. Chem. B 105,10210 (2001)
21. E. Bekyarova, K. Kaneko, M. Yudasaka. K. Murata, D. Kasuya, S. Iijima, Adv.
Muter. 14,973 (2002).
22. K. Murata,K. Kaneko, H. Kanoh, D. Kasuya, M. Yudasaka, S. Iijima, J. Phys. Chem.
in press.
23. A.C.Dillon, K. M Jones,. T. A Bekkedahl,. C. H.Kiang D. S.Bethune, M. J.Heben,
Nature, 386,377 (1997) Nauture, 410,734 (2001)
25. S.R. Challa, D.S. Sholl, J.K. Johnson, Phys. Rev. B 63,245419 (2001).
26. Q. Wang, S.R. Challa, D.S. Sholl, J.K. Johnson, Phys. Rev. Lett. 82,956 (1999).
27. K. Kaneko, N. Setoyama, and T. Suzuki, CharacterizationofPorous Solidr tII,J.
Rouquerol, F. Rodriguez-Reinoso, K. S. W. Sing, and K. Unger eds. Elsevier (1994),
p.593.
28. J. J. Beenakkaer, V. D. Bornman, and S.Yu. Krylov. Chem. Phys. Lett. 232,379
(1 995).
29. H. Tanaka, M. El-Merraoui, T. Kodaira, K. Kaneko, Chem. Phys. Lett. 352,334
(2002).

42
30. H. Tanaka, K. Murata,K. H. Kanoh, D. Kasuya, M. Yudasaka, S. Iiljima, Kaneko, J.
Phys. Chem. to be submitted.
31. G. J. Tjatjouls, D. L. Feke, J. A. Mann, Jr., J. Phys. Chem. 92,4006 (1988).
32. R. P. Feynman, Statistical Mechanics, Benjamin, Reading, MA (1972).
33. G . Stan,M. W. Cole, J. Low Temp. Phys. 41,611 (1980).
34. P. Tarazona, U. M. B. Marconi, R. Evans, Mol. Phys.,60,573 (1 987).
35. P. I. Ravikovitch and A. V. Neimark, J. Phys. Chem. 105,6817 (2001).
36. Mustapha El-Merraoui, M. Aoshha, K. Kaneko, Langmuir 16,4300(2000).
37. K. Murata, Mustapha El-Merraoui, K. Kaneko, J. Chem. Phys. 114,4 196 (200 1)
38. K. R. Matranga, A. L. Myers, E. D. Glandt, Chem. Eng. Sci., 47, 1569 (1992).
39. W. A.Steele, M.Bojan, J. A h . Colloid Interface Sci. 77-76, 153 (1998).
40. T. Ohba and K. Kaneko, J. Phys. Chem. 106,7171 (2002).
4 1. T. Takaishi, Y . Sensui, Trans. Furaday SOC.53,2503 (1 963).
42. J. W. Pichardson Jr., J. J. Pluth, J. V. Smith, Acta Crystullogr. C43, 1469 (1987).

43
 ADSORPTION IN MICROPOROUS MATERIALS: ANALYTICAL
EQUATIONS FOR TYPE I ISOTHERMS AT HIGH PRESSURE

A. L. MYERS
Department of'Chemical and Biomolecular Engineering
University of Pennsylvania, Philadelphia PA 19104 USA
E-mail: [email protected]

Existing analytical equations for Type I isotherms such as the Langmuir equation and
its modifications fail to describe experimental adsorption isotherms at high pressure.
All of these equations predict that the amount adsorbed increases monotonically with
increasing pressure. Experimental adsorption isotherms attain a maximum value in the
amount adsorbed and then fall to zero. At the critical temperature of the gas, the
maximum in the isotherm occurs at a pressure of about 10 bars and the zero occurs at a
much higher pressure of several hundred bars. For subcritical gases, the maximum occurs
at lower pressure and for supercritical gases the maximum occurs at higher pressure.
The observed behavior of high-pressure adsorption isotherms can be predicted from the
low-pressure (subatmospheric) portion by separating the Type I behavior expected for
absolute adsorption from the experimental behavior observed for excess adsorption.

1 Introduction
Tabulated data for experimental adsorption isotherms are fitted with analytical equations
for the calculation of thermodynamic properties by integration or differentiation. These
thermodynamic properties expressed as a function of temperature, pressure, and composition
are input to process simulators of adsorption columns. In addition, analytical equations for
isotherms are useful for interpolation and cautious extrapolation. Obviously, it is desirable
that the isotherm equations agree with experiment within the estimated experimental error.
The same points apply to theoretical isotherms obtained by molecular simulation, with the
requirement that the analytical equations should fit the isotherms within the estimated
statistical error of the molecular simulation.
Type I isotherms [3]are characterized by an asymptotic approach to a saturation capacity
with increasing pressure. This class of isotherms is most commonly observed for gases or
vapors (water is an exception) adsorbed in zeolites or activated carbon. A typical set of
Type I adsorption isotherms is shown in Fig. 1. Several questions may be asked about sets
of isotherms like these. How is the saturation capacity measured? Is the saturation capacity
a constant or does it decrease with temperature as suggested by Fig. l?
Variation of the pressure P with respect to temperature T at constant loading n is given
by the equation [l]:
IE
nI-[ = -E
where is the differential enthalpy of desorption (isosteric heat). This derivative of the
adsorption isotherms in Fig. 1 along a horizontal line becomes ambiguous at high loading
(n).The differential enthalpy for this system measured by a calorimeter is plotted on Fig. 2.

44
For type I1 isotherms with multilayer adsorption, the differential enthalpy for the second and
higher layers approaches the enthalpy of condensation of the liquid. Does the differential
enthalpy for type I isotherms approach some limit at saturation?
The intention of this paper is to seek answers to these questions.

5
30 C
20 c
4

n 70 C
g3
8
-E2

0
0 0.2 0.4 0.6 0.8 1.o
Pressure (bar)
Fig. 1. Adsorption isotherms of C,& on NaX zeolite. The 2OoC isotherm
was measured experimentally [4]. The other two isotherms were calculated
from Eq. (1) using the differential enthalpies in Fig. 2.

45

35
0 1 2 3 4
Amount adsorbed (moykg)
Fig. 2. Ditferential enthalpy of CzH4 adsorbed on NaX at 2OOC. Points are
experimental calorimetric data [4]. Solid line is polynomial fit of data.

45
2 Adsorption isotherms
A very useful equation for type I adsorption isotherms is 141:

where H is the Henry constant, m is the saturation capacity, and the Ci are virial coefficients
terminated after three or four terms, which are usually sufficient to fit adsorption isotherms
over several decades of pressure. Unlike most adsorption equations, this equation is implicit
in the pressure. This disadvantage is more than offset by the direct connection between the
differential enthalpy and the temperature dependence of the virial coefficients. The main
advantage of this equation is that it can be integrated analytically for the grand potential
(spreading pressure), which is needed for mixture calculations as shown in this mixture
section of this paper.
Experimental measurements yield excess adsorption; molecular simulations calculate ab-
solute adsorption. The relationship between the two variables is given by:

ne = n - V,p (3)
V, is the specific pore volume of the material; typical values are 200-400 cm3/kg for zeolites
and up to 1000 cm3/kg for activated carbon. n is the actual number of molecules contained
in the micropores; the excess adsorption ne subtracts from n the number of molecules which
would have been present in the micropores at the bulk density in the absence of adsorption.
The (oversimplified) case when absolute adsorption is described by the Langmuir equation
and the gas obeys the perfect gas law ( p = P / R T ) has been worked out in detail for the
isotherms and thermodynamic functions (enthalpy, entropy, etc.) [2].
The key step in this development is the recognition that a Type I adsorption equation like
Eq. (1) applies to absolute adsorption n. Absolute adsorption refers to the actual number
of molecules present in the micropores and increases monotonically with pressure to an
asymptote called the saturation capacity m. Experimental excess adsorption isotherms pass
through a maximum and then decrease with pressure.
The adsorption isotherms in Fig. 1 may be considered absolute adsorption as a function
of gas-phase fugacity f of ethylene. For P < 1 bar, the term V,p in Eq. (3) is negligible
compared to n. Therefore a fit of these isotherms with Eq. (1) using constants reported
previously [4] provides the absolute amount adsorbed n ( f ) . Given an equation for the
absolute isotherm and the pore volume of faujasite (340 cm3/kg), one can calculate the
excess isotherms at high pressure. Taking the fugacity f as an independent variable, the bulk
properties of gaseous ethylene derived from the SRK equation [5] were used to determine
the pressure P and density p = P / ( z R T ) , where z is the bulk compressibility factor of
gaseous ethylene. The excess functions for three temperatures are plotted on Fig. 3. For
subatmospheric pressure, the values of absolute and excess adsorption coincide. At higher
pressure, the absolute adsorption approaches a saturation value (4.534 mol/kg) while the
excess adsorption passes through a maximum and eventually decreases to negative values at
pressures above 100 bar. The isotherm for -30 C terminates at the dew point of ethylene
vapor, 19.4 bar. The maximum excess adsorption decreases with temperature. At the
pressure where the isotherms intersect, the temperature coefficient of adsorption is zero.
Above this pressure, excess adsorption increases with pressure.
The maximum in the excess adsorption isotherm occurs at the point where the densities
in the micropore and the bulk gas are increasing at the same rate with respect to pressure, so

46
that an increase in pressure has no effect upon the amount adsorbed. Zero excess adsorption
occurs at the pressure where the bulk and micropores densities are equal. Since the pore
density at saturation is p = m/% = (4.534/340)= 0.0133 mol/cm3, zero excess adsorption
corresponds to the same value for the bulk density of ethylene. The pressures corresponding
to this bulk density are 175 and 345 bar at 20C and 70"C, respectively. The density of
0.0133 mol/cm3 inside the micropores is 1.7 times the critical density or 0.7 of the liquid
density of ethylene at its normal boiling point (-110C). It is interesting that the micropore
volume may be calculated from the bulk density at which excess adsorption is zero using
V, = m/p.

0.1
0 . m 1 o.oO01 0.001 0.01 0.1 1 .o 10 100 500
Pressure (bar)
Fig. 3. Adsorption isotherms of CzH4 on NaX.The -30C isotherm termi-
nates at the dewpoint of ethylene vapor (19.4 bar)

The absolute adsorption isotherms for -30C and 70C were calculated from the 20C
isotherm using the integrated form of Eq. (1) and the differential enthalpy plotted on Fig. 2.
The reasonable approximation was made that the differential enthalpy is independent of
temperature. No other assumptions were needed to calculate the excess adsorption isotherms
on Fig. 3.
Examination of Fig. 1 answers the questions raised in the introduction about the deter-
mination of the saturation capacity. The saturation capacity for absolute adsorption cannot
be extracted from supercritical isotherms measured at sub-atmospheric pressure such as the
70C isotherm. Raising the pressure would bring the isotherm into the region where the dif-
ference between absolute and excess adsorption is no longer negligible. The Henry constant
(the slope of the adsorption isotherm at the limit of zero pressure) is difficult to extract from
sub-critical isotherms such as the -30C isotherm because the linear region occurs at very
low pressure. The optimum temperature for the determination of both the Henry constant
and the saturation capacity is near the critical temperature of the gas (9C for ethylene),
which is the 20C isotherm on Fig. 1.

47
3 Gas Mixtures
The principles of phase.equilibrium do not apply to excess adsorption variables at high
pressure where the excess adsorption passes through a maximum. Under these conditions,
the pressure is no longer a singlevalued function of excess adsorption so that ne cannot serve
as an independent variable for the determination of partial molar quantities such as activity
coefficients. Additional complications which arise at high pressure are: (1) the selectivity
for excess adsorption (S1,= (n:/v1)/(n;/~z)) approaches infinity BS $ + 0; and (2) the
differential enthalpy of the ith component has a singularity at the pressure corresponding to
maximum nt. For excess variables, the differential functions are undefined but the integral
functions for enthalpy and entropy are smooth and well-behaved [l].
The principles of solution thermodynamics can be applied to absolute adsorption vari-
ables without any of these complications. For absolute variables, which arise naturally in
molecular simulation, the pressure is a singlevalued function of n, the differential functions
exhibit no singularities, and the selectivity approaches a finite value as P ---t 00. Absolute
adsorption may be determined experimentally by measuring excess adsorption in the usual
way (volumetric or gravimetric method) at sub-atmospheric pressure where the difference
between absolute and excess adsorption is negligible.
The fugacity equations are solved using absolute variables. As discussed previously, the
single-gas isotherms at sub-atmospheric pressure provide the absolute isotherm in the form
f;(np). Given the temperature of the isotherms, the independent variables are P and y1 in
the bulk gas. For a binary mixture the fugacity equations are written:

The fugacity fi = Pyi&(P, yi) is determined from an equation of state (EOS) for the pure
bulk gas. The adsorbed-phase activity coefficients are functions of the grand potential (Q)
and composition, where $ = -R/RT. The form of Eq. (2) allows the grand potential to be
calculated analytically [4]:

The standard-state properties are measured at equal values of the grand potential:

$l(n;) = $z(%) (7)

Given a bulk EOS to determine the fugacity functions and noting that ( q + z z ) = (y1+yz) =
1, the three Eqs. (4), (5), and (7) can be solved for the three unknowns: ni, n:, and 21.
Following methods described previously 141, the absolute total (n)and individual amounts
adsorbed (nl,nz) are calculated from the standard-state values n;l and ni. Finally, using
an EOS to determine the bulk density p(P,yl), the absolute variables are converted to
experimental excess variables by [l]:

nt = ni - V , p y i (8)
Excess isotherms are shown on Figure 4 for binary mixtures of ethylene and ethane
adsorbed on NaX. Both curves are for 20C and a vapor composition of 10 mole percent eth-
ylene. The individual excess isotherm for ethylene is the difference between two isotherms.
Ethylene is preferentially adsorbed and the individual isotherm for ethane reaches a max-
imum at a pressure of 3 bar. The individual isotherms show the same behavior as the
singlegas isotherms shown on Figure 3: a maximum value in the amount adsorbed followed
by a steep decline to zero.

5
total

1 C2H6

0.1
0 .001 0.01 0.1 1 lo 30
Pressure (bar)
Fig. 4. Adsorption of mixtures of C& and c2H6 on NaX at 20 C for
10 mole % ethylene in gas phase. The individual isotherm for CzH4 is the
difference between the total and the individual isotherm piotted for C2H6.
The solid lines on Figure 4 take into account the nonideal behavior of adsorbed mixtures
of ethylene and ethane in NaX. This system is highly nonideal because of the interaction of
the quadrupole moment of ethylene with the sodium cations of NaX. Activity coefficients at
iniinite dilution are unity at the limit of zero pressure and 0.27 at high pressure. The dashed
lines on Figure 4 were calculated for an ideal adsorbed solution (IAS) and the resulting error
in the individual isotherm for ethane at 30 bar is 20%.

4 Discussion
The high pressure adsorption of single gases and mixtures can be predicted from the low
pressure (subatmospheric) data for the same systems. The optimum temperature for mea-
suring the adsorption of single gases is near their critical temperature where both the Henrys
constant and the absolute saturation capacity can be determined accurately.
The absolute adsorption isotherm as a function of gas-phase fugacity is obtained directly
from molecular simulations based on the grand canonical Monte Carlo (GCMC) method.
Since the difference between absolute and excess adsorption is negligible at subatmospheric
pressure, the low-pressure portion of the absolute isotherm can also be determined from
experiment. Eq. (2) is suitable for extrapolating the absolute isotherm from low to high
pressure and Eq. (3) provides the conversion to excess adsorption. Experiments are needed
to test these predictions of adsorption at high pressure.

49
References
[l] Myers, A. L., A.I.Ch.E. J., 2002,48, 145.
[2] Myers, A. L., and Monson, P. A., Adsorption in porous materials at high pressure:
Theory and experiment, Langmuir, 2002,in press.
[3] Ruthven, D.M.,Principles of Adsorption and Adsorption Processes, Wiley Interscience,
New York, 1982.
(41 Siperstein, F.R. and A.L. Myers, A.I.Ch.E. J., 2001,47, 1141.
[5] Smith, J. M., Van Ness, H.C., and Abbott, M.M., Introduction to Chemical Engineering
Themodynamics, 6th Edition, McGraw-Hill, New York, 2001.
 NEW SORBENTS FOR DESULFURIZATION OF TRANSPORTATION FUELS

RALPH T.YANG, ARTURO HERNANDEZ-WDONADO, AKIRA TAKAHASHI, AND

FRANCES H.YANG
Department of Chemical Engineering, University of Michigan, Ann Arbor, MI 48109-2136, USA.

New sorbents for desulfurization of liquid transportation fuels were developed using n-complexation.
Vapor phase adsorption isotherms were investigated to understand the interaction between
benzendthiophene and various kinds of sorbents: Ag-Y, Cu-Y, Na-Y, H-USY, Na-ZSM-5, activated
carbon, and modified activated alumina Liquid breakthrough experiments were conducted to test
their true sulfur removal effectiveness and capacities. Compared with Na-Y, Cu-Y and Ag-Y
adsorbed significantly larger amounts of both thiophene and benzene at low pressures, due to
x-complexation with C d and A$. Molecular orbital calculations confirmed the relative strengths of
n-complexation: thiophene > benzene and Cu' > A$. The experimental heats of adsorption for
x-complexation are in excellent agreement with theoretical molecular orbital predictions. The
sorbent capacities for thiophene at the low pressure of 2 . 3 ~ 1 0atm
~ were 0.92 moleculdCu+ and
0.42 moleculdAg', and followed the order: Cu-Y & Ag-Y >> Na-ZSM-5 > activated carbon > Na-Y
> modified alumina & H-USY. For liquid phase experiments, Cu(1)-Y, Ag-Y and Na-Y zeolites
were used to removal low concentration thiophene from mixtures including benzene and/or n-octane,
all at m m temperature and atmospheric pressure. Sulfur-free (i.e., below the detection limit of 4
ppmw sulfur) fuels were obtained with Cu(1)-Y and Ag-Y, but not Na-Y. Breakthrough and
saturation adsorption capacities obtained for an influent concentration of 760 ppmw sulfur (or 2000
ppmw thiophene) in n-octane follow the order Cu(1)-Y > Ag-Y > Na-Y and Cu(1)-Y > Na-Y > Ag-Y,
respectively. Regeneration of the adsorbent was accomplished by using air at 350'C followed by
re-activation in helium at 450'C. The observed adsorption behavior, in general, agrees well with the
studies performed for pure component vapor phase adsorption of thiophene and benzene with the
same adsorbents.

Introduction

Removal of sulfur-containing compounds is an important operation in petroleum refining,

and is achieved by catalytic processes at elevated temperatures and pressures [I]. The
hydrodesulhkation (HDS) process is highly efficient in removing thiols and sulfides, but
less effective for thiophenes and thiophene derivatives. New legislation will require
substantial reductions in the sulfur content of transportation fuels. For example, the new
US.EPA s u l k standards require that the sulfur contents in gasoline and diesel fuels for
on-board transportation will be 30 and 15 ppm, respectively, decreased fiom the current
levels of several hundred ppm. Faced with the severely high costs of compliance, a
surprisingnumber of refiners are seriously considering reducing or eliminating production
of on-board fuels [2]. The new challenge is to use adsorption to selectively remove these
sulfur compounds from liquid fuels. Since adsorption would be accomplished at ambient
temperature and pressure, success in this development would lead to a major advance in
petroleum refming. However, success would depend on the development of a highly
selective sorbent with a high sulfur capacity, because the commercial sorbents are not
desirable for this application. First results on sorbents based on n-complexation for
desulfurization have been reported recently by Yang et al. [3,4] which were shown to be
superior to all previously reported sorbents.
During the last decade, there have been several published accounts on using
adsorption for liquid fuel desulfurization. Commercially available sorbents (i.e., zeolites,
activated carbon and activated alumina) were used in all of these studies. Weitkamp et al.

51
[5]reported that thiophene adsorbed more selectively than benzene on ZSM-5 zeolite.
Based on this study, King et al. [6]studied selective adsorption of thiophene, methyl- and
dimethyl-thiophenes (all with one ring) over toluene and pxylene, also using ZSM-5.
They showed that thiophene was more selectively adsorbed, both based on fixed bed
breakthrough experiments. However, the capacities for thiophene were low (only 1-2%
wt. adsorbed at 1% thiophene concentration). Both vapor phase and liquid phase
breakthrough experiments were done in these studies, and the results from two phases
were consistent. The pore dimensions of ZSM-5 are 5.2-5.6A. Hence organic sulfur
compounds with more than one ring will be sterically hindered or excluded. Zeolites with
larger pores, as well as larger pore volumes, will be more desirable than ZSM-5 as the
selective sorbents. Indeed, results of Salem and Hamid [7]indicated that 13X zeolite as
well as activated carbon had much higher sorption capacities for s u l k compounds.
Based on the data of Salem and Hamid [7],the capacity for sulfur compounds by 13X
zeolite was approximately an order of magnitude higher than that of ZSM-5, when
compared with the data of King et al. [6]extrapolated to the same conditions. Modified
activated alumina (Alcoa Selexsorb), which contains proprietary modifier to provide
optimum adsorption of a number of polar organic compounds, has been used in an
adsorption process by Irvine [8].No direct comparison has been made among these
commercial sorbents. Their experiments were mostly done in fixed bed adsorbers, by
measuring the breakthrough capacities. Based on the literature, the large pores zeolites
(NaX or Nay) are about the same as activated carbon and alumina, in terms of adsorption
of thiophene.
Based on the principles of n-complexation, we have already developed a number of
new sorbents for a number of applications. These include sorbents for: (a)
olefdparaffm separations [9- 121,(b) diene/olefm separation or purification (i.e., removal
of trace amounts of dienes fiom olefins) [ 131,and (c) aromatics/aliphatics separation and
purification (i.e., removal of trace amounts of aromatics fiom aliphatics [141.Throughout
this work, we have used molecular orbital calculations to obtain a basic understanding for
the bonding between the sorbates and sorbent surfaces, and further, to develop a
methodology for predicting and designing n-complexation sorbents for targeted molecules
(e.g. Ref. 11).
First results on n-complexation sorbents for desulfurization with Ag-Y and Cu(1)-Y
zeolites have been reported recently [3,4]. In this work, we included the known
commercial sorbents such as Na-Y, Na-ZSM5, H-USY, activated carbon and activated
alumina (Alcoa Selexsorb) and made a direct comparison with Cu(1)-Y and Ag-Y which
were the sorbents with n-complexation capability. Thiophene and benzene vapors were
used as the model system for desulfiuization. Although most of these studies can be
applied directly to liquid phase problems, Cu-Y (auto-reduced) and Ag-Y zeolites were
also used to separate liquid mixtures of thiophenehenzene, thiopheneh-octane, and
thiophenehenzene/n-octane at room temperature and atmospheric pressure using
fixed-bed adsorptiodbreakthrough techniques. These mixtures were chosen to understand
the adsorption behavior of s u l k compounds present in hydrocarbon liquid mixtures and
to study the performance of the adsorbents in the desulfurization of transportation fuels.
Moreover, a technique for regeneration of the adsorbents was developed in this study [4].

52
Experimental

Adsorbent Preparation
Various kinds of sorbents were investigated in this work. Four as-received sorbents:
Na-type Y-zeolite (Si/A1=2.43, Strem Chemical), H-type ultra-stable Y-zeolite (Si/AI=l95,
TOSOH Corporation), activated carbon (Type PCB, Calgon Carbon Corporation) and
modified activated alumina (Selexsorb CDX, Alcoa Industrial Chemical), were used in
this study. According to the product datasheets, Selexsorb CDX is formulated for
adsorption of sulfur-based molecules, nitrogen-based molecules, and oxygenated
hydrocarbon molecules. Na-Y and H-USY were in powder form (binderless). Since
activated carbon was in granular form and activated alumina was in pellet form, they were
crushed into powder form for evaluation.
Cu(1)-Y was prepared by ion exchange of Na-Y zeolites with Cu(NO3)Z followed by
reduction of Cu2' to Cu'. First, as-received Na-Y was exchanged twice using excess
amounts (10-fold cation-exchange-capacity(CEC) assuming that one Cu" compensates
two aluminum sites) of 0.5 M Cu(NO& at room temperature for 24 hours. After the
exchange, the zeolite suspension was filtered and washed with copious amount of
de-ionized water. The product was dried at 100 "C overnight. Several groups have
reported reduction of Cu" to Cu' in zeolite in He (i.e., auto-reduction). In this study,
reduction of Cu2' to Cu' was carried out in He only at 450 "C for periods in the 1 to 18
hours range. Ag' ion-exchange Y-zeolite (Ag-Y) was prepared at room temperature for 24
h in the same manner as Cu" exchange, using 5-fold excess AgN03 (O.1M). 13X
(Si/Al=1.25, Linde) was used for the preparation of Cu-X (10 fold CEC solution of
Cu(NO&, ion-exchanged at 65 "C for 24 hrs, three times) and Ag-X (5-fold CEC solution
of AgN03, ion-exchanged at RT for 24 hrs, twice). Na-type ZSM-5 (Na-ZSM-5) was
prepared at room temperature by Na'-exchange of Nh-ZSM-5 (Si/Al=lO, ALSI-PENTA
Zeolite GmbH).
Vapor Phase Isotherms and Heat of Adsorption
The objective of this study is to compare the strength of adsorptive interaction between
adsorbents and thiophenehenzene. Extremely low partial pressures at less than l o 5 atm
would be necessary to meet this objective if isotherms were measured at ambient
temperature, because the isotherms at ambient temperature are fairly flat and are difficult
to compare. However, it is very difficult to obtain and control such low partial pressures
experimentally. Therefore, single component isotherms for benzene and thiophene were
measured at 90, 120 and 180 "C using standard gravimetric methods. A Shimadzu
TGA-50 automatic recording microbalance was employed. Isosteric heats of adsorption
were calculated using the Clausius-Clapeyron equation from isotherms at different
temperatures.
Fired Bed A&orptiodBreakthrough Experiments
All adsorption/breakthrough experiments were performed in vertical custom made quartz
adsorbers equipped with a supporting glass frit. Initially, the adsorbents were loaded
inside the adsorber (between 1 or 2 grams), and heated in situ (250 - 450'C) while
flowing either helium or nitrogen upwards. After activation treatment, the zeolite
adsorbent under study was allowed to cool down to room temperature under. Next, a
sulfur-free octane or benzene solution was allowed to flow downwards through the

53
sorbent at a rate of 0.5 cm3/min.After wetting the adsorbent for about 30 minutes, the feed
was changed to a mixture of CgHI8(n-octane) and/or c6H6 (benzene) containing different
concentrationsof C4H4S(thiophene) also at a 0.5 cm3/minrate. Samples were collected at
regular intervals until saturation was achieved, which .depended on the adsorption
dynamics and amount of adsorbent.
All the samples collected during the breakthrough experiments were analyzed using a
Shimadzu GC (Gas Chromatography) unit equipped with a polar column, an automatic
multi-sampler, and a FID detector. The minimum thiophene concentration detection was
around 10 ppmw or 4 ppmw on a sulfur basis.

Molecular Orbital Computational Details

Molecular orbital (MO) studies on the n-complexation bonding for benzene and sorbent
surfaces had been investigated recently [3]. In this work, similar MO studies were
extended to thiophene and zeolites. The Gaussian 94 Program [ 151 in Cerius2 molecular
modeling software [161fiom Molecular Simulation, Inc. was used for all calculations. MO
calculations for thiophene and sorbent surfaces were performed at the Hartree-Fock (HF)
and density functional theory (DFT) level using effective core potentials (ECPs) [17-191.
The LanL2DZ basis set [20] is a double-t; basis set containing ECP representations of
electrons near the nuclei for post-third-row atoms. The reliability of this basis set has been
confirmed by the accuracy of calculation results as compared with experimental data.
Therefore, the LanL2DZ basis set was employed for both geometry optimization and
natural bond orbital (NBO) analysis.
The restricted Hartree-Fock (RHF) theory at the LanL2DZ level basis set was used to
determine the geometries and the bonding energies of thiophene on AgCl and CuCI. The
simplest models with only a single metal chloride interacting with a thiophene molecule
were chosen for n-complexation studies. The optimized structures were then used for
bond energy calculations according to the following expression:
Eads = Eadsorbate -k Eadsorbent - Eadsorbent-adsorbate
where is the total energy of the bare
is the total energy of thiophene, Eadsorbent
adsorbent i.e. the metal chloride and Ea~~t-adsorb.te is the total energy of the
adsorbate/adsorbentsystem. A higher value of Endscorresponds to a stronger adsorption.
Natural Bond Orbital (NBO)
The optimized structures were also used for NBO analysis at the B3LYP/LanL2DZ level.
The B3LYP [21] approach is one of the most useful self-consistent hybrid (SCH)
approaches [22], it is Becks 3-parameter nonlocal exchange functional[23] with nonlocal
correlation functional of Lee, Yang and Parr [24]. The NBO analysis performs population
analysis that pertains to localized wave-function properties. It gives a better description of
the electron distribution in compounds of high ionic character, such as those containing
metal atoms [25]. It is known to be sensitive for calculating localized weak interactions,
such as charge transfer, hydrogen bonding and weak chemisorption. Therefore, the NBO
program [26] was used for studying the electron density distribution of the adsorption
system.

54
Modelsfor Ag-Zeolite (Agz) and Cu-Zeolite (CuZ)
The zeolite models selected for this study are similar to the ones used by Chen and Yang
[33], with the molecular formula of (HO)3Si- 0 -A1(OH)3, and the cation Ag' or Cu'
sits 2 - 3 A above the bridging Oxygen between Si and Al. This is a good cluster model
representing the chemistry of a univalent cation bonded on site I1 (SII) of the faujasite
framework (Z). Once the optimized structures of AgZ and CuZ are obtained at the
B3LYPLanL2DZ level, then a molecule of thiophene (C4&S) or benzene (C6&) is
added onto the cation of the zeolite model, and the resulting structure is hrther optimized
at the B3LYPLanL2DZ level.

Results and Discussion

Vapor Phase Adsorption

BenzenejThiophene Adsorption Isotherms

Figure 1 and Figure 2 show the isotherms of benzene and thiophene on Ag-Y and Cu-Y.
Curves are fitted with Dubinin-Astakhov (solid line) and Langmuir-Freundlich (dotted
line) isotherms. Compared with Figure 3, these sorbents adsorbed significantly more
thiophenehenzenethan Na-Y at pressures below 10" atm, and nearly the Same amounts at
high partial pressures. This result was a clear indication of n-complexation with Ag'and
Cu'; since Na' could not form n-complexation bonds. However, the difference of
thiophenehenzene adsorption amount did not reflect the relative strengths of
a-complexation between Cu' and Ag" because the Cu' exchange was not complete.
Neutron activation analyses of the sorbent samples showed that the Ag' exchange was
100% but the Cu+ exchange was only 46%. According to the EPR analysis, described
elsewhere [171, only a half of the Cu2+was auto-reduced to Cu' after our heat treatment at
450 "C for lhr in He. On a per-cation basis, it is seen that Cu' could adsorb higher
thiophene adsorption amounts. In fact, 0.92 thiophene molecule per Cu' was obtained at
2xlO-' atm at 120C. This amount was due to Cu' since the amount adsorbed by NaY at
the same pressure was negligible. At the same pressure, only 0.42 thiopheneIAg' was
obtained. This result indicated strong a-complexation bonds between both Cu' and Ag",
and that the bond with Cu' was stronger.

Comparison of Thiophene Adsorption on All Adsorbents

Thiophene adsorption isotherms on all sorbents are compared in Figure 4. It is clearly
seen that Ag-Y and Cu-Y could adsorb significantly larger amounts of thiophene even at
very low pressures.

55
 1 1
4.0

3.5
1.5 ._.-
3.0

2.5

2.0
1.o
1.5

1.o

--
0.5

0.0
1.E45 1.E.M 1.E- 1. E m l.E-01
(-1

Figure 1. Pure component equilibrium isotherms Figure 2. Pure component equilibrium isotherms of
of benzene and thiophene on Ag-Y (SVAk2.43) at benzene and thiophene on Cu-Y (Si/AI=2.43) at 90
120 OC and 180 "C. Fitted curves are not shown for "C and 120 "C.
benzene adsorption at 180 "C because the artificial
crossovers to the curves for thiophene at 180 OC are
observed.
3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0
l.wE.05 1.wE.M l.wE.03 1.oQE42 l.wE-01
1.E.05 1.E.M 1.E- 1.E-02 1.E-01
PuUd Pmsun(um) P N W Raun(am)

Figure 3. Pure component equilibrium isotherms Figure 4. Comparison of equilibrium adsorption

of benzene and thiophene on Na-Y(SVAl=2.43) at isotherms ofthiophene at 120 "C.
120 "C and 180 "C.

Heat of Adsorption
Heats of adsorption were calculated using the Clausius-Clapeyron equation fkom
isotherms at two different temperatures, and are shown in Table 1. All the heats of
thiophenehnzene adsorption had the tendency to decease as the loading increased. This
is a common phenomenon for the sorbents such as ion-exchanged zeolites that have
heterogeneous sites. The heats of adsorption on activated carbon, in particular, ranged
widely fkom 23.9 kcaYmol (at OSmmoVg loading) to 8.0 kcaYmol (at 3mmoVg). Ag-Y
and Cu(1)-Y exhibited the higher heats of adsorption than Na-Y for both benzene and
thiophene because of .rr-complexation. More importantly, the heats of adsorption for
thiophene were higher than that of benzene. These experimental results can be explained
by molecular orbital calculation and NBO analysis, which will be discussed shortly.
At the low loading of 0.5 mmoVg, Na-ZSM-5 showed nearly the same heats of
adsorption as Na-Y for both thiophene and benzene. The different pore dimensions
(5.2-5.6 A for ZSM-5 vs. 7.4 A for Na-Y) apparently had no influence on the heats of
adsorption. It is not clear why the amounts adsorbed on Na-Y decreased sharply at very
low pressures while that on Na-ZSM-5 maintained.
Table 1. Heat of Adsorption (kcal/ml)calculated from isothcm at difhent temperatures.

Na-Y Ag-Y Cu-Y BUSY Na-ZSM5 ActivuedC.rbon AaivakdALumi~

(SilAk2.43) (SUAk2.43) (SilAk2.43) (Si/Al=195) (SilAkIO) T y ~ PCB

e Selcxsorb CDX

Thiophc~c 19.1-19.6 21.3-21.5 20.8-22.4 7.9-11.2 18.6-19.2 8.0-23.9 16.1-17.5

(0.5-2.0) (1.5-1.7) (2.0-3.0) (0.1-0.3) (0.45-0.60) (0.5-3.0) (0.2-1.O)

Benzea~ 17.0-18.2 19.0-20.1 19.3-21.8 6.6-13.1 16.5-17.9 13.1-16.1 16.8-19.6

(1.5-2.0) (1.5-1.8) (1.8-2.5) (0.1-0.9 (0.45-0.65) (1.0-3.0) (0.6-1.0)

*) Numbers in parcnthtses indicate the adsorption ~moullls
(mmoVP, for calculation.
**)The hcpts of adsorptiondmeased with loading in all caacs.

Bond Energies, Geometries and NBO Results

The energies of adsorption are summarized in Table 2. The theoretical calculations
indicate that the n-complexation strengths follow the order CuZ > AgZ and more
importantly,thiophene > benzene. This trend is in agreement with the experimental data,
in Table 4. In fact, the molecular orbital results on CuZ and AgZ are in excellent
agreement with the experimental data. Both chloride and zeolite models were used as
the anion in the theoretical calculations, while only zeolite framework was the anion in the
experiment. It is known fiom our previous work that the anion has a large effect on the
n-complexation bonds [11,27]. The bond energies on the zeolites (Z)are significantly
higher than that on the chlorides (Table 2). This result indicates that the zeolite anion is
more electronegative than the chloride anion, which has been already revealed by Chen
and Yang in their ab initio molecular orbital calculations [27]. In the optimized structures
of thiophene-MC1complexes, the distance between the thiophene molecule and Cu ion is
about 0.3 A shorter than that of thiophene and Ag ion for chloride. The NBO analysis
for thiophene adsorption is summarized in Table 3. There is some donation of electron
charges from the n-orbital of thiophene to the vacant s orbital of metals known as ci
donation and, simultaneously, back donation of electron charges fiom the d orbitals of
metals to x* orbital (i.e., anti-bonding x orbital) of thiophene or x back-donation. It
appears that the o donation is more predominant for thiophene and the x back-donation is
more important for benzene (published elsewhere). Comparing the two anions, zeolite
anion and chloride anion, the NBO results show that both ci donation and d z* -
backdonation are significantly stronger with the zeolite anion bonded to Agf or Cu'. The
charge transfer results again confirmed the experimental data that the relative strengths of
the n-complexation bonds follow the order: thiophene > benzene and Cu' > Ag'.

57
 Table 2. Summary of energies of adsorption for thiophene and benzene in kcaUmol calculated
from molecular orbital theory (Z denotesZeolite anion)
E,,(Thiophene) E,,(Benzene)
CUCl 13.5 12.4
AgCl 9.0 8.6
CUZ 21.4 20.5
A@ 20.0 19.1

Table 3. Summary of NBO analysis* of n-complexationbetween thiophene and MCVMZ

C+M interaction M+C interaction
(adonation) -
(d x'backdonation) Net Change
Mx 91 e 91+e
CUCl 0.037 -0.022 0.015
AgCl 0.022 -0.014 0.008
CUZ 0.112 -0.063 0.049
A@ 0.101 -0.086 0.015
~ ~~ ~~ ~ ~~

* q I is the amount of electron population change in valence s orbitals of the metal, and e is the
total amount of electron populationdecrease in valence d orbitals of the metal.

Liquid Phase Aakorption

Figures 5 and 6 show breakthrough curves obtained for 2000 ppmw thiophene (760
ppmw s u l k basis), for n-octane as solvent (the breakthrough curve for Cu(1)-Y is shown
in a separate figure for clarity since the abscissas are quite different). All adsorbents
showed remarkable selectivity towards C41&S, indicating that CgHIg adsorption is not
competitive. Saturation adsorption capacities calculated from the breakthrough curves
were 1.05 and 0.90 mmoVg for Na-Y and Ag-Y, respectively. However, for Na-Y, the
breakthrough of thiophene molecules occurred earlier, at about 2.84 cm3/g compared to
22.50 cm3/g in Ag-Y. This was evidence of weak adsorbate-adsorbent interactions on
Na-Y, which did not have the ability for n-complexation as in the case of Ag-Y. This
agrees very well with the pure vapor phase adsorption data reported above. The saturation
adsorption amount in Na-Y was higher than that Ag-Y due to pore volume differences and
difference in the densities of zeolites.
For the same feed conditions described above, Cu(1)-Y showed again the highest
selectivity and capacities among the adsorbents studied. The saturation capacity was 2.55
mmoVg, which was more than twice the amount found for the other adsorbents, indicating
superior interaction with the thiophene molecules. For about 2 grams of Cu(I>Y, it took
more than 300 minutes for the thiophene molecules to break through the adsorbent at a
feed rate of 0.5 cm3/min (refer to figure 6). Saturation was reached after 600 minutes,
which was remarkable for such a small amount of adsorbent. A large amount of Cu2+ions
must have been reduced to Cu'. As mentioned earlier, Takahashi et al. and others [14]
have reported 50% auto-reduction of copper under helium/vacuum atmospheres after just
1-2 hours. The adsorbent used in this part of the work was exposed to helium at 450'C for
no less than 18 hours. Possibly longer activation time increased the amount of reduced
copper ions, while the adsorption behavior already indicates that the auto-reduction
process yields promising results.

58
 Figure 7 shows breakthrough curves for Cu(1)-Y for an influent containing 500 ppmw
thiophene (190 ppmw sulfur> in n-octane. The saturation capacity was reduced to 1.28
mmoVg, which was about 50% of the amount obtained previously with the 2000 ppmw
thiophene feed. This indicates that the equilibrium adsorption isotherm was not
"rectangular" in shape at low concentrations and rather showed a noticeable decrease in
adsorbed amount as one decreased the concentration. Despite this, the observed saturated
amount was not low, when taking into account that the thiophene concentration was 75%
less than the case discussed previously (i.e., 2000 ppmw).
Figure 7 also shows breakthrough curves after Cu(1)-Y adsorbent regeneration
(second cycles). Under an atmosphere of nitrogen at 350'C, the regenerated adsorbent did
not recover the original capacity. The new capacity for the adsorbent at saturation was
0.80 mmoVg, which was more than a 30% reduction fkom the original capacity. In fact, the
color of the adsorbent remained black, which indicated the presence of copper thiophene
complexes. Meanwhile, regeneration under air at 350'C followed by reactivation under
helium at 450'C recovered almost all of the original capacity. For this case, the observed
saturation capacity was about 1.20 mmoVg, which was only a 5% reduction from the
original capacity.
1 "
0"

0.8 0 0

0.2

0
0 10 20 30 40 50 60 70 0 20 40 60 80 100 120 140 160 180 200
cmVg (cumdative efauent volumdsorbentweight) cd/g (cumdative effluent volumdsorbent weid

Figure 5. Breakthrough of thiophene in a fixed-bed Figure 6. Breakthrough of thiophene in a

adsorber with Ag-Y 0 or Na-Y (0) adsorbents, fixed-bed adsorber with Cu(1)-Y adsorbent, with a
with a liquid feed containing 2000 ppmw (Ci) of liquid feed containing 2000 ppmw (Ci) of
thiophene in octane, at room temperature. thiophene in n-octane, at room temperature.
 0.4

o o
0.2 o . 6

0 50 100 150 200 250 300 350 400 0 5 0 1 0 0 1 5 0 # ) 0 2 5 0

cmVg (cumdative d u e n t vohmdsorben~weight cmV.3 (clumllative dfiucst v o l d s & t weight)

Figure 7. Breakthrough of thiophene in a fixed-bed Figure 8. Breakthrough of thiophene in a

adsorber with fresh (0) and regenerated (0,[7, fixed-bed adsorber with Cu(I)-Y adsorbent, with a
Cum-Y adsorbent, with a liquid feed containing 500 liquid feed containing 500 ppmw (C,) of thiophene,
ppmw (Ci) of thiophene in natane, at room 20 wt?hbenzene and 80 wt% n-octane, at room
temperahue. (a) Adsorbent regenerated in nitrogen temperature.
at 350C followed by re-activation in helium at
450C. 0 Adsorbent regenerated in air at 350C
followed by re-activation in helium at 450C.

For the final part of this study, it was desired to use mixtures with compositions
similar to that of transportation fuels. Gasoline contains about 20-30 wtoh aromatics, many
thiophenic compounds and 70-80% alkanes such as n-hexane and n-octane. The aromatic
contents in diesel and jet fuels are < 20%. Thus, a mixture containing 20 wt% benzene, 80
wt% n-octane, and 200 ppmw sulfiu (ca. 500 ppmw thiophene) was used to simulate
gasoline. Figure 8 shows a breakthrough curve for thiophene in such mixture after
adsorption at room temperature on Cu(1)-Y. The sulfur adsorption capacity was 0.44
mmoVg or about 1.4 wt% sulfur. The results are promising when compared to other
adsorbents used in previous studies. Ma et al. studied fixed-bed adsorption of thiophene
compounds fiom diesel and jet fuels using an undisclosed transition metal compound ( 5
wt% loading) supported in silica gel [28,29]. They obtained a saturation adsorption
capacity of 0.015 g of sulfur per cm3 of adsorbent which can be compared directly to our
results. Assuming that the density of the Cu(1)-Y is close to that of Na-Y (- 1.3 g/cm3),
which is lower than the actual value because sodium is lighter than copper, then the
observed saturation capacity in our case is approximately 0.018 g of sulfur per cm3 of
adsorbent. Ma et al. also showed that breakthrough occurs at about 20 cm3 effluent
volume for about 3.2 cm3 of the metal loaded silica gel compared to 30 cm3 effluent
volume for about 0.75 cm3 volume of Cu(1)-Y. Therefore, our adsorbent is capable of
processing more fuel with very low sulfur streams with less adsorbent material. It should
be mentioned that jet fuel and other fuels contain heavier thiophene compounds such as
benzothiophenes, dimethylthiophenes, and dimethylbenzothiophenes and these should
adsorb strongly in Cu(1)-Y. Breakthrough results on gasoline and other transportation
fuels will be published elsewhere shortly.
Conclusions

In this work, vapor-phase benzene/thiophene adsorption isotherms were investigated to

develop new sorbents for desulfurization. Among the sorbents studied, Cu(I>Y and Ag-Y
exhibited excellent adsorption performance (capacities and separation factors) for
desulfurization. This enhanced performance compared to Na-Y was due to the
x-complexation of thiophene with Cu' and Ag'. Molecular orbital calculations confirmed
the relative strengths of n-complexation: thiophene > benzene and Cu' > Ag'.
This work has also demonstrated that copper (auto-reduced) and silver exchanged
Y-type zeolites are excellent adsorbents for removal of thiophene tiom aromatic and/or
hydrocarbon mixtures, based on fixed-bed adsorption experiments. Both adsorbents were
capable of reducing sulfur content to values < 4 ppmw sulfur for long periods of time.
Cu(I>Y provided the best adsorption capacity both at breakthrough point and at saturation,
surpassing all other adsorbents by more than 50%. Regeneration of the copper based
adsorbents can be accomplished in air at 35072 which recovered almost all of the original
adsorption capacity. More studies will be needed in order to fully understand the effect of
copper loading and to include heavier thiophenes, which are abundant in liquid fuels.
Breakthrough results on gasoline and other transportation fuels will be published
elsewhere shortly.

Acknowledgements

-
Supports from NSF and DOE are acknowledged. A U.S. Patent is pending filed with U.S.
and foreign Patent Offices.

References

1. Farrauto, R. J.; Bartholomew, C . H. Fundamentals of Industrial Catalytic

Processes, Chapman and Hall, New York, 1997.
2. Parkinson, G., Diesel Desulfurization Puts Refiners in a Quandary. Chemical
Engineering, 2001, February issue, 37.
3. Yang, R. T.;Takahashi, A.; Yang, F. H., New Sorbents for Desulfurization of Liquid
Fuels by n-Complexation. Ind Eng. Chem. Res., 2001,40,6236.
4. Yang, R.T.; Takahashi, A.; Yang, F.H.; Hernandez-Maldonado, A. Seldctive Sorbents
for Desulfurization of Liquid Fuels. U.S. and foreign Patent applicationsfiled, 2002.
5. Weitkamp, J.; Schwark, M.;Emest, S. Removal of Thiophene Impurities from
Benzene by Selective Adsorption in Zeolite ZSM-5. J, Chem. SOC.Chem. Commun.,
1991,1133.
6. King, D. L; Faz,C.; Flynn, T. Desulfurization of Gasoline Feedstocks for Application
in Fuel Reforming. SAE Paper 2000-01-0002, SOC.Automotive Engineers, Detroit ,
2000.
7. Salem, A. S. H.; Hamid, H. S. Removal of Sulfur Compounds fiom Naphtha
Solutions Using Solid Adsorbents. Chem. Eng. Tech., 1997,20, 342.
8. Irvine, R L.,Process for Desulfirizing Gasoline and Hydrocarbon Feebtocks, U. S .
Patent 5,730,860(1998).

61
9. Yang, R.T.; Kikkinides, E. S. New Sorbents for Olefin-Param Separations by
Adsorption via p-Complexation AZChE J., 1995, 4I , 509.
10. Rege, S. U.; Padin, J; R; Yang,, R T. Olefm-Paraffin Separations by Adsorption:
Equilibrium Separation by n-complexation vs. Kinetic SeparationAZChE J., 1998,44,
799.
1 1 . Huang, H.Y.; Padin, J.; Yang, R. T. Ab Initio Effective Core Potential Study of
Olefin/Paraffin Separation by Adsorption via n-Complexation: Anion and Cation
Effects on Selective Olefin Adsorption J. P h s . Chem. B., 1999,103,3206.
12. Padin, J.; Yang, R T. New Sorbents for Olefin-Parafin Separations by Adsorption
via n-Complexation: Synthesis and Effects of Substrates Chem. Eng. Sci., 2000, 55,
2607.
13. Jayaraman, A; Yang, R. T.; Munson, C. L.; Chinn, D. Deactivation of
n-Complexation Adsorbents by Hydrogen and Rejuvenation by Oxidation, Znd Eng.
Chem. Res., 2001,40,4370.
14. Takahashi, A.; Yang, R T. New Adsorbents for Purification: Selective Removal of
Aromatics, AIChE J., 48,1457 (2002).
15. Frisch, M. J. et al., Gaussian 94, Revision B.3, Gaussian, Inc., Pittsburgh, PA., 1995
16. Bowie, J. E., Data Visualization in Molecular Science: Tools for Insight and
Innovation; Addison-Wesley Pub. Co.: Reading, Mass., 1995; chapter 9.
17. Hay, P. J.; Wadt, W. R Ab Inifio Effective Core Potential for Molecular
Calculations. Potentials for the Transition Metals Atoms Sc to Hg. J. Chem. Phys.
1985,82,270
18. Wadt, W. R.; Hay, P. J. Ab Inifio Effective Core Potential for Molecular
Calculations: Potentials for Main Group Elements Na to Bi. J. Chem. Phys. 1985, 92,
284
19. Hay, P. J.; Wadt, W. R. Ab Initio Effective Core Potential for Molecular
Calculations: Potentials for K to Au Including the outermost core orbitals. J. Chem.
Phys. 1985,82,299
20. Russo, T. V.; Martin, R. L.; Hay, P. J. Effective Core Potentials for DFT
Calculations. J. Phys. Chem. 1995,99, 17085
21. Becke, A. D. Density Functional Thermochemistry. 111. The Role of Exact
Exchanges. J. Chem. Phys. 1993,98,5648
22. Becke, A. D.; A New Mixing of Hartree-Fock and Local Density-Functional
Theories. J. Chem. Phys. 1993,98, 1372
23. Becke, A. D.; Density-Functional Thermochemistry. 11. The Effect of the
Perdew-Wang generalized-gradient correlation correction. J. Chem. Phys. 1992, 97,
9173
24. Lee, C.; Yang, W.; Parr, R.G. Development of the Colle-Salvetti Correlation-Energy
Formula into a Functional of the Electron Density. Phys. Rev., 1988, B37,785
25. Reed, A. E.; Weinstock, R.B.; Weinhold, F. Natural Population Analysis. J. Chem.
Phys. 1985, 83 (2). 735
26. Glendening, E. D.; Reed, A. E.; Carpenter, J. E., Weinhold, F. NBO Version 3. I ,
1995.
27. Chen, N.; Yang, R. T. Ab Inifio Molecular Orbital Study of Adsorption of Oxygen,
Nitrogen, and Ethylene on Silver-Zeolite and Silver Halides. Ind. Eng. Chem. Res.
1996,35,4020.

62
28. Ma, X.; Sun, L.; Yin, Z.; Song, C. New Approaches to Deep DesulfUrization of
Diesel Fuel, Jet Fuel, and Gasoline by Adsorption for Ultra-Clean Fuels and for Fuel
Cell Applications. Am. Chem. SOC.Div. Fuel. Chem. Prepr. 2001,46,648.
29. Ma, X.; Sprague, M.; Sun, L.; Song, C. Deep Desulfurization of Liquid
Hydrocarbons by Selective Adsorption for Fuel Cell Applications. Am. Chem. Sac.
Div. Pet. Chem. Prepr. 2002,47,48.

63
 OPTIMIZATION OF CONTINUOUS CHROMATOGRAPHY SEPARATIONS

ZIYANG ZHANG AND M. MORBIDELLI

Swiss Federal Institute of TechnologyZurich, Laboratoriumfir Technische Chemie/LTC,
ETH-HiinggerbergIHCI, CH-8093 Ziirich, Switzerland
E-mail: [email protected]

M.MAZzOTrI
ETH Zurich ,Institut f i r Verfarenstechnik, Sonneggstrasse3, CH-8092 Zurich, Switzerland
E-mail: manotti@ivuk mavt.ethz.ch

Two recent developments of the simulated moving bed chromatographic separation units, i.e. the
Varicol and the PowerFeed processes, are addressed. The performances of these three processes are
compared with reference to two chiral separation systems taken from literature, using a multiple
objective optimization technique based on a genetic algorithm. The performance of each process has
been optimized in a wide range of operating conditions, and from their comparison a good assessment
of their relative potential has been made. The optimization results have been discussed in the frame of
equilibrium theory and the N I ~ S of optimal design of the Varicol and PowerFeed processes have beem
discussed.

1 Introduction

Simulated Moving Bed (SMB) is an established technology for performing continuous

chromatographic separations covering all scales of possible interest in applications,
particularly in optical enantiomer separations. The SMB unit has been originally devised as
a practical realization of a true moving bed (TMB) unit where the two phases move
countercurrently. A schematic diagram of a typical 4-section SMB is shown in Figure 1(a).
The countercurrent movement of the solid and the fluid is simulated by moving
synchronously the inlet (feed and eluent) and outlet (raffinate and extract) ports by one
column in the direction of the fluid flow, with a predetermined period or switching time, 5.
The design and optimization of this unit can be done in the frame of equilibrium theory,
using the so called triangle theory [ 1,2] or using more detailed simulation models in
connection with various optimization strategies [3-71.
In order to make SMB units more economically efficient and competitive, several new
operation modes have been introduced. These include supercritical fluid SMB [&lo],
temperature gradient SMB [ 111 and solvent gradient SMB [12-14]. The basic idea is to
change the adsorption strength of the solute on the stationary phase in the different sections
by creating along the unit a gradient of either pressure or temperature or solvent
composition, respectively. Another direction which has been taken to improve SMB
performance is related to the idea of somehow forcing its dynamics. In this context, the SMB
unit is not regarded as an approximationthrough appropriate discretization of the TMB unit,
but is considered as a unit with many degrees of freedom that can be optimized to improve
its performance. The first step in this direction is the Varicol unit [ 151, where the inlet and
outlet ports are shifted asynchronously. An alternative operation has been proposed by
Kloppenburg and Gilles [ 161 and more recently by Zang and Wankat [171, who considered
fluid flow rates changing during the switching period as shown in Figure l(b). This will be
referred to in the following as PowerFeed operation. It is worth noting that, these two
ideas, Varicol and PowerFeed, can be regarded in some sense as having a common root, i.e.

64
changing within the switching period the flow rates, of the solid and the fluid, respectively.
In this sense, as mentioned above, they do not try to better approximate the TMB unit from
which the SMB is derived.

t -
----
SMB
Powdd

sectiw 2, Q s

Figure 1. (a) Operating diagram of a four section SMB unit; (b) Fluid flow rates schemes in the SMB and
PowerFeed operations during one switching period, t,
In this paper, we present and compare the optimal performances of the SMB,Varicol
and PowerFeed operations, using a multiobjective optimization technique, on two chiral
separation systems from the literature. The aim is to provide a clear picture, although
inevitably confined to the cases examined, of the relative potentials of these three operation
modes.

2 Optimization of SMB,Varicol and PowerFeed processes

Optimization of the SMB,Varicol and PowerFeed processes is very complex due to the
large amount of continuous and discrete parameters involved. These parameters include
fluid flow rates, switching time, unit configuration (total number of columns, column
distributionand column dimension), feed conditions, size of the packing particles and so on.
These parameters might have different values in the subintervals of one switching period in
the Varicol and PowerFeed operations. Depending upon the specific application, the best
process performance may be achieved by maximizing the product quality (mostly in terms
of the purity of either extract or raftinate stream or both) under fixed cost and productivity,
or by minimizing the cost and at the same time maximiig the productivity under some
given specification on product quality, or by other combinations of practical interest. Note
that in most cases, two of the objective hctions to be optimized are conflicting, as for
example: productivity increases if the unit operates at higher system fluid flow rates which,
however, imply a decrease in column efficiency, and therefore in the product purities.
Furthermore, various practical constraints on the column or entire unit pressure drop,
switching time, pump flow rates and product quality etc. complicate the optimization
problem.
In our earlier works [6,7],we have developed a new optimizationprocedure based on a
genetic algorithm, i.e. the non-dominated sorting genetic algorithm (NSGA), that allows to
handle these complex optimization problems. A more detailed description of this global
search and Optimizationtechnique is available elsewhere [18, 191.

65
3 Results and discussion

3.1 Comparison of SMB and Varicolfor diflerent total number of columns

An important aspect in comparing SMB and Varicol processes is the effect of the total
number of columns, Ned.It is in hct expected that as N,, increases, the discretizationof the
movement of the solid in the SMB improves and thereforethe main advantage of Varicol, i.e.
the possibility of better tuning the distribution of the columns in the sections by using non
integer values, becomes less effective. We consider in this work a single objective
optimization problem using as a model system the chiral separation reported by Biressi et al.
[3], aiming at investigating the effect of NW1on the performance of these two processes,
according to the following statement of the optimization problem:
Maximize J = PE[QI, ml, m2, m4, XI
Subject to P R 2 90% and APdt 5 70 bar
= ~dp=30 pn, N,,=4
Vsotid=120ml, R=l cm2,F=0.7 ml/min, C T ~ gA, -8
where the extract purity, PE is the objective to be maximized under the following
constraints: minimum 90% raffiate purity, PR,maximum 70 bar pressure drop along the
entire unit, Munit,given total solid volume, Vlolid,given production rate (i.e. faed feed flow
rate, F and total feed concentration, C,) and given packing particle diameter, d,. Having
fixed also the column cross section a,changes in NCdimply changes in column length. The
optimization variables are the flow rate in section 1, Q1,the flow rate ratios, ml,m2 and m4
[2], and the unit configuration represented by the parameter, x. For the SMB this can be
represented in the form of n1/n2/n3/n4where njrepresents the number of columns in section
j. For the Varicol operation the decision variables are the same, but now x can attain a much
larger number of values, which is in fact determined by the number of subintervals
considered in each switching interval. Thus for example, the complete configuration of a
4-subinterval Varicol unit with NC.l=5 can be described by the configuration sequence
2/1/1/1-1/2/1/1- 1/1/2/1-1/1/1/2 fiom the first subinterval to the last.
The equilibrium stage model reported by Zhang et al. [7], which accounts for the
influence of fluid flow rate on the column efficiency, has been used for the optimization
simulations.
The optimization results for a SMB and a 3-subinterval Varicol unit are compared in
Table 1 for various values of NWI.

Table 1. Optimization results of SMB and Varicol processes for various values ofNml

Process I N
, ,L Qi rnl rnz mc x mun
ti PR PE
I (crn) (mumin) (barj % %
I 4 30 30.328 1.550 0.867 0.754 1/1/1/1 52.79 90.02 86.49
5 24 30.747 1.567 0.843 0.738 1/2/1/1 52.14 90.00 91.68
SMB 6 20 35.126 1.512 0.900 0.594 1/2/2/1 60.47 90.07 96.04
7 17.14 34.735 1.584 0.888 0.525 1/3/2/1 57.86 90.02 97.13
8 15 36.595 1.643 0.900 0.444 1/3/3/1 59.86 90.06 97.29
4 30 29.502 1.747 0.837 0.680 * 48.40 90.01 91.55

varicol 5 24 35.761 1.538 0.895 0.663 ** 61.81 90.00 94.22

011~11-012/111-1111111;** 1/1/2/1-112/111-1/2/1/1

66
 It is seen that PEincreases with increasing number of columns, particularly at the lower
values of N d . PEincreases in fact by almost 10%from 86.5% for Nc0,=4to 96.0% for NcOl=6.
A further increase in total number of columns has a smaller influence on PE.In addition,
from a practical point of view, it is worth noting that the short switching time values caused
by the short column lengths may lead to difficulties in the operation of the recycle pump.
With respect to column configuration x, the results in Table 1 indicate that all the additional
columns tend to distribute equally between sections 2 and 3. This is because, once the lower
bound on ml and the upper bound on m4 are satisfied, sections 2 and 3 are the most
important in determining PEand PR.The mivalues in Table 1 indicate that indeed in the
cases under examinationboth such constraints are satisfied. In Table 1 the performances of
a 3-subinterval Varicol with 4 and 5 columns are also shown. It is seen that Varicol
outperforms SMB both for NCoI=4,and to a lower extent, for NeoI=5. For Nco1>5, the
optimization technique has not been able to find a Varicol configuration which improves PE
over the corresponding SMB value. This means that, as expected, for increasing total
number of columns, the Varicol configuration is not worthwhile anymore.
It is worth noting that in the case of N,,=4, the following five possible configurations
have been considered for Varicol: 0/1/2/1, 0/2/1/1, 1/2/1/0, 1/1/2/0 and l/l/l/l. The first
four of such configurationshave only three sections, which actually were not considered in
the cases where NmI>4.The configuration 0/1/2/1~0~/1/1-1/1/1/1 was found to be optimal
for the 3-subinterval Varicol, which improves PEby about 5% over the corresponding SMB
with a 1/1/1/1 configuration. It is interesting to note that in the first two subintervals there is
no column in section 1; the fourth column moves in fact from section 3 in the first
subinterval to section 2 in the second and eventually to section 1 in the last subinterval. This
corresponds to the time averaged configuration 0.33/1.33/1.33/1, which compared to the
configuration 1/1/1/1 ofthe SMB, indicates that in this particular case Varicol improves the
performance of the separation by reducing the length of section I and increasing that of
sections 2 and 3.

3.2 Comparison of SMB and Varicolfor direrent product purity requirements

Another situation of interest in applications is one where the product purities are fixed, and
the objectives for optimal process operation are to reduce operating costs and increase
production. Hence, in this case, for a fixed target product purity of both extract and raffiate
streams, we seek the optimal process parameters for a SMB and a 4-subinterval Varicol unit,
which maximize production using minimum amount of eluent, for another chiral separation
system taken fiom literature [6,15]. The optimization problem is represented
mathematically as follows:
Max J I = F CQz, F, D,t XI
Min Jz = D [Q2, F, D,4, XI
Subject to PE= x f 0.002, x = 0.90,0.95,0.99
P R = Xf0.002,~=0.90,0.95,0.99
Q1= 27.5 mYmin, NWI = 5 , Lcol= 0.1 m, NNTP=80
where Q1,the column flow rate in section 1, is set at 27.5 ml/min to fix the maximum column
pressure drop, and the total solid used is also set by fixing total number of columns, NmIand
column length, L,I. The decision variables are the column flow rate in section 2, Q2, feed
flow rate, F,eluent flow rate, D, switching time, t, and column configuration parameter, x.
Note that the two variables F and D appear also in the objective functions. The Pareto

67
optimal solutionsare shown in Figure 2 for purity requirementsfor both extract and rafiinate
streams of 90%, 95% and 99%.

0
P. and PI = M% PrandPm=SB%
.....................................................

....................................................... 8 ....................................................
................................................... 0

.... 5.8

................... ................................
0 vuisol
5.4
.o..............

.T
............ ........

2.2 2.3 2.4 2.5 2.8 2.7 2.0 1.5 1.8 1.7 1.8 1.9 2 0.4 0.8 0.8 1 1.2
C. r U m b F, mVhlln F, mumln

Figure 2. Optimal solutions for the SMB and Varicol processes with 9004 95% and 99% purity requirements.
In all cases, using 5 columns, the optimal column configurationsare found to be 1/2/1/1
and 1/1/1/2-1/1/2/1-1/1/2/1-1/2/1/1 for the SMB and 4-subinterval Varicol unit,
respectively. It can be observed that for fixed purity specifications, both the SMB and the
Varicol processes require to increase the eluent consumption in order to increase the feed
flow rate. Secondly, the Varicol process consumes less eluent, D than the SMB process for
the same feed flow rate, F;or equivalentlyfor the same eluent consumption, D,the Varicol
process can treat more feed, F.However, the extent of improvementdepends on the purity
specifications. The more stringent the purity requirement, the larger the improvement
achieved by Varicol over SMB. For example, at D = 5.6 mymin, the improvement in
production rate, F of Varicol over SMB is lo%, 25% and 127% for a purity requirement
both in the extract and in the d m a t e streams of 90%, 95% and 99%, respectively. Finally,
it is seen fkom Figure 3 (for the case of purity requirement of 95%)
42 - 4.2
SYB VIriCol
4 .............. ............ 0............ ?..........................
~ 4 ..............................................................
. e
3.8 ............................................................. em 3.8 .......................................................

3.6 ............................................................ .......................................................

E 3.4 .............................................................
A A A
3.4 .... .... ~L
...... . . ~..... d ..... .... *----
.

3 2 ...................................................................... 32 ........................................................

f g............0...........................
0 0 0 0 0
3 ............x............ x.............x .....x ..... x... r ......
3 ................... I......
X
2.8 4 1 2.8 A
1.5 1.8 1.7 1.8 1.0 1.5 1.8 1.7 1.8 1.9 2
F, mllmln F. mllmln

Figure 3. Flowate ratio parameters ml to nu for the SMB and Varicol processes with 95%purity requirement.

that the optimal values of the flow rate ratios in sections 2 and 3, m2 and m3, in the SMB and
in the Varicol process are very similar (although the corresponding optimal performances
are different) and change very little as the feed flow rate increases. This is consistent with
the triangle theory, which indicates that the optimal operating point in the (mz, m3) plane is
independent of the feed and eluent flow rates [1,2].

68
3.3 Comparison of SUB, Varicol and PowerFeed
The separation problem examined in this case requires the simultaneous maximization of
the rafiate (PR)and the extract purity (PE)for a given feed flow rate, F,eluent flow rate, D
and fixed configuration of the unit, for the same chiral separation considered in section 3.2.
This optimization problem in the case of PowerFeed operation can be represented
mathematically as follows:
Max Ji = PR[Qz,~,.*., Qz,s, Fi, Fs-I,DI,*-,b - 1 , t XI
MaX Jz = PE[Qz,i, .**, Qz,sy FI, *-) Fs-I,Di,--, Dsi, 4, XI
Subject to PR2 90%
P E 90%
~
Q1=27.5ml/min, N,I=~ or 6, D,,=6.24 ml/min, Fa,, NNTp
where Q2s, Fi and DI represent the flow rate of section 2, the feed flow rate and the eluent
flow rate in the P subinterval,respectively. The average feed and eluent flowrates (Fave and
Dave respectively) are fixed; therefore the flow rates in the last subinterval are not
independent since they are determined by the corresponding values in the previous
subintervals. Note that in this case it has been assumed that the flowrates Q2, F, D and
consequently Q3, Q4,R and E, change in S subintervals. The problem can be simplified by
changing in time less flow streams, e.g., only F,which implies the change of Q3 and R.It is
to be noted that in the problem above both the SMB and the Varicol operation modes have
three decision variables, i.e. Qz,t,and x. The same stage in series model described by Zhang
et al. [6]has been adopted to simulate SMB, Varicol and PowerFeedprocesses, with a slight
revision, which enables the column flowrates to change in time.
A comparison of the SMB, Varicol and PowerFeed operations has been conducted for
two sets of operatingconditions, one with FBve=1.62ml/min and NNTP=80and the other with
F,,,=2.2 ml/min and Nm=60. Two different PowerFeed configurations, one varying F in 4
subintervals ( S 4 ) and the other varying Q2,F and D in 3 subintervals(S=3), are considered
for the two cases, respectively. In the first configurationthe decision variables reduce to Q2,
FlyF2, F3, t,and x, while in the second they are Qz,,, Q ~ J Q , F2, DI, D2, t, and x. The
, ~ JF1,
optimization results of a 5-column PowerFeed unit are compared to those of the
corresponding SMB and Varicol units in Figure 4(a) and 4(b) for the two sets of operating
conditions, respectively. It is interesting to observe that firstly, in both cases for the SMB,
Varicol and PowerFeed units,we obtained Pareto solutions. Secondly, each Pareto has at
least one discontinuity due to change in column configuration x. This is due to the fact that
different purity requirement requires different column configuration. Thirdly, the 5-column
Varicol and PowerFeed processes perform better than the corresponding 5-column S M B
process. Finally, as the difficulty of the separation increases, as shown in Figure 4(b), the
PowerFeed performs better than the Varicol, and is even comparable to the 6-column S M B
at the two ends of the Pareto curve.
A better understanding of these results can be achieved by reasoning in terms of the
flowrate ratio parameter, m as shown in Figure 5, taking as an example the separation case
shown in Figure 4(a). It can be observed from Figure 5(a) that m2 and m3for the 5-column
SMB, Varicol and PowerFeed units are almost constant, although they tend to decreases
slightly as PRincreases. This is in agreementwith triangle theory as discussedby Bang et al.
[6].For the 5-column PowerFeed unit, the m3 values changing in time are shown in Figure
5(b). It is interestingto note that the m3of the SMB process, goes very closely to the average
m3 of the PowerFeed process, m3ave and to a slightly less degree to the m3of the Varicol

69
 process due to the column configurationchange of the Varicol operation, as shown in Figure
5(a). Therefore, the optimal design of the Varicol and PowerFeed processes can be replaced
by the optimal design of the corresponding SMB process given by the equilibrium theory,
followed by the search of the optimal ports switching and flow rates variation schemes
within the switching period, respectively.

0.99 --" O O O OD=-

0.98 -. 0.98

0.87 ..
0.96 -- 0.96

2" 0.95 -.
0.94 -. 0.94

0.83 -.
0.92 -. 0.92

0.91 -.

0.94 : : : ~ ; : : i i
. ,i
I 0.9
0.9 0.91 0.92 0.93 0.94 0.85 0% 0.97 0.88 0.99 1 0.9 0.82 0.94 0.96 0.98 1
PRI' PR
Figure 4. Optimal solutionsfor the SMB, Varicol and PowerFeedprocesses (x changes with increasing PR:5-col
SMB and PowerFeed: (a) and (b) 1/2/111-+1/1/2/1; 6-column SMB: (a) 1/212/1+211/2/1, (b) 1/2/2/1; 5-col
Varicol: (a) 111/~1-111/2/1-112/1/1-1/2/111+1/1/211-1/1/2/1-1/1/211-1~/1/1~1/112/1-1/11211-1/11211-2111111,
(b) 1~11211-112/1~1-1/2/111+1/1/2/1-1/1/211-1/2/1/1).
3.9 3.8
(a) comparison of the m2 and m3 (b) m,in lhe 4 subintervals oflhe P o w i f f e e d process
3.7 3.7

3.6
3.5
" w W ' A S ~ $ = M 3.5
E 3.3
U
E" 3.4
(.3.1
E 3.3

2.8
3.2
A SMB, m3 AVarkol n0 rFvwerFeed.ave tr3
2.7 3.1
0 SMB,nQ Varkol nQ FuwerFeed, nQ

2.5 3
0.8 0.91 0.92 0.93 0.94 0.95 0.88 0.97 0.88 0.99 1 0.9 0.91 0.92 0.93 0.94 0.95 0.88 0.97 0.88 0.99 1
PR PR
Figure 5. Comparison of (a) the average m 2 and m3 among the 5-column SMB, the Varicol and the PowerFeed
processes and (b) Values of m3 in the 4 subintervals of the PowerFeed process corresponding to the points in
Figure 4(a).

4 Conclusions

Our results show that the Varicol and the PowerFeed operations improve the performance of
the SMB process, particularlywhen the total number of columns is small and the separations
are difficult. The optimal design of the Varicol and PowerFeed processes can be replaced by
the optimal design of the corresponding SMB process given by the equilibrium theory,
followed by the search of the optimal ports switching and flow rates variation schemes
within the switching period, respectively.

70
References
1. Storti G., Mazzotti M., Morbidelli M. and Carra S., Robust design of binary
countercurrent adsorption separation processes, AICHE J. 39 (1993) pp. 471-492.
2. Mazzotti M., Storti G. and Morbidelli M., Optimal operation of simulated moving bed
units for nonlinear chromatographic separations, J. Chromafogr. A 769 (1997) pp.
3-24.
3. Biressi G.,Ludemann-HombourgerO., Mazzotti M., Nicoud R.M. and Morbidelli M.,
Design and optimization of a simulated moving bed unit: role of deviations from
equilibrium theory, J. Chromatogr. A 876 (2000)pp. 3-15.
4. Klatt K.U., Hanisch F. and Dunnebier G., Model-based control of a simulated moving
bed chromatographic process for the separation of hctose and glucose, J. Process
Contr. 12 (2002) pp. 203-219.
5. Ruthven D.M. and Ching C.B., Counter-current and simulated counter-current
-
adsorption separation processes, Chem. Eng. Sci. 44 (1989) pp. 1011 1038.
6. Zhang Z., Hidajat K., Ray A.K. and Morbidelli M., Multiobjective optimization of
simulated moving bed system and Varicol process for chiral separation, AZChE J.
(2002) in press.
7. Zhang Z., Mazzotti M. and Morbidelli M., Multiobjective optimization of SMB and
Varicol processes using genetic algorithm,J. Chromafogr.A (2002) in press.
8. Mazzotti M., Storti G.and Morbidelli M., Supercritical fluid simulated moving bed
chromatography,J. Chromafogr.A 786 (1997) pp. 309-320.
9. Di Giovanni O., Mazzotti M., Morbidelli M., Denet F., Hauck W. and Nicoud, R.M.,
Supercritical fluid simulated moving bed Chromatography 11. Langmuir isotherm, J.
Chromafogr.A 919 (2001) pp. 1-12.
10. Denet F., Hauck W., Nicoud R.M., Di Giovanni O., Mazzotti M., Jaubert J.N. and
Morbidelli M., Enantioseparation through supercritical fluid simulated moving bed
(SF-SMB) chromatography,Ind. Eng. Chem. Res. 40 (2001) pp. 4603-4609.
11. Migliorini C., Wendlinger M., Mazzotti M. and Morbidelli M., Temperature gradient
operation of a simulated moving bed unit, Ind. Eng. Chem. Res. 40 (2001) pp.
2606-2617.
12. Jensen T.B., Reijns T.G.P.,Billiet H.A.H. and van der Wielen L.A.M., Novel
simulated moving-bed method for reduced solvent consumption, J. Chromafogr.A 873
(2000) pp. 149-162.
13. Antos D.and Seidel-MorgensternA., Application of gradients in the simulated moving
bed process, Chem. Eng. Sci. 56 (2001) pp. 6667-6682.
14. Abel S., Mazzotti M. and Morbidelli M., Solvent gradient operation of simulated
moving beds I. Linear isotherm, J. Chromafogr.A 944 (2002) pp. 23-29.
15. Ludemann-Hombourger O., Nicoud RM. and Bailly M., The Varicol process: a new
multicolumn continuous chromatographic process, Sep. Sci. Technol. 35 (2000) pp.
1829-1862.
16. Kloppenburg E. and Gilles E.D., A new concept for operating simulated moving-bed
processes, Chem. Eng. Technol. 22 (1999) pp. 8 13-817.
17. Zang Y. and Wankat P.C., SMB operation strategy-Partial feed, Znd. Eng. Chem. Res.
41 (2002) pp. 2504-25 1 1.
18. Bhaskar V., Gupta S.K. and Ray A.K., Applications of multi-objective optimization in
chemical engineering,Rev. Chem. Eng. 16 (2000) pp. 1-54.
19. Srinivas N. and Deb K., Multiobjective function optimization using nondominated
sorting genetic algorithms, Evol. Compuf.2 (1 995) pp. 22 1-248.

71
 ADSORPTION TECHNOLOGY FOR GAS SEPARATION

SHIVAJI SIRCAR
Department of Chemical Engimering, Lehigh University
11 1 Research Drive, Iacocca Hall, Bethlehem, Pa 18015-4791, USA.
E-mail: [email protected]

Separation and purification of gas mixtures by selective adsorption on micro-meso porous

solid adsorbents such as zeolites, activated carbons, silica and alumina gels, polymeric
sorbents, etc., has found numerous commercial applications in the chemical,
petrochemical, environmental, medical, and electronic gas industries. Table 1 lists some
of the key uses of this technology [I]. Two generic cyclic process concepts called
Temperature Swing Adsorption (TSA) and Pressure Swing Adsorption (PSA) are
generally employed. Each of these concepts have numerous variations depending on (a)
the product specifications, (b) the energy of separation, (c) the sequences and the modes
of operation of the steps of the process, (d) the types of adsorbent used, etc. The gas
purification applications typically use the TSA processes except for gas drying and
solvent vapor recovery applications where both the TSA and the PSA processes are used.
The bulk gas separation applicationsexclusively use the PSA processes.

Table 1. Key Commercial Applications of Gas Separation and Purification by Adsorption Technology

T 'lut R val
Trace Organicand inorganic Impurity Removal Air Separation (4 and NZfrom Air)
hDrYing Hydrogen and Carbon Dioxide Roduaion from
Air PollutionControl Steam-Methane Reformer W G a s
Nuclear Waste Management Roduction of Ammonia SynthesisGas
Solvent Vapor Recovery Hydrogen Recovery fromRefinery off Gases
Electronic Gas Purification Methane-CarbonDioxide w o n from Landfill Gas
Carbon Monoxide-HydrogenSeparation
-
N o d Isoparaffn Scparahon
rn Alcohol Dehydration

The interest and growth in the research and development of adsorptive separation
processes have been phenomenal. Table 2 lists the number of U.S. patents cited by the
Derwent Chemical Patent index between the years of 1980 and 2000 under the keywords
given in the table [I]:
Table 2. Results of U.S. Patent Search Between 1980 and 2000

Kevwords NumberofP@nts m r d s m a l t s
Gas Separation by Adsorption 3050 Air Separation by PSA 391
Adsorption for Air Pollution 1164 Hydrogen Purification by PSA 185
Ressun swing Adsorption 608 Gas Drying of PSA 32
Temperature Swing A m t i o n 60

Gas adsorption has become the state-of-the-art technology for (a) trace impurity removal,
(b) small to large scale (1 40,000 SCFH) gas drying, (c) small to medium scale (0.0 1-1 00
TPD) production of O2 (90+%) and N2 (99+%) from air, and (d) small to large scale

72
(1-100 MMSCFD) production of high purity H2 (99.999+%) from steam-methane
reformer and refinery off-gases. Research and development on adsorptive processes has
primarily been directed towards (a) producing purer products (single or multiple) from a
feed gas at higher recoveries (b) lowering the capital and operating costs of separation, (c)
designing novel hardware and process control systems, and (d) increasing the scale of
applications. Some of the key achievements include (a) lowering the specific power ( 4 2
KW/T/D) for production of 90% O2 from air below that of conventional cryogenic
distillation air separation route, (b) direct production of high purity N2 ( 4 0 0 ppm 02)
from air, and (c) production of high purity H2with high recovery (90+%). The primary
reasons for this spectacularp w t b in this area are given below [2]:

(a) There is an extra degree of thermodynamic freedom for describing adsorption

systems compared to those for conventional gas separation methods like distillation
and absorption. This introduces an immense flexibility in the design and operation
of adsorptive separation processes.
(b) Numerous families of porous adsorbents are available which offer multiple choices of
core adsorptive properties (equilibria, kinetics, and heats) for a given separation
application.
(c) A successfuladsorptive process is generally a good marriage between the optimum
adsorbent and the efficient process design.
(d) There can be many different paths (combinationsof materials and processes) for
achieving the same separation goals.

The above reasons are also the key driving forces for promoting innovations in this area.

The points discussed above can be demonstrated by the case of simultaneous production
of O2 and N2 enriched gases from ambient air. Air can be fractionated by selectively
(thermodynamic) adsorbing N2 over O2 and Ar on a zeolite [3], or selectively (kinetic)
adsorbing O2over N2 and Ar on a molecular sieve carbon [4,5].

Furthermore, many different process schemes for air separation can be developed using
different zeolites and molecular sieve carbons having different adsorptive properties.
For example, a zeolite like Na-mordenite having a moderate N2 selectivity of -4 over O2
at ambient conditions can be used in a four-step PSA process [3]. The cyclic steps
would include (a) adsorption of N2at near ambient conditions by flowing air through a
packed bed of the zeolite, while producing the O2enriched product gas at feed air pressure
(PA),(b) rinsing the adsorber cocurrently with a stream of essentially pure N2 and venting
the air-like effluent, (c) evacuating the adsorber counter-currently to pressure PD and
withdrawing a N2 rich product gas, a part of which is used in step (b), and finally (d)
counter-currently pressurizing the adsorber from PD to PA with a part of product gas
generated by step (a).

The N2 rinse step is needed to displace the co-adsorbed and the void 0 2 remaining in the
column at the end of step (a) so that the desorbed gas in step (c) is essentially pure N2.
This step can be eliminated if the N2 selectivity over O2by the zeolite is high (say >8) as
in the case of CaX zeolite. The desorbed gas, in this case, can be hctionated in order to

73
reject the N2lean earlier part, and collect the N2rich latter part as the N2product gas [3].
Higher Nz selectivity of an adsorbent allows the desorbed gas fractionation concept to be
practical, which leads to a simpler three-step process scheme, while meeting the product
specificationsof the former four-step cycle.

The cyclic steps of one of the PSA processes using the molecular sieve carbon as the
adsorbent consist of (a) flowing compressed air through a packed bed of the carbon so that
0 2 can diffuse and adsorb into the carbon pores faster than N2and Ar and produce a N2
rich product gas at feed air pressure (PA), (b) pressure equalizing the adsorber with a
companion adsorber, (c) counter-currently depressuring the adsorber to near ambient
pressure to produce the O2 enriched gas, (d) pressure equalizing with another adsorber,
and finally (e) repressurizingthe adsorber to PAwith feed air [4].

Table 3 shows an example of the comparative performances of these three processes. All
of them can produce a 99+%N2enriched product gas. The two zeolite processes also
produce a 85-90% O2 enriched product gas. The O2 product purity of the carbon sieve
process is, however, low. This shows that different adsorbents can be married with
different process schemes to obtain similar product purities but different process
performances (recovery, productivity, product pressure, etc.).
Table 3. ComparativePerformancesof Various Air Fractionation Processes

Process Adsorbent Purity Productivity Recovery Pressure

(%) (SCfh/P) (W (Psi@
(a) Na-Mordenite 84. I 31 63.0 2
(b) Ca-X 90.0 20 24.0 2
(c) Carbon Molecular Sieve 33.8 144 98.1 0

Process Adsorbent Purity Productivity Recovery Pressure

(a) Na-Mordenite 99.0 122 65.0 0
(b) Ca-X 99.0 83 30.0 0
(c) Carbon Molecular Sieve 99.0 92 49.4 104

TRENDS IN FUTURE ADSORPTIVE PROCESS DEVELOPMENTS

Two areas of development have attracted considerable attention in recent years. They are:

(a) Rapid PSA cycles for bulk gas separation

(b) Novel adsorbet designs

RaDid PSA
The concept is to use faster cycle times (seconds) than those used by the conventional
PSA cycles (minutes) in order to obtain a step change in the productivity of the process
(volume of productholume of adsorbenthour). Some of these designs use a
conventional PSA cycle but they are operated faster by appropriate changes in the
mechanical designs [6]. Others, propose novel process schemes in order to
accommodate fast cycle times [7-91. An example of the second case is a RPSA process
for air separation where a single adsorber vessel is packed with two or more (even
numbers) shallow layers of small zeolite particles (-0.5 mm). The layers are separated

74
by screens which act as pressure drop devices. The cycle steps consist of (a)
simultaneouslypressurizing and adsorbing N2from air on one layer of the adsorbent while
producing an O2 enriched product gas, and then (b) simultaneously depressurizing and
back purging the layer with a part of the O2 enriched product gas produced by a
companion layer [9].

Using a 5A zeolite as adsorbent and a total cycle time of 10 seconds (compared to

conventional PSA cycles of 60-240 seconds), the RPSA process could produce a 27.5%
O2enriched gas stream with an O2recovery of -64.1% from feed air at a pressure of 2.22
atm [3]. The 0 2 productivity rate was -2300 sfi?/ft3/hrwhich was an order of magnitude
higher than that of a conventional PSA. The process was found to be suitable for
producing 2340% 02 from air for enhanced combustion applications in cupolas,
metallurgical furnaces, etc. [lo].

Novel Designs
Conventional adsorber designs include vertical (length/diameter >1) or horizontal
(lengtlddiarneter 4)packed beds. The maximum permissible gas flow rates through
these vessels is governed by pressure drop and the possibility of local fluidization and
channeling [ 1 13. Some of these problems are solved by novel designs such as (a) radial
bed and (b) rotary bed adsorbers.

Radial Bed Adsorbers (PSA and TSA)

The adsorbent is placed in an annular section between two co-axial cylinders. The walls
of the cylinders are perforated for gas flow in a radial direction. The entire assembly is
enclosed inside the adsorber vessel with gas inlet and outlet conduits. Many different
designs are patented [ll-131. The adsorbers allow faster cycle times, lower pressure
drops, and higher gas throughputs without fluidization. However, the equipment design
is more complex and costly.

Rotarv Bed Adsorbers (TSA)

The rotary bed adsorber (also called adsorption wheel) provides a truly continuous TSA
system. It uses a shallow wheel-shaped adsorption bed that continuously turns about an
axis inside a fixed supporting frame. A section of the wheel is continuously used for
adsorbing impurities 60m a gas while the other section is continuously regenerated by
heating it with an impurity free gas. The adsorbent is made from a honeycomb-shaped
alumina substrate that can be coated with layers of silica gels, activated carbons, or
zeolites [14]. It has been used for gas dehumidification, solvent vapor recovery, VOC
removal, and deodorization of a gas stream.

NOVEL ADSORPTIVE GAS SEPARATION CONCEPTS

Two novel hybrid concepts for gas separation using adsorption technology have emerged
in recent years. They include (a) adsorbent membranes, and (b) simultaneous adsorption
and reaction.

Adsorbent Membranes

75
They consist of a thin layer ( 4 0 pm) of a nanoporous (3-lOA) carbon filmsupported on a
meso-macroporous inorganic solid (alumina) or on a carbonized polymeric structure [151.
They are produced by pyrolysis of polymeric films. The following two types of
membranes are produced:

Molecular Sieve Carbon (MSC) Membranes

The MSC membranes are produced by carbonization of PAN, polymide, and phenolic
resins. They contain nanopores, which allow some of the molecules of a feed gas
mixture to enter the pore structure at the high pressure side, adsorb, and then diffise to the
low pressure side of the membrane, while excluding the other molecules of the feed gas.
Thus, separation is based on the difference in the molecular sizes of the feed gas
components. The smaller molecules preferentially d i f i e through the MSC membrane
as shown by Table 4 [16,17].
Table 4. Separation Performance of Various MSC Membranes

Selective Surface Flow (SSF) Membranes

The SSF membranes, which are produced by carbonization of PVDC, contain nanopores
that allow all of the molecules of a feed gas mixture to enter the pore structure. However,
the larger and more polar molecules are selectively adsorbed on the carbon pore walls at
the high pressure side, and then they difiiise selectively to the low pressure side. The
smaller molecules are enriched at the high pressure side. These membranes can be used
to enrich H2 from mixtures with CI-C4hydrocarbons or from mixtures with C02 and CH4.
They can also be used to separate C&-H2S and H2S-H2 mixtures. Table 5 compares
performances of SSF carbon and polymeric PTMSP membranes for H2 enrichment from
FCC off gas [15]. Clearly, the SSF membrane is much superior for this application.
Table 5. Separation Performancesof SSF and PTMSP Membranes

Commnent Reiections Hi& Pressure Product Comwsitions (%)

Gas Components SSF PTMSP SSF PTMSP
C3H8 98.2 86.0 1.16 6.33
c3H6 98.8 86.0 I .49 12.20
CZH6 94.1 76.0 2.30 6.33
c
2H4 93.3 72.0 2.03 5.19
CH4 52.0 46.0 41.46 32.73
Hz 40.0 40.0 51.56 36.19

SimultaneousAdsomtion and Reaction

76
The conversion of reactants to products, as well as the rate of product formation, of an
equilibrium controlled reaction can be increased by removing a product from the reaction
zone, according to the Le Chatelier's principle. Adsorption has been used to achieve this
goal by using admixtures of catalysts and adsorbents in packed bed reactors. Process
concepts called "pressure swing reactors" have been proposed when the adsorbent is
periodically regenerated by using the principles of PSA [181.

Recently, a novel scheme called "Sorption Enhanced Reaction Process (SERP)" was
developed for direct production of essentially carbon oxide free ( 4 0 ppm) H2 enriched
gas stream (90+%) containing CH., as the primary impurity by steam-methane reforming
(SMR). The process could be operated at a much lower temperature (400-5OO0C) than
that needed by the conventional SMR reactor (800-900C), and yet produce high
conversion of CH4 to H2 [19-201. The cyclic steps of the SERP consisted of (a)
sorption-reaction, where a gaseous mixture of CH., and H20 is passed through the reactor
and a stream of COXfree H2 enriched gas is directly produced at feed gas pressure (PA),
counter-current depressurization of the reactor to near ambient pressure and venting the
effluent gas, (b) counter-current evacuation of the reactor to pressure PD and purging the
reactor with steam at PD while discarding the C02 rich effluent gas, and finally (c)
counter-currently pressurizing the reactor fiom PD to PA with steam. The reactor is
maintained at 400-5OO0Cthroughout the entire cycle.

Table 6 shows an example of the cyclic steady state performance of the SERP concept
using an admixture of a SMR catalyst (noble metal on alumina) and a C02 chemisorbent
(KzC03promoted hydmtalcite) in the reactor [20]. The reactor temperature was 4 9 O O C .
The feed H20:CH4ratio was 6: 1. The concept can directly produce -95% H2 product
(dry basis) with a CH., to H2 conversion of 73%. The trace impurities in the product gas
contained less than 40 ppm COP The corresponding product gas composition
(thermodynamic limit) of a SMR reactor operated without the C02 chemisorbent will be
-67.2% H2, 15.7% CH.,, 15.9% C02, and 1-2% CO (dry basis), and the CHI to H2
conversion will be only 52%. Thus, the SEW concept may be attractive for direct
production of a CO free H2 stream for fuel cell applications.
Table 6. Performance of the SEW Concept for H2 Production

-
Feed 1-
Press. (psia) Feed Purge Steam Hydrogen Product
26.2 0.60 1.88 0.25

Hvdroeen Product Puritv IDrvl Methane Conv.

H2 (9'0 cH4 (YO CO m ToH dro en 9'0
94.4 5.6 40 Not detected 73

Removal of Bulk COi from a Wet HiPh TemDerature Gas

Another interesting example of using a chemisorbent (Na20 supported on alumina) in a

PSA process is direct removal and recovery of bulk C02 from a wet high temperature feed
gas without pre-drying or cooling the feed [21]. A gas stream at 200 C containing 10
mole % C02 and an inert component (dry basis), and which is saturated with water, can be
treated to simultaneously produce a COz depleted stream (<3% C02) and a COz enriched
gas (99+%)without removing the water. The PSA cycle steps consist of (a) adsorption at

77
-125 Psia, (b) co-current C02 rinse at that pressure with recycle of effluent gas as feed,
-
(c) counter-current evacuation to 2.5 Psia with simultaneous steam purge, and (d)
counter-current pressurization with a part of C02depleted product gas from step (a). The
effluent gas from step (c) is partly used as the purge gas in step (b) after recompression,
and the balance is withdrawn as the recovered C02 product gas. The inert gas and COZ
-
recoveries from the feed gas are, respectively, 100 and 78% [21]. This application may
be attractive for green house gas emission control (C02sequestration from a hot flue gas
or C02 removal fkom combustion gases for power generation).

REFERENCES

1. S. Sircar, Adsorption, 6,359 (2000).

2. S. Sircar, "Applicationsof gas separation by adsorption for the hture," Adsorption
Sci. Tech., in press.
3. S. Sircar, M. B. Rao, and T. C. Golden, "Fractionation of air by zeolites," in
"Adsorption and its applications in industry and environmentalprotection," A.
Dabrowski (ed.), Elsevier, New York, pp. 395423 (1999).
4. K. Knoblauch, H. Heimbach, and B. Harder, U.S.Patent 4,548,799 (1985).
5. N. C. Lemcoff and R. C. Gmelin, U.S. Patent 5,176,722 (1 993).
6. B. G. Keefer and D. G. Doman, WIPO International Publication No. W097/39821
(1997).
7. R. Jones, G. E. Keller, and R. C. Wells, U.S. Patent 4,194,892(1980).
8. S. Sircar, U.S. Patent 5,071,449 (1991).
9. M. Suzuki, T. Suzuki, A. Sakoda, and J. Izumi, Adsorption, 2,111 (1996).
10. S. Sircar, Adsorption, 2,323 (1996).
11. U. Von Gemmingen, Linde reports on Science and Technology v. 54 (1994).
12. M. Poteau and S. Eteve, U.S. Patent 5,232,479 (1993).
13. J. Smolarek, F. W. Leavitt, J. J. Nowobilski, E. Bergsten, and J. H. Fassbaugh, U.S.
Patent 5,759,242 (1998).
14. T. Hirose and T. Kuma, "Honeycomb rotor continuous adsorber for solvent recovery
and dehumidification," 2nd Korea-Japan Symp. on Separation Technology (1 990).
15. S. Sircar and M. B. Rao, "Nanoporous carbon membranes for gas separation," in
Recent advances on gas separation by microporous ceramic membranes," N.
Kanellopoulos (ed.), Elsevier, New York, p. 473 (2000).
16. Carbon Membranes Ltd., Israel, trade literature.
17. C. W. Jones and W. J. Koros, Carbon, 22,1419 (1994).
18. G. G. Vaporciyan and R. H. Kadlec, AJChE J., 33, 1334 (1987).
19. S. Sircar, J. R. Hufion, and S. Nataraj, U.S. Patent 6,103,143 (2000).
20. W. E. Waldron, J. R. Hufton, and S. Sircar, AIChE J., 47, 1477 (2001).
21. S. Sircar and C.M.A. Golden, US.Patent 6,322,612 (2001).

78
 CARBON COMPOSITE MEMBRANES

MOTOYUKI SUZUKP
The UnitedNations University,5-53-70Jingumae, Shibuya-ku, Tokyo 150-8925, JAPAN
E-mail: [email protected]

AKIYOSHI SAKODA, SANG-DAE BAE, TAKESHI NOMURA

Institute of Industrial Science, Universityof Tokyo, 4-61 Komaba, Meguro-Ku, Tokyo 153-8505,
JAPAN.
E-mail: [email protected]

YUAN-YAO LI
Department of Chemical Engineering, National Chung-Cheng University,Chia-Yi 621, TAIWAN,
R 0.C.
E-mail: chmyyl6Jccu.edu.m

A novel membrane named carbon whisker membrane (CWM) is introduced for an advanced
filtration process. The C W M is a tubular ceramic membrane with a layer of vapor-grown carbon
fibers (VGCF) grown on the surface of the membrane. We employed chemical vapor deposition
(CVD) of methane with the presence of catalysts to fabricate the CWM and examined its
performances on filtration processes. It was found that, in the process, the VGCFs on the
membrane acted as buffers for preventing impurities such as micro-particles or bio-materials in
the liquid not to direct contact with the membrane body so that the problem of fouling which
commonly occurred in the conventional membrane separation processes can be solved. As a result,
the life time of the membrane and the flux of filtrate were enhanced. Moreover, the VGCFs layer
has a self-cleaning hnction, that is, the flexible fibers can remove attached particles easily during
the filtration or a simply cleaning process after the filtration. This phenomenon is similar to the lotus
effect known in the environmental nanotechnology. In addition to the CWM, we report another
type of carbon membrane called carbon-coated ceramic membrane. The membrane consists of a
tubular porous ceramic substrate which is made by sintering of granular ceramics and a thin carbon
film was coated on granular ceramic by CVD method. By controlling the thickness of deposited
carbon film, we enabled to control the pore size of the carbon-coated ceramic membrane.
Moreover, the carbon thin film changes the properties of the membrane such as hydrophobicity,
thermal conductivity and electrical conductivity. These membranes were all designed for solving
water treatment issues.

1 Introduction

In the past few decades, membrane technologies have emerged as an important means of
separation in a variety of industries. Membrane separation technology offers the
advantages of, at least, high selectivity, low energy consumption, moderate cost to
performance ratio, and compact and modular design in comparison with other separation
processes. In addition, inorganic membranes such as ceramic and carbon membranes
possess characteristics of high thermal stability, chemical compatibility, and good
mechanical strength over membranes made of organic polymers. It is therefore that
inorganic membranes have come into wide use in the water/ wastewater treatment
industries. However, fouling/cake generation problem is one of the challenges for the
membrane separation technology. The particles, colloids, macromolecules or
bio-materials in the water permeation process will deposit on the membrane surface and

79
generated secondary film. As a result, membrane will be blocked and permeate flux will
diminish with time [l-31.

h t u s effect is one of the significant concepts in environmental nanotechnology. In

nature, fine wax crystals with around 1 nm in diameter coated on the leaves of lotus plants
and the fine hydrophobic structure repels water. As a consequence, the water droplets
bead up and the surfaces stay dry even during a heavy rain. Self-cleaning of lotus leaves
means that, if the surface of leave inclines, the droplets pick up small particles of dirt on
the surface as they roll. The leaves thereby keep clean and fresh. The lotus effect has
now become well known and biomimic materials based on lotus effect have extensively
been developed in the industries [4]. The nanoparticles can be mixed with paints so that
painted objects, say outside wall of houses, have a rough surface with the presence of
nanoparticles. While rain falls, the nanoparticles on the wall act as an interface between
dirt and objects resulting lotus-effect and a self-cleaning function. Aerosol with
nanoparticles can also be sprayed on paper, leather, textiles and masonry to become
advanced materials with a self-cleaningproperty.

A layer of carbon whiskers grown on the surface of objects can also play a self-cleaning
role. This is because that the whiskers avoid dirt or particles directly attach on the
objects and, because of minute contact surface between whiskers and dirt, the flexible
whiskers can easily remove adsorbed dirt or particles fiom the contact surface while
altering the flow rate, direction and form of the flow. Depending on the size and density
of the grown whiskers, the carbon whisker layer is capable of keeping substrate clean from
dirt with different sizes and also bio-materials which was known as an antibacterial
property. If the substrate is a porous membrane and a filtration process is conducted, the
carbon whisker layer will play characters of self-cleaning and fouling reduction. It is
therefore that the permeant flux and separation efficient will be enhanced.

We report two types of membrane namely, carbon whisker membrane (CWM) and
carbon-coated ceramic membrane. The CVD method and iron catalysts were used to
fabricate a layer of carbon whiskers grown on the surface of the tubular ceramic
membrane. This novel membrane is called carbon whisker membrane (CWM). We
demonstrate the formation of the CWMs with various whisker sizes and membrane pore
size as well as their performance on the filtration process using PMMA particles. A
comparison of cleaning process with conventional membranes will also be discussed.
Carbon coated ceramic membrane were also fabricated by CVD method. By controlling
experimental factors such as concentration of carbon source, pyrolysis temperature or
deposition time, we demonstrate that it is possible to adjust the thickness of the carbon
thin film so that the pore size of the membrane can be regulated. It is therefore that the
membranes can be used for microfiltration, ultrafiltration or even nanofiltration. Figure
1 shows the cross-section of the different type of the membrane. Figure 1(A) is ceramic
membrane made by sintered granular ceramics. We used this ceramic membrane as a
substrate for fabrication of C W M or carbon-coated ceramic membrane. Figure 1(B) is
activated carbon membrane. The membrane is made by spin dipping & coating of a
layer of carbon source on the surface of the ceramic membrane. By thermal activation
process, a layer of porous and activation carbon thin film can be formed. Figure 1(C) is
carbon-coated ceramic membrane fabricated by CVD method. Clearly, the membrane is

80
different fiom activated carbon membrane. Figure 1(D) is the carbon whisker membrane
(CWM).

(A) Ceramic Membrane (C) Carbon-Coated

Ceramic Membrane

Cerami; Substrate/ Carbon- Ceramic

Membrane Film Substrate

(B)Activated Carbon Membrane (D) Carbon Whisker Membrane

pore Carbon L,ayer Carbon Whisker

CeramidSubstrate Carbon Ceramic

Film Substrate
Figure 1 Cross-sections of (A) Ceramic Substrate, (B) Activated Carbon Membrane, (C) Carbon-Coated
ceramic Membrane, and (D) Carbon Whisker Membrane

2 Experimental

To grow carbon whiskers on the membranes, we employed our recipes fkom the previous
studies [5, 6]. We used methane as carbon source to make carbon whiskers and nitrogen
as carrier and diluting gases. Femc sulhte was employed as catalyst precursor to make
Fe catalysts. The recipes offer a cost-effective production and safe process. The
experiments were carried out with a tubular quartz reactor and a furnace. A porous
ceramic tube from Kubota Co., Japan was firstly coated with aqueous ferric sulfate
solution using dipping & coating technique and then dried at room temperature overnight
prior to be suspended in the center of the reactor. A designed pipe line allowed to reverse
feed flow in the tubular reactor was employed for the fabrication. The temperature was
then ramp up to 1000C with nitrogen as a purging gas to form Fe catalyst particles fiom
the precursor. After the temperature reaches the steady state, methane was then introduced
into the reactor and started to pyrolysis and deposition. In order to ensure a consistent
carbon deposition profile for a uniform layer of carbon whiskers, the direction of the feed
flow was reversed periodically after a desired deposition time. The CWM can be
therefore fabricated.

Filtrations for CWM were carried out with a cross flow membrane separation system. It
consists of a cross flow membrane module, a feed pump, a recirculation pump and process
pipes. The feed steam contained 1000 ppm PMMA with a mean particle size of 0.8

81
micrometer for the experiments. Backflushing technique to regenerate membrane was
employed for one minute in every 20 minutes filtration process in order to remove
attached particles. For the fabrication of carbon-coated ceramic membrane, the materials,
apparatus and processes were all the same but the presence of catalyst particles. The
measurement of the pore size of the membranes were conducted by gas permeability.
The experiments were carried out at room temperature with different gases such as oxygen,
nitrogen, argon and helium 171.

3 Results and discussions

3.I Carbon whisker membrane

Figure 2 shows the top-view and side-view of the CWMs under the observation of SEM.
As can be seen, the whiskers were grown on the substrate randomly to form a network
structure on the surface of the membrane. The results show that the diameter of carbon
whisker is about 100 nanometers and the thickness of the carbon whisker film is about 2
micrometers. However, the diameter, density and layer thickness of carbon whiskers is
strongly dependant of experimental conditions such as deposition time, deposition
temperature, concentration of methane, feed flow rate and catalyst concentration.
Different experimental or operational conditions will change the characters of CWMs.
Figure 3 shows the CWMs with different size of carbon whiskers. In three CWMs, the
one on the top has a thickest and longest carbon whisker layer but smallest pore size of
membrane. The diameter of visible carbon whiskers is about a few minimeters while the
length is about 1-2 centimeter. It was known that, if the size of carbon whiskers
increases, the pore size of the membrane decrease. This is because that, while the CVD
of methane is in process, carbon species not only deposited on the catalyst particle to grow
carbon whiskers but also deposited on the surface of the substrate to form a carbon film on
the substrate. As the deposition time increased, diameter and length of carbon whisker
increase and the carbon film thicken, which causes the decrease of the pore size of the
membrane. The middle CWMs has carbon whiskers which vaguely can be seen. The
diameter of whiskers is in micrometer scale while the length can reach to a few minimeters.
The last CWM has to be characterizedby SEM observationjust like Figure 1 showed.

(A) Top-view of CWM

Figure 2 (A) Top-view and (B) Side-view of Carbon whisker membrane

82
 Figure 3 Carbon whisker membranes with different size of carbon whiskers. The one on the top has a thicker
and longer carbon whisker layer but smaller pore size of membrane compare to the rest of the two CWMs.
Figure 4 shows the filtration performance between carbon whisker membrane (CWM) and
carbon membrane (CM). The results clearly reveal that the permeant flux of CWM is
higher than that of CM even after a few cycles of filtration. Because of the large amount
and network structure of carbon whiskers, particles will be trapped in the carbon whisker
layer before contacting porous membrane substrates during filtration processes. As a
result, the pores of substrate will not be blocked, which will not cause the fouling
phenomenon and not reduce the life-time of the membrane. In addition, if we used pulse
flow for the feedings or changed the direction of the feedings, the flexible whiskers might
remove the attacheaadsorbed particles away as a self-cleaning function. Figure 5 shows
the result of backflushing cleaning process for the membrane after the filtration. We
found that particles attached on the CWM can be easier to remove than that of the CM.
Within the same cleaning time, particles (white color) were still attached on the CM while
no obvious particle was found on the CWM. These results lead to suggest that the CWM
possess a higher flux permeability, a self-cleaning character and a easier regeneration
process of the membrane.
. -

0.013
-
0.012

o.oll
0.010
t
0
PMMA: lOOOppm (0.8 pm),AP 0.1 kgf/cm*
Backflushing :dirtiled water, lmin,2 kgflem'
'
10 20 30 40 50 60 70
I

80
Time Wn.01
Figure 4 Flux versus time for crossflow membrane filtration with backflushing cleaning

83
 Figure 5 Observation of PMMA remod on membranes after backflushing cleaning procedue

3.2 Carbon-coatedceramic membranes

Carbon-coated ceramic membranes were made by the CVD of methane without the
presence of the catalysts. Figure 6 shows the ceramic membrane and carbon-coated
ceramic membranes with different pore size. We simply controlled the deposition time
so that the thickness of the carbon film is different resulting the different pore size in each
membrane. As can been seen, the top membrane has a brighter color and, under the
observation of the surface by SEM, it was found to be a dense membrane. The pore size
can be reduced to nano-scale. The second membrane &om the top has submicro pores
which can be used as an ultrafilter. It is no doubt that the deposition time to fabricate
ultrafilter is less than that of the nanofilter. If only a little carbon deposited on the
ceramic membrane, the pore size of the membrane might just change slightly. However,
the properties of the membrane altered dramatically such as hydrophobicity, electrical
conductivity and thermal conductivity because of the nature of carbon. These changes
create a new research field for a better separation process by membrane technology such
as membrane filtration with assistance of electrically conductivity, hydrophobicity or
magnetic field separation [8]. Gas permeability can also be used for examining the pore
size of the membrane. Figure 7 shows nitrogen flux against transmembrane pressure
through Membranes A, B, and C. It is clear that a higher pressure drop increase the
nitrogen flux for all the membranes. Under the same condition of the pressure drop, the
nitrogen flux through Membrane A is the highest. This result suggests Membrane A has
a bigger pore size so that more nitrogen can flow through. Therefore, the pore size of the
membrane A is biggest followed by Membrane B and then Membrane C.

4 Conclusions

We reported our recent developed membranes called carbon whisker membrane (CWM)
and carbon-coated ceramic membrane. The CWM performed a better permeant flux in
the filtration process in comparison with the membrane without whiskers. Also, C W s
have a self-cleaning function which can increase the separation efficiency during the
filtration and increase the lift-time of the membrane. The carbon-coated ceramic
membranes with various pore sizes can be made for the purpose of nanofiltration,

84
ultrafiltration or microfiltration. It was suggested that the unique properties of the
carbon coated ceramic membranes such as good hydrophobicity, thermal conductivity and
electrical conductivity should be taken into account while designing an advanced
membrane separation process.

Figure 6 Carbon-coated ceramic membrane with different pore sizes

0.6

0.5

5 0.4
I r a

B
-
E
' 0.3
a
2 0.2
E
0.1

0
0 25 50 75 100 125 150 175 200
P w u n drop (mmHg)

Figure 7 Nitrogen flux against transmembrane pressure through MembranesA, B, and C

References

1. Flemming, H.C.,Schaule, G.,Schmitt, J. and Tamachkiarowa, A., Biofoulng- the

Achilles heel of membrane processes, Desalination, 1997, 1 13,215-225.
2. Vrouwenvelder, H.S., van Paassen, J.A.M., Folmer, H.C., Hofjnan, J.A.M.H.,
Nederlof, M.M.and van der Kooij, D., Biofouling of membranes for drinking water
production, Desalination, 1998, 1 18, 157- 166.
3. Abd El Aleem, F.A., Al-Sugair, K.A., Alahmad, M.I.,Biofouling problems in
membrane processes for water desalination and reuse in Saudi Arabia, International
biodeterioration& biodegradation, 1998,4 1, 19-23.

85
4. Bemhard, S., Edwin, N., Markus, O., 2002, Geometyrical shaping of surfaces with a
lotus effect, Patent US2002164443, EP1238717, DElOllO589, JP2002321224.
5. Li, Y.Y.,Bae, S.D., Sakoda, A. and Suzuki, M., 2001, Formation of vapor grown
carbon fibers with sulfuric catalyst precursors and nitrogen as carrier gas, Carbon,
39(1), 91-100.
6. Li, Y.Y., Bae, S.D., Sakoda, A. and Suzuki, M., 2000, Fabrication and
characterization of carbon whisker, The 2nd Pacific Basin Conference on Adsorption
Science and Technology, May 14-18,Queensland, Australia, p376-380.
7. Li, Y.Y.,Nomura, T.,Sakoda, A. and Suzuki, M., 2002, Fabrication of carbon coated
ceramic membranes by pyrolysis of methane using a modified chemical vapor
deposition apparatus, Journal of Membrane Science, 197 (1-2), 23-35
8. Li, Y.Y.,Sakoda, A. and Suzuki, M., 2001, A review of fabrication methods of
carbon membranes and applications related to their hydrophobic and electrically
conductive properties, Journal of Institute of Industrial Science, University of Tokyo,
53(3), 70-74.

86
Invited Papers
This page intentionally left blank
 ON THE DOMINANT ROLE OF ADSORPTION EFFECTS IN
HETEROGENEOUSCATALYSIS

JOERl F. DENAYER, GIN0 V. BARON

Dienst Chernische Ingenieurstechnieken, Vrije UniversiteitBrussel, Pleinlaan 2, I050 Brussels,
Belgium E-mail: [email protected]

DIRK DEVOS, JOHAN A. MARTENS AND PIERRE A. JACOBS

Centrefor Suflace Chemistry and Catalysis, K. U.Leuven, KasteelparkArenberg 23, 3001 Leuven,
Belgium E-mail:[email protected]. be

Adsorption effects in heterogeneous liquid phase reactions are seldom considered or recognized as
important. Liquid phase adsorption is also poorly understood for typical reaction mixtures and hardly
any models are available for its description. In this contribution we will briefly review some of the
methods available for measuring gas and liquid phase adsorption (batch tests, gas and liquid
chromatography, various detection methods) and describe a few applications in heterogeneous
catalysis with zeolites. Concentrations in the micropore system of such zeolite catalysts can differ
strongly from the external liquid bulk phase concentrations, and adsorption and diffusion phenomena
essentiallycontrol reaction rate and product selectivity.A good understandingof these phenomena can
help in selecting the best solvent or catalyst support for the reaction, understand deactivation and may
be instrumental in correct modeling of the reaction. Two examples from hydrocarbon conversion and
partial oxidation are presented.

1 Introduction

Physisorption is often included in gas phase reaction models, and its importance well
recognized, but the parameters are often estimated from reaction data, leading to poor fits
and biased parameters, often leading to erroneous mechanisms. Independent measurement
of reactant and product adsorption in near reaction conditions allows for elegant modeling
of complex reaction systems. For a variety of heterogeneously catalysed gas and liquid
phase reaction systems, we have systematically determinedthe adsorption properties of the
catalyst in close to reaction conditions. Given the high temperatures, high pressure,
multi-phase conditions, only a few of the usual techniques are really applicable. These are
mainly gas or liquid chromatography, both in tracer and perturbation mode and batch
adsorption experiments which we currently run in a robotic system as these measurements
are very labor intensive. These data are then used to model quantitatively the reaction
system, and allowing to extract intrinsic rates of reaction, or used to analyse the behaviour of
the reaction when changing zeolite composition, type, pore size or solvent used. Two
examples from our work are briefly described next as an illustration of the methodology.

2 Epoxidation of a-alkenes with Ti-zeolites

Ti-substituted zeolites (Titanium Silicalite-1 or TS-I, Ti+) are truly heterogeneous

catalysts, and they can use H202as the oxidant. As TS-I hardly decomposes any H24,
yields on peroxide basis are usually excellent. This has qualitatively been ascribed to the
hydrophobic nature of TS-I,disfavoring peroxide accumulation in the intraporous volume.
A second important characteristicof TS-1 epoxidations is that they are generally fastest in

87
methanol. This preference has been explained by a model in which methanol coordinateson
the active Ti sites [1,2]. The possible link between physisorption and solvent effects in
olefin epoxidation, however, has not extensively been studied. In the case of alkene
oxidation with TS-I ,the slower reaction of larger alkenes has been attributed to diffusion
limitation of these molecules. These measurements were however performed at high
conversion where deactivation was dominating.
Tracer liquid chromatographic methods were used to study liquid phase adsorption on
TS-1, Ti-Beta and Ti-MCM-41 and completely elucidate the reaction properties [3].
Partition coefficients for alkenes, alkanes, epoxides, and other, polar products are strongly
dependent on the carrier solvent. Linear a-olefins are concentrated inside the TS-1
micropores, particularly when methanol is the solvent. This agrees well with the superior
initial rates of olefin epoxidation with TS-1 in methanol. Sorption also governs the relative
reactivities of olefin substrates, especially in competitive experiments. Thus, in truly initial
conditions, I-hexene is less reactive than I-octene or I-nonene. For the latter substrates,
however, deactivation is fast, especially in methanol. This process is related to the strong
adsorption of higher 1,2-epoxyalkanes in TS-1 in methanol. Deactivation due to
competitive epoxide adsorption is slower in acetone, making this a more suitable solvent
than methanol for 1-nonene epoxidation with TS-1. Physisorption effects play a dominant
role in the small pore TS-1 catalyst, due to the close interaction of substrates such as alkenes
with the pore wall. Wider-pore catalystssuch as Ti-Beta and especially Ti-MCM-41, do not
adsorb olefins as selectively and hence intraporous olefin concentrations and reaction rates
are much lower.

3 Hydroeonversion of n-alkanes

Bifhctional zeolite catalysts such as platinum loaded acid zeolite catalysts are applied
in several petroleum refinery operations, designated as hydroconversion processes:
isomerisation of light naphtha, iso-dewaxingand hydrocrackingof heavy fractions [4]. Most
experimental investigations in academic laboratories are typically performed with pure
model components or simple mixtures thereof as feedstock, and using reaction conditions
under which the hydrocarbon compoundsare in the vapor phase. Industrialhydroconversion
processes are mostly run under three phase, or even in some cases under liquid phase
conditions and with feedstocks that are extremely complex mixtures of large numbers of
different hydrocarbon compounds [4].
In earlier work, adsorption equilibria of a broad range of hydrocarbon components
were determined on ultrastable Y (USY)zeolites under catalytic conditions. It was found
that Henry adsorptionconstants increase exponentiallywith the alkane carbon number [5,6].
This adsorptive discrimination of long alkanes in favor of shorter ones is most pronounced
in the Henry regime and decreases when the pores are more filled up with hydrocarbon
molecules at increasing vapor pressures [7]. USY type zeolite contacted with liquid alkanes
does not show any differences in affinity towards alkanes of different molecular weight [8].
These investigations showed that on USY type zeolite, adsorption selectivity depends
strongly on the loading of the micropores and on the aggregation state of the alkane
contacted with the zeolite.
The conversions of n-alkanes, cyclo-alkanes and their mixtures over WUSY zeolites
under vapor phase reaction were successhlly modeled using independently determined
adsorption equilibria and fundamental reaction networks based on alkylcarbenium ion
chemistry [9-121. In these models, the intrinsic reaction rate of an individual reaction step

88
such as a branching rearrangement via a protonated cyclopropane is the same in all
molecules and at all positions in the carbon chain. Intrinsic reactivity differences between
alkene intermediates are accounted for by one single parameter, reflecting differences in
protonation enthalpy [lo]. Under vapor phase reaction conditions on an W S Y type
catalyst, there is always preferential conversion of the heaviest alkane owing to its
preferentialadsorption. In absence of adsorption selectivityin the liquid phase, it is however
expected that the relative reactivities of the two hydrocarbons reflect the intrinsic reaction
kinetics of vapor phase experiments.
Figure 1 shows the result for gas phase conversion of an equimolar heptane - nonane
mixture over Pt/USY type zeolite catalyst CBV720 (SVAI = 13, Zeolyst), and demonstrates
it is indispensable to use appropriate expressions for the multicomponent adsorption
equilibria. For vapor phase conditions resulting in significant pore filling, the Langmuir +
interaction model is appropriate.

0.9

0.8

0.7

; o.s
P
g
OS
0.4

0.3

0.2

0. t

0
0.0 0.1 02 0.3 0.4 0.5 0.9 0.7 0.8 0.0 1.0
Avenp. convenion

Figure 1 Conversion of heptane and nonane from Figure 2 Conversion of heptane and nonane From
their binary equimolar mixture in vapor phase their binary mixture in liquid phase plotted against
conditions at 230 O C and different spacc times the average conversion (symbols: experimental
(symbols: experimental data points, full curves: data points, full curves: Langmuir adsorption
Langmuir adsorption model, dotted curves: model, dotted curves: no selective adsorption).
Langmuir + Interaction model).

Figure 2 shows the selectivity for the longer alkane is strongly reduced in liquid phase
and that a non-competitive model for adsorption is needed in that case to represent the
results. Using a Langmuir type model based on gas phase data rather approximates the gas
phase results and is completely in error for the prediction of liquid phase conversion.
Both in vapor and in liquid phase reaction conditions, nonane is more reactive than
heptane. The reactivity difference is however much more pronounced in the vapor phase. In
USY zeolite micropores exposed to the vapor of the two n-alkanes, the heaviest alkane
molecule is preferentially adsorbed, resulting in a higher apparent reaction rate. When the
alkane mixture is fed in the liquid phase, the competing alkanes are adsorbed in a
non-selective manner in the micro- and mesopores of USY.Consequently, in liquid phase
conditions the relative reactivity of the n-alkanes corresponds to the relative intrinsic
reactivities.

89
4 Conclusion

These two examples show that adsorption effects can be dominant over the other
phenomena in a spectacular way and need to be taken into account directly through
independent measurement. Understandingthese phenomena can help in selecting, designing
or optimizing catalysts or reaction conditions.

5 Acknowledgements

This research was financially supported by FWO Vlaanderen (G.0127.99). J. Denayer

is gratehl to the F.W.0.-Vlaanderen, for a fellowship as postdoctoral researcher. The
involved teams are participating in the IAP-PA1 programme on Supramolecular Chemistry
and Catalysis, sponsored by the Belgian Federal Government.

References

1. Bellussi, G., Carati, A., Clerici, M.G., Maddinelli, G., and Millini, R., J. Cuful. 133,
220 ( 1992).
2. Clerici, M.G., and Ingallina, P., J. Cuful. 140,71 (1993).
3. Langhendries G., Baron G.V., De Vos D.E. and Jacobs P.A. (1999) J. Catal., 187,
453-463
4. Maxwell, I.E. and Stork, W.H.J, Stud. Surf.Sci. Catal. 137,2001,747.
5. Denayer, J.F.M., Baron, G.V., Jacobs, P.A., Martens, J.A., J. Phys. Chem. B 102 (1 7),
1998,307.
6. Kiselev, A.V., Adv. Chromatogr. 4, 1967, 1 13.
7. Denayer, Joeri F. M.; Baron, Gino V., Proc. Fundamentals of. Adsorption 6., Ed.
Meunier, F., 1998, Elsevier, 99- 104.
8. Denayer, J.F.M., Bouyermaouen,A., Baron, G.V., Ind. Eng. Chem. Res. 37 (9), 1998,
369 1.
9. Denayer, J.F.M., Baron, G.V., Martens, J.A., J. Catal. 190 (2), 2000,469.
10. Thybaut, J. W., Marin, G. B., Martens, J. A., Jacobs, P. A., Baron, G. V., J. Catal. 202,
200 1,324.
11. G. G. Martens, G. B. Marin, J. A. Martens, P. A. Jacobs, G. V. Baron, J. Catal 195,
2000,253.
12. Denayer, J.F.M., Baron, G.V., Martens, J.A., PCCP 2(5), 2000, 1007.
13. Denayer, J.F.M., De Jonckheere, B., Hloch, M, Marin, G.B., Vanbutsele, G., Martens,
J.A., Baron, G.V., J. Catal210,2002,445.
 SUPERCRITICAL ADSORPTION:PARADOX, PROBLEMS, AND INSIGHTS

LI ZHOU
High Pressure Adsorption Laboratory, School of Chemical Engineering, Tianjin Universiv,
Tianjin 300072. P R China
E-mail: [email protected]

Abstract. Paradoxes, problems and ideologies in the study of supercritical adsorption were
discussed. A macroscopic interpretation of supercritical adsorption was presented basing on a
general model that derived at from the Gibbs definition and a straightforward method of determining
absolute adsorption. The model does not include any assumption, but relies on experimental data
and keeps the formal continuity of adsorption theory. It was shown to apply for wide ranges of
temperature and pressure, and bore an impact to the characterizationof adsorbents.

1. Problems of supercritical adsorption

1.I Is supercritical ahorption an artificial topic?

The adsorption of gases at above-critical temperatures is referred to as supercritical
adsorption. The first paradox about supercritical adsorption is it is an artificial topic and
attracts onlyf a v researchers, which was heard at comments on manuscripts devoting to
the topic. It did not, in deed, attract must interest before, but many works were dedicated
to it in recent years following the need of clean fuels, such as natural gas and hydrogen.
Storage of these fuels on-board vehicles constitutes the bottleneck of the technology, and
adsorptive storage seemed promising [I ,2]. The storage temperature of interest is much
higher than the critical temperature of fuel gases, therefore, supercritical adsorption is the
theoretical base of adsorptive storage. It is also the theoretical base of adsorptive
separation processes [3,4]. Besides, it was suggested for the characterization of porous
materials [ 5 ] . Because of the importance of supercritical adsorption, a special session was
organized for it on FOA7.

I .2 Relation and difference between sub- and super-critical adsorption

The adsorption below and above the critical temperature looks different as shown by
experimental isotherms. There are five types of isotherms below the critical temperature
depending on the porous structure of adsorbents [6],however, there is only one type of
isotherms at above-critical temperatures no matter what kind of adsorbents was tested.
The supercritical isotherm shows type-I feature for its initial part, but there must be a
maximum followed by a negative increment as long as pressure is high enough or
temperature is low enough. Each type of isotherm reflects a definite mechanism o f
adsorption, therefore, supercritical adsorption must have just one mechanism. All
adsorption isotherm equations available, for example, the Langmuir or the BET equation,
were derived at from a prescribed mechanism and explained the experimental isotherms
satisfactorily. Therefore, clarification of the unique mechanism of adsorption at above
critical temperatures is necessary for setting up a pertinent model of supercritical
adsorption.

91
I .3 Application ofpotential theory to supercritical adsorption
Facing the fast development of industrial techniques, progress in theoretical studies of
supercritical adsorption is left far behind the need of engineering requirement. The
Polanyi-Dubinin potential theory [7,8] is universal in nature, but cannot be applied
directly to interpret supercritical adsorption data. Theoretically, the amount adsorbed at
above-critical temperatures might be predicted basing on a characteristic curve, which is
independent on temperature and, thus, could be constructed basing on the isotherms at
subcritical temperatures. However, the characteristic curve was defined as the plot of
adsorbed volume versus the adsorption potential. If we use the curve to predict the amount
adsorbed at above-critical temperatures, we must specify the value of adsorption potential.
The adsorption potential was defined as the compression energy of conveying a mole gas
from the gas phase to the adsorbed phase, i.e., A =
c Vdp . The density of the adsorbed

phase, pa,and the pressure corresponding to f i in the integral is, however, not known.
Therefore, there is not a way to spec!& the adsorption potential at above-critical
temperatures and, hence, the knowledge of potential theory available cannot directly solve
the problems encountered at above-critical temperature.

1.4 Classical versus new models of supercritical adsorption

To explore the mechanism of adsorption, a mathematical model was usually
presented for experimental isotherms. If a model fits experimental data well, the
assumption underlined the model would be a tentative "theory". A pertinent model is also
required by process simulation of PSA. Therefore, modeling supercritical isotherms
attracted much research interest and different methodologies were presented. Classical
models of adsorption isotherms, including Langmuir, BET, and the potential theory
models such as Dubinin-Astakhov equation, correlate the adsorption data with adsorption
condition through few (two to three) parameters. Such models are simple in form and have
been applied successfully to subcritical adsorption, therefore, were tried to apply for
supercritical region [8,9]. However, the difference between sub- and super-critical
adsorption was overlooked previously. It was well known that all isotherm equations were
initially derived at for absolute adsorption, but the isotherms obtained experimentally are
excess quantities. The difference between the excess and the absolute adsorption is
negligible for the subcritical condition, therefore, the absolute adsorption models work
well also for the excess isotherms. However, the difference between the excess and the
absolute adsorption becomes large at high pressures and the classical models do not work
well for the excess supercritical isotherms. Typically, they cannot describe isotherms with
maximum. Then new models that can describe isotherms with maximum such as
Ono-Kondo equation [lo], models basing on density function theory [11,12], and the
models derived from equations of state of gases [13,14] were proposed. The DFT
modeling method is similar to GCMC in that the state of adsorbate at the minimum of
grand potential energy was searched and, hence, a lot of computation work is required.
Computational models may describe the excess isothenns well, but deviated off the
classical way of modeling adsorption isotherms. The philosophy underlined the two kinds
of modeling is different. There is a thermodynamically distinct interface between the
adsorbed phase and the bulk gas according to Gibbsian point of view, and the density
profile in the normal direction of solid surface is not continuous. However, such interface

92
was omitted, and a continuous density profile in the normal direction of solid surface was
assumed in the computations basing on minimizing the grand potential energy. Rome is
there, and many ways lead to Rome. The results obtained by different methods may
contribute to better understanding supercritical adsorption, therefore, microscopic or
macroscopic way of study should not expel each other.

1.5 Volumetric versus gravimetric method of adrorption measurement

Most adsorption data were collected by volumetric method until microbalance of
high sensitivity appeared few years ago. It can hardly say which method is superior to the
other, and both methods need the value of the skeleton volume of sample adsorbent. This
volume has to be subtracted fiom the whole volume of the sample container to obtain the
volume of void space, which is used for the calculation of the amount adsorbed. The
skeleton volume of sample adsorbent was directly used in the calculation of buoyancy
correction in gravimetric method. This volume was usually determined by helium
assuming the amount of helium adsorbed was negligible. If, however, helium adsorption
cannot be omitted, error would yield in the skeleton volume and, finally, in the calculated
amount adsorbed. However, the effect of helium adsorption would be much less for
volumetric method if the skeleton volume is considerably less than the volume of void
space, but the volume of void space cannot affect buoyancy correction. In this respect,
helium adsorption would result in less consequenceon volumetric method especially when
the skeleton volume was determined at room temperature and pressures less than 15 MPa.
The skeleton volume (or density) was taken for a parameter in modeling process in some
gravimetric measurements. However, the true value of skeleton volume (or density) can
hardly be more reliable basing on a fitted parameter than on a measured value. Therefore,
one method of measurement cannot expel the other up to now, and the consequence of
helium adsorption in the measured amount adsorbed should be estimated appropriately.

2 A classical description of supercritical adsorption

The Gibbs definition of adsorption is valid not only for subcritical adsorption, but
also for supercritical one:
n = n, - Vapg (1)
The experimentally measured amount adsorbed, n, is an excess quantity concentrated in
the so-called adsorption space over the bulk gas phase, and the so-called absolute
adsorption, n , corresponds to the total quantity of adsorbate in the mentioned space.
According to Gibbs, there is a thermodynamically distinct interface between the bulk gas
and the equilibrium adsorbed phase, the density of which (pa)is remarkably higher than
that of the bulk gas phase @B). V, is the volume of the adsorbed phase, and the product
V& accounts for the difference between n and nt.
All isotherm equations were derived at for absolute adsorption. Therefore, there are
more choices for the expression of nt. However, the initial part of supercritical isotherms
looks like Type-I, and a simple model proposed for this kind of adsorption on
heterogeneous surface [151 was applied
n = n$ - exp(- bp4)l- rapg (2)
Such a model bears prominent merits: 1. It keeps the formal continuity of isotherm

93
equations available to date; 2. It keeps the simplicity of isotherm equation because there
are only three parameters to be determined experimentally: &', b and q. This equation is
similar to the Langmuir-Freundlich equation and so does the parameters. Parameter np is
a saturation quantity of n, since n,' if p=m. Parameter b is related to the energy
change of adsorbate after adsorption, and parameter q is an index of surface heterogeneity.
There are two unknowns in Eq.(l): n, and V,. Either one is determined, the other could be
evaluated from the equation. Previous efforts were to determine n, basing on an assumed
V, [161 or the density of the adsorbed phase (pb) [171. However, a value of nt determined
basing on experimental data would be preferable.

3 A straightforward method of determining absolute adsorption

Because n, is the totai mass confined in the adsorbed phase, it must vary with the
experimental condition, therefore, should be determined as a function of temperature and
pressure. A straightforward method was proposed by the author [ 18-19]. It is known from
Eq.1 that n = n, if Vs, can be neglected comparing to n. Therefore, we can use the
experimental values of n that comply with the constraint to formulate the model of
absolute adrorption. The experimental data experienced twice transformations to reach a
linear plot as was usually done for the establishment of a model for a set of data. The
experimental data were utilized to the utmost in the transformation processes, and the data
that do not comply with the constraint were sifted out. A plot of In[ln(dk)] versus lllnp 0,
in kPa) was thus constructed. Parameter Swas used to adjust the magnitude of n in order
to avoid evaluating the logarithm of negative numbers. Its value could be set at 1, 10 or
100. A model with two parameters were obtained from the linear plots for the absolute
adsorption isotherms:

[(
n, = exp exp a + -
:PI1
(3)

The value of parameter a and p is a function of temperature and, therefore, n, is a function

of temperature and pressure. The volume of the adsorbed phase, V, could be determined
via Eq.( 1) from the measured n and the Eq.(3)-determined n,. So, Eq.(2) could be used to
model an isotherm of supercritical adsorption. It must emphasized that Eq.(2) is the model
of supercritical isotherms, and there are only three parameters since the value of n, (or V a
could be determined the other way if it is available.
Perfect fit was observed at experimental isotherms of different adsorption systems in
large ranges of temperature and pressure [20,21]. Shown in Fig.1 and Fig2 are only
examples. The model for absolute adsorption isotherm was obtained basing on the data at
relatively low pressure, but the model fits the data at high pressure as well. It is concluded
that the adsorption mechanism of supercritical adsorption does not change as pressure
increases although maximum or even negative (excess) adsorption was observed.

4 Continuity of the model and its impact on the characterization of adsorbent

Eq.(2) together with Eq.(3) describe isotherms not only for supercritical, but also for
subcritical if the isotherms show type-Ifeature. It was shown recently [22] that this model
works well for the isotherms distributed densely around the critical temperature as shown

94
in Fig.3. Although 298 K is lower than the critical temperature (304.2 K), maximum is
shown on the isotherm. Therefore, the difference between the excess and the absolute
adsorption cannot always be neglected for subcritical region. The evaluation of absolute
adsorption is especially important for the characterization of adsorbents basing on the
adsorption isotherm of C02at 273 K [23] and the isotherm shows type-I feature. Shown in
Figd is a calibration isotherm of COz at 273 K for a sample of activated carbon, where
dots are the data measured, and the solid line is the absolute adsorption isotherm
generated by Eq.(3). Since 273 K is below the critical temperature (304 K), the density of
COz gas is so low that the product Vfig is much less than the value of n for most
conditions. As consequence, most experimental points locate on the absolute adsorption
isotherm. More points would also locate on the linear plot shown in Fig.5, which renders
the formulation of absolute isotherms more reliable. However, the difference between the
absolute and the excess isotherm becomes considerable as pressure approaches saturation.
This difference is important because both pore volume and the specific surface area are
determined by the highest amount adsorbed. The volume of micropores was usually
estimated by dividing the amount adsorbed at relative pressure of pips = 0.95 by the
density of saturated liquid if the isotherm is type4 [24]. The assumption underlined is the
full-filled pore space with liquid adsorbate. However, the amount of adsorbate determined
as such does not correspond to the total mass fillkd in the space, but only a part of it,
although the major part. It is the absolute adsorption that corresponds to the total mass in
the adsorbed phase, therefore, the volume of pores should be determined basing on it. As
is seen in Fig.4, the absolute isotherm intersects the condensation line when pressure
reaches saturation, which proves a convention that the adsorbed phase has the same state
as saturation liquid. It is thus concluded,pore volume should be evaluated by dividing the
saturated absolute amount by liquid density. For example, the pore volume of the sample
shown in FigA would be 1.52 cm3/g basing on the absolute isotherm compared to 1.27
cm3/g basing on the excess isotherm. The relative difference is as large as 20%. The
surface area was usually determined by the DRK plot if the calibration isotherm is type-I
[24]. The difference in the DRK plot between the excess and the absolute adsorption is
also considerable as shown in Fig.6 for the same sample. The intercept of the two plots is
1.43 and 1.49 respectively yielding a relative difference of 15% in the surface area.

5 Is there a border of supercritical adsorption?

It is well known that the saturation pressure, ps, is the border of subcritical
adsorption, beyond which another phenomenon, condensation, happens. It seems there is
not a similar border for supercritical adsorption because no matter how high the pressure
is, adsorption isotherms could always be recorded. However, such border must exist
considering the cause of adsorption. It is the interaction between gas and solid that causes
the density difference between gas phase and adsorbed phase. The strength of the
intermolecular (atomic) attraction force is limited, therefore, the density difference
resulted must be limited, and the pressure difference corresponds to the density difference
must be limited. So, there is no reason to think of supercritical adsorption (or
high-pressure adsorption as synonym)as borderless.
A proof for the existence of the border can be found in a parameter of the
above-mentioned model of supercritical adsorption isotherms. Parameter np in Eq.(2)
indicates a saturated amount of absolute adsorption, which is the maximum adsorbate

95
contained in the adsorbed phase, therefore, presents the border of adsorption. The linear
plots of isotherms provide another proof. Shown in Fig.7 is such a plot for the adsorption
of hydrogen on activated carbon AX-2 1 for 77- 298 K [181. All plots intersect at one point,
which defmes a limit state of supercritical adsorption. The representation of isotherm in a
system of n versus pgprovides h t h e r proof of existing border of supercritical adsorption.
As Menon [25] pointed out, such isotherm has a linear section after the maximum. This
was proven by the adsorption of C02 on activated carbon for the near-critical region [22]
as shown in Fig.8. Applying Eq.(l) to the linear section of the isotherm, one would
conclude that both V. and n, must be constant if isotherm is linear. It means the value of 4,
V,, and does not change although pressure keeps increase for the region of linearity.
However, such state of the adsorbed phase is not thermodynamically stable. The density
of bulk gas & keeps increase while the density of the adsorbed phase pa keeps constant.
The metastable state would be broken if pgbecomes larger enough than A.As a symbol of
the breakage of the metastable state, the recorded "isotherm" begins to deviate from
linear.
Determination of the border of supercriticaladsorption is very important for the study
of supercritical adsorption. For example, the fugacity at the border can be used to replace
pr to define adsorption potential or used to draw the 4 plot. In fact, most theories
available for subcritical adsorption would be possible to apply for supercritical adsorption
if the border could be properly defined.

Acknowledgements

This work is subsidized by the Special Funds of Major State Basic Research Projects
(G2000026404) and supported by the National Natural Science Foundation of China
(#I29936100).

References

1. Matranga K. R., Mayers A. L. and Glandt E. D. Storage of Natural Gas by

Adsorption on Activated Carbon. Chem. Eng. Sci. 47 (1992) pp. 1569-1579.
2. Noh J. S., Agarwal R. K. and Schwarz J. A. Hydrogen Storage Systems Using
Activated Carbon. lnt. J. Hydrogen Energy 12 (1987) pp. 693-700.
3. Yang R.T. Gas Separation ly Adsorption Processes (Butterworths, London, 1987).
4. Ruthven D. M., Farooq S. and Knaebel K. S. Pressure Swing Adsorption (VCH
Publishers, New York, 1994).
5. Nguyen C. and Do D. D. Adsorption of Supercritical Gases in Porous Media:
Determination of Micropore Size Distribution, A Pbs. Chem. B 103 (1999) pp.
6900-6908.
6. IUPAC Commission on Colloid and Surface Chemistry Including Catalysis, Pure
Appl. Chem. 57 (1985) p. 603.
7. M. Polanyi, Verh. Deut. Physik. Ges. 1914, 16, 1012; 1916, 18,55.
8. Dubinin M. M. and Astakhov V. A. Trans. from lzyestiya Akademii Nauk SSSR,
Seriya Khimicheskaya, 1971, No. 1 : 5- 1 1.
9. Ozawa S., Kusumi S. and Ogino Y., Physical adsorption of gases at high pressure, IV.
An improvement of the Dubinin-Astakhov adsorption equation. J. Colloid &
Interface Sci. 56 (1976) pp.83-91.

96
10. Aranovich G. L. and Donohue M. D. Adsorption of supercritical fluids, J. Colloid
& hte$ Sci. 180 (1996) pp.537-541.
11. Jiang S. Y., Zollweg J. A. and Gubbins K. E. High-pressure adsorption of methane
and ethane in activated carbon and carbon fibers, J. Phys. Chem. 98 (1994)
pp.5709-57 13.
12. Chen J. H.,Wong D. S. H., Tan C. S., Subramanian R., Lira C. T. and Orth M.
Adsorption and desorption of carbon dioxide onto and from activated carbon at high
pressures. Ind Eng. Chem. Res., 36 (1997) pp.2808-28 15.
13. Aranovich G. L. and Donohue M. D. Surface compression in adsorption systems,
Colloi& andSurfaces A, 187-188 (2001) pp.95-108.
14. Usthov E. A., Do D. D., Herbst A., Staudt R. and P. Harting, Modeling of gas
adsorption equilibrium over a wide range of pressure: A thermodynamic approach
based on equation of state, J. Colloid & Inteface Sci. 250 (2002) pp.49-62.
15. Zhou, L., Zhang, J.-Sh. and Zhou, Y.-P. A Simple Isotherm Equation for Modeling
the Adsorption Equilibria on Porous Solids over Wide Temperature Ranges,
Langmuir 17 (2001) pp.5503-5507.
16. Sipersteh F., Talu 0. and Myers A. L. Gas storage: absolute adsorption versus
excess adsorption, Proceedings FOA7 (2001) pp.3 1 1-318.
17. Staudt R., Saller G., Tomalla M. and Keller J. U. A note on gravimetric
measurements of gas-adsorption equilibria. Ber. Bmsenges. Phys. Chem. 97 (1 993)
pp.98-105.
18. Zhou L. and Zhou Y .-P. Linearization of Adsorption Isotherms for High Pressure
Applications, Chem. Eng. Sci. 53 (1998) pp.253 1-2536.
19. Zhou L. and Zhou Y.-P. A Mathematical Method for the Determination of Absolute
Adsorption from Experimental Isotherms of Supercritical Gases, Cn J. Chem. Eng., 9
(2001) pp.110-115.
20. Zhou Y.-P., Bai Sh.-P., Zhou, L. and Yang, B. Studies on the Physical Adsorption
Equilibria of Gases on Porous Solids over a Wide Temperature Range Spanning the
Critical Region-Adsorption on Microporous Activated Carbon, Cn J. Chem. 19
(2001) pp.943-948.
21. Zhou L., Zhou Y.-P., Bai Sb-P. and Yang B. Studies on the Transition Behavior of
Physical Adsorption from the Sub- to the Supercritical Region: Experiments on Silica
Gel, J. Colloid & Inteflace Sci. 2002 (in press).
22, Bai Sh.-P. Studies on the adsorption behavior of C 0 2 for the near-critical region,
Dissertation for PhD degree (2002) School of Chemical Engineering, Tianjin
University, Tianjin, China.
23. Cazorla-Amoros D., Alcaniz-Monge J., De la Casa-Lillo M. A. and Linares-Solano
A. C02As an Adsorptive to Characterize Carbon Molecular Sieves and Activated
Carbons, Langmuir 14 (1998) pp.4589-4596.
24. Gregg S. J. and Sing K. S. W. Adsorption, Surface Area and Porosity (Academic
Press, London, 1982).
25. Menon P. G. Adsorption at High Pressures, Chem. Rev. 68 (1 968) pp.277-294.

97
 p.-, , , , , .. ,,

D
PYR

Fig. 1 A comparison of the Fig.2 Model fitting the data Fig3 M,,,.J~~~ the
absolute with the excess of C& on activated carbon. adsorption of co2on
adsorption isotherms of Points: experimental; activated carbon for the
C& on activated carbon Curves: model predicted near-critical region. points:
experimental; Curves:
model predicted.
0

ti1::
I

Fig.6 A comparison of the

Fig.4 A calibration isotherm Fig5 The linear plot of DRK plot between the
of C02at 273 K formulating the absolute
excess and the absolute
adsorption of COzat 273 K
adsorption

-t

P#=+rn*

Fig.7 Linear plots of the Fig.8 Linear section on the

hydrogen adsorption on isotherms of C@ on
activated carbon AX-2 1 activated carbon

98
Contributed Papers
This page intentionally left blank
 MICROWAVE DRYING FOR PREPARATION OF MESOPOROUS CARBON

HAJIME TAMON, TETSUO SUZUKI AND SHIN R. MUKAI

Department of Chemical Engineering, Kyoto University, Kyoto 606-8501,Japan
E-mail:[email protected]

TAKUJI YAMAMOTO
Research Institute for Green Technology,
National Institute of Advanced Industrial Science and Technology,
Tsukuba 305-8565, Japan
E-mail: [email protected]

Resorcinol-formaldehyde hydrogels were synthesized by sol-gel polycondensation of resorcinol with

formaldehyde in a slightly basic aqueous solution. RF cryogels, RF xerogels, and RF xerogels (MW
gels) were respectively prepared from RF hydrogels by freeze drying, hot air drying, and microwave
drying. Carbon cryogels, carbon xerogels and carbon M W gels were subsequently obtained by
pyrolyzing RF drygels in an inert atmosphere. Freeze drying and microwave drying were effective to
prepare mesoporous RF drygels and carbon gels. RF cryogels and carbon cryogels showed high
mesoporosity over wide ranges of the molar ratio of resorcinol to catalyst (WC) and the ratio of
resorcinol to water (WW) used in sol-gel polycondensation. Although RF xerogels had a few
mesopores, carbon xerogels had no mesopores. RF M W gels and carbon M W gels showed
mesoporosity if appropriate values of WC and W W were selected.

1 Introduction

Organic aerogels (RF aerogels) are prepared by the sol-gel polycondensation of resorcinol
(1,3-dihydroxybemene) (C6H4(OH)2) (R) with formaldehyde (HCHO) (F) and
supercritical drying with carbon dioxide ( C 0 3 [3]. Carbon aerogels are also obtained by
pyrolyzing RF aerogels in an inert atmosphere at 1323 K [4]. RF and carbon aerogels
have high porosity (> 80 %), and high surface areas (400 - 900 m2/g). The authors have
experimentally elucidated the influence of the amounts of resorcinol, water, and basic
catalyst used in the sol-gel polycondensation on porous structures of EW and carbon
aerogels [5, 61. Carbon cryogels are also prepared via sol-gel polycondensation of
resorcinol with formaldehyde, freeze drying, and pyrolysis [7-101.
The cryogels are unique materials with high surface areas and large mesopore
volumes, and they are expected to be used as catalysts, adsorbents, electric double layer
capacitors, and materials for chromatographic separation. The drying time of freeze
drying is, however, extremely long. Hence, the objective of the present work is to study
the applicability of microwave drying to the preparation of mesoporous carbons.

2 Experimental

Resorcinol-formaldehyde (RF) solutions were prepared from resorcinol (Wako Pure

Chemical Industries, Inc., research grade), formaldehyde (Wako Pure Chemical Industries,
lnc., research grade, 37wt??formaldehyde stabilized with 8 wt?? methanol), sodium
carbonate (C) (Wako Pure Chemical Industries, Inc., research grade) and distilled water
after ion exchange (W). The solutions were poured into glass tubes (inner diameter: 4 mm,

99
length: 4 cm), and gelled by curing them at 298 K for 1 day, at 323 K for lday and at 363
K for 3 days to obtain RF hydrogels. The synthesis conditions are listed in Table 1.

Table 1. Porous properties of RF drygels and carbon gels.

carbon cryogel 7.0 674 0.73 0.16

200 0.250 0.5 RF cryogel 6.2 496 1.14 ND
carbon cryogel 6.2 696 1.29 0.14
200 0.375 0.5 RF cryogel 5.5 554 1.45 ND
carbon cryogel 5.5 662 0.94 0.14
200 0.500 0.5 RF cryogel 4.2 472 1.05 ND
carbon cryogel 2.5 609 0.56 0.13
I00 0.I25 0.5 RF cryogel 7.0 492 1.17 ND
carbon cryogel 6.2 754 0.78 0.15
100 0.250 0.5 RF cryogel 6.2 542 1.06 ND
carbon cryogel 3.6 625 0.82 0.11
100 0.375 0.5 RF cryogel 2.4 558 0.79 ND
carbon cryogel 1.9 577 0.35 0.15
100 0.500 0.5 RF cryogel 2.1 548 0.72 ND
-___ carbon cryogel- .- 1.6 599 0.31 0.13
200 0.125 0.5 RF xerogel 1.8 I52 0.27 ND
carbon xerogel 1.8 49 0.07 ND
200 0.375 0.5 RF xerogel 1.8 255 0.31 ND
carbon xerogel - 1.6 121 0.06 0.03
200 0.125 0.5 RF MW gel 2.1 361 0.58 0.I4
carbon MW gel 1.6 626 0.47 0.I4
200 0.250 0.5 RF MW gel 2.2 323 0.47 0.21
carbon MW gel 2.I 582 0.48 0.41
200 0.375 0.5 RF MW gel 2.4 422 0.56 0.33
carbon MW gel 1.8 549 0.48 0.09
200 0.500 0.5 RF MW gel 2.6 465 0.68 0.25
carbon MW gel 2.0 608 0.47 0.14
100 0.250 0.5 RF MW gel 1.8 407 0.42 ND
carbon MW gel 1.5 407 0.14 0.15
WC:mole ratio of resorcinol to basic catalyst, R/W: ratio of resorcinol to water,
WF: mole ratio of resorcinol to formaldehyde, ND: not detected.

RF hydrogels were immersed into 10 times volume of t-butanol, and RF cryogels

were prepared by drying RF hydrogels at 263 K for 24 hours after pre-freezing at 243 K
for 6 hours [7]. RF xerogels were prepared by drying RF hydrogels in an oven kept at 323
K for 48 hours. RF MW gels were prepared by drying RF hydrogels in a microwave range
(SANYO Electric Co., Ltd.; EM-LAl) for 10 minutes. The changes of the weight of RF
hydrogels during drying were measured and the drying time of 10 minutes is enough to
obtain RF drygels.
Carbon gels were prepared by pyrolyzing RF drygels in a conventional furnace [3].
Nitrogen gas flowed through a quartz reactor containing RF drygels set in the h a c e at
2.0 x lo4 m3/min during the pyrolysis. The h a c e was heated to 523 K at 250 wh and
kept at 523 K for 2 hours. Then the furnace was heated to 1273 K at 250 wh and
maintained at 1273 K for 4 hours. After the pyrolysis, the reactor was cooled to a room
temperature with its own thermal mass.
The porous properties of RF drygels and carbon gels were determined by nitrogen
adsorption. The adsorption and desorption isotherms were measured at 77 K using an

100
adsorption apparatus (BEL Japan, Inc.; BELSORP28). BET surface area, SBET, mesopore
size distribution, and kcroporosity were evaluated. The pore size distribution and the
mesopore volume, V-, were determined by applying the Dollimore-Heal method [ 11 to
the desorption isotherm, and the micropore volume, V,,,i,, was evaluated by the t-plot
method [2]. The cross sections of RF drygels and carbon gels were observed using a
scanning electron microscope (JEOL, Ltd.; JSM-6340FS).

3 Results and Discussion

3. I Porous Structure of RF Drygels

Figure 1 shows the pore size distributions of RF drygels, and Table 1 also shows the
porous properties of RF @gels. r w denotes the peak value of pore size distribution. One
can see the development of mesoporosity in RF cryogels, and the values of rw and V,,
of RF cryogel decrease as decreasing WC or as increasing WW. On the other hand, SBET
shows no obvious dependence on R/C and WW. The above dependencies are considered
to reflect the structure of RF hydrogels. This is because solvent inside the pores is
removed retaining the porous structure of RF hydrogels. From Fig. 1 (b) and Table 1, it
can be seen that pore size distributions of RF xerogels have sharp peaks at around 2 nm,
and SBET and Vmes are much smaller than those of RF cryogels. Figure 1 (c) and Table 1
also show the mesoporosity of RF MW gels. The pore size distributions of RF MW gels
have sharp peaks at around 2 nm.Although it is difficult to detect obvious depencence of
the porous properties on the value of WW,the values of S ~ E Tand V,,, are respectively
estimated to be 65 - 85 % and 38 - 67 % of those of RF cryogels.

R/W=OJ25 ---D-- R/W=O350 ____ &-.. R/W=O375 - - V - . R / W = m

Figure 1. Pore size distributions of (a) RF cryogels, (b) RF xerogels and (c) RF MW gels synthesized under
the condition of WC = 200.

From the above results, it can be concluded that freeze drying is strongly
recommended to obtain mesoporous RF drygels among the drying methods examined in
this work. It is also possible to prepare mesoporous RF drygels by microwave drying.
However, it is difficult to prepare mesoporous RF drygels by hot air drying.
RF aerogels have previously been prepared over wide ranges of WC (12.5 IWC I
800) and R/W (0.125 I R/W I0.500) [6]. By comparing the drying methods from the
viewpoint of the ranges of R/C and R/W to obtain mesoporous RF drygels, supercritical

101
drying is considered to cover the widest ranges, followed by freeze drying, microwave
drying and hot air dtyiig.

3.2 Porous Structure of Carbon Gels

From Figure 2 (a) and Table 1, one can see that the values of V,, of carbon cryogels are
slightly smaller than those of RF cryogels and also see that micropores are formed on
carbon cryogels, which is caused by the shrinkage of mesopores during pyrolysis.
Interestingly, SeErincreases to some extent because of micropore formation during
pyrolysis. Figure 2 (b) and Table 1 show that carbon xerogels scarcely have mesopores
because mesopores of RF xerogels have been collapsed by pyrolysis. Figure 2 (c) and
Table 1 also indicate that the mesopores of RF MW gels have shrunk by pyrolysis. It
should be noted that carbon MW gels prepared under the condition of R/C = 200 also
show mesoporosity although rpcd and V,,, are smaller than those of carbon cryogels.
These results suggest that appropriate values of R/C and R/W should be selected to obtain
mesoporous carbon MW gels. The values of& and V,,, of carbon MW gels synthesized
under the condition of R/C = 200 are confirmed to be 83 - 99 % and 37 - 84 % of those of
carbon cryogels respectively.
The cross sections of RF drygels and carbon gels were observed by SEM and the
following results were obtained. Carbon cryogels are also composed of primary particles.
On the other hand, primary particles composing RF xerogels and RF MW gels are melted
by pyrolysis. Carbon xerogels have smooth surface and no mesopores, which fact supports
the results of nitrogen adsorption. Although the porous structure of carbon MW gels are
partially melted, the cross-linked structure is partially maintained and retaining
mesoporosity.
From the above results, freeze drying and microwave drying are recommended to
prepare mesoporous carbon gels. It is confirmed that freeze drying is available to prepare
mesoporous carbon cryogels over wide ranges of R/C (50 I R/C I 200) and R/W (0.125 I
R/W I0.500).When microwave drying is used, the value of R/C should be kept close on
200 to obtain mesoporous carbon MW gels.

- M
8
-M
8
-M
8

;1'
I
n'
I 5
I

4
I

3 4 3- 4
-
a

-
b
M
-
M
-
Q
>
Q
a 0
3a o
>
4 0
1 10 1 10 1 10

Figure 2. Pore size distributions of (a) carbon cryogels, (b) carbon xerogels and (c) carbon MW gels
synthesized under the condition of WC = 200.

102
 4 Conclusion

1. Freeze drying is effective to retain porous structure of RF hydrogels. It is possible to

prepare mesoporous RF drygels and carbon gels over wide ranges of R/C and R/W.
2. It is difficult to obtain mesoporous RF drygels by hot air drying. The mesopores of
RF xerogels are collapsed by pyrolysis.
3. Microwave drying is also available to prepare mesoporous RF drygels. If the
appropriate values of R/C and R/W are selected, it is possible to obtain mesoporous
carbon gels.

5 Acknowledgements

This research was partially supported by Japan Society for the Promotion of Science,
Grant-in-Aid for Scientific Research (B), No. 14350416 (2002), Industrial Technology
Research Grant Program in '01 from New Energy and Industrial Technology
Development Organization (NEDO) of Japan, and Hosokawa Powder Technology
Foundation (2000).

References

1. Dollimore, D. and Heal, G.R., An improved method for the calculation of pore-size
distribution fiom adsorption data, J. Appl. Chem., 14 ( 1964) pp. 109- 1 14.
2. Lippens, B.C. and de Boer, J.H., Pore system in catalyst V. The t- method, J. Cutul.,
4 (1965) pp. 3 19-323.
3. Pekala, R.W., Alviso, C.T., Kong, F.M. and Hulsey, S.S., Aerogels derived fkom
multifunctional organic monomers, J. Non-Crysr. Solids,145 (1992) pp. 90-98.
4. Pekala, R.W. and Alviso, C.T., Carbon aerogels and xerogels, Muter. Res. SOC.Proc.,
270 (1992) pp. 3-14.
5. Tamon, H., Ishizaka, H., Mikami, M. and Okazaki, M.,Porous structure of organic
and carbon aerogels synthesized by sol-gel polycondensation of resorcinol with
formaldehyde, Carbon, 35 (1 997) pp. 79 1-796.
6. Tamon, H., Ishizaka, H., Araki, T. and Okazaki, M., Control of mesoporous structure
of organic and carbon aerogels, Carbon, 36 (1998) pp. 1257-1262.
7. Tamon, H., Ishizaka, H., Yamamoto, T. and Suzuki, T., Preparation of mesoporous
carbon by fieeze drying, Carbon, 37, (1 999) pp. 2049-2055.
8. Tamon, H., Ishizaka, H., Yamamoto, T. and Suzuki, T., Influence of freeze-drying
conditions on the mesoporosity of organic gels as carbon precursors, Carbon, 38
(2000) pp. 1099-1105.
9. Yamamoto, T.,Nishimura, T., Suzuki, T. and Tamon, H., Control of mesoporosity of
carbon gels prepared by sol-gel polycondensation and freeze drying, J. Non-Clyst.
Solids,288 (2001) pp. 46-55.
10. Yamamoto, T., Sugimoto, T., Suzuki, T., Mukai S.R. and Tamon, H., Preparation
and characterization of carbon cryogel microspheres, Carbon, 40 (2002) pp.
1345-1351.

103
COMPUTER SIMULATION OF TRANSPORT IN CYLINDRICAL MESOPORES

S.K.BHATIA AND D.NICHOLSON

Department of Chemical Engineering, The Universityof Queensland, Brisbane QLD 4072,
Australia

The transport of a sub-critical Lennard-Jones fluid in a cylindrical mesopore is investigated here,

using a combination of equilibrium and nonequilibrium as well as dual control volume grand
canonical molecular dynamics methods. It is shown that all three techniques yield the same value
of the transport coefficient for diffiscly reflecting pore walls, even in the presence of viscous
transport. It is also demonstrated that the classical Knudsen mechanism is not manifested, and that
a combination of viscous flow and momentum exchange at the pore wall govern the transport over a
wide range of densities.

1 Introduction

lntense world-wide activity in applicationsof newly developed templated porous materials,

carbon nanotubes and and a variety of related materials [ 1,2], has led to renewed interest
in the long-standing problem of modeling transport in nanopores. Generally this is
represented in a diffusional m e w o r k with a concentration dependent diffusivity [3-51;
however, there is much controversy regarding the underlying mechanisms. It has for
some time been considered that transport in confined spaces, such as in nanopores,
comprises of a purely diffusive component, as well as a hydrodynamic or viscous
component, [6,7]. The diffusive component is often considered dominant and arbitrarily
treated as activated surface flow [8], Knudsen diffusion [4,6,7,8] or slip flow [9],
particularly at low coverage, and in pores of near-molecular width, where the
hydrodynamic approaches [lo] predict a vanishing transport coefficient contrary to
experiment or simulations [3-51.
It has been argued [ 1 13 that the transport coefficient obtained by dual control volume
grand canonical (DCV-GCMD) simulation, in which the flux in a finite capillary under the
action of a chemical potential gradient is measured, represents the combined effects of
diffusive and viscous flow, while nonequilibrium molecular dynamics (NEMD), in which
the steady state flux in an infinite capillary under the action of a constant force is
measured, should yield only the viscous component. On the other hand it has also been
surmised [5] that equilibrium molecular dynamics (EMD) should yield only the diffusive
component since bulk flow is absent in this method. More recently it has been found
[ 121 that all three techniques yield the same transport coefficient in micropores, where
viscous flow is considered negligible. This therefore still leaves open the question
regarding the differences between the methods in larger pores where viscous transport is
significant which, along with elucidation of the transport mechanism, forms the subject of
the present study. A more detailed account of this work is reported elsewhere [13,14].

2 Molecular Dynamics Simulations

The simulations conducted model the flow of Lennard-Jones (LJ) methane at 150 K and
170 K in a cylindrical silica pore of radius 1.919 nm, having infinitely thick pore walls
comprising spherical LJ sites. For methane we use the established LJ parameter values

104
&jkB= 148.2 K,q =0.381 m. For the solid W parameters we use 4 k B = 290 K, a, =
0.29 nm, obtained by fitting argon isotherms at 87 K in MCM-41 of various pore
diameters [15], using grand canonical Monte Carlo (GCMC) simulation. The
Loren&-Berthelot d e s are used to estimate solid-fluid LJ interaction parameters. A
cut-off separation of 1.5 nm corresponding to about 3.94q is used in computing
fluid-fluid potentials.
In the molecular dynamics calculations the trajectories of methane molecules in the
pore are followed using the equation of motion with appropriate temperature control. A
diffuse reflection condition is applied at the pore wall. For the EMD simulations a
collective transport coefficient obtained from autocorrelation of the fluctuating axial
streaming velocity via a Green-Kubo relation [5]
Dm = N lirnC<
r-hw
u,(O)u,(t) > dt

where udt) = C&id. For the NEMD simulations a constant axial acceleration is applied
to the particles, and a transport coefficient computed from the measured flux as
D, = k,Tg&nr, where 1 is the axial number flux and is the methane density in the
pore. For the DCV-GCMD simulations a three-zone method is used [5,16], with the two
end wnes maintained at different chemical potentials. An effective Fickian transport
coefficient is then computed h m the measured flux through D,,a = -jUAp where I!
is the length of the central gradient zone and Apthe applied density difference.

3 Comparison of Methods

The different transport coefficient obtained at

160 M
A A
A% various densities from the three MD techniques
were essentially identical. Figure 1 [13] depicts
the transport coefficients from EMD and NEMD
at the two temperatures, showing essentially no
difference. Clear evidence of an asymptotic
non-zero transport coefficient at low densities is
also seen, with only weak density dependence in
this region. The inset depicts the comparison of a
01 . . . ' . NEYD- inKI purely
A
I

below,
no-slip viscous theory, to be discussed
with the NEMD results at 150 K, showing
o 2 4 6 a 1012
good correspondence at moderate and high
adsorbate density (nm") densities, and failure at low densities where the
Figure 1. Variation of transport
latter predicts a vanishing transport coefficient.
coefficient with density. Inset Despite the latter deviation the general trends are
depicts the comparison of a purely similar, in particular a nearly constant transport
no-slip viscous theory with the coefficient below a density of about 4 nm-3 (after
NEMD Esllitq 1 so K. an initial increase tkom a value of zero for the
theory), which corresponds to the monolayer
region. The good agreement at high density and similarity in Lend does suggest the role
of viscous effects in the transport, though an additional mechanism is also signified by the
quantitative disagreement at low densities.

105
 Our results clearly indicate that the EMD
transport coefficient does include the viscous
part, despite the absence of imposed bulk flow,
N<
'O0

.
-
640
0
-
contradicting earlier assertions [5,11] discussed 0
2e 480
above. This is rationalised if cross-sectional
equilibrium of the streaming velocity and density 5
320
profiles is attained. Evidence of the latter is g
provided in the inset in Figure 2 for a pore $ 160
density of 5.95 nm-3at 150 K. E
A large number of DCV-GCMC runs were 0
also conducted at 150 K, for various pair of 0 160 320 480 640 800
density values in the two end sections, and the predicted D, x l
Rn Og (m2/s)
effective Fickian transport Coefficient, D,,,
obtained. A similar coefficient is also estimated Figure 2. Comparison of measured
from h e EMD and NEMD values of Dm,that C O e f i c i e n ~obtained using
correspond to a chemical potential gradient DCV-GCMD, with those predicted
based on coefficients obtained using
driving force, following
EMD ( 0 ) and NEMD (0). Inset
jt 1 a hf shows that equilibrium density profiles
Dt,g-= --4 = -
4 ~ ~ D ~ ~' l( npp) (T dp
- ) (2) are attained in EMD and NEMD.

which matches that from DCV-GCMD, as seen in Figure 2 [ 131, over the wide range of
densities covered. Herefis the bulk hgacity of methane in equilibrium with adsorbed
density p, and the associated derivative is estimated from isotherms obtained by GCMC
simulation. Thus, it is clear that neither NEMD nor DCV-GCMD probes any new
mechanism beyond that captured by EMD simulations, even for mesopores.

4 Transport Model

Our theoretical calculations indicated viscous flow to be dominant at high densities, as

depicted by the agreement of the theory in the inset in Figure 1. A viscous flow model
may be used over length scales larger than the mean free path, which is largely satisfied
for mesopores. To obtain the theoretical transport coefficient we solved the Navier
Stokes equation

assuming cross-sectional equilibrium with a no-slip boundary condition, which leads to

Here R, represents the pore radius measured from the center of the surface atoms, and
is the mean pore density. The radial density profile p(r) is obtained from simulations,
while the local viscosity is evaluated using the method of Chung et al. [ 171, at a density
locally averaged over a sphere of radius of/2 [ 181. The radius r, in eq. (4) represents
the position of the minimum of the fluid-solid potential, which is essentially the location

106
of the reflection boundary.
While both theory and simulation yield agreement at high densities, as seen in Figure
1, the measured non-vanishing transport coefficient at low densities does suggest a
significant degree of slip not considered in the theory. To capture this we consider the
surface boundary condition
duL
kpouo= -r;l- at r = r,
dr
where po is the local density at the potential minimum, u, is the slip velocity and k a
friction coefficient. Solution of eq. (4) with the above boundary condition yields the
transport coefficient

in place of eq. (4), which has the important feature that it predicts an asymptotic
non-vanishing transport coefficient in the low-density
region. The variation of the ffiction coefficient k
with local density p, at the surface of friction (the
location of the potential minimum) can be obtained by
substituting the values of D,,and density profile p(r)
obtained from GCMC simulation at density into
eq. (6). Figure 3 [13] depicts the results, showing a
relatively constant value of the friction constant k of
about 1.2- 1.6 N.sec.mole-' at I50 K and 1.3- 1.7
N.sec.mole-' at 177 K, for local density ,q, below a
0 20 40 80 80 100
critical value (about 85 m-3at 150 K, and 70 run3 at
density at potential minimum (nm? 177 K). Subsequently it increases steeply to very
Figure 3. Variation of friction large values, approaching the no-slip condition at high
factor with density at potential densities.
minimum, obtained at temperatures The constancy of the friction coefficient is a
of I50 K and 177 K. Inset depicts strong indicator of the importance of slip flow at the
predicted and measured streaming
pore wall in rnicropores. Further, calculations
velocity profile at adsorbed density
of 5.95 nm-3 and temperature of showed that the above values of the friction constant
150 K. are closely consistent with momentum transfer
arguments. For this we consider the frictional force
as arising from the momentum loss ondiffise reflection at the wall leading to
kp,u, = mu ,
Z (7)

where Z ( = p, V / 4) is the collision frequency under conditions of local equilibrium.

Substitution of the kinetic theory result V :=-/, yields k = -/,
providing the estimates k = 1.78 N.sec.mole-' at 150 K and 1.94 N.sec.mole-' at 177 K, in
remarkably good agreement with the values obtained from the simulation results.
Further support is obtained from comparison of the predicted streaming velocity profile
from solution of eq. (4) with that obtained from NEMD simulation. The inset in Figure 3
depicts one such comparison, obtained for an adsorbed density of 5.95 nmJ at 150 K and

107
imposed acceleration of 0.07 nm/ps*, showing excellent agreement except for some
deviation in the inner low density region where the flux is negligible.
The abrupt increase of the wall fiction coefficient to very large values leading to the
no-slip condition at a critical density is most likely due to the high frequency of
interparticle collisions in this region. Diffuse wall reflection then dissipates the collective
axial momentum in this region, and the no-slip condition is approached as seen in figure 3.

5 Conclusion

It is clear from the above results that all three simulation techniques yield identical
results for the transport coefficient in pores with diffusely reflecting walls. Further, a
combination of momentum transfer at the wall and viscous transport in the fluid suffices to
explain the transport behavior of pure component fluids in mesopores.

6 Acknowledgements

This research has been supported by a grant from the Australian Research Council under
the Large Grants Scheme.

7 References

1. M.E. Davis, Nature 417,813 (2002).

2. S . Ijima, Nature 354,56 (1991).
3. S.K. Bhatia, Proc. R SOC.Lond A446,15-37 (1994).
4. S.K. Bhatia, in Advances in Transport Processes LY,edited by AS. Mujumdar and
R.A. Mashelkar (Elsevier, Amsterdam, 1993).
5. D. Nicholson and K. Travis, in Recent Advances in Gas Separation by
Microporous Membranes, edited by N. Kanellopoulos (Elsevier, Amsterdam,
2000).
6. E.A. Mason, A.P. Malinauskas and R.B.Evans, J. Chem.Phys. 46,3 199 (1967).
7. R. Jackson, Transport in Porous Catalysts (Elsevier, Amsterdam, 1977).
8. J. Ktirger, and D.M. Ruthven, Dimion in zeolites and other Microporous solids
(John Wiley, New York,1992).
9. D. Nicholson, JMemb Sci. 129,209-219 (1997).
10. J.H. Petropoulos, and G.K. Papadopoulos, J. Memb. Sci 101,127 (1995).
11. K. D. Travis and K.E. Gubbins, Molecular simulation 25,209 (2000).
12. G. Arya, H-C. Chang. And E. Maginn, J. Chem. Phys. 115,8112 (2001).
13. S.K.Bhatia and D. Nicholson, Phys.Rev. Lett., accepted (2002).
14. S.K.Bhatia and D. Nicholson, J. Chem. Phys., submitted (2002).
15. M. Kruk. and M. Jaroniec, Chem. Muter. 12,222 (2000).
16. R.F. Cracknell, D. Nicholson and N. Quirke, Pbs. Rev. Lett. 74,2463 (1995).
17. T.H. Chug, M. Ajlan, L.L. Lee and K.E. Starling, Ind Eng. Chem. Res. 27,671
(1988).
18. 1. Bitsanis, T.K. Vanderlick, M. Tirrell and H.T. Davis, J. Chem. Phys. 89, 3152
(1 988).

108
 MULTICOMPONENT MASS TRANSFER DIFFUSION MODEL FOR THE
ADSORPTION OF ACID DYES ON ACTIVATED CARBON

K.K.H. CHOY, J.F. PORTER AND G. MCKAY

Department of Chemical Engiwering, Hong Kong University of Science and Technologx Clear
WaferBay, Kowloon, Hong Kong.
E-mail: [email protected]

The ability of activated carbon to adsorb two acid dyes, namely, Acid Blue 80 (ABSO) and Acid
Yellow (AY117) from wastewater has been studied in both single component and multicomponent
systems. Two single component systems and one binary system have been studied experimentally by
measuring the equilibrium isotherm and the concentration versus time decay curves in batch kinetic
systems. For the equilibrium isotherm studies, two single component isotherms (AB80 & AYI17)
were analysed using the Redlich-Peterson (RP) equation. For the batch kinetic studies. the effects of
initial dye concentration and activated carbon mass on the rate of Acid Blue 80 and Acid Yellow 117
removal have been investigated. A two resistance mass transport model based on film and surface
diffusion control, Homogeneous Surface Diffusion Model (HSDM), has been applied to model the
concentration decay curves in single component system. A multicomponent HSDM was developed to
predict the binary component system decay curves from the single component HSDM and the
IAST-RP model.

1 Introduction

Modeling in multicomponent adsorption systems is an extension to that of single

component adsorption. Many models have been reported in the literature for the prediction
of concentration versus time decay curves in single component batch adsorption system.
However, there are very few research papers on the topic of multicomponentmass transport
studies for liquid phase adsorption, therefore, it is a valuable contribution and novel
development to adsorption research.

2 Methods

A multicomponent HSDM for acid dyelcarbon adsorption has been developed based on
the ideal adsorbed solution theory (IAST) and the homogeneous surface diffision model (H
SDM)to predict the concentration versus time decay curves. The IAST with the Redlich-P
eterson equation is used to determine the pair of liquid phase concentrations, Cs,land C.s,2,
fiom the corresponding pair of solid phase concentrations,q5.1and qs.2,at the surface of the
carbon particle in the binary component.
The concentration of solute inside the typical particle at distance r from
the centre and at time t is defined by qi (ri, t) and the variation of 9 with distance and time
is governed by the diffusion equation for each componentj:

A Dimensionless variables are applied by defining q = Xfli, D t and ,y = L in

=s
R2 R
equation (1) to form equation (2) with following boundary condition (3) and (4).

109
 do,r )= 0 = ul(x,o)(31,
The material balance equation for the multicomponentsystem and the rate of change of
solid phase concentrationof each componentj are:
' 1 =',.I - ( $ ) 9 J- (5)y E('J)= 3fpJ(xJ9rJ)xja J (6)

The mathematical problem presented by the model requires the simultaneoussolution of

equations (2) to (7). The starting point is a semi-analytical solution of equation (8) [ 11,
previously only applied to single component systems, that satisfiesthe boundary conditions

By considering a short time interval, ro.rl, r2 rn,in the integration of equation (8) forj
,

component and it is assumed that the time intervals (rI- T , - ~ ) are sufficiently short that
f(?) is effectively constant in each interval to solve equation (8) by a numerical method.

The IAS theory dictates that the spreading pressure should be constant for each
component in a given system (v,= vZ= ...= y,,)[2].

For a given pair of qe,,and qe,2,the values of sI,s2 amd qT cam be calculated.
1 s (12)
-=l++ s
9r 9z.1 4r.2
To calculate the spreadingpressure, numerical computer integration is used to solve the
' equation (1 3).

For a given (guess values) pair of qe,1 and qe,2,the values of S1,S2 and qTcan be calculated.
By applying the Redlich-Peterson equation (14) (Redlich & Peterson, 1959) in the equation
(12), it gives equation (15).
1 S. S,

Since 'y, = 'y, ,a numerical computer program is used to optimize the ct1value until
the spreadingpressures of component 1 and 2 are equal. Hence, the value of C,, and Ce,2can
be calculated from the c;., and c,q2by equations (16) and (17) while q I= c:,~ and
c8,*= c;,,.Consequently:

110
3 Results and Discussion
3. I Equilibrium Isotherm Studies
The equilibrium isotherms were measured for Acid Blue 80 (AB80) and Acid Yellow
1 17 (AY 1 17) on Activated Carbon F400.The Redlich-Peterson isotherm (equation 14) is
used to relate equilibrium concentrationsbetween liquid phase and solid phase loading. The
values of KR, aR andpare 28.32dm3/g, 103.6 (dm3/mmole)0.965, 0.965 for AB8O on activated
carbon and 55.40dm3/g, 200.2 (dm3/mmole)0.910, 0.910 for AY117 on activated carbon.
Those isotherm parameters have been used in the multicomponent HSDM to correlate the
concentration decay curve.

3.2 Batch Kinetic Studies

The effects of initial solute concentration and adsorbent mass have been studied on the
diffusion processes for the multicomponentHSDM. The HSDM was used to correlate two
acid dyes (AB80 & AY 117) in the single component system in order to determine the
relation between the single component and multicomponent systems. A comparison of the
theoretical and experimental data for the AY 1 17 in ABSO+AY 1 17 binary systems with
different carbon masses is illustrated in Figure 1. The correlated results are good for all
adsorbent mass variations for the AB80+AY 1 17 system. There is some deviation in the first
30 minutes for the three binary component systems. This is due to the single component
external mass transfer coefficient, k j l and kJ2, (see Tables 1&2) which are used in the
multicomponent HSDM model; indicating that the values of the external mass transfer
coefficient in the multicomponent system are slightly different to those of the single
component system. The Ds,iand SSE values for AB80+AY I I7 binary system with different
carbon masses are shown in Table 1. It was found that parameters D.,I and Ds,2are almost
constant for different carbon mass systems. These results are consistent with the findings in
the single component HSDM and the values of the surface diffusion coefficient in the binary
system of the acid dyes are different to those values in the single component systems (shown
in Table 2).

Figure 1: Effect of Activated

Carbon Masses on the Adsorption
of AY117 in AB8WAY117 using
MulticomponentHSDM (IASRP).

0.5 4 1
0 200 400 600 800

Time (min)
eO.6 WO.88 A I . 2 g 0 1 . 7 g X2.2g

111
 Figure 2: Effect of Initial Dye
.oo
1
Concentrations on the Adsorption
of AB80 in ABSO+AY117 using
0.90
MulticomponentHSDM (IASRP).
uo 0.80
3 0.70
0.60

0.50 1 1 1

For the AB80+AY 1 17 system, the D, values of AB80 in the single component system are
smaller than the Ds,fvalues of AB80 in the AB8O+AY 1 17 binary system while the D, values
of AY 117 produce an opposite trend, the D, values of AY 1 17 in the single component
system are greater than the Ds,2values of AY 1I7 in the AB80+AY 1I7 binary system. The
D, values of AB80 (DS,Jare greater than the D,,values of AY 117(Ds,2)in the AB8O+AY 1 17
system. This may be because the AB80 dye has a higher molecular mobility in the
AB80+AY 1 17 mixture. This implies a higher adsorption affinity towards the surface of the
activated carbon than the AY 117 dye.

The effect of different initial dye concentrations is shown in Figures 2 for a wide range of
initial dye concentrations for AB80 in ABSO+AY1 17 binary system. A good correlation
between experimental and theoretical results is also obtained in the binary systems. The
fitted parameters, kl, k2,D , , and Ds,2are given in Table 1. According to the table, the
surface diffusion coefficient is dependent on the initial dye concentration of each system, as
in the single component system, parameters Ds,, and Ds,2are strongly dependent on the
initial dye concentration. The Higashi expression [3] has been used to correlate the
relationship between the multicomponent surface diffisivities, 0,.I and Ds,2, and the
fractional coverage, 0" and BB, of the acid dyes in the binary adsorption system.
4.OE-I0 Figure 3: The Plots of Surface Diffusivity against
Fractional Coverage Expression for AY117 in
3.OE-I0
Single Component System and Binary System
(AMWAY1 17).
- 2.OdlO
a*
I.OE-10

112
The expressionsare shown in equations 18 References.
and 19 and the plots are shown
- in Figure 3: [l] McKay, G., Application of Surface
Diffusion Model to the Adsorption of
Dyes on Bagasse Pith, Adsorption 4
(1998) pp. 361-372.
[2] Radke, C.J., Prausnitz, J.M.,
Thermodynamics of multi-solute
where D,and 0 2 are the self-diffiivities adsorption from dilute liquid solutions.
for the component 1 and 2 in the mixture. It AlChE Journal 18(4) (1972) pp.
was found that the surface diffusivitiesare 76 1-768.
lineirly proportion to the M o n a 1
coverage expressions and are similar in [3] Higashi, K. Ito, H. and Oishi, J. Surface
trend with the Diffusion Phenomena in Gaseous
findings in the single component HDSM Diffusion: I, Suflace Di@iuion of Pure
(see Table 3). GUS5 (1963) pp. 846-853.
Table 1: Extend Masa T r a d e r and Solid-phuc Diffusion
Coeffidentafor the Adaorplioa of ABM)cAY117 011 Aetiv.
ted Carboa using tbe Multieomment HSDM (IASRP).
SOP

Tabk 2: External Mass Traoafer md Sdiipiuse Diffusion

Cocfiieienta for tbe Adaorplh of A W and AY117 on Act
ivated Carbon in Siagk Component System using tbe HSD

i.
1.7
1.7 I f

100
I00
I 7.36xlC"
934x10'" I :::: I
1.69xIO-~~
2.26~10-'~
9.11~10'
9.35xIQ"
9.34~10'"
6.002
7.749
3.363
3.070
6.471 '
1 I
1 . 2 3 ~ 1 0 ' ' ~ 1.498
1 . 7 1 ~ 1 0 ~ ' ~ 3.512
2 . 5 8 ~ 1 0 ' ~ ~ 6.002
3 . 5 5 ~ 1 0 " ~ 7.749
1.86~10' 3.363
1 . 8 9 ~ 1 0 " ~ 3.070
1 . 7 9 ~ 1 0 " ~ 6.471
1.7 100 9.34x10'" 3.512 1.71~10''~ 3.512

Table J:Slope and y-Intercept Valucs of the Correlated Li

nee in the Plots of tbe Surface Diffwivity against Fnetiona
I Coverage Erprrssioa for Smgk Component and Binary S

some y-intcmpt
ABW (sin&) 2.646~10' 2.686~10'
AdBhK80
, ABW+AY117
BW+AY117 1.449xI1T'~
1.449~IO.'~ 1.524~10"~
1.524~10"~
MYdlow AYll7(riagk) 2.702~10' 2.8%x109
117 ABW+AYI17 2.525~10~ 2.566xlO9

113
 SORPTION THERMODYNAMICS OF NITROUS OXIDE / LSX ZEOLITE
SYSTEMS

MARTIN BULOW, WNGMIN SHEN AND SUDHAKAR R.JALE

BOC PGS Technology, 100 Mountain Ave., Murray Hill, NJ07974, USA
E-mail: [email protected]

Sorption thermodynamic functions of nitrous oxide, N20, are described for zeolites NaLSX and
CaLSX in shepes of clay-bound beads. They were determined by the Sorption Isosteric Method (SM)
over complete ranges of sorption-phase concentration and compand with those fix carbon dioxide,
CQ,RpoltedearliaforthesameNaLSXsorbent.

1 Introduction

Nitrous oxide is a greenhouse gas. Its earth-atmospheric concentration (currently =2ppm)

increases steadily by c (0.2-0.3)% p.u. Ihis is caused mainly by antmpogenic activities and
emissions limn chemical pmsses, eg., synthesis of adipic acid for Nylon-66, automotive power
genedon and wastewater tmtment N@ is very stable in &,its lifetime amounts to c. 150 years.
Remod of N o knn air streams in tiOnt of airseparation units (ASV) [I] is critical as that of
HzOand (2%. Excess of N@ in ASU may lead to plugging of tubes and heat exchangers and to
-
3
4
c onof products, specificallyof noble gases. As the Concentration of N a in air increases
fhther, the current regime of air-pqmilication units (PPV)in h t of ASU may become
inadequate, since N o cannotbe removed easily by existing technologies. Bmkduwgh curves for
NzO in a CaA-zeolite bed at 8,OOO-ppmN20and variouS C q concenlmtions in air [2] indicate that
C Q displaces NzO as the mass-tmnsfm h t progresses along the adsoher bed. This finding
questions the usage of CaA as single s o h t for N a removal. Although N@ and C Q have
identical molecular weights, soxption interaction of NzO with zeoliteNaX modifications, especially
NaLSX [3], deviates h m that of CQ due to d-i in specific Properties of these gases such
as dipole and quadnrpole moments. It is importslnt to undemtand so@m thmodynamia of NzO
on PPU sorben& and to assess difkences in Sorption properties between N a and COz. Such
knowledge will allow for development of novel materials and @on processes to purify ASU-
fdgas.

2 Sorption Isosteric Method and Sorbents

The Sorption Isosteric Method (SIM)was used to determine thermodynamic data of N2O
sorption by zeolites NaLSX and CaLSX. This technique and its utilization for an
investigation of sorption-thermodynamic properties of COz on identical NaLSX beads
were described in detail in refs. [4-81.
Sorbentsused representbeads of NaLSX (98.99% Na) and CaLSX (97.29 % 0, both of
(8 x 12) mesh. CaLSX is prepared by Na vs. Ca2+exchange of NaLSX beads with c. 12 %
attapulgite binder. Ihe appamt XRD crystallinity ofNaLSX beads amounts to c. (78+5)% as
amparedto arefenmce sample without binder. Dry weights ofNaLSX and CaLSX beads utilized
by SIM amount to 92727 g (bulk densii of NaLSX: 0.67 and 8.7033 g, respectively. Prior
to the z e o b rlre a c t i care~v
~ in ti^ S I M - ~ ~ ~ sample
~ W Wc e at ~ <1 0 - ~ t o ~
and 673K, during 4 8 h m ibr NaLSX (beads wm nearly dry at loading), and over 7days for
CaLSX (fully hydrated beads).

114
3 Results and Discussion

3. I Sorption Isosteres of N 2 0 on NaLSX and CaLSX Zeolites

&@on isosteresmeasured for the N2GNaLSXand N2ocaLsX systems over complete ranges
of sopion-phase concentmtiom,n,are presentedin Figs. 1 and 2, respectively. Slopes
- i ' i ' i ' i ' i ' l ' i ' l ' i ' i ' l
n. mom:
rn 0.0592 7
0 0.1167 :
A 0.2200 :
v 0.4599 -
+ 0.7225 .
1' x 1.0304
f 1.3083 -
1.7314 :
-
0
L
0 2.1365
0
.cI
D
A 2 . 7 ~
D
v 3.0571
Q
0 3 . m
W
0
4.2140
4.7400
-:
4 5.2701 -
I 5.5905 -
0 6.20M
D 6.9337.
I 7.7254
E 8.6133 Y
.
i '
'.
i ' i ' i ' i ' l ' i ' l ' i ' l ' i ' i ~
3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4

1000/(T, K)
Figure 1. Sorption isosteres for NzO on NaLSX zeolite (clay-bound beads).
of these isosteres change with n if the sorption heat depends on concentration. Except for
a few isosteres for concentrations close to saturation capacity, the isosteres determined at
actual experimental conditions appear to be liiear, indicating that no sorption-phase transition
takes place. Therefore, they are characteristicof well-defined sorption-equilibriumprocesses
of N20on NaLSX and CaLSX zeolites. They are attributed to the correspon-ding sorption-
phase concentrations set via dosing procedures. Since the boiling point, 184.67 K, and the
triple point, 182.33 K, of an N20-bulkphase at 1 at, are close to each other, the related phase
transitions can be. observed h m isosteres as N20 concentmtion approaches and exceeds the
sorbent-samation capacity. This feature is obvious h m r.ks. isosteres in Figs. 1 and 2. From
particular slopes of two segments of the isostere for highest n, 8.6133 molikg, shown in Fig.
1, the latent heats of vaporization, 16.55 kl/mol at 184.67 K, and fusion, 6.54 kJ/mol at
182.33 K, are obtained. These quantities agree well with handbook data [9].
Values of isosteric sorption enthalpy, -AH,standard sorption entropy, AS3 and
standard Gibbs fiee sorption energy, AGO, are calculated as dependences on n, c j , [4-81.
To calculate AGO, the boiling-point temperature of N20 is chosen as reference state. This
choice provides a check for thermodynamic consistency of experimental data since AGO

115
should become zero as n exceeds the saturation capacity at boiling-point temperature.
Values of AGO at any other temperature can be calculated from those of -AHand AS
using the Gibbs equation. Thereafter, equilibrium mass distributions, i.e., pressures that
correspond to the quantities n, -AH,AYand dG3 can be determined.

1000/(T,K)
Figure 2. Sorption isosteresfor N20on CaLSX zeolite (clay-bound beads).

3.2 Sorption Thermo&namics of N20 and C02 on N a N and C a mZeolites

Sorption thermodynamic data for N 2 0 on NaLSX and CaLSX obtained h m isosteres as
functions of n, are compared with those for C02 on NaLSX, cf.,Figs. 3-5. For NaLSX, the
-AH vs. n dependence for C02 proceeds always above that for N20. As concentration
exceeds the saturation capacities for the two gases, the enthalpy for C02 sublimation, viz.,
- 25.23 kJ/mol, and the enthalpy for N2O evaporation, - 16.55 kJ/mol, are derived from
the corresponding isosteres. This finding characterizes two bulk-phase transitions that take
place and confirms correctness of experimental data. At low n, the -AH values for N 2 0
decrease sharply with increasing n, unlike for C02, for which the plot -AHvs. n shows a
plateau at n o 1.5 m o m . Sorption thermodynamics suggests that interaction between N20
and NaLSX at low coverage is much weaker than that for C02, although the molecular
weights of the two gases are identical, their quadrupole moments differ slightly only
(3.65 x e.s.u. and 4.3 x e.s.u. for N2O and C02, respectively [lo]), and N2O has
a dipole moment, 0.167 D.
Standard sorption entropy values, ds, that refer to a gas-phase pressure,
p = 760 torr, as standard state, show wave-like concentration dependences for the two
gases on NaLSX, cf.,Fig. 4. Compared to their standard gas phases, a remarkable entropy
loss takes place over the entire ranges of n. Particularly, C02 experiences less freedom
than N20, due to its stronger interaction with sorption sites and, probably, denser
sorption-phase packing.

116
 c

;
=i
Y
i
?

lo
o
1
0
~ .
1
l .
2
l
3
. l
4
. l
5
. l
8
.
7
l .
8
, .
I
9
~ . {
n, mollkg
Figure 3. Isosteric sorption heats of N20and C02on CaLSX and NaLSX.

40- . I . I . 1 . 1 . 1 . 1 . 1 . 1 .

60 - -a- N,O I CaLSX

-0- CO, I NaLSX

Y
e--o

-120 -

-1401
0
n I
1
v

2
I - I
3
= I
4
. I
5
- I
6
* I
7
I
8
*
0
n, moMkg
Figure 4. Standard sorptionentropies of N20 and C02on CaLSX and NaLSX
Dependences dG" vs. n of N20 and C02 for NaLSX, as r e f e d to both the boiling
ternperatwe of each gas and the gas pressure, p o = 760 ton; as standard state are shown
in Fig. 5, together with related data for the N20-caLSX system. Values Mi0 for all
systems change k n negative ones to zefo as n increases, approaches and exceeds the
micropore-saturation capacities.This proves thennodynamic consistency and correctness of
the SIM data,

117
 0

-5

-2 5

-30

- 3 5 : . 1 . I . , . , . I . I . , . I . I
0 1 2 3 4 5 6 7 8 9
n , mol/kg
Figure 5. Standard Gibbs free sorption energies of NzO and Cot on CaLSX and NaLSX,
referred to the boiling temperatures as standard.

A large difference in AGO values exists between the two gases for n c c. 2 molkg on
NaLSX, which disappears as n increases. This proves that COz is sorbed preferentially over
NzO. Asp and, thus, n increase, this selectivity disappears. If the quantities AGO for NzOand
COZare referred to one and the same temperature, e.g., 298 K, sorption isotherms can be
obtained directly and compared to each other at this temperature.
Sorption thermodynamics for NzO-CaLSX is characterized by quantities -AH,AS" and
AGO as functions of n, which are shown in Figs. 3-5 for comparison with those for the
NaLSX systems. Fig. 3 proves that CaLSX compared to NaLSX, provides sorption sites
for NzO, which exhibit stronger sorption interaction at n < 1.5 mol/kg. This indicates
that NzO-Caz' interactions are considerably stronger than the NzO-Na+ ones, in LSX
supercages. As n increases fiom ca. 1 to 5 molkg, the isosteric sorption heats of NzO
on NaLSX and CaLSX zeolites become close to each other. Most probably, non-
specific van der Wads-type interactions between N 2 0 and the lattices prevail, and they
are nearly the same for NaLSX and CaLSX. Fig. 4 shows AS" vs. n dependences for NzO
on CaLSX and NaLSX. Interestingly, these curves differ significantly for the two cationic
modifications of LSX. For NaLSX, NzOprobes at least two different (cationic) sorption sites.
For CaLSX, the curve shape expresses a more smooth distribution of energetically
heterogeneous sorption sites, which reminds qualitatively that for the Nz-CaA system [6,7]. The
shape of the B" vs. n dependence for COz on NaLSX is reminiscent of that for the NzO-
NaLSX system and witnesses preferred sorption of Cot. Standard Gibbs free sorption
energies as basic sorption-equilibrium quantities of NzO and COZ, are referred to
184.67 K and 194.65 K, respectively. The "more negative" values for NzO on CaLSX if
compared to those for NaLSX at n<2molkg, indicate that CaLSX attracts N 2 0
significantly stronger. At n > 2 molkg, values -AGO of N20 for the two zeolites become
close to each other because non-specific sorption interaction between NzO and the LSX
lattice is similar in the two materials. This situation is reflected by the isotherm courses for
NzO on the two sorbents, cJ,Fig. 6, which were derived from SIM data.
Figs. 3-6 elucidate differences in sorption thermodynamics of systems, NzCX~LSX,NzO-

118
NaLSXand CQ-NaLSX. Generally, the -AH data obey a sequence, C&NaLSX >N2G
CaLSX >N2o-NaLsx, over almost the entire ranges of n. However, the cormponding -0
sequence varies withe which can be seen fiom Fig. 5. At n o 1.5 molkg,the absolute values
8 . ......., .......

-0- CO, I NaLSX

-0- N,O I NaLSX

m
0 4- -
5
i 3- -
-
-
0 ......... .................... ,,.,..I ...............
n
10" 10' loo 10'
10' lo3 10' 105
p, torr
Figure 6. Sorption isotherms of N 2 0 and C02 on CaLSX and NaLSX at 298.15 K.
of dG" obey the order N20aLSX > CGNaLSX > N20-NaLSX. This indicates that sorption of
N20 by CaLSX is stronger than that of C Q by NaLSX, in that @on. In the mge(2 o n
4.5) m o m the ditkmces for the three systems dkqmr, but they reoccur in a partly r e v 4
order at n o 4.5 m o b : N20aLSX > N20-NaLSX > COTNaLSX.
From these dependences A G O vs. n, sorption isotherms can be obtained for any tempemure
and pressure values that are meaningM physically, as exemplified in Fig. 6. Isotherms therein do
not only illustrate differences in sorption properties for the thm systems, NzO-C~LSX,N20-
NaLSX and CQ-NUX, but make it obvious that - at low equilibrium pressure - the sorption
capacity of N20 on CaLSX is comparable with that of C02on NaLSX. These capacity values are
much higher than those for N20on NaLSX. Moreover, at very low equilibrium pressure, < 0.5 torr,
the sorption capacity of CaLSX for N20 is higher than that of NaLSX for C02. This particular
feature makes the CaLSX mIite a preferred N20-selective sorbent for the removal of trace
amounts of N@ h m air in h n t of ASU [lo]. For none of the systems investigated at cryogenic
tem- a Henry region could be identified h m the sorption isostereS, which is due to strong
sorption interactions and m&odical reasonsas well.
A quantitative comparison of sorption-thermodynamicproperties h r N20 as described here
with l i t e m data for a CaA-tYpe zeolite [2] meets difficulties since no such thermodynamic data
could be derived reliably h m the latter source. Prefmntial interaction of e m - h e w o r k Ca"
cations with C@ compared to that with N20as identified in [2] agrees qualitatively with the results
of this investigation.

4 Conclusions

Equilibrium sorption thermodynamic properties of N20 sorbed by NaLSX and CaLSX are

119
investigatedby SIM. These are the first comprehensive data sets ever measured for N20. They
are compared with those for the C&NaLSX system. Soption of N20 on NaLSX is much
weaker than that of CQ on the same zeolite. At low pressures, < lotorr, NaLSX has a very
high sorption capcity for CQ, but has a low capacity for N@. CaLSX has a higher sorption
capacity for N20at pressure less than 0.5 torr, compared with that of NaLSX for CQ, and it has
nearly the same sorption capacity at high pressures. Highly exchanged CaLSX zeolites are
recommended for use as N20-SeleCtive sorbents. By combining NaLSX and CaLSX in certain
adsorber-bed layer anangements for TSA PPU proceses, CQ and N20 could be removed
subsequently and completely h m air streams.

5 Acknowledgements

The authors are grateful to Drs. F.R. Fitch and A.F. Ojo, Murray Hill, for helpful
discussions. They thank the BOC Group for the permission to publish this work.

6 References

[l] Golden T.C., Taylor, F.W., Johnson, L.M., Malik, N.H., and Raiswell, C.J.,
Purification of Air, US Patent No. 6,106,593; August, 22,2000.
[2] Mayinger, F., and Egg&-Steger, R, J Energy. Heat & Mars Traqfiw, 15 (1993) 165.
[3] Ojo, A.F., Fitch, F.R., and Biilow, M., Removal of Carbon Dioxidefiom Gas Streams,
US Patent No. 5,531,808; July 02, 1996.
[4] Biilow, M., and Shen, D., in Fundamentals of Adsorption-6 (Ed.: F. Meunier),
Elsevier, Paris, 1998, p. 87.
[5] Shen, D., and Biilow, M., Micropor. Mesopor, Materials, 22 (1998) 237.
[6] Shen, D., Engelhard, M., Siperstein, F.,Myers, A.L., and BUlow, M., in Adsorption
Science h Technology (Ed.: D.D. Do), Proc. 2d Pacific Bash C o d Adsorption Science
Technol., World Scientific, Singapore, 2000, p. 106.
[7] Shen, D., BUlow, M., Jale, S.R., Fitch, F.R., and Ojo, A.F., Micropor. Mesopor.
Materials, 48 (2001) 2 1 1.
[8] Biilow, M., Shen, D., and Jale, S.R., Appl. Surf Sci., 196 (2002) 157.
[9] The Mathewn Company, Matheson Gar Data Book 4* Edn., East Rutherford, 1%6, p. 83
(CW; P. 387 ( N 2 0 ) .
Eyring, H., Henderson, D., and Jost, W., Physical Chemistry: An Advanced Treatise,
Vol. 1V:Molecular Properties, Academic Press New York, 1970, p. 381.
Shen, D., Huggahalli, M., Biilow, M., Jale, S.R., and Kumar R., TSA Process for the
Removal of Dinitrogen Oxide, Hydrocarbons and Other Trace Impuritiesfrom Air,
US Patent No. 6,391,092; May 21,2002.

120
 ACTIVATED CARBON MEMBRANE WITH CARBON WHISKER

SANG-DAE BAE AND AKIYOSHI SAKODA*

Institute of Industrial Science, University of Tokyo
4-6-1 Komaba,Meguro-ku, Tokyo,Japan
E-mail: [email protected]
A novel composite membrane (an activated carbon membrane with carbon whiskers) applicable to water
treatment was fabricated by a combination of conventional carbonization and thermal deposition methods. The
carbonization was performed at 1050-1100 c using poly-vinylydenchloride (PVdc) and ply-vinylchloride
-
(PVC) microsphens, and the thermal deposition was accomplished after femc sulfate (Fe2(S04)j nH20) was
applied as a coat on and within a ceramic support. The membrane consists of three parts, each with a different
function for water treatments: a carbon whisker layer, a carbon layer, and an activated carbon layer. Those
structural characteristics were confirmed with a scanning electron microscope (SEM) and by measuring
electrical resistance and water and gas permeability. Water treatment experiments, using phenol and
polyethylmethacrylate(PMMA) as model pollutants, indicated that this membrane was able to remove dissolved
organics with low molecular weights and suspended solids simultaneously. Also, the carbon whisker layer
prevented the fouling of fine particles on the membrane effectively with performing the frequent back washing.
These results indicate that the novel composite membranes can clarify the water within a relatively short time,
and are usefbl for small scale distributed drinking watedwaste water treatments.

1. INTRODUCTION

Watedwastewater treatment usually consists of several component operations such as

biological treatment, coagulation, sand filtration, membrane filtration, and activated
carbon adsorption. Raw water usually contains solid particles ranging in size from
nanometer to millimeter, as well as dissolved pollutants. This means that any single unit
operation cannot remove all pollutants in one stage. On the other hand, simultaneous
removal of particulate and dissolved pollutants can be achieved if we can combine the
physical filtration and adsorption processes. From this standpoint, we first developed an
activated carbon membrane (Sakoda er ul., 1996, 1998). However, during the membrane
filtration operation, we cannot prevent the deposition and accumulation of fine particles
on and within the membrane, resulting in a cake layer. As a promising method to solve
this fouling problem, we then developed a novel membrane called the carbon whisker
membrane (Li er ul., 200 1a, 200 1b, 2002a), which basically contains vapor-grown carbon
fibers (whiskers) on the external surface of a carbon-coated ceramic support. The
filtration experiments using polyethylmethacrylate (PMMA) as model particles indicated
that the carbon whiskers can prevent direct contact between the membrane body and
PMMA particles, and that fouling would occur with greater dificulty than on as
membranes without carbon whiskers (Li,submitted). Unfortunately, however, the water
flux of this carbon whisker membrane was not large enough for practical use, for the
following reason. To solve this fatal problem, we propose a new methodology for
preparing a new activated carbon membrane with carbon whiskers, as shown in Figure I.
We also characterizethe membrane and the water treatments that can employ it.
For further details of this preceding, see this paper. (Bae ef ul., 2002)

2.EXPERIMENTAL

Ceramic tubing (inner diameter = 9 mm, outer diameter = 13 mm, pore size = 2.3
pm) was kindly provided by Kubota Co., Ltd., Japan, and used as the membrane support.
The polymer latex containingpoly-vinylydenchloride(PVdC) and poly-vinylchloride

121
 Carboaiaed and aggregated polymer
microspheres with micropores for
PerRntatioa adsorption
(particle siae=O. 1p mPore size=0.7nrn)
Figure 1 Structure of activated carbon membrane with carbon whiskers

(PVC) microspheres of 0.10-0.15 pm in diameter (Asahi Kasei Co. Ltd., Japan) was used
as a pre-coat reagent and precursor of the activated carbon (Sakoda et ul., 1996).
Preliminary experiments revealed that thermal decomposition occurs around 300 c, and
that a temperature above 700 C is needed for the complete carbonization of the polymer
latex employed. The ferric sulfate (Fe2(S04), nH20) was used as the catalytic
precursor in order to make the carbon whiskers (Li et ul. 2001a, b).
A ceramic tube, with one end plugged by a polymer paste, was slowly dipped into
polymer latex and taken out while being rotated at 600 rpm. By this dipping method, an
aggregate of the polymer micro spheres was formed at the surface and subsurface parts of
the ceramic tube (Sakoda et ul., 1996). The sample was dried at room temperature for
30 minutes. The dried sample was slowly dipped into 0.5 M ferric sulfate solution and
dried at room temperature for 8 hours (Li et ul., 2001 a, b).
A 15-cm-long ceramic sample was put in a quartz tube with an i.d. diameter of 25cm,
placed in an electric oven, then dried at 300 c for 10 c Smin- and kept at 300 c for
30 minutes. The temperature was then raised again to 1000-1100 C at a rate of
10 C-min-. The temperature was reach at 1000-1100 c, we started to input the
nitrogedmethane = (80/20) volume % mixture gas at 500 ml*min-,25C in flow rate, in
order to form carbon whiskers on the activated carbon layer. The hydrocarbon
deposition time and temperature were set at 20 minutes and 1050 or 1100 C,respectively.
Mem. Component of Dipping time CVD time CVD
polymer in polymer (min.) temperature
la tex(w t%1 latexkec.) (c 1
PVdC : PVC
A 45 : 55 20 20 1100
B 70 : 30 20 20 1050
C 7 0 : 30 10 20 1050
D - - 20 1100

Table 1 Preparation conditions for the MembranesA, B and C

Note: Membrane D is the carbon whiskers membrane reported in the previous work
(Li, et ul., 200 1a,b)

122
 In this study, we made three novel membranes, labeled Membrane A, B, and C, with
different formation conditions. Table 1 lists the preparation conditions of the
membranes. Membrane D is a carbon whisker reported in Li et al. (2001), and is used
for comparison.
The morphology of each membrane surface was observed using a scanning electron
microscope (SEM).
Pure water flux (PWF) measurements were carried out using a 15-cm-long
membrane in a cross-flow apparatus consisting of a feed reservoir, a membrane module, a
back washing reservoir, and a regeneration reservoir, at 25 'c . Seven gases (hydrogen,
helium, methane, nitrogen, oxygen, argon, and carbon dioxide) were used for the gas
permeability measurements. Gas was introduced to the module, and permeation flux
was measured under a 100 mmHg pressure drop by a bubble flow meter at room
temperature.
Filtration experiments were carried out using this prepared membrane in a cross-flow
apparatus, which was the same as the PWF measurement apparatus. PMMA particles
(1000 mg-l-', 0.8 pm in diameter) were used as model particle pollutants. When the
PMMA solution was introduced by the feed magnet gear pump, the flux of the same
object declined to a certain level. In this time, membranes needed regeneration, we used
the technique of back washing. To carry out the back washing, we changed the flow
direction by niddle valve and used pure water at a pressure of 2 kg*cm'*for 1 minute.
The membrane developed in this work and a commercial activated carbon F-400
(Calgon), which is widely used in various water treatments, were used for adsorption
experiments. The adsorption isotherms of phenol were measured according to the
following batch adsorption method. The sealed vial was placed in a constant-
temperature water bath kept at 25 "c. The samples were well stirred, using a magnetic
stirrer overnight (Sakoda er ul., 1991).

3. RESULTS AND DISCUSSION

Figure 2 shows the SEM pictures of the prepared Membranes A, B, and C.

Membrane D is also shown. It is obvious that Membranes A, B, and C have finer carbon
whiskers, with diameters of approximately 0.6, 0.4, and 0.2 pm, respectively, while the
whiskers of membrane D are larger, 2 pm (Li er ul., 2001). As the SEM pictures show,
after carbonization of polymer latex and CVD of methane, the novel membranes had thin
and dense carbon whiskers compared with the conventional ones.
The results of gas permeability, hydrogen, helium, methane, nitrogen, oxygen, argon,
and carbon dioxide permeation under pressure of 100 mmHg with Membranes A to C.
For all samples, the permeation rates were found to be proportional to the molecular
weight powered to -0.5. This means that for all membrane samples, the flow in the
pores become a Knudsen-type flow and each membrane sample had a linear plot,
indicating the absence of pinholes.
In the pure water flux (PWF) of Membranes A to D that these membranes have
different ranges of pore size. The PWF of Membranes, A, B, C, and D were 0.7,2.4,2.8,
and 0.03 [m day-' atm-'1, respectively. These data indicate that the PWF of the
membranes developed in this work were much better than those of carbon whisker
membranes repeated in the previous works. The resultant PWF over 0.5 [m day-'
atm-'1 at the pressure of 50 [kPa] is large enough for practical applications, since
commercial membranes have PWF typically of 2 [m day-' atm-'1.

123
 2um

Figure 2 SEM pictures of membranes prepared in this work, A, B and C

Note: Membrane D is the carbon whiskers membrane reported in the previous work (Li, e
/a/.,2001a,b)

In the virgin state, Membrane C had the highest flux of all membranes. Meanwhile,
Membrane A had a better cleaning recovery rate than did membranes B and C.
Moreover, in membrane A, the first cleaning process had a greater 99% higher cleaning
recovery rate than that of the second cleaning process. This suggests that the flux of a
membrane, as well as the diameter and density of carbon whiskers, can be controlled by
manufacturing conditions.
Thus we concluded that the carbon whisker layer with high surface density and
narrow space between the carbon
whiskers has the practical effect of 1000
preventing fouling.
Figure 3 shows the adsorption
isotherms of the activated carbon
membrane and activated carbon, F-400.
The adsorption capacity of the activated 100
carbon membrane for thus particular 8
adsorbate was similar to that of F-400,
suggesting the possibility that the o Membrane A
adsorption capacity is increased by a Membrane B
A F-400
somehow developing larger micropores 10
within the carbonized microspheres. In 1 10 100
spite of their different compounds of
C|mg/I|
polymer latex, Membranes A and B had
almost the same amounts of adsorption. Figure 3 Adsorption isotherms of phenol
to Membranes A and B, and F-400 at 251:

124
4. CONCLUSION

This study described the manufacturing of a novel composite membrane (an

activated carbon membrane with carbon whiskers) and characterized several variations of
the membranes. The resultant PWF is large enough for practical applications. The
introduction of the polymer latex in the manufacturing membrane process, not only makes
use of femc sulfate solutions preventing of the subsurface, but also acts as a precursor to
the activated carbon layer. The use of the polymer latex allowed us to improve the fine
and high density of carbon whiskers to the extent that they can prevent blocking and/or
fouling on or in the membrane, while allowing for easy cleaning with the back-washing
regeneration technique. Also, this novel membrane increased the adsorptivity of
dissolved organics through the carbonization of the polymer latex. Considering these
results, we are convinced that the novel composite membrane can be used widely in the
practical and commercial fields of water treatments.

-
Acknowledgements Ceramic tubing and polymer latexes used in this work were kindly
provided by Kubota Co. Ltd. and Asahi Kasei Co. Ltd., respectively.

REFERENCES

Bae, S.D.,Masaki, S., Sakoda, A. and Suzuki, M. (2002), Activated Carbon Membran
e with Carbon Wiskers and its Application to Water Treatments, Water Research
(submitted)
Endo M., Takeuchi K., Kobori K., Takahashi K., Kroto H.W. and Sarkar A. (1995),
Pyrolytic carbon nanotubes from vapor-grown carbon fibers, Carbon, 33,873-88 I.
Hagg May-Britt (1998), MEMBRANES IN CHEMICAL PROCESSMCj A Review of
Applications and Novel Developments, SEPARATION AND PURIFICATION
METHODS, 27(1), 51-168.
Iijima S. (1991), Helical microtubules of graphitic carbon, Nature, 354, pp 56-58.
Li, Y.Y.,Bae, S.D., Sakoda, A., Suzuki,- M. (2001a), Formation of vapor grow carbon
fibers with sulfuric catalyst precursors and nitrogen as carrier gas, Carbon, 39,9 1- 100.
Li, Y.Y.,Bae, S.D., Nomura, T., Sakoda, A. and Suzuki, M. (2001b), Preparation of
Custom-Tailored Carbon Whisker Membrane by Chemical Vapor Deposition, in: K.
Kaneko (Ed.), Fundamentals of Adsorption, 7,279-286.
Li, Y.Y., Bae, S.D, Nomura, T., Sakoda, A., Suzuki, M (2002a), Carbon Whisker
Membrane, Adsorption (Submitted) .
Li, Y.Y.,Nomura, T., Sakoda, A., Suzuki, M. (2002b), Fabrication of carbon coated
ceramic membranes by pyrolysis of methane using a modified chemical vapor
deposition apparatus; Journal of Membrane Science, 197,23-35.
Pierson Hugh O., HANDBOOK OF CHEMICAL VAPOR DEPOSITION (1992);
Principles, Technology and Application, Noyes Publications, New Jersey.
Sakoda, A., Suzuki M., Hirai R.,and Kawazoe K. (1991), Trihalomethane adsorption on
the activated carbon fibers, Water Research, 25(2), 2 19-225.
Sakoda, A., Nomura T. and Suzuki M. (1996), Application of Activated Carbon
Membrane to Water Treatments: Decolorization of Coke Furnace Wastewater,
Ahorption, 3(1), 93-98.
V. Kuberkar, P. Czekaj, R.Davis (1998), Flux Enhancement for Membrane Filtration of
Bacterial Suspensions Using High-Frequency Back washing, Biothchnology and
Bioengineering, 60,77-87.

125
 MESOPOROUS SILICA WITH LOCAL MFI STRUCTURE

SAJO P. NAIK('), ANTHONY S.T. CHIANG(')*,ROBERT W. THOMPSON'", F. C. HUANG"'

AND HSIEN-MING KAO(~)

("Department of Chemical and Material Engineering, National Central University, Chung-

Li, Taiwan ROC 320.
(2'Department of Chemical Engineering, WorcesterPolytechnic Institute,Worcester,MA
USA 01609.
'3)GraduateInstitute of Environment Engineering,National Central University,Chung-Li,
Taiwan ROC 320.
")Department of Chemistty, National Central University, Chung-Li, Taiwan ROC 320

A new type of mesoporous silica has been prepared via a dual-template, three-step hydrothermal-
flocculation-steaming (HFS) synthesis procedure. This material showed 780 m'/g of BET surface
area and 0.6 mug of primary mesopores narrowly distributed around 4.2 nm. More importantly
however, is that it showed short-range MFI zeolite crystallinity as demonstrated by FTIR and XRD
analysis, and hydrophobicity as demonstrated by water and n-hexane adsorption.

1 Introduction

When the MCM-41s type ordered mesoporous materials were first introduced,'' .*. it was
anticipated that they could function as catalysts for bulky molecules. Subsequent studies
however revealed that their mild acidity and inferior hydrothermal stability could not
filfill such promises. The stability and acidity of these materials may be improved if
zeolite-like order, such as the MFI zeolite structure of ZSMJ and silicalite, could be
introduced into the mesopore walls by using simultaneously the large surfactant as
template for mesopores and the small organic templates as that for zeolitic micropores.
This dual template strategy has been tested'. ' 4. and shown to exhibit enhanced acidity
and steam stability compared to the corresponding MCM-41. However, a composite
material of MCM/MFI instead of a homogeneous one was obtained. Other approaches

surfactants?
achieved.
-'
include the post-treat with structure directing cations under dry condition for creation of
zeolitic micro ores: *a or the hydrothermal treatment of zeolite seeds after adding
In most case, better hydrothermal stability and stronger acidity could be

However, in the post-treatment approach, the surface area decreases quickly with the
formation of zeolite phase. The mesoporous material eventually becomes an aggregate of
zeolite nanocrystals. The zeolite seed hydrothermal ap roach, on the other hand, produces
materials with only indirect hints of zeolitic nature! Neither the FTIR nor the XRD
spectrum showed strong evidence for the formation of zeolite structure.
From our earlier experiences: we have learned that the zeolite seed hydrothermal
approach is rather difficult to reproduce. Consequently, a three-step-synthesis procedure
involving the preparation of zeolite nanoprecursors (NPs) by a short hydrothermal step,
the flocculation of these NPs using a surfactant, and the steaming of the NPs/surfactant
composite to produce the final material was developed. We have recently"' demonstrated
that aggregates of less than 30 nm silicalite nanocrystals can be prepared from this
procedure. We further discovered that the nature of the as-collected NPs was very much
dependent on the stirring time of NPsKTAMeBr flocculants. Under identical steaming
condition, the 3 h-stirred NPs were converted into nanocrystals of silicalite-1, whereas,

126
 *. 11% the 36 h-stirred NPs were converted
into a mesoporous structure having
very high surface area.

2 Experimentals
* 47%
A sol with molar composition 0.25
-70 -80 -90 -190 -110 -1u) -130 -140
TPAOH/ITEOS/80H20 was prepared
as per the procedures described
-'*
prevoius~y,'~* which was then
m-1 ~ ~ ~ ~ R r p c e b o l n o l ( ~ ) - t ~ m m ) ) . R rsealed in a polypropylene (PP) bottle
BpldnL.
and heated in an air oven at 80 "C for
18 h. After cooling the sol to room
temperature, the NPs were harvested
by slowly adding an ethanolic
surfactant solution under stirring.
The white flocculants resulted were
filtered after stirred at room
.-
b temperature for 36 h, dried at room
{ temperature for 24 h, then at 70 "C
I

9 for further 3 h. The dried powder was

:.
t
- pressed into pellets and steamed at
I50 "C for 24 h in a Teflon lined 190
ml autoclave, where a ceramic
honeycomb is placed with 0.19 g of
water at the bottom. The as collected
and steamed samples were then
5 Ill 15 m 25 34 calcined in flowing air with a heating
2 eldegree rate of 2 "C/min and a holding time
Figare2XRDprttaosaf(.)uEdledcdNPh(b).Itadirrct~~.(ioa, Of at 550 OC. The samples were
(c) rtamed and calcined then examined by XRD (Shimadzu
LAB-X-700), FTIR (Jasco-4lo), Nitrogen adsorption (Micromeritics ASAP 20 10) and
29
Si MAS-NMR (Bruker DSX-300). The room temperature adsorption of water and n-
hexane were measured with a G-Cahn-200 microbalance.

3 Results and discussions

Showed in Figure 1 are the NMR spectra of the surfactant-collected precursor before and
after 150C steaming. For both samples, there was approximately an equal distribution of
Q3 and Q4 silicon environment. The steaming produced only a small increase of Q2
species, suggesting the hydration of surface silica species. The NMR spectra are very
similar to that obtained by Kremer et al.'" recently.
As showed in Figures 2, the as-collected NPs and the directly calcined one showed
only a XRD band in the 20-25" 20 range. This band is consistent with the XRD pattern of
nanometer-sized MFI crystallites recently simulated by Schlenker and Peterson,14' where
a broad band was predicted to result from the broadening of the 12 distinct peaks in this
range. The steamed sample, on the other hand, showed distinct MFI XRD peaks.

127
 For nanometer size MFI zeolite,
XRD may not be the best technique to
determine its crystallinity. It has been
reported that IR crystallinity appeared
before XRD peaks were observed"
. For example, the peak at 550-570
cm-' is known to appear due the
presence of five-member rings present
in MFI zeolitic structures."'
The as collected NPs showed a
weaker hump at 550-590 cm-' and a
stronger absorption at 620 cm" which
could come from the Q2 silica
observed in the solid state NMR The
absorption in 550-590 cm-' range
disappeared if the NPs were directly
calcined, but became stronger if the
calcination was done after steaming.
This highlights the importance of the
steaming step in stabilizing the
zeolitic structure.
The nitrogen adsorption isotherms provide M e r evidence of the type of material
obtained by the steaming of NPs. As observed in Figure 4, both the direct calcined and
the steamed and calcined samples were mesoporous materials. For comparison, the
isotherm of a typical MCM-41 sample with 3.0 nm mesopores was also given. The
steamed sample showed a rather sharp uptake at -0.4P, corresponding to the existence of
uniform mesopores about 4.2 nm in size (from KJS modified BJH calculations" ). The
total pore volume, as well as the mesopore size of this sample, is obviously larger than
that of MCM-41. This suggests that the mesopores of the steamed sample are different
fiom that of MCM-41, and most likely come from the inter-particle void between nano-
crystals. For direct calcined NPs, the amount adsorbed below 0.3P0is higher than steamed
one, but a large hysteresis due to the texture mesopores existed above 0.45P0. However,
according to the BJH calculation, the primary mesopores are about 2.5 nm in size, which
is smaller than that of MCM-41.
The micropore volume, mesopore volume and the surface area of the mesopores can
be calculated from the isotherms. The results of these calculations are also summarized in
Table 1. It is known that well prepared MCM-41 material is free of any micropore. The
mesoporous silica obtained from NPs, on the other hand, do show small but noticeable
micropore volume. The larger mesopore size and higher mesopore volume compared to
MCM-41 confirm that obtained mesoporous silica has it unique structure.
The water and n-hexane adsorption isotherms of the zeolitic mesoporous materials
obtained are compared to that of a 450 nm colloidal silicalite-1 in Figure 5. The water
adsorption isotherms are distinctively type Ill, whereas the n-hexane isotherms are type 1.
The lowest water isotherm was for the colloidal silicalite-I, where the first point
measured for the n-hexane isotherm was already at 80 mg/g. The amount of n-heme
adsorbed reached 250 mg/g at high pressure, which roughly corresponds to the filling of
silicalite-1 micropores.
The steamed NPs is more hydrophobic then the directly calcined one, but slightly
inferior to the colloidal silicalite-1. However, due to its huge mesopore volume, the
loading capacity for hexane is much higher than that of colloidal silicalite-1. It could

128
accommodate more than 350 mg/g of n-hexane at -0.3P0, which is high even compared to
activated carbon.

Table 1 Texture properties from nitrogen adsorption analysis

PR. PRO
F w 4 Adsorption isotherms for nitrgcn at 77JK on Figure 5 Room tempenbre adsorption*otbcrms of water
surf8et.ntcoouce~NPa a R a (a) direct cd&ubiw, (filkdsymbols)and beune ( o w symbols).
(b)steaming .nd calcination,(c) MCM41 mfrrmce materid.

4 Conclusions

We have successfully synthesized a zeolitic material having high surface area, a narrow
mesopore size distribution, and XRD crystallinity of the MFI structure. The size of the
mesopore was -1 nm larger than that of MCM41 materials synthesized from the same
surfactant. This excluded the suspicion of a heterogeneous MCM4l/MFI mixture. We
have further found that the stirring time after the addition of surfactant and the humidity
under which the steaming is carried out have profound effect on the final product. By
indirect proof, it was proposed that the material obtained might be similar to the
delaminated zeolite obtained in other zeolite systems. In any case, the three-step
procedure seemed to give better control over the process of forming this material.

5 References

K.D. Schmitt,
1. J.S. Beck, J.C. Vartuli, W.H. Roth, M.E. Leonowicz, C.T. Kresge,
C.T-W Chu, D.H. Olson, E.W.Sheppard, S.B. McCullen, J.B. Higgins, J.L.
Schlenker, J. Am. Chem. SOC.114 (1992) 10834.

129
 2. C. Kresge, M. Leonwicz, W. Roth,J. Vartuli, US patent 5098684 (1992)
3. L.M. Huang, W.P. Guo, P. Deng, Z.Y. Xue, Q.Z. Li, J Phys. Chem. B 104 (2000)
2817.
4. A. Karlsson, M. Stocker, R. Schmidt, Micropor. Mesopor. Muter. 27 (1 999) 18 1.
5. D.T. On,S. Kaliaguine, Angew Chem h t Ed. 40 (2001) 3248.
6. K.R. Kloetstra, H. van Bekkum, J.C. Jansen, Chem Commun, (1997) 2281.
7. S.S. Chen, Y.W. Chen, A.S.T. Chiang, Proc.of Pacific Basin Conf. On Ads. Sci. and
Tech., Ed. D.D. Do, World Scientific Inc., Australia (2000) 130.
8. Y. Liu and T. J. Pinnavaia, Chem. Muter. 14 (2002) 3 .
9. J. Liu, X. Zhang, Y. Han, F.S.Xiao, Chem. Muter. 14 (2002) 2536.
10. S. P. Naik, J. C. Chen and A. S . T. Chiang, Micropor. Mesopor. Muter. 54 (2002)
293.
1 1. S.P. Naik, A.S.T. Chiang, R.W. Thompson, F.C. Huang, submitted to Chem. Muter.
(2002).
12. S.S. Chen, A.S.T. Chiang, presented in 7* International C o d On Fundamental of
Adsorption, Nagasaki, Japan, May 20-25,2001.
13. S.P.B Kremer, C.E.A. Kischhock, M. Tialen, F. Collignon, P.J. Grobet, P.A. Jacobs,
J.A. Martens, A&. Func. Muter. 12 (2002) 286.
14. J.L Schlenker, B.K. Peterson, J. Appl. Cryst. 29 (1996) 178.
15. T. Armaroli, M. Trombetta, A. Alejandre, J. Solis and G Busca, Topics in Cutuhsis
15 (2001) 63.
16. Y.S. Lin, N. Yamamoto, Y. Choi, T. Yamaguchi, T. Okubo, S.1. Nakao, Micropor.
Mesopor. Muter. 38 (2000) 207.
17. R. Ravishankar, C.E.A. Kirschhock, B.J. Schoeman, P. Vannopen, P.J. Grobet, S.
Storck, W.F. Maier, LA. Martens, F.C.De Schryver, P.A. Jacobs, J. Phys. Chem. B
102 (1998) 2633.
18. M. Kruk, M. Jaroniec, ASayari, Lungmuir, 13 (1997) 6267.

130
 INFINITE DILUTION SELECTIVITY MEASUREMENTS BY
GAS CHROMATOGRAPHY

SASIDHAR GUMMA AND ORHAN TALU

Department of Chemical Engineering, I960 E. 24IhStreet,Stilwell Hall 455
Cleveland State Vniversiw, Cleveland,OH 441 15, USA
E-mail: [email protected]
[email protected]

Adsorption is a widely used separation process in the chemical process industry. Selectivity is a key
property that determines the ease and efficiency of any separation. Although, selectivity is a function
of both composition and pressure, most of the design calculations are carried out using a Langmuirian
approach with constant selectivity, using only pure component data The conventional techniques to
characterizethe selectivitybehavior by measuring mixture equilibrium data require a large investment
of time and elaborate experimental setup. Moreover selectivity is very sensitive to experimentalerrors.
In this article we present an experimental method to measure the selectivity in binary gas
adsorption under infinite dilution conditions over a range of pressures and thereby characterize its
behavior with respect to both pressure and composition. The method is f k t , efficient and robust. We
use the methane-ethane binary gas mixture on silicalite as a demonstration. The experimental results
obtained using this method are compared with the predictions from two models to check their validity.

1 Introduction

Adsorption equilibrium has an extra degree of freedom compared to conventional

vapor-liquid equilibrium. This extra degree of freedom increases the difficulty in
experimental measurements. It is difficult to find enough experimental data on binary
equilibria in the adsorption literature. On the other hand, pure component isotherm
measurement is so common that commercial push-button systems are available in the market
for over a decade [11.
The variation of total and partial amount of ethane adsorbed as a function of
composition at 270 kPa for a methane-ethane mixture is shown in Fig. 1. Two models,
Langmuir for mixtures using Innes and Rowley correlation [2] and IAST [3] are used to
predict the data from pure component isotherms. Both the models do reasonably well in
predicting both the partial and total amount adsorbed. The pure component methane and
ethane isotherms fix the end points of total amount adsorbed. The partial amount of ethane is
also restricted between its pure component value and zero. The two models simply predict
the curves in between the end points fixed by these thermodynamicrestrictions.
Selectivity is a key variable that affects the adsorption process and is essential for
design. The variation of selectivity of ethane with pressure and composition is shown in a
3D graph in Fig. 2. Pure component data yields only the line AB at zero pressure, which is
the ratio of Henrys constants. Using only this information it is not possible to accurately
estimate the variation in selectivity. The two models differ substantially with respect to
selectivity predictions. In the Langmuirian approach the selectivity is constant and is given
by the ratio of Henrys constants (along a horizontal plane through AB). Selectivityby IAST
approaches the same limit at zero pressure but rapidly decreases with pressure.

131
2 Conventional Measurement Techniques

There are several experimental techniques to characterize binary adsorption. The

conventional techniques like volumetric or volumetric-gravimetric are full information
techniques that yield both the total amount adsorbed and surface composition and hence the
selectivity. In these techniques the error in selectivity goes to infinity

6
0 01 01 03 04 RS 01 07 08 08 ?

Yol

Figure 1. Partial and total amounts adsorbed for CH&&, mixture at 308 K and 270 kPa

Figure 2. Selectivity of CzH6 from CH4-C2&, mixture at 305 K

as the composition approaches either of the ends. In the combined volumetric- gravimetric
technique an additional error source is introduced if the molecular weights of the gases are

132
 close to one another. Both these techniques are cumbersome, time-consuming and require
very specialized equipment and expertise.

3 Gas Chromatographic Techniques

There are several reports in the literature that measure binary adsorption equilibria using gas
chromatography [4,5,6]. In GC techniques the adsorbent is equilibrated with a continuous
flow of carrier gas (gas 1). Then a pulse of gas 2 is injected at the column inlet. A peak of the
gas 2 is eluted at the exit of the column after some time. Net retention time (or volume) is
calculated from the first moment of the peak after correcting for void volume (by measuring
the retention time of a non-adsorbing species). If the canier gas is inert (i.e. helium) the net
retention time is related to the pure component Henry's constant. Typical binary
measurements reported so far use a mixture of the two gases as carrier and introduce a small
perturbation in composition. The net retention volume is related to the thermodynamic
properties by [4]
V D dn dn

In the above equation PIand p2 are the partial pressures of the gases, nl and n2 are partial
amounts adsorbed and m is mass of the adsorbent. Thus, the experiments yield the
derivative information of the partial isotherms at various conditions. Some mathematical
technique is needed to integrate the experimentaldata and obtain the partial isotherm [5,6].

Although GC techniques are very fast compared to the conventionalmethods, any error
in the experimental data is magnified and propagated in the calculationsdue to integration.
There are also problems associatedwith polynomial fitting for the retention volume in terms
of composition in order to obtain binary isotherms [5,7].

4 Experimental

We used a typical GC setup and a mass spectroscopic detector. The only modification
involves controllingcolumn pressure between 20 - 1000 kPa. The following table lists the
main features of the experimental system used. All previous attempts to use the GC
technique for binary measurementswere conducted at constant (atmospheric)pressure. The
pure component isotherms are obtained form a conventionalvolumetric technique.
Table 1. Experimental system and conditions used
~

Column 5.35 mm ID, 65 mm long

Adsorbent Silicalite(HISIV 3000, UOP),1.6mm pellets
Adsorbent activation temperature 573 K
Experimental temperature 305 K
Exaerimental aressure 20-233kPa

133
5 Results

When the adsorbent is equilibrated with gas 1 and an infinitely small pulse of gas 2 is
injected into the carrier gas stream, Eq.( 1) can be written as [7]

The selectivity of 2 ( 2,1) at these conditions is given by Eq.(3). The quantity n l ( P } in the
above equation is the pure componentamount adsorbed for gas 1 at total column pressure P.
Experimentalmeasurementsare required for V ,(obtained fiom the infinite dilution system)
and data for pure component isotherm (obtained independently using a volumetric
technique) to calculate selectivity (LHS of Eq.3). A similar equation can be written for the
infinite dilution of gas 1.
Figure 3 shows the variation of selectivity with pressure at infinite dilution conditions
for both the gases. The broken horizontal line is the result from the Langmuir model which
is same at infinite dilution of either gas. The selectivitiespredicted by IAST are different for
the two gases as shown by the solid lines. The points are experimental data.
In contrast to the previous attempts, our GC technique does not involve polynomial fits
and integrations. It is especially developed to measure selectivity at infinite dilution of one
of the components as a h c t i o n of pressure.
This data is also shown in figure 2 as filled circles. As already stated, the solid line AB
is obtained fiom the pure Component data only. After measuring the pure component data,
instead of going through cumbersome measurements of binary selectivity at finite
composition using existing techniques, we characterized it by measuring selectivity at
infinite dilution conditions in a fast and efficient way using the proposed method. Given the
line AB and the infinite dilution curves as the limits, most models will be able to estimate the
3D mesh with reasonable accuracy.
We needed less than two days to measure the data shown. This is much shorter
compared to any conventional technique, which requires almost a day per point for finite
binary compositions. Moreover, the infinite dilution extremes can not be measured using the
conventional techniques.

6 Conclusions

We proposed a novel method which can be used to for fast measurements of infmite dilution
selectivity in binary gas adsorption. This experimental system can handle a wide range of
pressure variation. Using the data obtained one can perform a quick characterization for
selectivity. From the design point of view this method can be effectively employed to check
validity of the model being used.
The major limitation of this technique is that it may not be suitable to use it for highly
selective systems like methane-butane mixture on silicalite. For these systems the retention
times may be long and a significant dispersion in the peak might occur.

134
 15

14
0 50 I DO 150 100 I50 D

P (kPa)

Figure 3. Infinite dilution selectivity variation with pressure at 305 K

7 Acknowledgements

We thank the Cleveland State Universitys Doctoral Dissertation Research Expense Award
Program (DDREAP) for partial support of this work.

References

1. Talu,O., Needs,status,techniques and problems with binary gas adsorption

experiments, Advances in Colloid and Interface Science 76-77 (1998) pp.227-269
2. Innes, W.B. and Rowley,H.H. Journal of Physical Chemistry, 51( 1947), pp. 1 154
3. Myers,A.L. and Prausnitz,J.M., Thermodynamics of Mixed Gas Adsorption, AIChE
J , 11 (1965), pp. 121-127
4. Van Der Vlist, E. and Van Der Meijden, J., Determinationof adsorption isotherms of
the components of binary gas mixtures by gas chromatography, Journal of
Chromatography,79( 1973), pp. 1 13 -
5. Hyun,S.H. and Danner, R.P., Determination of gas adsorption equilibria by the
concentration pulse technique, AIChE Symp. Ser., 34( 1982),pp. 1861-1877
6. Tezel, F.H., Tezel, H.O. and Ruthven,D.M., Determination of Pure and Binary
Isotherms for Nitrogen and Krypton, Journal of Colloid and Interface Science, 149
(1992), pp. 197-207
7. Harlick, P.J.E. and Tezel, F.H., A Novel Solution Method for Interpreting Binary
Adsorption Isotherms from Concentration Pulse ChromatographyData, Adsorption, 6
(2000), pp.293-309
8. Kohl,S., Measurement of the effect of pressure on infinite dilution adsorbed phase
activity coe%cients by gas chromatographic techniques, Diplomarbeit, University of
Seigen (1996)

135
 ADSORPTION PROPERTIES OF COLLOID-IMPRINTEDCARBONS

MIETEK JARONIEC AND ZUOJIANG LI

Department of Chcmishy.Kent State University, Kent OH 44242, USA
E-mail: [email protected]

This work is focused on the synthesis and adsorption properties of colloid-imprinted carbons (CIC)
exhibiting bimodal pore size distribution. It is shown that the p @ e s such as the specific surface
area, pore volume. pore openings and the d e p of graphitization can be tailored by selecting the
proper size of the colloidal silica and by adjusting the synthesis conditions. In addition, their surface
properties can be adjusted by proper post-synthesis modification. Nitrogen adsorption isotherms
wen measured for a colloid-irnprintcdcarbon with bimodal distribution of spherical pores. Analysis
of adsorption data shows that the imprinting of the mesophase pitch particles with colloidal silica
affords carbons with uniform spherical mesopores and negligible amount of micropores.

1 Introduction
Since the discovery of the supramolecular self-assembly of silica and surfactant species [11,
an impressive progress has been achieved in the area of novel mesoporous materials, their
synthesis, characterization and applications [2-4]. The supramolecular self-assembly of
various inorganic and organic species into ordered mesostructures became a powerful
method for the synthesis of mesoporousmolecular sieves of tailored -work composition,
pore structure, pore size and desired surface functionality for advanced applications in such
areas as separations, adsorption, catalysis, environmental cleanup and nanotechnology.
While the self-assemblyprocess can be successfully used to obtain various organic-inorganic
nanocomposites such as ordered silica-surfactant nanostructures, it is inherently unsuitable
for the synthesis of ordered mesoporous carbons and carbon composites. Although porous
carbons have a very long history and have become adsorbents of great industrial importance,
some attempts to obtain ordered carbon mesostructures have appeared just recently [5-14).
One of such attempts explored by Ryoo and co-workers employs the ordered mesoporous
silicas (OMS),e.g., MCM-48 and SBA-15, as templates [9,11-141. The templating method
involves the filling of the ordered porous structure of a silica material with a carbon
precursor followed by carbonization and silica dissolution, and affords ordered carbons of
pore sizes below 6-7 nm. Another templating method, which is more appropriate for
preparation of macroporous carbons, employs siliceous colloidal crystals [5-81. The colloidal
crystal templating, which affords carbons with ordered macropores of size greater than 50
nm, involves the formation of colloidal crystals, infiltration of the crystal's interstitial space
with a fluid-type carbon precursor-and its solidificationfollowed by removal of the template.
Although the both templating methods are attractive and worthy of further
exploration, they do not allow for the synthesis of carbons with uniform and tailored pore
sizes in the range from 7 to 50 nm. Recently, a new approach to the synthesis of
mesoporous carbons, which employs the idea of colloidal imprinting, was proposed [ 151.
This approach seems to be very promising for the synthesis of mesoporous carbons with
uniform spherical pores, high pore volume and relatively high surface area. The current
work refers to the recently reported synthesis and adsorption properties of colloid-
imprinted carbons (CIC) [15,16] and focuses on the preparation and adsorption properties
of the CIC samples with bimodal pore size distribution. It is shown that the carbons with
spherical mesopores of two different sizes can be synthesized by using a suitable binary
mixture of colloidal silicas.

136
2 Experimental

2.1 Synthesis of colloid-imprintedcarbom

The colloids used in the synthesis of the CIC sample with bimodal pore size distribution
were Ludox AS40 from Aldrich and Bindzil30/360 from Eka Chemicals. The synthetic
mesophase pitch AR (softening point 237 "C) from Mitsubishi was used as carbon
precursor. The carbon sample with bimodal pore size distribution was prepared by
adapting the previously reported procedure (see Figure 1) [HI, which involves the
grinding of the pitch precursor, pretreating it in silicone oil for 30 min at 250 "C followed
by washing with toluene and acetone. Next, the mesophase pitch particles were dispersed
in ethanol and the resulting mixture was added gradually under vigorous stirring to a
binary mixture (wt. ratio -1:l) containing the Ludox AS40 and Bindzil30/360 colloidal
silicas. The stirring was continued for five hours to allow a complete evaporation of the
solvent. The pore imprinting procedure was canied out at 260 "C in nitrogen atmosphere
for 30 minutes, followed by heating at a rate of 2 "C/min to 600 "C and 5 "Chin to 900
"C, and an isothermal treatment at 900 "C for 120 minutes in nitrogen. The carbonized
sample was treated with 3M sodium hydroxide solution (95 "C) until silica particles were
totally dissolved.

Heating and Silica

imprinting dissolution
Mixture of silica
colloids and pitch _____,
particles Carbonization

Figure 1. Scheme illustrating the synthesis of mesopomus carbons by colloid imprinting process.

2.2 Nitrogen aalsorption measurements

Nitrogen adsorption/desorption isotherm for the CIC sample was measured by using a
volumetric gas adsorption analyzer from Micrornetrics (Norcross, GA) at -196 "C. Prior
adsorption measurements the carbon sample was degassed under vacuum at 200 "C for
two hours. The BET specific surface area was calculated using the standard Brunauer-
Emmett-Teller (BET)method in the relative pressure range of 0.04-0.15. The total pore
volume was evaluated from the amount adsorbed at a relative pressure of 0.99. The t-plot
method was used to estimate the amount of micropores. The pore size distribution (PSD)
was calculated from the adsorption branch by the Barrett-Joyner-Halenda (BJH) method,
in which we used the statistical film thickness (t-curve) evaluated from nitrogen
adsorption isotherm on the BP 280 carbon black by fitting it to the multilayer range of the
t-curve calibrated for the MCM-41 samples [ 171.

3 Results and Discussion

One of the advantages of the colloidal imprinting method is the ease of controlling the pore
size as well as pore size distributionsby choosing proper colloids [HI.
As shown in [I61 the

137
high temperature treatment of the colloid-imprinted carbons (so-called graphitization) did
not reduce the accessibilityof porous structure but even caused its improvement evidenced
by increasing the pore size openings. The current work shows that the use of a binary
mixture of colloidal silicas afford mesoporous carbons with bimodal pore size distribution.
The CIC sample synthesized by using a 1:l mixture of Ludox AS40 and Bindzil30/360
colloidal silicas had the BET specific surface area of 425 m2/gand the total pore volume of
1.08 cm3/g. The nitrogen adsorptioddesorption isotherm and the corresponding pore size
distribution for this CIC sample are shown in Figures 2 and 3, respectively.

0.0 0.2 0.4 0.6 0.8 1.0

Relative Pressure

Figure 2. Nitrogen adsorption isotherm for the CIC sample synthesized by using 1 I and 24 nm silica colloids.

As can be seen from Figure 2 the adsorption branch of this isotherm exhibits two
distinct steps that reflect the capillary condensation inside smaller or larger mesopores at
relative pressures about 0.79 and 0.9, respectively. The condensation in the relative pressure
range of 0.95-0.995 reflects condensation in secondary mesopores or small macropores,
which resulted fiom the imprintingof agglomeratesof colloidal particles. To our knowledge,
this kind of isotherm has not been reported for porous carbon materials. The pore size
distribution for this mesoporous carbon shown in Figure 3 exhibits two distinct peaks
located about 1 1 nm and 24 nm, which correspond to the particle size of Bindzil30/360 and
Ludox AS40 colloidal silicas, respectively.
It should be noted that the t-plot analysis shows a very small amount of micropores in
the sample studied, which is due to the use of the mesophase pitch as the carbon precursor.
It is known that many precursors, e.g., sucrose and polyfurfiuyl alcohol, infiltrate well
siliceous templates but after Carbonization give meso- or macroporous carbons with
complementarymicroporosity. This is not the case for the mesophase pitch, which is used to
synthesize carbon fibers that are nonporous materials.

138
 0.06

0.04

0.02

0.00
10 20 30 40 50 60
Pore Sue (nm)

Figure 3. Pore size distribution for the CIC sample synthesized by using I I and 24 nm silica colloids.

Our previous papers [15,16] and the current work show that the imprinting of
mesophase pitch particles with colloidal silica is an efficient technique to prepare
mesoporous carbons with uniform spherical pores as well as carbons with bimodal pore size
distributions. These carbons exhibit negligible amount of micropores, which can be further
eliminated during graphitization process. If micropores are need, they can be created by
controlled oxidation analogous to that used in the preparation of activated carbon fibers.
The possibility of tailoring the size of uniform spherical mesopores is of great importance
for catalysis, adsorption and other advanced applications such as the manufacture of
high-quality electrochemical double-layer capacitors, fuel cells and lithium batteries.

4 Acknowledgements

The authors gratefully acknowledge the donors of the Petroleum Research Fund
administered by the American Chemical Society for the partial support of this research.

References

1. Kresge C. T., Leonowicz M. E., Roth W.J., Vartuli J. C. and Beck J.S., Ordered
mesoporous molecular-sieves synthesized by a liquid-crystal template mechanism.
Nufure359 (1992) pp. 710-712.
2. Pang J. B., Qiu K. Y. and Wei Y ., Recent progress in research on mesoporous materials.
J. Znorg. Mufer. 17 (2002) pp. 407-414.
3. Ying J. Y., Mehnert C. P. and Wong M. S., Synthesis and applications of
supramolecular-templated mesoporous materials. Angew. Chem. Inf. Edit. 38 (1 999)
pp. 56-77.
4. Ciesla U. and Schuth F., Ordered mesoporous materials. Microporous Mesoporous
-
Mafer. 27 (1 999) pp. 13 1 149.
5. Zakhidov A. A., Baughman R.H., Iqbal Z., Cui C., Khayrullin I., Dantas, S. O., Marti J.
and Ralchenko V. G., Carbon structures with three-dimensional periodicity at optical

139
 wavelengths. Science 282 (1 998) pp. 897-90 1.
6. Yu J. S., Yoon S. B. and Chai G.S., Ordered uniform porous carbon by carbonization
of sugars. Carbon 39 (2001) pp. 1442-1446.
7. Lei Z., Bang Y., Wang H., Ke Y.,Li J., Li F. and Xing J., Fabrication of well-ordered
macroporous active carbon with a microporous framework. J. Muter. Chem. 11 (2001)
pp. 1975-1977.
8. Kang S., Yu J. S., Kruk M. and Jaroniec M., Synthesis of an ordered macroporous
carbon with 62 nm spherical pores that exhibit unique gas adsorption properties. Chem.
Commun. (2002) pp. 1670-1671.
9. Ryoo R., Joo S. H. and S. Jun, Synthesisof highly ordered carbon molecular sieves via
template-mediated structural transformation. J. Pjys. Chem. B 103 (1999) pp.
7743-7746.
10. Lee J., Yoon S., Hyeon T., Oh S. M. and Kim K.B., Synthesis of a new mesoporous
carbon and its applicationto electrochemical double-layer capacitors. Chem. Commun.
(1 999) pp. 2 177-2178.
1 1. Jun S., Joo S. H., Ryoo R., Kruk M., Jaroniec M., Liu Z., Ohsuna T. and Terasaki O.,
Synthesis of new nanoporous carbon with hexagonally ordered mesostructure. J. Am.
Chem. SOC.122 (2000) pp. 10712-10713.
12. Ryoo R., Joo S. H., Kruk M. and Jaroniec M., Ordered mesoporous carbons, A h .
Muter. 13 (200 1) pp. 677-68 1.
13. Joo S. H., Choi S. J., Oh I., Kwak J., Liu Z., Terasaki 0. and Ryoo R., Ordered
nanoporous arrays of carbon supporting high dispersions of platinum nanoparticles.
Nuture 412 (2001) pp. 169-172.
14. Lee J. S., Joo S. H. and Ryoo R., Synthesis of mesoporous silicas of controlled pore
wall thickness and their replication to ordered nanoporous carbons with various pore
diameters. J. Am. Chem. SOC.124 (2001) pp. 1 156-1157.
15. Li Z. and Jaroniec M., Colloidal imprinting: a novel approach to the synthesis of
mesoporous carbons. J. Am. Chem. SOC.123 (2001) pp. 9208-9209.
16. Li Z., Jaroniec M., Lee Y. J. and Radovic L. R., High surface area graphitized carbon
with uniform mesopores synthesised by colloidal imprinting method. Chem. Commun.
(2002) pp. 1346-1347.
17. Choma J., Jaroniec M. and Kloske M., Improved pore-size analysis of carbonaceous
adsorbents. Adsorption Sci. & Technol. 20 (2002) pp. 307-3 15.

140
 ON THE ROLE OF WATER IN THE PROCESS OF METHYL MERCAPTAN
ADSORPTION ON ACTIVATED CARBONS

S.BASHKOVA, A. BAGREEV AND T.J. BANDOSZ

Department of Chemistry City College of New York and Graduate School of CUNY
138kStreet and Convent Ave, New York IVY 10031, USA
E-mail:[email protected]

Adsorption of methyl mercaptan in moist conditions was performed on numerous samples of

activated carbons of various origins. Methyl mercaptan adsorption was tested by a dynamic method.
The amount of products of surface reaction was evaluated using thermal analysis. The results
revealed that the main product of oxidation, dimethyl disulfide, is adsorbed in pores smaller than 50
A. There is apparent competition for adsorption sites between water (moist conditions) and
dimethyl disulfide. The competition is won by the latter molecule due to its strong adsorption in the
carbon pore system. Although dimethyl disulfide has to compete with water for the adsorption sites
it can not be formed in a significant quantity without water. Water facilitates dissociation of methyl
mercaptan and thus ensures the efficient removal process.

Introduction

The amount of methyl mercaptan (MM) adsorbed (and converted to dimethyl disulfides
(DMDS)) depends on the surface pH [1,2], and the presence of various impregnants, such
as potassium iodide, potassium iodite, potassium carbonate or ammonia [3, 41. It has also
been pointed out in the literature that different functional groups on the carbon surface
or/and metal ions such as iron can catalyze oxidation of mercaptans to disulfides [3-61. As
we have found recently, there is an indication of a competition for high-energy adsorption
sites between dimethyl disulfide and water molecules when adsorption occurs in the
presence of moisture [l, 21. This happens as a result of big differences between water and
DMDS in the strength of adsorption forces and their incompatibility (DMDS has very low
solubility in water) [7].
An objective of this paper it to describe the results of our further investigation of the
competition for adsorption sites between water and dimethyl disulfide molecules during
methyl mercaptan adsorption on activated carbons. Moreover, we attempt to indicate the
apparent borderlines between the conditions of adsorption processes leading to different
adsorptiodoxidation paths. Those working conditions have a significant effect on the
feasibility of methyl mercaptan removal.

Methods

Materials: Adsorption of methyl mercaptan was performed on numerous samples of

activated carbons of various origins. Among the carbons studied were BAX-I500 (wood
based -Westvaco), S208 (coconut shell - Waterlink Bamabey and Sutcliffe), Centaur
- -
(catalytic carbon Calgon), BPL (bituminous coal Calgon), PCB (coconut shell-Calgon),
Maxsorb (mesophase pitch-Kansai), and polymeric based synthetic carbon, SCN [8]. To
broaden the spectrum of pH and surface properties, carbons were oxidized, acidified with
HCl, impregnated with NaOH, and modified with urea [9]. We do not refer to any specific
sample because our objective here is to show the general behavior of the system under

141
study.
CHSH aborption: The carbons were studied as adsorbents of methyl mercaptan in the
dynamic tests described elsewhere [l]. Briefly, 6 cm3 of adsorbent samples were ground
(1-2 mm particle size) and packed into a glass column (bed depth 80 mm, diameter 9 nun)
and prehumidified with moist air (relative humidity 80 % at 25 "C) for one hour. Dry or
moist air (relative humidity 80 % at 25 "C) containing 0.3 % (3,000 ppm) CH3SH was
then passed through the adsorbent bed at 0.5 L/min. The breakthrough of CH3SH was
monitored using a Micromax monitoring system (Lumidor) with an electrochemical sensor,
previously calibrated for MM. The test was stopped at the breakthrough concentration of
50 ppm. The adsorption capacities of each adsorbent in terms of mg of CH3SH per g of
carbon were calculated by integration of the area above the breakthrough curves, and from
the CH3SH concentration in the inlet gas, flow rate, breakthrough time, and mass of
adsorbent. For each sample the breakthrough test was repeated at least twice. The
determined capacities agreed within 4 %.
p H of carbon surface: 0.4 g of carbon powder was placed in 20 mL of water and
equilibrated during night. Then the pH of suspension was measured.
Sorption of nitrogen: Nitrogen isotherms were measured using a ASAP 2010
(Micromeritics) at -196 OC. Before the experiment the samples were heated at 120 "C
and then outgassed overnight at this temperature under a vacuum of l o 5 Torr to constant
pressure. The isotherms were used to calculate the surface area and pore (DFT [lo]) and
characteristic energy of adsorption, (Eo) (Dubinin-Radushkevichmethod [ 1 I]).
Thermal analysis:Thermai analysis was carried out using TA Instruments Thermal
Analyzer. The instrument settings were: heating rate 10 deg/min in nitrogen atmosphere
with 100 mL/min flow rate.

Discussion

0.8
E
0
B
g 0.6
.-
P
3RI 0.4
P
g 0.2
v)

0
2 4 6 8 1 0 1 2
pH of carbon surface

Figure 1. Dependence of the amount of MM adsorbed on the surface pH.

The dependence of the amount of methyl mercaptan adsorbed normalized to the pore
volume of carbons studied on the surface pH is presented in Figure 1. In the case of

142
microporous carbons the volume of micropores is used whereas for carbons with a large
contribution of mesopores the volume in pores smaller than 50 A was chosen. It was done
assuming that only those pores are active in the adsorption process. This choice was
justified knowing that: 1) at a small concentration of adsorbate in a gas mixture it is likely
that only micropores and small mesopores are active in the adsorption process [12]; 2)
molecular simulation studies of water adsorption suggested that at 80% humidity (our
conditions) 50 A and smaller pores are filled with water molecules [13]; 3) if the
normalization is done using only the volume of micropores the density of adsorbed
species is higher than liquid density of DMDS. Moreover, a certain pore volume has to
be a limiting factor for the adsorption capacity since the products of adsorptiodoxidation
should be stored there [ 1,2, 141. Although the data is scattered due to the complexity of
the system the maximum boundary of the adsorption capacity can be noticed. Moreover,
when experiments were done at dry air the capacity was almost constant. Similar
phenomenon was described for adsorption of hydrogen sulfide on activated carbons 1141.
In the case of MM adsorption on activated carbon the presence of water facilitates its
oxidation to DMDS owing to adsorption and oxidation of thiolate ions [9]. It was found
that for the process to be efficient the pH of the carbon surface should be greater than 7.5.
1.o
@T
E
-.-s 0.8

c)

F 0.6
$
n
m
.S2 0.4
3Q.
1: 0.2
0

ZI 0 . O l - L L F b Y
0.0 0.2 0.4 0.6 0.8 1.0
H20specific adsorption [g/cmT

Figure 2. Dependence of the DMDS adsorption on the amount of water adsorbed on

carbons.
As mentioned elsewhere a typical DTG plot for exhausted carbon after MM
adsorption consists of two peaks [l, 2, 91. One, low temperature, at about 80 "C,
represents desorption of water, and second, with maximum at about 200 "C,represents
desorption of dimethyl disulfide. Following the assumption that either H2O or DMDS are
adsorbed only in pores smaller than 50 A, the data was normalized based on that volume.
Figure 2 shows the relationship between the normalized amount of DMDS and water. The
correlation coefficient and slope are equal to 0.89 and - 0.99, respectively. The slope
represents the density of DMDS (1.06 g/cm3). The small discrepancy is likely related to
the fact that not all pores are filled by oxidation products owing to the existence of some
physical hindrances (blocked pore entrances). The thin line represents theoretical limit of
adsorption assuming real density of DMDS and H 2 0 . The fact that almost all points are
located below this line validates our hypothesis about the "active" pore volume. It is
important to mention here that all points represent equilibrium data. If equilibrium

143
conditions, for instance for adsorption of water, are not fblfilled the amount of DMDS is
usually small and the point "moves" from the established dependence line.
How the adsorption occurs can be evaluated based on Dubin-Radushkevich approach
[ 1 11. The amount of vapor, A, adsorbed at pressure P and temperature T can be calculated
from DR equation [ 1 I]:

where V, E, ..dL and PS are the volume of micropores, characteristic energy of

adsorption of standard gas (benzene), the affinity coefficient for vapor adsorbed, density
of liquid adsorptive and its saturation pressure at temperature T, respectively. The
required for our calculation parameters of adsorbents were determined from nitrogen
adsorption isotherms at - 193 "C [9]. The affinity coefficients were estimated from
parachor values [151. They are equal to 0.580 for MM and 1.018 for DMDS.
Interesting data for BAX and S208 is collected in Figure 3 where the predicted
amounts of MM, DMDS and water adsorbed are plotted. Prediction of water adsorption
was done following the approach of Lodewyckx and Vansant [ 161 based on application of
Dubinin-Astakhov (DA) equation using affinity coefficient equal to 0.063 for activated
carbons with low content of the oxygen surface groups. For both carbons at relative
pressure P/P,s equal to 0.8 (our experimental conditions) the amount of water, which can
be adsorbed is close to the amounts of adsorbed MM and DMDS. What is even more
interesting, the amount of DMDS adsorbed at very low relative pressure f/P,S = 0.05
corresponding to inlet concentration of MM (3000 ppm) is practically the same as the
amount of water adsorbed at P/Ps = 0.8. This can be considered as another favorable
factor to promote competition for adsorption sites, which results in replacement of water
with much stronger adsorbed DMDS molecules. Our calculations based on statistical
molecular approach showed that, however the initial heat of DMDS adsorption on
activated carbons is 3 times higher that that for water (60 kl/mol compared to 20 kl/mol),
the Henry constant is about seven orders of magnitude greater for DMDS than that for
water molecule (4.61 x 10" compared to 3.83 x I d ) [9].
I I i I I I I 1
t I
I
1 I
1 i
1

0.8

0.e
s
B 0.4
B
B
02

0
o 02 0.4 0.e 0.8 I

Figure 3. The predicted isotherms (with relative pressure units) for MM, DMS and HzO
on two carbons, BAX (A) and S208 (B).

144
 The results presented in this paper indicate the dual role of water in the process of
methyl mercaptan removal on activated carbons. It is demonstrated that DMDS,formed
during surface oxidation of MM, has to compete with water for adsorption sites. However,
the competition exists, DMDS is always the "winner" owing to its strong adsorption on
carbons. On the other hand, the formation of significant amount of DMDS would not be
possible without the presence of water in the system. Water facilitates dissociation of
methyl mercaptan Ieading to its oxidation by oxygen, mainly from air. Since dissociation
is an important step, the pH of the carbon surface has to be greater than the estimated
threshold (about 7.5).

Acknowledgements

Acknowledgement is made to the donors of the Petroleum Research Fund, administered

by the ACS, for support of this research (Grant # ACS-PRF#35449-AC5). The authors are
gratefbl to Ms. Thiri 00for her experimental help.

References

1. Bashkova, S., Bagreev, A. and Bandosz, T.J., Adsorption of methyl mercaptan on

activated carbons. Environ. Sci. Technol. 36 (2002) 2777.
2. Bashkova, S., Bagreev, A. and Bandosz, T.J.,Effect of surface characteristics on
adsorption of methyl mercaptan on activated carbons. Ind Eng. Chem. Res. 41 (2002)
4346.
3. Turk, A., Sakalis, E., Lessuck, J., Karamitsos, H. and Rago, O., Ammonia injection
enhances capacity of activated carbon for hydrogen sulfide and methyl mercaptan.
Environ. Sci. Technol. 23 (1989) 1242.
4. Shin, C.S., Kim, K.H., Yu,S.H.and Ryu, S.K.,Adsorption of methyl mercaptan and
hydrogen sulfide on impregnated activated carbon. Presented at 7~ International
Conference on Fundamentals of Adsorption, Nagasaki, Japan, May 20-25,2001.
5. Tanada, S.,Boki, K. and Matsumoto, K., Adsorption properties of methyl sulfide and
methyl disulfide on activated carbon, zeolite, and silicate, and their porous structure.
Chem. Pharm. Bull. 26 (1978) 1527.
6. Dalai, A.K., Tollefkon, E.L., Yang, A. and Sasaoka, E., Oxidation of methyl
mercaptan over an activated carbon in a futed bed reactor. Ind. Eng. Chem. Res. 36
(1 997) 4726.
7. Nuzzo, R. G., Zegarski, R. B. and Dubois, L.H., Fundamental studies of the
chemisorption of organosulfur compounds on Au (111). Implications for molecular
self-assembly on gold surfaces. J. Am. Chem. SOC.109 (1987) 733.
8. Lahaye, J., Nanse, G., Bagreev, A. and Strelko, V., Porous structure and surface
chemistry of nitrogen containing carbons from polymers. Carbon 37 (1999), 585.
9. Bagreev, A., Bashkova, S. and Bandosz. T. J., Dual role of water in the process of
methyl mercaptan adsorption on activated carbons. Langmuir, in press.
10. Lastoskie, C.M., Gubbins, K.E. and Quirke, N.J.,Pore size distribution analysis of
microporous carbons: a density functional theory approach. J Phys. Chem. 97 (1993)
4786. Olivier, J.P., Modeling physical adsorption on porous and nonporous solids
using density functional theory. J. Porous Materials 2 (I 995) 9.

145
1 1. Dubinin, M.M. and Radushkevich, L.V., Equation of the characteristic curve of
activated charcoal. Proc. Akud Sci. USSR 55 (1947) 33 1.
12. Everett, D. H.and Powl, J.C., Adsorption in slit-like and cylindrical micropores in the
Henrys law region. J. Chem. SOC.,Faradqy Trans., 1.72 (1976) 619.
13. McCallum, C.L.,Bandosz, T.J., McGrother, S.C., Muller, E.A. and Gubbins, K.E., A
molecular model for adsorption of water on activated carbon: comparison of
simulation and experiment. Lungmuir 15 (I 999) 533.
14. Adib, F., Bagreev, A. and Bandosz, T.J., Analysis of the relationship between H2S
removal capacity and surface properties of unimpregnated activated carbons.
Environ. Sci. Technol.34 (2000) 686.
15. Dubinin, M.M., Porous structure and adsorption properties of active carbons. In
Chemistry and Physics of Carbon, P. J . Walker, Ed., M. Dekker: New York, 2 (1 966)
51.
16. Lodewyckx, P. and Vansant, E.F., Water isotherms of activated carbons with small
amounts of surface oxygen. Carbon 37 (1999) 1647.

146
 STUDIES ON THE ADSORPTION PROPERTIES OF ION-EXCHANGED LOW
SILICA X ZEOLITE

H. JIANG, W. TANG. J. P.ZHANG, B. Y.ZHAO AND Y . C . XIE'

College of Chemistry and Molecular Engineering. Peking University. Beijing 100871. China
E-mail: yxie 63Dku.edu.cn

It was found that two kinds of water exerted different influence on the nitrogen adsorption capacity
of IiLSX. One was so called "residual water", which were left in zeolite after heating hydrated
LiLSX the other was so called "adsorbed watef'. which were introduced from outside into fully
dehydrated LiLSX. In the "adsorbed water" samples, the nitrogen adsorption capacity was duced
not so much as that in the "residual water" samples. This can be explained by the heterogeneity of
adsorption sites in IiLSX. The influence of Ii+and Cak exchange degree on the nitrogen adsorption
capacity of IiNaLSX and CaNaLSX were compared. After Li+ exchange degree reached a thnshold
value at about 2/3, the nitrogen adsorption capacity of LiNaLSX started to increase rapidly; while in
the case of CaNaLSX, the threshold appead at about 1M. This is due to that the active adsorption
sites in LiNaLSX are the 32 Li+ at SIII sites, which account 1/3 of the total Li+; while the active
adsorption sites in CaNaLSX are the 32 Caz+at SII sites, which account 2/3 of the total Ca".

1 Introduction

Oxygen is an important gas in industry. It has numerous application such as steel and
nonferrous metallurgy industry, chemical industry, pulp and paper industry, glass melting
furnace, welding, wastewater treatment and city garbage burn out etc.. The traditionally
method of oxygen production is by cryogenic distillation of air. This method needs low
temperature and high pressure equipments which is complicate and high cost. The
cryogenic equipment starts up and shuts down very slowly (several days). Pressure Swing
Adsorption (PSA) as an alternative source of oxygen production first became a viable
option in the early 1970's. Now PSA is more economic and convenient for small and
middle size oxygen production, especially when high purity oxygen is not required [ 11.
Synthetic zeolites 5A and faujasite X (13X
or CaX)have been typically used as adsorbents,
which adsorb NZ more strongly than 0 2 . The
extraframework cations in the zeolites play a
major role in the adsorption selectivity. Because
the quadruple moment of NZ is about 3 times
more than that of Oz,the interaction between the
extraframework cations and nitrogen is much
stronger.
In 1989, Chao [2] reported that LiLSX
(lithium ion-exchanged low silica X zeolite,
having a Si/Al ratio close to 1.0) showed an
unexpected high capacitiy and selectivity for
nitrogen over oxygen. He found a Li' exchange
threshold value in LiNaLSX at about 2/3. Below
Figure 1. Structure of faujasite zeolite and
that threshold value, the adsorption capacity of
the distribution of cation sites

To whom correspondenceshould be addressed.

147
N2 was rather small; while above that, it increased drastically.
To explain this point, the site population of Li' in LiNaLSX and LiLSX was
investigated by neutron dihction and 'Li MAS NMR [3,4,5]. LSX has faujasite
structure (shown in Fig.l), in which five kinds of cation site (SI,SI', SII, SII' and SIII)
exist. G.Engelhardt et.al. [3] reported that in fully exchanged LiLSX, SI', SII and SIII
sites were occupied each by 32 Li' per unit cell, among which only Li' at SIII sites can
adsorb N2. In the case of LiNaLSX, only SI' and SII sites are occupied by Li' up to about
213 Li' exchange. At Li' exchange between 2/3 and 1, SI' and SII sites have already each
been fully occupied by 32 Li', so that the population of SIII sites increase, resulting in the
rapid increase of N2 adsorption.
RT.Yang et.al. [6] studied the influence of residual water on the adsorption
properties of LSSX. They removed water out of hydrated LiLSX by heating at different
temperatures. Their results showed that very small amounts of water in LiLSX have a
significant effect on the adsorptive capacity. The nitrogen adsorption capacity dropped
from -17.4 N2 mo1eculedu.c. (per unit cell) for the hlly dehydrated LiLSX to <2 N2
molecules/u.c. when the sample contained about 32 water molecules/u.c.. They pointed
out that this also could be well correlated to the Li' population of SIII sites in LiLSX. Li'
at SIII sites preferred to attract H 2 0 molecules over N2 molecules, because of the much
stronger interaction between H 2 0 and Li'. When H 2 0 molecules were adsorbed on Li' at
SIII sites, they would block the adsorption of other molecules. This explains why it took
only 32 H 2 0 molecules to significantlydiminish the N2 adsorption capacity.
In PSA oxygen production, dehydrated LiLSX loaded in adsorbent beds might absorb
water h m air to be deactivated gradually. Does this kind of adsorbed water influences
the adsorption properties of LiLSX in the same way as the residual water does? Our
experimentalresults show that the two kinds of water have different effect.
We found that CaNaLSX reported by C.G. Coe et.al. [7] also had a Ca2' exchange
threshold but at a different value at about 1/3. We tentatively explain the difference of
threshold in the view of the different site population of Li' and Ca2' in LiNaLSX and
CaNaLSX.

2 Experimental

Sodium, potassium LSX (NaKLSX) was synthesized following the procedure reported by
Kuhl [8]. On the basis of Kuhl's work, the synthesis conditions such as the ageing
temperature, reactants composition etc. were optimized further so that it was easy to
repeatedly prepare NaKLSX with crystallinity near 100% in our lab [9]. The product was
characterized by XRD and elemental analysis. XRD pattern indicated that the synthesized
NaKLSX was pure faujasite-type zeolite without detectable crystalline impurities or
amorphous materials. Elemental analysis showed that the synthesized NaKLSX had SUAl
molar ratio close to 1.O.
LiLSX was prepared by five static ion exchange of the synthesized NaKLSX at 100
C with a 5.6-fold equivalent excess of 2.0 M LiCl [lo]. After ion exchange, LiLSX with
Li/Al greater than 98% was obtained.
The adsorption isotherms of N2 and O2were measured by constant volume method at
ambient temperature. The adsorption properties deduced from the isotherms further
confirmed the high crystallinityand Li' exchange degree of LiLSX prepared in our lab.

148
3 Results and Discussions

3. I Influence of water ahorption on LiLSX

Influence of residual water on the adsorption properties of LiLSX has been reported by
RT.Yang et.al. [6]. We repeated the experiment in our lab. Fig.2 (a) shows the N2
capacity versus the residual water in LiLSX. At first, as the water content goes up, the N2
capacity drops sharply from 30mYg to 3.8mVg, then it descends much more slowly. The
curve reveals a turn-point at about 32 H20per unit cell. This result is in good agreement
with that of R.T.Yang.
In the industrial application, dehydrated LiLSX is loaded in the beds of PSA
equipment and water in the air can be adsorbed gradually on LiLSX, resulting in the
decrease of its adsorption performance. The adsorbed water may exert different
influence on the adsorption properties of LiLSX fiom the residual water. So we
introduced water into fully dried LiLSX at ambient temperature step by step and measure
the change of Nz adsorption capacity. Seen fiom Fig2 (b), when water content was below
about 7 molecules per unit cell, the nitrogen capacities were almost the same for the two
kinds of samples. Yet above that water content, the nitrogen capacity was always higher
for the samples with adsorbed water. After heating the samples with adsorbed water
in sealed tube, the nitrogen capacity dropped again to almost the same amount as that of
the samples with residual water, as shown in Fig2 (c).

o P 40 60 80 im izo 140 160

y\glsrcalent(mdeduC)
Figure 2. Nitrogen adsorption capacity versus water content. (a) m samples with residual water
(b) V samples with adsMbed water (c) A samples with adsorbed water, heated in sealed tube
This can be explained by the heterogeneity of adsorption sites and the migration of
the adsorbed water in LiLSX. As mentioned above, Li+ at SIII sites are the strongest
adsorption sites in LiLSX, but there are many other site which can adsorb water also.
TPD results obtained by B. Hunger et.al. [I I] showed four desorption peaks for water
adsorbed on L U X , the desorption energy of which were 50, 58, 69, 85 KJ/mol,
respectively. The water content corresponding to the maximum desorption energy (85
KJ/mol) peak was about 29 moIecules/u.c., quite near 32 molecu1edu.c.. Though the
authors didnt make designation to this peak, according to the above results, we

149
tentatively attribute it to water adsorbed on Lit at SIII site inside the supercages. When
water molecules were introduced into a dehydrated LiLSX particle, from the view of
thermodynamics, it was most favorable for them to be adsorbed on the strongest
adsorption sites, namely Li at SIII sites, thus caused the sharp decrease of nitrogen
capacity. But from the view of kinetics, part of the water molecules might adsorb on the
weaker sites before they could diffuse into the inner part of the particle. Therefore, some
Li at SIT1 sites located in the inner part of a LiLSX paricle remained intact, leading to the
higher nitrogen capacity. Heating the samples with adsorbed water caused the
migration of water molecules from the weaker adsorption sites to Li at SIII sites, so that
the nitrogen capacity decreased again to that of the samples with residual water. The
adsorbed water introduced from outside influences nitrogen adsorption not so seriously as
the residual water does. This point provides for us a round understanding of the water
effect in the 4 PSA industrial application.

3.2 Relationship between N2 arlsorption capacity and Cd content in CaNaLSX

As mentioned above, Li content in LiNaLSX showed a threshold value at about 2/3,
above which the nitrogen adsorption capacity increased rapidly [lo], as shown in Fig.3
(a). Ca2 exchanged LSX can also be a good adsorbent for N2 [7].We noticed that N2
adsorption capacity of CaNaLSX as a function of Ca2 exchange degree gives a threshold
value at 3 1% (about 113) [7],different from that of LiNaLSX, as shown in Fig.3 (b).
1 (b) /

..
j 31
o! , , .i/ , . , , , . ,
0 P 40 eQ 80 im
Exchangeoeg- (%I
Figure 3. Comparison of threshold value between LiNaLSX and CaNaLSX.
This difference of threshold values is owing to the difference of their cation
population and distribution. K. Cheetham et.al. [I21 employed X-ray and neutron
dithction to study the structure of CaLSX. They observed that SI and SI1 sites were
respectively occupied by nearly 16 and 32 Ca2 per unit cell. Applying this result to the
N2 adsorption on CaLSX, we can draw the following conclusions. Ca2 at S1 sites which
are located in the double six-ring units, are inaccessible to N2 molecules. Ca2 at SII sites,
which are on the wall of the supercage, are located above the six-membered oxygen ring,
so that they can contaet and adsorb Nz molecules. In the case of CaNaLSX, when Ca2 are
exchanged into LSX, the first 16 Ca will occupy SI sites, because the energy of Ca at
S1 site is lower. The following exchanged 32 Ca2 are located at SII sites. Therefore,
when Ca2 content exceeds 16 Ca2 per unit cell (i.e. 113 of the total Ca in CaLSX), the
adsorption capacity of N2 increases rapidly. Thus, the site population of Ca2 is in
accordance with the threshold value.

150
4 Acknowledgement

We thank the Major State Basic Research Development Program (G2000077503) in

China for the fmancial support.

References

1. Kumar R., VSA process for oxygen production-a historical perspective. Separation
Science and Technology31 (1996) pp. 877-893.
2. Chao C. C.,Process for separating nitrogen from mixtures thereof with less polar
substances. US Pat. 48592 17 (1 989)
3. Feuerstein M., Engelhardt G., McDaniel P. L., MacDougall J. E. and Gaffney T. R.,
Solid state nuclear magnetic resonance investigation of cation siting in LiNaLSX
zeolites. Microporous and Mesoporous Materials 26 (1998) pp. 27-35
4. Plevert J., Di R e m F., Fajuia F. and Chiari G., Structure of dehydrated zeolite Li-
LSX by neutron diffiction: Evidence for a low-temperature orthorhombic faujasite.
J. Phys. Chem. B 101 (1997) pp. 10340-10346
5. Feuerstein M. and Lob0 R. F., Characterization of Li cations in zeolite LiX by solid-
state NMR spectroscopy and neutron diffraction. Chem. Muter. 10 (1 998) pp, 2197-
2204
6. Hutson N. D., Zajic S. C. and Yang R. T., Influence of residual water on the
adsorption of atmospheric gases in LiX zeolite: experiment and simulation. Ind Eng.
Chern. Res. 39 (2000) pp. 1775- 1780
7. Coe C. G., Kirner J. F., Pierantozzi R. and White T. R., Nitrogen adsorption with a
Ca andor Sr exchanged LiX zeolite. US Pat. 5 152813 (1992)
8. Kuhl G. H., Crystallizationof low silica faujasite. Zeolites 7 (1987) pp. 451-457
9. Jiang H., Tang W.,Zhao B. Y. and Xie Y. C., Studies on synthesis of Maximun
Aluminum X zeolite. ChemicalJ of Chinese Universities 23 (2002) pp. 772-776
10. Kirner J. F., Nitrogen adsorption with highly Li exchanged X zeolites with low Si/AI
ratio. U S Pat. 5268023 (1993)
11. Hunger B., Klepel O., Kirschhock C., Heuchel M., Toufar H. and Fuess H.,
Interaction of water with alkali-metal cation-exchanged X type zeolites: TPD and
XRD study. Langmur, 15 (1999) pp. 5937-5941
12. Vitale G., Bull L. M., Moms R. E. and Cheetham A. K., Combined neutron and X-
ray powder diffraction study of zeolite CaLSX and a *H NMR study of its complex
with benzene. J. Phys. Chem. 99 (1995) pp. 16087-16092

151
 CARBONIZATION OF ORGANIC WASTES USING SUPER-HEATED WATER
VAPOR AND THEIR ADSORPTION PROPERTIES

HIROYUKI YOSHIDA, NAOYA MIYAGAMI,AND MASAAKI TERASHIMA

Department of Chemical Engineering, Graduate School of Engineering, Osaka Prefecture
University, 1-1, Gakuen-cho,Saki City, JAPAN

E-mail: [email protected]

Various organic wastes, such as waste wood chip, sake lees, used tealeaves and so on, were
-
carbonized with the super-heated water vapor at 623 K. B y the 30 90 minutes treatment, the
organic wastes lost about 50 - 90 % of their original weight. The capability of gas adsorption has
been evaluated. The surface areas determined by nitrogen adsorption for the carbonized materials
were much smaller than that of the activated carbon Granular Shirasagi. The surface area
determined by carbon dioxide adsorption, on the other hand, of the carbonated materials were almost
the same order of magnitude to that of the activated carbons. These results show that the
carbonized materials have micro-pores whose diameter is less than 50 nm.
We have found that the waste wood and used tealeaves showed high adsorption capabilities for
ammonia gas in low equilibrium pressure (< 13.3 kPa). The amount of the adsorbed ammonia for
these carbonized materials were much higher that of the activated carbon Granular Shirasagi GS3 x
4/6. These results suggest that the carbonized materials from organic wastes could be utilized as
adsorbents for ammonia

Introduction

Conversion of organic wastes to usehl resources is a challenging task for the

engineers to contribute to the realization of the sustainable development in the 2 1st century.
The Japanese government, for example, plans to reduce the amount of the wastes
processed by landfill to 370,000,000 todyear in the year of 2010 [ 13. Since about 80%
of the wastes are organic wastes such as sludge, paper, wood chip, fiber, and excreta of
animals and so on, the conversion of the organic wastes to the useful resources is a key
factor to achieve the government plan. Production of methane and compost from the
waste foods and the animal excreta has been studied by the many researchers. Charcoal
production fiom the waste wood chips has also studied to utilized the organic wastes [2].
Super-heated water vapor has been widely used in many industrial processes such as
heat-exchange process and drying, and has also been used in the activation process for
activated carbon production. Recently, the super-heated water vapor has been utilized in
food industry for production of instant food and drying of vegetables and tea leafs. The
characteristics of the super-heated water vapor [3] are (1) it can heat the materials without
oxidation because it does not contain oxygen and carbon dioxide, (2) drying speed
becomes much faster than super-heated air due to heat emission of water molecules, and
(3) waste gas is easily recovered by condensing.
In this work, the various organic wastes were carbonized with the super-heated water
vapor at 623 K. The specific surface areas and adsorption characteristics of the
carbonized materials have been studied.

152
2 Methods

Preparation of carbonized materials

Various organic wastes were carbonized with super-heated water vapor using a rotary
drum super-heated water vapor generator (SJH-IOM, Johnson Boiler, Japan) and a rotary
kiln (JBT-I OM, Johnson Boiler, Japan). The organic wastes were processed at 623 K for
30 - 90 min. Process data and properties of various carbonized materials are
summarized in Table 1. An activated carbon (Granular Shirasagi GS3 x 416, Takeda
Chemical Industries Ltd., Japan) and an activated carbon prepared for alkaline gas
adsorption (GAH 4-8, Cataler Corp., Japan) were used as controls.
Measurements of specific surface area and micro-pore volume
Specific surface areas of various carbonized materials were measured by nitrogen gas
adsorption with BET methods using an automated surface area analyzer (micro-track type
4200, Nikkiso, Japan). For mesopores whose diameter were less than 50 nm, the surface
areas and pore volumes were measured by carbon dioxide adsorption. The carbon
dioxide adsorption at 298 K was measured with Bellsorp 28 (BEL Japan). The pore
volume was determined using Dubinin-Radushkevich equation [4], and the surface area
was determined by Medek's method [5].
Measurement of ammonia gas adsorption
Adsorption capacities of the carbonized materials for ammonia gas were anaIyzed by
batch-wise equilibrium experiment using a micro-balance. The experimental methods
were described in the previous paper in detail [6].
Table 1 Carbonized wastes and their process data

I Garbage 1 10.0 I 1.O I 90 J

3 Results

Surface areas determined by nitrogen gas adsorption (SNZ),and those determined by

carbon dioxide (SCo2)for the various carbonized materials are summarized in Table 2.
The carbonated materials showed much smaller SN2 than the activated carbon Granular
Shirasagi. The Sco2, on the other hand, of the carbonated materials were almost the same
order of magnitude to that of the activated carbon. These results show that the
carbonized materials have mesopores whose diameter is less than 50 nm.

153
 Table 2. Surface areas of carbonized materials measured by nitrogen, carbon dioxide,

Figure 1 shows a logarithmic plot of the amount of adsorbed ammonia against the
equilibrium pressure. The adsorption isotherms were well correlated by Freundlich
equation. The waste wood and used tealeaves showed high adsorption capabilities in
low equilibrium pressure (< 13.3 Wa). The amount of the adsorbed ammonia for these
carbonized materials were much higher than that of the activated carbon Granular
Shirasagi GS3 x 416. In the case of the activated carbon GAH 4-8, specialized for
alkaline gas adsorption, the straight line bent at the equilibrium pressure 13.3 kPa. This
result suggests that ammonia adsorbed on the mesopores of the activated carbon GAH 4-8
under low equilibrium pressure and then adsorbed on the macro-pores under high
equilibrium pressure. In the cases of the carbonized materials, ammonia gas adsorbed on
the mesopores, because the surface areas measured by nitrogen gas adsorption were very
small. in order to confirm these points, the adsorption data were analyzed by
Dubinin-Radushkevichequation [4]:
W=W&xp[-(RT/E)( In(PdP))] (1)
where W, Wo,E, PO, P, R, T are adsorption volume [cm3/g-adsorbent], saturated
adsorption volume [cm3/g-adsorbent], characteristic energy for adsorption FJ/mol],
saturated pressure Fpa], equilibrium pressure [kPa], gas constant, and temperature [K].
Figure 2 shows a plot of logW against (log(Pdp)) for the activated carbons. The
experimental data deviated from the straight line in the high-pressure range. These
results suggest that the ammonia adsorbed mainly on the macro pore in the high-pressure
range, and adsorbed on the mesopores in the low-pressure range. The similar plots for
various carbonized materials are shown in Figure 3. All the experimental data were well
correlated by straight lines, suggesting that ammonia adsorbed on the mesopores. The
micro-surface areas determined from ammonia adsorption are also listed in Table 2.
Elementary analysis would be useful to characterize unique carbons developed in this
work.

154
 Pressure [ma]
Figure 1.Ammonia adsorption on various carbonized waste materials
+
0 : Used tealeaves :
0:Scrapwood 'I:

-1
-1
3
-
M
0
-1.5 B
8
-2 -1.5

Figure 2 Logarithmic plot for

activated carbon

Figure 3 Log plot for carbonized

wastes (right)

4 Discussion

Various organic wastes could be successfully carbonized by the super-heated water vapor
-
at 623 K. As summarized in Table 1, the organic wastes lost 50 90 'YOof their original
-
weight by 30 90 min treatment. Observation by scanning electron microscope (data

155
 not shown) revealed that the macroscopic structures of the carbonized materials strongly
depend on the original structures. The measurements of specific surface area showed
that the diameters of the mesopores of the carbonized materials are smaller than nitrogen
molecule (about 50 nm), and are probably about 20-30 nm.
It should be noted that the adsorption capacities of the waste wood and the used tealeaves
for ammonium gas in the low equilibrium pressure are much higher than that of Granular
Shirasagi, and are compatible to that of GAH 8-4 which is specialized for alkaline gas
adsorption. These carbonized materials probably can be used as the substituted materials
for the activated carbon.
The carbonization of the organic wastes by the super-heated water vapor is an inexpensive
method to convert the organic wastes to useful resources. Effects of carbonization
temperature, heating rate, and other conditions on the characteristics of the carbonized
materials are extensively studying in our laboratory.

Acknowledgement
A part of this research was financially supported by 21" Century COE Program (24403,
E-1) from Ministry of Education, Culture, Sports, Science and Technology of Japan.

References

1. Notification No. 34 of the ministry of the Environment (2001)

2. Nakazaki K., Recycle of organic sludge by conversion into industrial raw materials,
In SOC.Chem. Eng. Japan (ed.), Processing of Wastes (in Japanese) (2001), pp.
65-74.
3. Iyota H., Nishimura N., and Nomura T., Reverse process of super heated steam
drying from condensation to evaporation, Nihon Kikai Gakkai Ronbunshu Part B (in
Japanese) 66 (2000) pp. 2681-2688.
4. Dubunin M. M., Adsorption in micropores, J. Colloid Interjhce Sci. 23 (1967) pp.
489-499
5. Medek J., Possibility of micropore analysis of coal and coke from the carbon dioxide
isotherm, Fuel 56 (1977) pp. 13 1-133
6. Yoshida H., Oehlenschraeger, Water vapor adsorption on basic anion exchangers, Ind.
Eng. Chem. Res., 40 (2001) pp. 4850-4856.

156
 FURTHER SUCCESSFUL APPLICATIONS OF THE NEW THEORETICAL
DESCRIPTION OF ADSORPTION/DESORPTIONKINETICS BASED ON THE
STATISTICAL RATE THEORY

WLADYSLAW RUDZINSKI
Department of Theoretical Chemistry, Faculty of Chemistry UMCSpl. Marii Curie-Sklodowskiej3,
Lublin, 20-031, POLANDph.: +48 81 5375633;fa: +48 81 5375685
e-mail: [email protected]

TOMASZ PANCZYK
Groupfor Theoretical Problems of Adsorption, Institute of Catalysis and Surjiace Chemistry,
Polish Academy of Sciences, ul. Niezapominajek8, 30-239 Krakow, POUND

It is shown how the new approach to adsorptioddesorption kinetics based on the Statistical Rate
Theory can be successfully applied to describe the kinetics of dissociative gas chemisorption on
solids.

1 Introduction

Since I9 16 when Langmuir published his fundamental paper on adsorption, the Theory of
Activated AdsorptiodDesorption Kinetics (TAAD) has, almost exclusively, been used for
the interpretation of adsorptioddesorption kinetics. However, contrary to the success of
Langmuir equation to represent the adsorption equilibria, a dramatic failure of TAAD was
observed to represent by the Langmuir kinetic equation and its further modifications, [1-
31 the monitored adsorptioddesorptionkinetics.
So, it became more and more obvious that the theoretical hndamentals of
adsorptioddesorptionkinetics must be reconsidered.
The breakthrough came at the beginning of the eighties of the 20 century with
the new theoretical approach called the Statistical Rate Theory (SRT), linking the rate of
adsorptioddesorptionkinetics to the chemical potentials of the molecules in the bulk and
the adsorbed phases. [4]
The Statistical Rate Theory (SRT) is based on considering the quantum-mechanical
transition probability in an isolated many particle system. Assuming that the transport of
molecules between the phases at the thermal equilibrium results primarily fiom single
molecular events, the expression for the rate of molecular transport between the two
phases I and 2, RI2,was developed by using the first-order perturbation analysis of
the Schrodinger equation and the Boltzmann definition of entropy.

In eq. (I), p1 and p2 are the chemical potentials of the molecules in the phases I
and 2, respectively, and R, is the exchange rate at equilibrium.
The SRT approach was applied first to describe the kinetics of adsorption in the
systems with well-defined solid surfaces. [5-91 However, in spite of the demonstrated
impressive success, that new SRT approach did not meet a general easy acceptance. This
was probably due to some inertia caused by using the classical TAAD approach during
the previous 70 years.

157
 Recently the new SRT approach has been generalized fiuther to take into
account the energetic heterogeneity of the actual adsorption systems and the possible role
of the interactions between the adsorbed molecules. 110-20) Most recently, the authors
have shown, that the SRT approach can be successfully applied to describe the multi-site-
occupancy adsorption of molecules which do not dissociate after being adsorbed. [14]
Compact simple analytical expressions were developed, and next used successfully to
correlate experimental data for adsorptioddesorption kinetics in various gadsolid systems.
The purpose of this presentation is to show, that the new SRT approach can be also
applied to represent the kinetics of dissociative gas adsorption on solids. That kinetics is
of a crucial importance in a variety of catalytic reactions occurring on solid surfaces.

2 Theory

Quantum-mechanical calculations of gas-solid interactions, and analysis of experimental

kinetic data suggest existence of a certain precursor state which is a weakly adsorbed
molecular species. In a next step the precursor species dissociates into adsorbed atoms.
Let us assume that the rate determining step of this kinetic process is the transition from
the precursor state to the adsorbed atoms.
We consider the adsorbed species passing the dissociation barrier as atoms, having
potential p*.They are in equilibrium with the bulk phase but not with the adsorbed atoms.
Then, pa # p* = % pg,where pa is the chemical potential of the adsorbed atoms, and pB is
the chemical potential of molecules in the gas phase. The SRT equation takes then the
form,

While assuming the Langmuir model for adsorption of atoms, and the ideal gas
expression for pg,from eq. (2), we obtain,
r 1

where K is a temperature dependent constant. At equilibrium, when de/dt = 0, eq. (3)

yields the following isotherm equation,
I &

- _
1+ ( ~ p ( ~ ) ) s e k T
where the superscript (e) refers to the equilibrium conditions.
The explicit form of the expression K', for the equilibrium exchange rate will
depend on the path of the dissociative adsorption in a particular adsorption system under
consideration. Namely it will depend on the character of the first step of that dissociative
adsorption. In general, that first step (molecular adsorption) may not yet be the precursor
state. By a fast surface diffusion, the adsorbed molecules may migrate to another state

158
from which they dissociate. Because K', - p"' we have to consider K', as the
following expression,
K', = K,p(e)(l-6g))" (5)
where (1 - 6:) is the fixtion of the adsorption sites available for molecular adsorption
in the' first adsorbed state, and n is the number of these adsorption sites occupied by the
adsorbing molecule consisting of s atoms. To solve the differential eq. (3), el,")has to be
related to 6. That relation will depend on the particular nature of chemisorption system
under consideration. For the purpose of illustration we will take into consideration the
dissociative adsorption of hydrogen molecules on the Fe(100) crystal face. For this
particular system, Christmann e t d . [21J reported kinetic data measured at different non-
equilibrium pressures, whereas the related quantum mechanical calculations can be found
in the recent paper by Kresse. [22] These calculations show that the most favorable
reaction path is when the center of hydrogen molecule approaching the surface is located
over the top site, and the molecule is oriented in such a way that after dissociation the
atoms go into two neighboring bridge sites. Thus 6, can be identified with the occupancy
of the top sites, because one adsorbed atom makes (eliminates) adsorption over one
neighboringtop site impossible. Thus 6, can be identified with 6 under the condition that
n=l.
So, we put n=l in eq. (5). Equation ( 5 ) can be then integrated analytically to yield,
r _
1 E_l -I -&

"I
sgn l-(Kp)2ekT (Kp)2ekT 2E
ln[[Kpeg - 1 b 2 -ZKpekT6+KpekT
- +C(6)
t(0) =

L J

When 6(t=O)=O, the integration constant C takes the form,

Using eq. (6) we have been able to fit successfully the kinetic isotherms reported by
Christmann, e t d . [21]. The results are shown in the Figure 1.
So far we have not considered yet the more complicated problem of the kinetics of
dissociative adsorption of asymmetric molecules like CO. The new level of complexity
here lies in the necessity of considering of the mutual blocking of the adsorption centers
for different atoms after dissociation of adsorbed molecule. Nevertheless we continue our
efforts along these lines, and we hope to publish some solution for that problem in our
future publications.
 - 0.4
- CD
- - 0.3 e!ao,-
- a>
- 0.2 8
- .-> Q)

- 0.1 9 CI

- h
0
0.01 0.1 1 10 100
time, min
figure 1. The bat-fit of the experimental hetic isotherms of hydrogen adsorption on Ni(100). measud by
christmann et.al. at p~7.10-'~ Ton, 2.5 .lo4 Tom and 7.5 .lo* Tom by applying the SRT eq. (6). and the
following best-fit parameters: K"exp(&") = 7.52.103Todn. &=2.58.10' Ton-'min". and N,= 16.02a.u.
where a.u. axe the arbitrary units used by christmanner.al. [21]to represent the adsorbed amount.

References

1. Rudzinski W. and Panczyk T., The langmuirian adsorption kinetics revised: A

farewell to the XX-th century theories?, Adsorption, 8 (2002) pp.23-34.
2. Rudzinski W. and Panczyk T., Phenomenological Kinetics of Real Gas-Adsorption-
Systems: Isothermal Adsorption, Journal of Non-Equilibrium Thermodynamics, 27
(2002) pp.149-204.
3. Rudzinski W. and Panczyk T.,Remarks on the current state of adsorption kinetic
theories for heterogeneous solid surfaces: A comparison of the ART and the SRT
approaches, Lungmuir, 18 (2002) pp.439-449.
4. Ward C.A. and Findlay R.D., Statistical rate theory of interfacial transport. IV.
Predicted rate of dissociative adsorption, J. Chem Phys. 76 (1982) pp.5624-5631.
5. Ward C.A. and Elmoseli M.B., Molecular adsorption at a well defined gas-solid
interphase: Statisticalrate theory approach, Sur$ Sci. 176 (1986) pp.457-475.
6. Elliott J.A.W. and Ward C.A., Chemical potential of adsorbed molecules from a
quantum statistical formulation,Lungmuir 13 (1997) pp.951-960.
7. Elliott J.A.W. and Ward C.A., Statistical Rate Theory and the Material Properties
Controlling Adsorption Kinetics. In RLJDZINSKI W., STEJXE W.A., ZGRABLICH
G, (eds.) Equilibria and Dynamics of Gas Adsorption on Heterogeneous Solid
Surfaces (Elsevier: New York, 1997), pp.285-333.
8. Elliott J.A.W. and Ward C.A., Temperature programmed desorption: A statistical
rate theory approach, J. Chem. Phys. 106 (1997) pp.5677-5684.
9. Elliott J.A.W. and Ward C.A., Statistical rate theory description of beam-dosing
adsorption kinetics, J. Chem Phys. 106 (1997) pp.5667-5676.

160
10. Rudzinski W. and Panczyk T. Surface Heterogeneity Effects on Adsorption
Equilibria and Kinetics: Rationalisation of Elovich Equation. In SCHWARZ J. and
CONTESCU C. (eds.) Su$aces of Nanoparticles and Porous Materials (Marcel
Dekker, 1999), pp.355-390.
11. Rudzinski W. and Panczyk T., The Kinetics Of Isothermal Adsorption On
Energetically Heterogeneous Solid Surfaces: A New Theoretical Description Based
On The Statistical Rate Theory Of Interfacial Transport, J. Phys. Chem. B. 104
(2000) pp.9149-9 162.
12. Rudzinski W. and Panczyk T., Kinetics of Gas Adsorption in Activated Carbons,
Studied by Applying the Statistical Rate Theory of Interfacial Transport, J. Phys.
Chem. B., 105 (2001) pp.6858-6866.
13. Rudzinski W. and Panczyk T., Remarks on the current state of adsorption kinetic
theories for heterogeneous solid surfaces: A comparison of the ART and the SRT
approaches, Langmuir, 18 (2002) pp.439-449.
14. Panczyk T. and Rudzinski W., Kinetics of Multi-Site-Occupancy Adsorption on
Heterogeneous Solid Surfaces: A Statistical Rate Theory Approach, J. Phys. Chem.,
106 (2002) pp.7846-7851.
15. Rudzinski W., Borowiecki T., Dominko A., and Panczyk T., New Method of
Estimating the Solid Surface Energetic Heterogeneity from TPD Spectra Based on
the Statistical Rate Theory of Interfacial Transport, Langmuir, 13 (1997) pp.3445-
3453.
16. Rudzinski W., Borowiecki T., Dominko A., and Panczyk T. and Gryglicki, J., On the
Quantitative Estimation of Surface Energetic Heterogeneity of Adsorbents and
Catalysts Surfaces from TPD Spectra Based on the Statistical Rate Theory of
Interfacial Transport: Variable Heating Rates as a Promisinig Way to Establish the
Values of All Unknown Parameters of Interest, Polish J. Chem., 72 (1998) pp.2103-
21 14.
17. Rudzinski W., Borowiecki T., Dominko A., and Panczyk T., A New Quantitative
Interpretation of TPD Spectra from Heterogeneous Solid Surfaces, Based on
Statistical Rate Ttheory of Interfacial Transport: the Effects of Simultaneous
Readsorption, Langmuir, 15 (1999) pp.6386-6394.
18. Rudzinski W., Borowiecki T., Panczyk T., and Dominko A., On the Applicability of
Anhenius Plot Methods to Determine Surface Energetic Heterogeneity of Adsorbents
and Catalysts Surfaces fiom Experimental TPD Spectra, A h . Coff.Znteface Sci., 84
(2000) pp. 1-26.
19. Rudzinski W., Borowiecki T., Panczyk T., and Dominko A. A quantitative approach
to calculating the energetic heterogeneity of solid surfaces fiom an analysis of TPD
peaks: Comparison of the results obtained using the absolute rate theory and the
statistical rate theory of interfacial transport, J. Phys. Chem. B, 104 (2000) pp. 1984-
1997.
20. Rudzinski W., Borowiecki T., Panczyk T., and Dominko A., Theory Of
Thermodesorption From Energetically Heterogeneous Surfaces: The Combined
Effects Of Surface Heterogeneity, Re-Adsorption, And Interactions Between The
Adsorbed Molecules, Lungmuir, 16 (2000) pp.8037-8049.
21. Christmann K., Schober O., Ertl G., and Neuman M., Adsorption of hydrogen on
nickel single crystal surfaces, J. Phys. Chem., 60 (1974) pp.4528-4540.
22. Kresse G., Dissociation and sticking of H-2 On the Ni(l1 l), (loo), and (1 10)
substrate, Phys. Rev. B, 62 (2000) pp.8295-8305.

161
 CHARACTERIZATION AND ETHYLENE ADSORPTION PROPERTIES OF
SILVER-LOADED FER ZEOLITE POTENTIALLY USED AS TRAP MATERIAL
OF COLD-START HYDROCARBON EMISSION FROM VEHICLES

Y.TERAOKA
Department of Molecular and Material Sciences, InterdisciplinaryGraduate School
of EngineeringSciences, Kyushu University,Kasuga, Fukuoka 816-8580, Japan
E-mail: [email protected]

H. ONOUE, H. FURUKAWA, 1. MORIGUCHI

Department of Applied Chemistry, Faculty of Engineering, Nagasaki Universiy,
Nagasaki 852-8521, Japan

H.OGAWA AND M. NAKANO

Nanyo Research Laboratoy, Tosoh Corporation, Shinnanyo, 746-8501. Japan

The adsorption properties of silver- and cupper-loaded zeolites for C2 and C3 hydrocarbons were
investigated to explore excellent materials for cold-start hydrocarbon trap. The adsorption property
and the stability of the adsorbents depended significantlyon the metal species and host zeolites. It
has turned out that silver-loaded femerite zeolite is the promising material with excellent olefin
selectivity, high adsorption capacity,desirable storage ability and hydrothermal stability.

1 Introduction

The air pollution by vehicle emissions especially in urban area is a serious problem to be
urgently solved, and regulation of the emission has been becoming more and more
tightened with respect particularly to NOx and non-methane hydrocarbons (NMHC). As
for the NMHC emission control, it is well known that the so-called cold-start hydrocarbon
emission, which contributes about 80% of the total emission, should be reduced. Just
after the ignition of an engine, the exhaust gas is too cold for the three-way catalyst
(TWC) to be active for removing CO, NOx and HC. The HC emission before the exhaust
gas heats up to the working temperature of the TWC is called as the cold-start HC
emission. One, and probably the most promising, countermeasure to reduce the cold-start
HC is to install the cold-stat HC trap on the upper stream to or on the same honeycomb
as TWC [I-51. Some of important requirements for trap materials (adsorbents) are as
follows:
0 They should have high adsorption capacity of HC during the cold condition.
0 They should store the adsorbed HC until the exhaust temperature becomes high
enough for TWC to work (200 C or higher).
0 They should not be deteriorated even under the hydrothermal condition in humid gas
stream at elevated temperatures like 850 C.
This paper reports adsorption property for C2-C3hydrocarbons and characterization
of Ag- and Cu-loaded zeolites to explore excellent materials for the cold-start HC trap.

162
2 Experimental
Three kinds of zeolites, which were kindly supplied by Tosoh Corporation, were used in
this study: W-ferrierite (FER) with SiOdA1203 molar ratio of 60.2, W - M F I with
SiOz/&o3=39.5 and NH4-b (BEA) with Si02/A1203=37. Ag- and Cu-loaded zeolites
were prepared by an impregnation method or an ion-exchange method using aqueous
solutions of metal nitrates, followed by the calcination in an open air at 500 "C for 1 h
(fresh sample). The hydrothermal treatment was carried out by exposing fresh samples to
moist air (10% HzO)at 850 "Cfor 5 h (aged sample). The metal loadings were expressed
on the basis of wt%; for example, 2.0wt% Ag-loaded FER(60.2) is expressed as Ag(2.0)-
FER(60.2).
The adsorption property was measured by a static method at 30 "C with a
conventional volumetric apparatus as well as by the temperature programmed desorption
(TPD) method. The details of the pretreatment and adsorption procedures were shown in
Results and Discussion section. Metal-loaded zeolite samples were characterized by
XRD, diffuse reflectance UV-Vis spectroscopy @RS) and electron spin resonance (ESR).

3 Results and Discussion

3.1 Adsorption of Cz and C, hydrocarbons on Ag-FER
Figure 1 shows adsorption isotherms of CzH.,. C2& and C& on the fresh Ag(2.0)-
FJZR(60.2). The sample was pre-evacuated at 500 "C for 0.5 h, and after cooling down to
30 "C,the first adsorption isotherm was obtained. After evacuation at the same

5 2.0
E
E
1 1.5
L
C

g 3
1.0

'$
0.5
$
2 0.0
0 0.1 0.2 0.3 0.4 0.5 0.6
Relative pressure, p/p,
-b 2.0
-
-in 2.0 -
E
. E 1.5 E 1.5
\
c
j
c)

1.0

0.5
e
8
-0
a 0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Relative pressure, p/po Relative pressure, p/po
Figure 1. Adsorption isotherms of (a) CZ& (b) Cz& and (c) C3& on fresh Ag(2.0)-FER(60.2)preevacuated
at 500 "C for 0.5 h. Open and closed circles cormpond to the Ist and 2nd adsorption measurements (see text
for the details).

163
temperature for 0.5 h, the second adsorption measurement was carried out. The adsorbed
amount in the second isotherm corresponds to the amount desorbed during the evacuation
at 30 "C or that of weakly (reversibly) adsorbed species (V,,,),and the difference between
the first and second isotherms does to the amount of strongly adsorbed species (V,) which
did not desorb by the evacuation at 30 "C; the amount in the first isotherm thus
corresponds to the total amount adsorbed (V,=V,+V,). The first and second isotherms of
C2Hs completely overlapped each other. This indicates that parafiic C2& reversibly
adsorbs at 30 "C and no strong adsorbed species is formed. For C2& and C&, on the
other hand, the discrepancy between the 1st and 2nd isotherms was observed, showing
that strongly adsorbed species is formed in addition to weakly adsorbed species. These
result clearly show that Ag(2.0)-FER(60.2) is an olefin-selective adsorbent.

0.0 0.1 0.2 0.3

Ag loading I mmol g-1
Figure 2. The amount of C2H4 adsorbed on fresh Ag-FER(60.2) (at p/p0=0.15) as a function of Ag loading.
The Ag loading is shown in unit of mmol g-'.

0 100 200 300 400 500

Temperature I "C
Figure TPD chromatograms of C2I% from Ag- and H-FER(60.2). The aged sample was subje :d to the
hydrothermal treatment by exposing to moist air (1 0%HzO)at 850 "C for 5 h.

The adsorption of C2)4 was measured on fresh Ag-FER(60.2) with various Ag

loadings and the adsorbed amounts at p/po=O. 15 were plotted as a function of the Ag
loading (Figure 2). As can be seen from Figure l(a), the 1st and 2nd isotherms are nearly
parallel above p/po=O.15, which shows that the V, is almost constant above p/po=O. 15. The
amount of the weakly adsorbed species (V,) was roughly independent on the Ag loading,
only showing a tendency to slightly decrease with an increase in the Ag loading. The
amount of strongly adsorbed species (VJ,on the other hand, increased almost linearly

I64
with increasing the Ag loading, and the values of V, were in the same order of the Ag
loading: the ratios of VJAg ranged between 0.68 and 0.86 except for the Ag-FER(60.2)
with the lowest Ag loading (VJAg d.43). It can thus be concluded that Ag loaded idon
the FER zeolite is responsible for the formation of strongly adsorbed CZ& species.
Desorption of C2& as a function of temperature was measured by means of TPD
technique (Figure 3). A sample was preevacuated at 500 "C for 0.5 h, cooled down to
room temperature in vacuo, and exposed to C2& (100 torr) for 0.5 h. After evacuation
for 0.5 h and replacing the atmosphere with a helium flow, the sample was heated at a rate
of 10 "C min-' and the desorbed CZ& was monitored by TCD. The static adsorption
experiment showed that H-FER(60.2) adsorbed a substantial amount of C2& strongly at
30 "C, but a majority of them desorbed below 250 "C as shown in Figure 3. As for fresh
Ag(2.0)-FER(60.2), on the other hand, a majority of the strongly adsorbed C2& desorbed
above 200 "C.
Another characteristic feature of Ag-FER as a C2& adsorbent is that the adsorption
property depends only slightly on the pretreatments and the existing state of Ag. The
amount (Table 1) and temperature behavior (TPD, not shown) of strongly adsorbed C2&
is roughly the same after the evacuation (fresh) and oxidation at 500 "C;combined use of
DRS and ESR indicated that Ag idon the FER zeolite was present as cluster ions (Ag,,,?
[6] and metallic silver (Ag? after the evacuation and Ag,* and Ag2O after the oxidation.
More importantly, it has tuned out that Ag-FER is tolerant against the hydrothermal
treatment at 850 "C. Although the amount of the strongly adsorbed species somewhat
decreased, the desorption behavior did not so much change as can be seen from Figure 3
and Table 1.

Table 1. Adsorbed amounts of Czfi on various metal-loaded zeolite adsorbents (30 "C, p/@.15)')
Adsorbed amount I
Sample S:)l
Pretreatment mmol g-' V,per
[metal content I mmol g-'I m2 g-'
V. V..
Fresh 0.16 0.81 0.86 252
Ag(2.0)-FER(60-2) Ox-5002) 0.15 0.79 0.81 ----
[0:185]
Aging3' 0.13 0.73 0.70 256
Cu(0.6)-FER(60.2) Fresh 0.10 0.91 1.10 ----
[0.090] Aging" 0.01 0.82 0.1 1 ----
Ag(5.4)-MFI(39.5) Fresh 0.35 1.10 0.70 347
[0.499] Aging3' 0.16 0.54 0.32 216
Ag( 1.8)-BEA(37.0) Fresh 0.15 1.21 0.9 1 487
[O. 1641 Aging3' 0.06 0.20 0.37 98
1) All the samdes were evacuated at 500 "C for 0.5 h before. the adsorption measurements.
2) Exposure to 0,(100 torr) at 500 "C for 0.5 h. 3) Hydrothermal treatment by exposing fresh samples
to moist air (10% HzO) at 850 "C for 5 h. 4) VJmetal content. 5) Specific surface area.

3.2 Effect of metal species

Adsorption properties for C2& of Ag(2.0)- and Cu(0.6)-FER(60.2) were compared. In
the fresh state, Cu-FER strongly adsorbs a substantial amount of Cz& with the VJCu ratio
close to unity and the strongly adsorbed species desorbed between 200 "C and 500 "C,
indicating that Cu in FER zeolite in a fresh state also serves as an active site for C2&
adsorption with the desirable capacity and storage ability. After the hydrothermal (aging)
treatment, however, the capability of strongly adsorbing CZ& was almost disappeared.

165
There results clearly indicate that Ag is superior to Cu as the active metal species for C2H4
adsorption with respect to the hydrothermal stability.

3.3 Eflect of host zeolites

In the fresh state, Ag(5.4>MFI(39.5), Ag( 1.8)-BEA(37.0) and Ag(2.O>FER(60.2) showed
the C2I& adsorption property characteristic to Ag species idon zeolites, namely, forming
strongly adsorbed species with the VJAg ratios between 0.7 and 0.9 and desorbing it
above 200 "C. As described above, the impact of the hydrothermal treatment on Ag(2.0)-
FER(60.2) was not so severe. For the other two samples of MFI and BEA zeolites, on the
other hand, the adsorption capacity of both strongly and weakly adsorbed species
considerably decreased. Powder X-ray diffraction measurements showed that the
hydrothermal treatment caused the partial destruction of the zeolite structure for MFI and
BEA systems, but not for the FER system. Such structural stability was also recognized
from the change of the specific surface area, S, (Table I). The S, values of fresh and aged
Ag-FER was almost the same, while those of Ag-MFI and Ag-BEA decreased
considerably after the hydrothermal treatment.

3.4 Characteristics of Ag-FER as cold-HC trap material

The results reported in this paper clearly show that Ag-FER(60.2) is the promising
material for cold-HC trap with the following characteristics.
(1) superior selectivity to olefins (C2H4, C3H6)over a parafin (C2H6)
(2) high adsorption capacity with the VJAg ratio in the range of roughly 0.7-0.9
(at p/po=O.15)
(3) desirable storage ability desorbing the strongly adsorbed species above 200 "C
(4) high hydrothermal stability
Selectivity (l), capacity (2) and storage ability (3) originate from the characteristics
of Ag itself, while the stability (4) of Ag-FER(60.2) is contributed by the structure
stability of the FER(60.2) and insensitivity of the adsorption property of Ag to its existing
states. The higher SiO2/AI2O3ratio of FER(60.2) is the most probable reason for the
zeolite-structure stability. The insensitivity of Ag is a favorable property as -a stable HC
trap material, and its origin will be elucidated by a future study.

References

1. Burk P. L., Hockmuth J. K., Anderson, D. R., Sung, S, Punke, A, Dahle, U, Tauster
S. J., et al., Stud. S u Sci.
~ Catal., 96 (1995) pp.9 13-939.
2. Nishizawa K., J. of Soc. of Automotive Engineers of Jpn. (Japanese), 50 (1996) pp.
61-65.
3. Ballinger, T. H., Manning W. A., Lafyatis, D. S., SAE Paper (1997) pp.27-31
(970741).
4. Lyfyatis, D. S., Ansell, G. P., Bennett, S. C.,Frost, J.C., Millington P. J., Rajaram, R.
R., Walker A. P., Ballinger, T.H., Appl. Catal. B, 18 (1998) pp.123-135.
5. Czaplewski, K. F., Reits T. L., Kim, Y. J., Snurr, R. Q., Microporous and
Mesoporuos Materials, 56 (2002) pp.55-64.
6. Bogdanchikova, N. E., Dulin, M. N., Toktarev, A. V., Shevnia, G. B., Kolomiichuk,
V. N., Zailovskii V. I., Petranovskii, V. P.,Stud Surf: Sci. Catal., 84 (1994) pp. 1067-
1074.

166
PRESSURE-DEPENDENT MODELS FOR ADSORPTION KINETICS ON A CMS

Youn-Sang Bae, YoungKi Ryu, and Chang-Ha Lee'

Dept. of Chem. Eng., Yonsei University, Seoul, Korea
Tel.: +82-2-2 123-2762, Fax: +82-2-3 12-6401, E-mail: leechk2vonsei.ac.h

An adsorption kinetic model was developed to evaluate the adsorption rates of five pure gases (Nz,
02, Ar, CO, and Ch) on a Takeda-3A CMS over a wide range of pressures up to 15atm. The kinetic
characteristics of adsorptionon the CMS were studied by using the adsorption equilibrium of five pure
gases measured at three different temperatures and their physical properties. Since the diffisional time
constants of all the components showed much stronger dependence of pressure than those expected by
the traditional Darken relation, a structural diffusion model was applied to predict the strong pressure
dependence. The proposed model successfully predicted the diffisional time constant up to high
pressure on the CMS.

1 Introduction

Carbon molecular sieve (CMS) is useful in air separation processes because of its
ability to selectively discriminate on the basis of molecular size and hence adsorb the
smaller oxygen molecule over nitrogen. The difference in the adsorption kinetics of
various gases aIlows the separation of gas mixtures into pure components using pressure
swing adsorption (PSA).
Generally, the kinetic rate constants on adsorbents increase with increasing surface
coverage. The reason is probably related to surface diffusion (Reid and Thomas, 1999).
Ruthven (1992) pointed out that the diffisivity of oxygen increased with adsorbate
loading on CMS more or less in terms of Darken's equation. The pressure-dependencesof
D/? are generally predicted by Darken-relation, but in some cases the pressure-
dependeces are so strong that it cannot be predicted by traditional Darken-relation. Hence,
in these cases, the model that can predict these strong pressure-dependencesis needed.
In this paper, the isotherms and diffisivities of five pure gases @I2, 02,Ar, CO, and
CH4)in CMS were studied in the range of 293-313K, 0-15atm.

2 Experimentals

The volumetric method was used to obtain the data of the adsorption equilibrium and
the adsorption kinetics. The adsorbent used in this study was Carbon Molecular Sieve
(Takeda Co.) and has an average pore size of 3A. The adsorbates were 99.99%-purity
gases. Prior to the measurements, the adsorbent was regenerated by evacuation at 423K
during 12hr. The CMS is loaded with an adsorbate in a stepwise procedure; equilibrium
and kinetic data are obtained in each step. In determining the size of each step, we
considered the pressure range in which linear isotherm can be applied.

167
3 Mathematical Models

3. I Equilibrium Models

bP
Langmuir isotherm: c, = c,
l+bP
b PI'"
L F isotherm: c, = c,
1 + b P""
bP
Toth isotherm: c, = c,
(1 + (bP)" ) I / "

D-R isotherm: P
C , = C, -exp[- a' x(In(-))2 ] (4)
P
O

3.2 Pressure-dependent Modelsfor Adsorption Kinetics

The exponential increases of the effective diffusional time constant might be explained
by the following Darken-relation and this relation was derived under the assumption that
the chemical potential gradient is the driving force of the diffusion (Ruthven, 1992).

Darken-relation :

If above Darken-relation (Eq.(5)) is combined with the Langmuir, L-F, Toth, and D-R
isotherms, the resulted models are as follows:
t

Darken-Langmuir: D, = D,, ( 1 + bP ) (6)

I

Darken-LF: D,= D,, [ 1 + bP"" ] (7)

Darken-Toth:

I
Da rken-DR. D, = Dfl
ln(4 / P)

168
4 ResultsandDiscussion

4.1 Adsorption Equilibrium

Adsorption isotherms of five gases at 293K for C M S determined from the volumetric
experiments are shown in Fig. 1. Adsorption capacities of each gas are as follows: Nz,4,
Ar < co << CH4.

- 4

8E CH, --- Langmuir

E3 .... ........ loth
01
c

0
3

5 2
D

i
$ 1

0
0 2 4 6 8 10 12 14 16 18
Pressure [atm]

Figure 1. Adsorplion isotherms of five pun, gases at 293K

and h e lib of the Langmuir, LF. and Tom lsomenns

4.2 Adsorprion Kinetics

In Figure 2, dimensionless pressure histories of NZ and 02 at the same temperature and
pressure condition were shown and they were compared with prediction results by
Brandani model (Brandani, 1998). The values of Dl? were obtained by nonlinear
regression using the least-square method.

I i I

3:L
0.2

Dodng d l P. (Simu1fi.d)

-
0.1 j
- Ad&wpIion cdl P. (Exp.)
0.0
0 200 400 em sm lorn 1200 14m 1-
[-I
Figure 2. Comparisonsof experinentai d~nensionless pressure histories and Me
predicad results by Brandani model at the case of & and 4adsorptions.

169
 In the adsorption rates of all the components on the CMS,the diffusional time constants
showed strong dependence of pressure. Reid and Thomas (1 999) suggest that the reason
for these increases in the adsorption rate with surface coverage is probably related to the
surface diffusion. These exponential increases of the effective diffusional time constant
might be explained by above Darken-based models.
As shown in Figures 3 and 4, the large deviations between experimental and predicted
results by Darken-based models (Eqs. (6)-(9)) were observed in the experimental range of
pressure. At each case, the experimental diffusional time constant showed much stronger
pressure dependence than the results predicted by the Darken-based models.

0.008 , 0.0025
Darkenlangmuir N2 Ar
c1

y
0.004 .
0.003-
Darken-LF
Darken-Toth
Darken-DR
-'2 0.0020

0.0015
Darken-Langmuir
DarkEn-LF
Darken-Toth
0
a 0.0010 0

a O.wO5 0

0
0 o o
O.oo00

1 10
1 10
p 11-
P 1aMI
Figure 3.Presswedependena,Ot D/?fW N, adwrptlons Figwe 4. Pressuredependenceof W? for Ar adswptim
and the predictions by Damen-based madsls (Eqs. (6)-(9)) and the prediions by Darkekbasedmodels (Eqs. (6)-(9))

Do (1 996) proposed a structural diffusion model, and derived the following relation for
the surface diffusivity on the assumption that the gradient of the isotherm is large in the
initial stage.
A
Do-rela tion : D, =
dC, i d P

Here, A is a constant.
Above four isotherms were applied to Do-relation (Eq. (lo)), and the following
relations were derived.

*
Do-Langmuir: D, = D,, ( 1 + bP )'

DO-LF:
* (1 + b P y
Djl = Djlo I -n
-
~

P"
Do-Toth: * I+n

D,= Dfl ( 1 + (bP)")"

Do-DR D,=Dd
* (-P).expEa'(InP)'-2dInP-lnP,]
* (1 4)
In(P/P,)

170
 In Figures 5 and 6, the prediction results by Do-based models (Eqs. (1 1)-(14)) were
shown for N2 and Ar. The similar prediction results were obtained at the other gas
adsorptions. From these results, it is concluded that Do-based models shows better
prediction results than Darken-based models.

0.006

0.0025

-
0.005 N, 303K-expt
0 Ar
-
y
L
0.004

0.003
0
- 0.0020

b 0.0015
Do-Langmuir
--- Do-LF
_ - OO-TOth
Do-DR
r 0.002 0 L 0
a % 0.0010
a O.OOO5
0.001
0 oooo
0.000
0.OoM)

I
1 10
1 10
p
P [am1
Figure 5. Pressuredependenceof Dlr' for N, adsorptions Figure6. Pressuredependenceof D/r' fw Ar adswptions
and the pmdidions by Do-basedmodels (Eqs. (11)-(14)) and the predidionsby Do-based models (Eqs. (11)-(14))

5 Conclusions

Adsorption kinetics was a hnction of atomic/molecular size, interaction (vertical or

lateral), atomic/molecular shape, polarity, pressure, and so forth. The
pressure-dependences of D/?could not be predicted by Darken-based models, but could
be well predicted by Do-based models.

6 Acknowledgements

The fmancial support of Korea Research Foundation (KRF-2001-005-E0031) is

gratefully acknowledged.

References

1. Ruthven, D.M., Diffusion of Oxygen and Nitrogen in Carbon Molecular Sieve, Chem.
Eng. Sci., 47,4305 (1992).
2. Brandani, S., Analysis of the Piezometric Method for the Study of Diffusion in
Microporous Solids: Isothermal Case, Adsorption, 4, 17 (1998).
3. Bae, Y.S.,* Kim K.I., and Lee C.H., Sorption Equilibrium and Kinetics in CMS by
Piezometric Method, Theory and Applications of Chemical Engineering, 7( l), 2422
(2001).
4. Do, D.D, A model for Surface Diffusion of Ethane and Propane in Activated Carbon,
Chem. Eng. Sci., 51,4145, (1996).
5 . Reid, C.R., and Thomas K.M., Adsorption of Gases on a Carbon Molecular Sieve
Used for Air Separation: Linear Adsorptives as Probes for Kinetic Selectivity, Langmuir,
15,3206 (1999).

171
 PREPARATIVE ENANTIOSEPARATION OF FLUOXETINE
BY SIMULATED MOVING BED
H.W.YU AND C.B. CHING
Chemical &Environmental Engineering Department, National University of Singapore,
Singapore, 119260; [email protected]

Based on the triangle method, a shortcut method was developed to establish the possible optimal
operating condition of a SMB unit because the theoretical optimal operating condition was diverted
from the actual complete separation region by too many unexpected disturbances. A new CSP was
used in the simulated moving bed for the enantioseparation of fluoxetine. Good separation results
were obtained. The effects of the difference between m2 and my on performance parameters were
studied. The results show that this method is useful for establishing the operating conditions. It can be
concluded that the new CSP is efficient for the enantioseparationof chiral drugs.

1 Introduction

When compared to the batchwise preparative chromatography, Simulated moving bed

(SMB) units exhibit a number of advantages. These advantages are primarily because of
the continuous nature of the operation and the efficient use of the stationary and mobile
phases, which allows a decrease in desorbent requirement and an improvement of the
productivity per unit time and per unit mass of stationary phase. In addition, high
performances can be achieved even at rather low values of selectivity and with a
relatively small number of theoretical plates. Due to these positive features, SMB is
particularly attractive in the case of enantiomer separations, since it is difficult to separate
enantiomers by conventional techniques. More recent applications related to chiral
technology were reported [l-31.
In this study, the enantioseparation of fluoxetine racemic mixture will be reported
using SMB technique. Fluoxetine is an antidepressant drug. (S)-fluoxetine is highly
desirable because it is a potent antidepressant and appetite suppressant and is free of
many undesirable side effects found with the racemic mixture. A high yield route to
enantiomerically pure (S)-fluoxetine, which could be used in the place of the racemic
mixture, may have commercial value.
The efficiency of chiral stationary phase (CSP) is crucial in chromatographic
technique. Recently, a new P-cyclodextrin phenyl isocyanate bonded chiral stationary
phase (CSP) was developed. This CSP is quite stable and can be used in most of HPLC
solvents. Many drug enantiomers that do not have enantioseparation effect on native f3-
cyclodextrin column in reversed phase were separated very well on this new CSP.
Design and optimization of operating conditions is very important in the operation
of SMB. Based on the equilibrium theory, triangle method was developed to design
operating conditions of SMB under non-linear condition in the past ten years [4-51. Due
to the uncertainty of the theoretical modeling parameters and the disturbance of operation,
the theoretical optimal operating condition is not robust in actual application using this
method. A shortcut method based on the triangle method is developed in this work to
establish possible robust optimal operating condition.

172
2 Experimental

2.1 Introduction of Q new CSP

This CSP was prepared with a pre-derived procedure. Perfunctionalised cyclodextrins

were first synthesized, purified and characterized and then chemically anchored on the
surface of aminized silica gel via the hydrolytically stable urethane linkage.

2.2 Chemicals
The eluent used was methanol and buffer (40:60). The concentration of triethylamine
acetate in buffer is 2% and the pH of buffer is 5.0 adjusted by addition of glacial acetic
acid. The feed was prepared by dissolving racemic fluoxetine in the mobile phase.
Fluoxetine was extracted from Prozace capsule, providbd by Lilly Company.

2.3 Instrument

The SMB system consists of 8 columns (250mmx1Omm) packed by the new CSP. The
columns are fed with either the feed or the eluent (inlet and outlet) via 5 or 8 port rotary
valves (VICI). The configuration tested is 2222 (two columns per zone. The
concentrations of the extract and the raffinate streams were analyzed for each stage using
a standard analytical chromatographic system. An analytical column (250mmx4.6mm)
packed by 5pm CSP was used to analyze the concentration of these samples. The
adsorbance wavelength was set at 225nm.

2.4 Adsorption Isotherm

In the study, Langmuir model was assumed to fit the adsorption equilibrium relationship
of fluoxetine on the column packed by the new CSP and the following approximated
extended Langmuir model was used to describe the adsorption behavior:

- 5.94cA
- 1 + 0 . 7 1 5 ~+ ~0 . 2 1 4 ~ ~
5.18~~
qB = 1 + 0 . 7 1 5 ~+~0 . 2 1 4 ~ ~

where A is (S)-fluoxetine and B is (R)-fluoxetine.

3 Results and Discussions

Now, let us design the operating conditions following the above process. Several runs
were carried out and the position of every operating point on the m2-1~13plane was shown
in Figure 1. Every sample was collected after the system was stable.
Operating point S is the theoretical optimal operating condition derived from
triangle method [5]. The purity of the extract and the rafinate at this experimental point is
83.4% and 99.8%, respectively. Experimental results show that the theoretical optimal
operation condition is not robust or distorts from actual complete separation region;
Points A and C are close to the actual complete separation region and point B is in the
actual complete separation region, When PACand PBAare extended to intersect at WO,a
triangle W ~ A PisB formed. This is assumed to be actual complete separation region. It
must be mentioned that A and C could be in the theoretical complete separation region
after considering the operational disturbance [6].

4) 41 6d 6.6 u u

mt

Figure 1. Schematic diagram on establishingoptimal robust operating condition

174
 To improve the robustness of the operation, the difference between m 3 and m2 for
point B is decreased to point H.Due to the increase in eluent flow-rate, the pressure of the
system is increased to an extent that the system cannot tolerate. Thus, the stability of the
operation is damaged, and good separation results are not achieved. This means that
although decreasing the difference between m3 and mz could improve the robustness of
the operation, the operation performance parameters will be sacrificed, sometimes even
the SMB system cannot be run smoothly. Unstable operation should be avoided in real
applications. Runs I, J and K are carried out to find the possible optimal operating
condition by increasing the difference between m3 and m2. At point K, the robustness of
the operation cannot hinder the disturbance of the operation, so the purity of the extract
and the raffinate are all decreased. Enrichment of the extract and the raffinate is increased
when the difference between m3 and m2 is increased. In other words, the solvent
consumption is decreased. Because of the decreasing of the robustness, these three
operation parameters all decreased slightly. With reference to the operation parameters of
B, I, J, K, operating point J can be considered to be the possible optimal operation point.

4 Conclusions

In this work, a new CSP was developed to separate a chiral drug using the
chromatographic technique, and a pragmatic method was developed to establish the
operating conditions of the SMB system based on the theoretical triangle method. A
triangle region determined from the experimental results was assumed to be an actual
complete separation region. In this region, possible an optimal operating condition of the
SMB was found. Good separation results can be obtained. It is shown that this method is
available to establish operating conditions and reduces the experimental efforts
effectively. It can be concluded that the new CSP is efficient for enantioseparation of
chiral drugs.

5 Acknowledgements

Authors express our appreciation to National University of Singapore in financial support

of this project.

175
References

1. Kiister E.,Gerber G. and Antia F.G., Enantioseparation of A Chiral Epoxide by

Simulated Moving Bed Chromatography Using Chiralcel-OD,
Chromatographia,40, (1995) 718, pp.387-393.
2. Cavoy E., Deltent M.-F., Lehoucq S. and Miggiano D., Laboratory-
developed Simulated Moving Bed for Chiral Drug Separationsand Design of the
System and Separation of Tramadol Enantiomers, J. Chromatogr. A 769, (1 997)
pp.49-57.
3. Strube J., Altenhoner U., Meurer M., Schmidt-Traub, H. and Schulte M.,
Dynamic Simulation of Simulated Moving-bed Chromatographic Processes for
the Optimization of Chiral Separations, J. Chromafogr.A, 769, (1997) pp.8 1-92.
4. Storti G., Mazzotti M., Morbidelli M. and Carra s., Robust Design of Binary
Countercurrent Adsorption Separation Processes, AZCHE J., 39(3), ( 1993) pp.
471-492.
5. Mazzotti M., Storti G. and Morbidelli M., Optimal Operation of Simulated
Moving Bed Units for Nonlinear Chromatographic Separations,J. Chromatogr.
A 769, (1 997) pp.3-24.
6. Yu H.W.and Ching C.B., Design and Optimization of a Simulated Moving Bed
Based on an Approximated Model Using a Novel Shortcut Method, AICHE
Journal vol. 48, (2002) No.10.2240-2246.

176
 OPTIMIZATION BASED ADAPTIVE CONTROL OF SIMULATED MOVING
BEDS

S. ABEL AND M . MAZZOTTI

ETH Zurich, Institute of Process Engineering, Sonneggstrasse 3, 8092 Zurich, Switzerland
E-mail: abel@ivuk mavt.ethz.ch; mazzotti@,ivukmavt.ethz.ch

G. ERDEM AND M. M O W
ETH Zurich, Automatic Control Laboratory, Physikstrasse 3, 8092 Zurich, Switzerland
E-mail: [email protected]; morariamt.ee.ethz.ch

M.MORBIDELLI
ETH Zurich, Laboratory of Chemical Engineering and Industrial Chemistry,8093 Zurich,
Switzerland
E-mail: [email protected]

In the recent years Simulated Moving Bed (SMB) technology has become more and more attractive for
complex separation tasks. To ensure the compliance with product specifications, a robust control is
required. In this work a new optimization based adaptive control strategy for the SMB is proposed: A
linearized reduced order model, which accounts for the periodic nature of the SMB process is used for
online optimization and control purposes. Concentration measurements at the raffinate and extract
outlets are used as the feedback information together with a periodic Kalman filter to remove model
emors and to handle disturbances. The state estimate from the periodic Kalman filter is then used for
the prediction of the outlet concentrations over a predefined time horizon. Predicted outlet
concentrations constitute the basis for the calculation of the optimal input adjustments, which
maximize the productivity and minimize the desorbent consumption subject to constraints on product
purities.

1 Introduction

SMB technology is based on the concept of a continuous chromatographic countercurrent

process. The real countercurrent is called the True Moving Bed (TMB), which is technically
not feasible without damaging the solid adsorbent particles. SMB technology overcomes
this problem by using fixed bed columns and simulating the solid flow by periodically
switchingthe inlet and outlet ports of the unit in the same direction as the fluid flow. Due to
this periodic process dynamicsthe S M B never reaches a real equilibrium, but a cyclic steady
state with a time dependent concentration profile inside the unit, which challenges the
control strategy. The goal of this work is to develop an automatic control for SMBs, which
can both find the optimal operating conditions for the plant on-line and run it at these
conditions despite possible disturbances or any kind of unexpected deviations of the
predicted behaviour. This will make it possible to exploit the full economical potential of
SMB technology.

The challenges in SMB control are not only the complex nonlinear process dynamics, but
also the long delays of the effect of disturbances. The required control strategy has to be able
to handle multivariable dynamics with time-delays and hard constraints. Model Predictive
Control (MPC) has been proven to be the most effective control strategy for this type of
problems [1,2]. Only recently, a few scientific publications have addressed the automatic

177
control of SMB by adopting different strategies [3,4]. In this work we will follow a new
model-based predictive control method that was introduced by Lee and Natarjan [5,6]. This
is called Repetitive Model Predictive Control (RMPC), and combines the concepts of both
MPC and Repetitive Control (RC).

2 Control concept

The suggested control strategy uses a simplified SMB model in order to optimize the
operating conditions of the plant on-line. The basic concept of the adopted approach is
illustrated in figure 1.
Objective FuncUon
+ Product Specifications

SMB Plant
puw,
Internal Flowrates Productivity,
_____*
Solvent
Consumption
~

I
I I
Concentration
Measurements

Figure 1. Scheme of the OptimizingControl concept.

The controller receives the on-line composition measurementof the product outlets (extract
and rafinate) as feedback data from the plant. These measurements are filtered through a
periodic Kalman filter and used together with the simplified SMB model results to estimate
the state of the system and to remove the possible model errors. The formulation of RMPC
is based on the assumption that possible errors or disturbances are likely to repeat and will
have a periodic effect on the output, which is the most likely correlation between
disturbances and output in a SMB unit. The estimated future concentration profile in the
SMB is used to optimize the future behaviour of the plant over a predefined prediction
horizon. The controller implements the calculated optimal plant input by changing the
external flow rates in order to control the internal flow rates, which are the manipulated
variables. Time lags, e.g. between online concentration measurements and optimizer or
between optimizer and SMB plant, are insignificant relative to the process dynamics and
sampling time for the planned scheme.

2. I Simplified SMB model

It is not possible to apply a detailed SMB model for on-line use together with the controller
without facing severe computational problems. Hence a simplified model is needed, which

178
is simple enough for efficient computation, but still captures the characteristic process
dynamics of an SMB. This model is then used, according to the control scheme shown
above, to estimate the internal concentration profiles and to calculate optimal operating
conditions as input for the plant. To obtain the simplified model from the equilibrium
dispersive SMB model, the model equations are linearized around a compositionprofile (in
space) calculated at cyclic steady state. Since at cyclic steady state concentration fronts
propagate along the columns and the composition profile (in space) also changes, different
composition profiles are used to linearize the model equations at different points in time
during one SMB cycle. It is worth noting here that one cycle corresponds to a number of
time periods between inlet-outlet switches equal to the number of columns in the SMB unit.
Finally the model order is reduced by using balanced model reduction as proposed by Lee
161. The obtained simplified model is then used to predict and to optimize the future
behaviour of the SMB plant.

2.2 Optimizationproblem
The idea of the control concept introduced in this work is to optimize the process over a
predefined prediction horizon, which has been chosen as 2 cycles. The control horizon in
which possible flow rate changes are calculated, is 1 cycle, though. The optimization
problem of a SMB unit with a constant switch time can be formulated as maximizing the
feed flow rate and minimizing the desorbent consumption (cost function). In practice,
economical considerations or others will guide the choice of the relative importance
attributed to these two optimization goals. Of course, this problem is subjected to certain
constraints. First of all the purity requirements in the extract and rafkinate stream has to be
fulfilled. In addition to this the system has a number of physical limitations such as
non-negative flow rates. The cost function together with the defmed constraints constitute a
Linear Program (LP) to be solved at each time step based on the new measurements
available. Calculated optimal internal flow rates, which fi~lfillthe constraints and optimize
the performance of the SMB unit, are implemented on the unit via changes of the external
flow rates.

3 Performance assessment of the controller

In order to evaluate the performance of the controller, various scenarios have been
simulated on a virtual platform. This means that instead of a real plant a SMB model based
on the equilibrium dispersive model is used [7]. In the following an example is given to
show the flexibility and performance of the controller.

3. I VirtualSMB system
An 8 column SMB with a 2-2-2-2closed-loop configurationis considered. Available for the
controller are the on-line concentration measurements at the two outlets of extract and
raffinate. The substances to be separated are A and B, where A adsorbs stronger than B.
Linear adsorptionbehaviour has been assumed (qi=Hiq). The Henry constants Hi have been
chosen as HA=4and H B = ~The . controller is able to adjust the external flow rates in order to
change the internal flow rates in the four sections of the SMB, which are the manipulated
variables. Since all 4 internal flow rates cannot be directly controlled via the external flow
rates, one internal flow rate has to be directly controlled by an additional pump, which

179
controls e.g. the recycle flow. The switch time is kept constant during the process.

3.2 Simulation result

To assess the performance of the developed controller the evolution of the purity over time,
in comparison to the uncontrolled case, is monitored. A minimum product purity of 99% is
given as product specifications. The controller is switched on after the plant has reached the
cyclic steady-state at the chosen start-up operating point, which is the operating point used
for linearization (ml=4.0, m2=2.1, m3=3.9,m4=2.1; see [7]for the definitions) and leads to
poor product purities. The chosen example shows the case of a sudden, rather large step
change in Henry constants. The controller is switchedon at cycle 20. Then it brings the plant
to fulfill specificationsand the plant is run at its optimal operatingconditions. A step change
in the characteristicadsorption parameters occurs at cycle 60, when HAand HBincrease by
10% and 15%, respectively. It can be observed, that the purity of both extract and raffmate
outlet decreases down to 85% for the uncontrolled plant. On the contrary the controller
reacts and recovers the purities within 5 to 8 cycles, then driving the plant to the new
optimum operating conditions.

0.91

0.86

--- extract
O.8ZL I
10 20 30 40 50 60 70 80 90 100 110
Cycles

Figure 2. Comparison of the SMB outlet purities in the controlled and uncontrolled case. The controller is
switched on after reaching steady state and a step disturbance in Henry's constants takes place at cycle 60.
(AHA+~OYO,A Hs=+15%).

4 Conclusion

An optimizing adaptive control strategy for the automatic control of the SMB process has

180
been proposed. It enables to operate the SMB unit at its optimal conditions. The developed
controller has been tested on a virtual platform where a linear adsorption isotherm was used.
It has been shown that the designed controller fulfills the process specificationsand at the
same time it optimizes the productivity and solvent consumption. It can adapt the operating
conditions to the changes caused by disturbances (e.g. caused by temperature changes),
operate under extreme model mismatch conditions and cope with irregularities andor aging
of the chromatographic system. One of the most important features of the controller is its
ability to find the real optimum of the process and adapt the operating conditions if
necessary.

5 Acknowledgements

The authors are gratefiil to Prof. Jay H. Lee and Prof. Hyun-Kun Rhee for helphl
discussions. The support of ETH Zurich through grant TH-23YOO-1 is gratefully
acknowledged.

References

1. Garcia C.E., Prett D.M. and Morari M., Model Predictive Contol - Theory and Practice
- A survey, Automatica 25 (1989) pp. 335-348.
2. Bemporad A. and Morari M., Model Predictive Contol: A survey, Lecture Notes in
Control and Information Sciences 245 (1999) pp. 207-226.
3. Kloppenburg E. and Gilles E.D., Automatic control of the simulated moving bed
process for C8 aromatics separation using asymptotically exact
input/output-linearization.A of Process Control 9 (2000) pp. 4 1-50.
4. Klatt K.U., Hanisch F. and Dlinnebier G., Model-based control of a simulated moving
bed chromatographicprocess for the separation of fructose and glucose. J. of Process
Control 12 (2002) pp. 203-219.
5. Natarajan S. and Lee J.H., Repetitive model predictive control applied to a simulated
moving bed chromatography system, Computers and Chem. Eng. 24 (2000) pp.
1127-1133.
6. Lee J.H., Natarajan S. and Lee K.S., A model-based predictive control approach to
repetitive control of continuous processes with periodic operations. Journal of Process
Control 11 (2001) pp. 195-207.
7. Migliorini C., Gentilini A., Mazzotti M. and Morbidelli M., Design of simulated
moving bed units under non-ideal conditions, Ind. Eng. Chem. Res. 38 (1999) pp.
2400-24 10.

181
 MONO-METHYL PARAFFIN ADSORPTIVE SEPARATION PROCESS

SANTI KULPRATHIPANJA, JAMES REKOSKE, MICHAEL GATTER, AND STEPHEN SOHN

UOP LLC, 25 E. Algonquin Rd, Des Plaines IL 6001 7, USA

A process for the recovery and purification of mono-methyl branched C~IJ- CISparaffins From
kerosene or n-paraffin depleted kerosene through an adsorptive separation has boen developed. Two
complimentary adsorptive separation-operatingmodes were demonstrated for efficient mono-methyl
paraffin production. The first method uses simulated moving-bed adsorptive separation with a
silicalite adsorbent and a mixture of CSand/or C6 n-paraffin and is0 and/or cyclo-parafin as the
desorbent. The second method, which provides enhanced mono-methyl paraffin recovery, uses the
pre-pulse technique in combination with the first operating method. Pulse tests (chromatographic
separation) with both techniques showed good performance in the laboratory. Both methods were
demonstrated successfi~llyin a continuous counter-current chromatographic separation pilot plant
using commercial n - p d i n depleted kerosene feedstock. Better than 90% normal and
mono-methyl paraffin purity with greater than 70% recovery of normal and mono-methyl paraffins
was achieved from the pilot plant. Detailed information on the process development including feed
composition, desorbent composition, temperature, diffusion rates, and model simulations will be
discussed.

1. Introduction

Surfactants are the most critical part of a detergent formulation. Surfactants with enhanced
water solubility relative to existing commercial surfactants are highly desirable for the
formulation of improved household detergents and cleaning products. Many existing
commercial detergents are composed of anionic surfactants derived from linear alkyl
benzene. It has been demonstrated that surfactants that are formulated from benzene
alkylated with mono-methyl parafin intermediates have a greater water solubility than
those formulated fiom benzene alkylated n-paraffin intermediates. To create these
enhanced water-soluble surfactants in commercially significant quantities, a raw material
source for Clo to CI3 mono-methyl parafin is required. This range of mono-methyl
paraffm is typically found in kerosene and n-paraffin depleted kerosene. N-Paraffin
depleted kerosene (Molex Raffinate) is a by-product obtained fiom the UOP Molex
process and is typically used in Jet Fuel Pool blending. Both kerosene and n-paraffin
depleted kerosene are composed of a mixture of n-paraffin, mono-methyl parafin,
di-/tri-branched paraffin, naphthenes, and aromatics. Previously, no commercial process
was available to separate mono-methyl paraffin from kerosene or n-paraffin depleted
kerosene. UOP has developed an adsorptive separation process for mono-methyl
parafin. This process utilizes the simulated moving bed system and is called MMP
SorbexTM.

2. MMP Sorbexm Development

2. I Silicalite Adsorbent
Early in the development of the process, we determined that silicalite, a MFI structure
type adsorbent, is suitable for the process. Silicalite is crystalline silica, which has a
novel topologic type of tetrahedral h e w o r k similar to aluminosilicate molecular sieves.

182
Oxygen rings containing 10-members define the straight channels along the b-axis with an
elliptical cross session of 5.7-5.8 Angstrom by 5.1-5.2Angstrom. These channels are
interconnected by zigzag channels along the a-axis defined also by 10-membered oxygen
rings with a nearly circular cross section of 5.4 Angstrom. Based on the pore openings
above, one would expect n-paraffin, mono-methyl paraffi, and low molecular weight
aromatics and naphthenes enter into the channels of silicalite.

2.2 Molecular Simulation

To confirm the silicalite pore fitting predictions above, diffusion simulations was carried
out using the Solids Diffusion module in the Accelrys Insight I1 MoIecular Modeling
package. (Accelrysis a wholly owned subsidiary of Pharmacopeia Inc., headquartered in
San Diego, CA.)

Table 1. Energetics and Predictions from Modeling Molecular Diffusion in Silicalite.

1-butyl-3-propylcyclopentane 14 -13 slow

2,ddimethyloctane 15 -17 slow
n-hexylcyclohexane 27 -15 very slow

3,3,5-trimethylheptane 45 +35 excluded

1,2,3-trimethybenzene 80 +50 excluded
2-ethyl-1,3-dimethylbenzene 80 +50 excluded
trans-l,2-di~ro~ylcvclohexane 125 +loo excluded
150 +I07 excluded I
The simulated diffusion model involves advancing an probe molecule through the
zeolites main channel and using molecular mechanics methods to calculate the energy of
the probe molecule at regular intervals as it interacts with the atoms in the sieve wall.
The energy difference between the low and high energy positions of the probe molecule in
the channel is defined as the energy barrier which can be used to predict the diffusion ease
of the probe molecule in the zeolite channel - the larger the barrier, the more restrictive
the diffusion. Another parameter that assists in predicting diffisivity is the sign and
magnitude of the maximum energy (Emax),which is a measure of how much the probe
molecule is stabilized (negative values) or destabilized (positive values) at the bottleneck
positions in the zeolite channel.
.The modeling results fiom simulating the diffision of pertinent molecules in silicalite
are summarized in Table 1. Note that the predicted diffusivity of the molecules, based
on the magnitude of the energy barrier, agrees with experimental data (see pulse test
session). The n-decane and 2-methylnonane, with energy barriers less than 10 kcal, have
good diffisivity whereas 2,ddimethyloctane and small naphthenes, with energy barriers

183
between 10 and 30 kcals, have limited difisivity in silicalite. Tri-substituted
paraffins/aromaticsand the bulkier naphthenes, which have energy barriers greater than 40
kcal and positiveLF,,values, are found to be too large to fit in silicalite. These results
suggests that, to a great extent, normal and mono-methyl paraffins will be extracted by
applying silicalite in an adsorptive separation, while naphthenes and aromatics will be
largely rejected.

2.3. Pulse Test

Pulse test is a laboratory technique used to evaluate the suitability of adsorbent and
desorbent for adsorptive separation process. This is a means to optimizing the
aforementioned factors. The pulse test procedure (1) begins with an injection of
feedstock into a desorbent stream that is flowing through a packed adsorbent column at
constant flow rate and temperature. To determine the on-stream column effluent as a
function of time or volume of desorbent passed, each feed component is monitored as it
emerges from the column using gas chromatographyor liquid chromatography.
For mono-methyl paraffin separation, two pulse test techniques, one with and one
without iso-octane pre-pulse, were developed (2,3). In each test the feed was a mixture
containing equal volumes of 3,3,5-trimethyl heptane, 2,ddimethyl octane, 2-methyl
nonane, n-decane, and 1,3,5-trimethyl benzene. The pulse test column had a volume of
70 cc and was held at a temperature of 12OOC in the experiments shown. The flow rate
through the column was 1.2 ml/min. The adsorbent was silicalite and the desorbent was
a 50/50 volume % mixture of n-hexane/cyclohexane. Test 1 was run without a pre-pulse
and test 2 was run with a pre-pulse of 40 ml of iso-octane injected into the test loop
immediately before the feed mixture was injected. Iso-octane pre-pulse diluted the
n-hexane concentration at the adsorption zone and increased the adsorbent selectivity for
mono-methyl paraffin.

A graphical representation of the results of this comparison test run is shown in Figures 1
and 2. Figure 1 shows a plot of the relative concentrations of the components versus
volume of effluent. The improved separation using the pre-pulse is also shows
graphically in Figure 2. A comparison of the two plots shows that use of a pre-pulse
resulted in a much better separation of mono-methyl paraffin (2-methyl nonane) from the
balance of feed components.

184
 30
20
10
n
"20 30 40 50- 60 70 80 90 100 110 120
Retention Volume (mi)
Figure 1. Chromatographicseparation (Pulse Test) of mono-methylparaffin using Silicalite
adsorbent, 50150 n-43 /cyclohexane desorbent

_r.
80
70 -
60- 3,3,5-TM-C,

50
- I2 ,QDM-C,

Retention Volume (ml)

Figure 2. Chromatographicseparation (Pulse Test) of mono-methylparatlin using silicalite adsorbent, 50150

n-Cb /cycIohexane desorbent and with iso-octane
pre pulse.

A third pulse test with pre-pulse of iso-octane was carried out using a feed representative
of that which would be processed in a commercial operation, n-parafin depleted kerosene
(Molex Raffinate). The feed contains mono-, di-, and tri-branched paraffins, naphthenes,
-
aromatics, and a trace (2 4 wt.%) of n-paraffins. Figure 3 shows a gas chromatogramof
the Molex Raffinate. As one can see, the feed chromatogram is rather complex and
contains at least several hundred compounds. The high number of compounds and its
complexity prevent the identification of individual compounds. The effectiveness of the
separation was therefore determined by collecting fractions of the pulse test effluent and
analyzing the normal and mono-methyl paraffin content of each fraction. Figure 4 shows
the pulse test of the relative concentration of all components present in the feed versus
effluent volume. The separation performance obtained from the Molex Rafinate feed is

185
similar the test 2 above. The highly branched paraffins, naphthenes, and aromatics were
rejected and the desired mono-methyl paraffins were extracted. The gas chromatogram -
mass spectrophotometry of the extracted mono-methyl paraffins is shown in Figure 5.
As expected from the diffusion simulations, trace quantities of naphthenes and other small
molecular compounds are present in the extracted product along with the desired product,
normal and mono-methyl paraffins.

Time I_,

Figure 3. Gas Chromatogram of N-Paraffin depleted kerosene (UOP Molex Railinate)

Retention Volume

Figure 4. Chromatographic separation of mono-methylparaffin from N-Parafin depleted Kerosene

using silicalite adsorbent, 50/50 nC6 kyclohexane desorbent, and iso-octane pre-pulse

186
 27 29 31 33 35 37 39 41

Time L

-
Figure 5. Gas Chromatogram Mass Spectrophotometryof extract product from Pulse Test

2.4. Sorbex Pilot Plant Demonstration

The process was demonstrated in a simulated continuous counter-current chromatographic
separation pilot plant. Both the primary method of operation and the pre-pulse technique
were demonstrated, with the pre-pulse technique showing improved recovery. Using
commercial n-paraffin depleted kerosene (Molex Raffinate) feedstock we routinely
demonstrated the ability to achieve better than 90% mono-methyl and normal paraffin
purity with greater than 70% recovery of mono-methyl paraffins.

3 Conclusion

Two adsorption techniques have been developed for the separation of mono-methyl
paraffins from kerosene or n-paraffin depleted kerosene using silicalite adsorbent. The
first one is using a mixture of C5and/or C6 n-paraffin and cyclo- and or iso-paraffin
desorbent. The second one is an enhanced mono-methyl parafin recovering process using
the pre-pulse technique in combination with the previous one. Pulse tests
(chromatographic separation) with both techniques showed good performance in the
laboratory. The process was demonstrated successfully in a simulated continuous
counter-current chromatographic separation pilot plant using commercial n-parafin
depieted kerosene feedstock. Better than 90% mono-methyl and normal paraffin purity
with greater than 70% recovery of mono-methyl paraffins was achieved from the pilot

187
plant. The simulated diffusion model involves advancing an organic molecule through
the zeolites main channel was investigated. Results confirmed the experimental
findings.
References:
1. R.W.Neuzil, U.S. Patent 5,382,747, 1995
2. Santi Kulprathipanja,Monomethyl paraffin adsorptive separation process, U.S.A.
Patent 6,222,088 B 1,200 1
3. Santi Kulprathipanja, Process for monomethyl acyclic hydrocarbon adsorptive
separation, U.S.A. Patent 6,252,127 B1,2001
4. Santi Kulprathipanja and James Johnson Liquid Separation,Handbook of Porous
Solids, Edited by F. Schuth, K. Sing, J. Weitkamp, To be Published by Wiley-VCH,
Germany in 2002

188
 CHROMIUM (VI) AND (111) SPECIES ADSORPTION FROM AQUEOUS
SOLUTIONS BY ACTIVATED CARBON FIBERS

0. ASTACHKINA, A. LYSSENKO, 0. MUHINA

Saint-Petersburg State University of Technology and Design,
St. B. Morskaya, 18,Saint-Petersbwg, Russia, I91186
E-mail: [email protected]

Prepared from rayon several activated carbon fiben (ACF) with different pore structure were used to remove
Cr(VI) and/or Cr(1Il) species from solutions. The adsorption experiments were carried out to determine the
influence of ACFlsolution contact time, pH, temperature, initial Cr(V1) and Cr(l1l) concentration on the
efficiency of chromium ions removal by ACF. It was found that for ACF with total pore volume more than
O.bm-/gthe porous texture has no great importance for the amount of chromium retained. For all non-oxidized
ACF the amounts of Cr species adsorbed and Cr(V1) reduced to Cr(l1l) after 48 h of ACF/solution contact are
very close. At the beginning phase of ACF/solution contact the latent period of Cr(VI) to Cr(I1l) reduction
was observed. Oxidized ACF has lower adsorption capacity to Cr(V1) species and higher to Cr(ll1) ions in
respect to non-oxidized ACF. The increase of initial Cr(V1) concentration increases the chromium species
removal but the increase of pH and temperature decreases it.
Key words - adsorption from solution, activated carbon fibers, chromium removal.

1 Introduction

Chromium compounds are widely used in many industries: metal finishing and
electroplating, leather tanning, pigments manufacturing, photography and catalysts
production [I]. The presence of chromium species in wastewater of all of these industries
is a problem because of the affect onto the human physiology. Chromium removal from
wastewaters by adsorption onto activated charcoals is an important process in the
environmental protection [2]. On the other hand the chromium species adsorption from
aqueous solution is one of the processes for chromium catalysts supported on activated
carbons production [3].
Usually the activated carbon granules are employed for chromium removal or for
catalysts preparation. However in some previous reports the perspectives and advantages
of activated carbon fibers (ACF) utilization for the same employment have been
documented [2,4]. It seems from the analysis of the articles that the use of ACF or
activated carbon cloth has a great potential.
In the present work the adsorption of Cr(V1) and Cr(II1) on ACF with different pore
structure as well as on the oxidized form of ACF was studied.

2 Materials and methods

The ACF were characterized by N2 adsorption at 77 K, methylene blue (MB), I2 and

benzene adsorption at 25 C.
The table 1 includes some preparation conditions and some properties of ACF.
Sample ACF-2 was oxidized with HN03 to obtain sample ACF-2ox. The Cr(V1) and
Cr(II1) solutions were prepared from K2Cr207and Cr(N03)3respectively. The retention of
chromium was determined by spectrophotometer method [5].

189
 For adsorption 0.1 g of ACF (previously washed up to pH 7.0 and dried) were
added to 100 ml of solution. The experimental variables were: time of contact between
ACF and solutions, initial concentration of chromium ions, pH, and temperature.
Table 1. Preparation conditions and some properties of ACF

SbN2 - specific surface area, measured by the N2 adsorption.

W,- volume adsorbed at a relative pressure - 0.95, corresponds to the (microporetmesopore)volume
Vd - microporevolume
,V - mesopore volume
M B a - Methylene Blue adsorption
12 - 12 adsorption

3 Results and discussion

3.1 Kinetics of Cr (VI) removal

To obtain the experimental data the suspensions of ACF in the solutions of chromium(V1)
with the initial concentrations 200 mg/l were shaken at 22-25C. As the reduction of
Cr(V1) to Cr(II1) was shown in some papers [6] in our investigation the variation of both
chromium ions concentration during adsorption was determined.
Table 2 presents the abatement of Cr(V1) concentration and the rate of all
chromium species adsorptions to wit - the adsorption capacity (AC) of different ACF,
when fig.l show the growth of Cr(II1) concentration. In all cases the initial pH was 4.7,
the final - 6.4. The experiment carried out showed that for ACF- 1, ACF-2, ACF-3 pore
structure and the value of surface area has no great importance.
On the other hand in the case of ACF-4 when the total pore volume is around 0.3
cm3/gand Sh,=500 - 600 m2/g during the initial phase of adsorption (time less then 2 h)
the AC decreases twice. At the same time for all non oxidized ACF the final (reached
after 48 h) amounts of chromium species adsorbed and Cr (VI) reduced to Cr(l11) are very
close. Oxidized sample ACF-~OX, possessing the similar texture as ACF-3, has notably
lower adsorption and reduction capacity in respect to non-oxidized sorbents.
It is important to mention that at the beginning phase of ACF contact with
solutions the latent period of Cr(IV) to Cr(1II) reduction can be distinguished (fig. I).
This phenomena becomes more visible in the range ACF-4>ACF-3 >ACF-2 >ACF-1 and
can be attributed to the self-catalicale reduction-oxidation reactions between carbon and
Cr(V1) species and/or oxygen-containing groups on carbon (-OH, -COH) and Cr(V1) ions.
The similar effect was described in our previous paper for the platinum and gold species
adsorption on ACF [7].
The data presented in table 3 and fig2 show the abatement of Cr(V1) concentration,
AC of ACF and the growth of Cr(II1) concentration at pH 2.5-2.9. A large increase in the
amount of Cr(V1) transformed to Cr(II1) and in the amount of chromium removed is
produced as the pH decreases. Furthermore the amount of Cr(1ll) (fig.2) rises up to

190
maximum in the initial phase of ACF/solution contact (time up to 2 h) and than falls to
the minimum (time up to 48 h). It indicates that the Cr(II1) species could be adsorbed onto
ACF when the ACF are oxidized during the adsorptiodreduction of Cr(V1) ions. The
presence of oxidized form of ACF is expected to facilitate the Cr(1II) elimination but they
are not effective for Cr(V1) removal.
The ACF-2ox/Cr(VI) solution contact confirm low capacity of oxidized fibers for
chromium (VI) ions adsorption.
Table 2. chromium (vi) species adsorption at initial pH - 4.7 as a function of time.

A. Cr(V1) concentration, mg/l

B. Total mount of chromium (VI) and (Ill) removed - AC, mg/g

r
X a

0 0,s 1 1.5 2 24 48
a h
-
Figure 1. Chromium (111) concentrationvariation as a function of time. Initial pH 4.7; temperature 20 "C.
Table 3. chromium (vi) species adsorption at initial PH - 2.5 as a function time.

A. Cr (VI).concent&on, mg/l .
B. Total mount of chromium (VI) and (Ill) removed - AC, mg/g

0
A
m
x ax
X X

0 0.5 I 1.5 2 24 48
81me. h
-
Figure 2. Chromium (III) concentrationvariation as a function of time. Initial pH 2.5; temperature 20 'C.

191
3.2 Kinetics of Cr(II4 removal
As from the previous experiments it was impossible to determine the adsorption capacity
in regard only to Cr(1II) ions theirs adsorption onto different ACF was examined from
Cr@IO3), solution with initial concentration of Cr(II1) - 100 mg/l. Fig.3 shows the
amount of Cr(II1) species removed at pH 3.2 - 3.9 temperature 20-25 "C and at different
ACF/solution contact time.

tACF-4
+ACF-3
t ACF-2
JC-ACF-1
Jlt ACF-ZOX

0 1 2 3 4 5
time, h

Figure 3. Chromium (111) removal by ACF at 20 "C as a hnction of time.

According to the experiments carried out the maximum adsorption capacity to Cr(lI1)
ions can be affirmed for oxidized sample ACF-2ox. The Cr(lI1) removal by non-oxidized
ACF is more than twice less. The adsorption equilibrium for all ACF was reached in a
period of time close to 3 h.

3.3 The influence of temperature on Cr(VI) species adsorption

Figure 4 compares the amount of total chromium removed by ACF-1 at different
temperatures for saturation time 48 h, pH=2 and initial concentration of Cr(V1) in solution
200 mg/l.

0 # ) 4 0 6 0 8 0
T
Figure 4. Chromium (VI) removal by ACF-I as a hnction of temperature (T, "C).

The inverse relationship between the amount of chromium ions removed and the
temperature was found. This fact can be explained in the consideration that the amount of
Cr(V1) reduced to Cr(II1) increases with the increase of the temperature and that the
adsorption of Cr(II1) ions on ACF-1 is relatively low. So, more the Cr(lI1) ions appears in
the solution during ACF and Cr(V1) species contact less the total amount of chromium
removed.
Additionally it can be mentioned that the three states of chromium (Cr20;2, Cr203,
C i 3 ) on surface of ACF were detected by X-ray photoelectron spectroscopy. According

192
to the data obtained the direct adsorption of Cr(V1) and Cr(Il1) species (if they are both in
the solution) as well as Cr(V1) reduction to Cr203 and to Cr (111) ion with the subsequent
adsorption can be supposed.

3.4 The influence of concentration

The relationship between the initial concentration of Cr (VI) species in the solutions and
adsorption capacity of ACF-1, ACF-2 and ACF-3 at pH 2.5, temperature 20 'C, after 48 h
is presented in figure 5.

2 0 200 400 600 800 lo00

initial concentration,mgh

Figure 5. Chromium (VI) adsorptionon ACF-las a function of initial concentration

The amount of chromium increases with the increase of initial concentration of

Cr(V1). The adsorption capacity of 800 mg/g can be reached when the initial Cr(V1)
concentration is 1000mg/l. The adsorption capacity of oxidized sample ACF-2ox at the
same conditions was not more than 310 mg/g. The amount of Cr(II1) removed by ACF-
20x from the solution with Cr(II1) initial concentration 1000 mg/l was only 140 mg/g.
Thus, the factors which affect to Cr(V1) and Cr(II1) removal by ACF are not only
temperature and initial concentration but also the oxidationheduction ability of fibers and
theirs oxidation state before and during adsorption. The ACF used in the study can be
useful to remove Cr(V1) and Cr(II1) from aqueous solution. To ameliorate the rate of
chromium removal it seems better to use at the same time the oxidized and non-oxidized
ACF. The most usefid porous texture of ACF is when the total pore volume is more than
0.4 and less 0.5 cm3/gand contains 70-80 % of micropores.

References
1 . Lalvani S.B., Wiltowsk T., Hubner A., Weston A., Mandich N., Carbon 36 (1998) p.
1219.
2. Morozova A. A., Zh. Prikl. Khimii 68 (1995) p. 770.
3. Brown P.N., Jayson G.G., Thomson G., Wilkinson M.C., Carbon 27 (1989) p. 821.
4. Lyssenko A., Carbon-based media for water purification, The international magazine
for technical textile users 38 N24 (2000) p. 33.
5. Chand S., Agarwal V.K., Kumar P., Indian J., Enviror HLTH 36 Ne3 (1994) p. 15 1.
6. Aggarwal D., Goyal M.,Bansal R.C., Carbon 37 (1 999) p. 1989.
7. Lyssenko A., Simanova S., Abstracts, At 7th International conference of
Fundamentals of Adsorption, Nagasaki, Japan, May (200 1) p. 38.

193
 TREATMENT OF COMPLEX WASTEWATERS BY BIOSORPTION AND
ACTIVATED CARBON : BATCH STUDIES

C. GERENTE, Z. REDDAD, Y. ANDRES, C. FAUR-BRASQUET, AND P. LE CLOIREC

Ecole des Mines de Nantes, GEPEA, UMR CNRS 6144, BP 20722, 44300 Nantes Cedex 03,
France

E-mail: [email protected]

In order to treat effluents characterized by metallic and organic pollutants, an association of two
different adsorbents, sugar beet pulp and granular activated carbon, is investigated. In a first step,
equilibrium data are determined for each adsorbent and mono-component solutions. Then. multi-
metallic and organic-metal solutions are tested to determine some inhibitions or special selectivities.
Finally, it is shown that the association of sugar beet pulp for metal removal, and activated carbon
for organic elimination, is efficient to treat complex wastewaters.

1 Introduction

Industrial wastewaters are often complex aqueous mixtures containing various pollutants:
heavy metal ions, organic molecules, dyes, etc. and some of them are toxic or undesirable
to many living species.
Previous investigations were mainly focused on the use of low-cost sorbents [ I ] as a
replacement for costly methods of removing heavy metals from solution. Commonly, it
concerns inorganic materials like fly-ash [2] as well as chitosan [3,4], biomass [5,6],
sewage sludges [7], peat [8] ... For these latter of biological origin, the term of
biosorption is used to encompass contaminant uptake via physico-chemical mechanisms
such as adsorption or ion exchange. The low-cost biosorbent used in this paper is a by-
product of the sugar industry: the sugar beet pulp. This material is very cheap (1 00 per
metric tonne) and its production reaches 14.106 tomes of dry matter each year in the
European Community [9]. Sugar beet pulp is a natural polysaccharide and composed of
20 % and more than 40 % of cellulosic and pectic substances respectively. These latter
contain polygalacturonic acids which cany carboxyl functions and consequently exhibit
good capacities to retain metal ions [10,1 I]. Activated carbon, within its different forms,
is commonly used in water treatment for organic pollutant removal [12,13,14].
Nevertheless, some applicationshave showed its ability in metal removal [ 151.
In this work, the treatment of a synthetic wastewater composed by metals ions (Cu2+,
Ni2+ and Pb2+) and organic molecules (benzoic acid, benzaldehyde and phenol) is
investigated with a mixture of two sorbents, sugar beet pulp and granular activated carbon.
In a first step, equilibrium data are determined in a batch reactor for each adsorbent and
mono-component solutions. Then, the pollutants are mixed to treat binary and ternary
systems of metal ions, or a combination of Cu2+with organics. A second part focuses on
an association of the activated carbon with the polysaccharide for the treatment of a
solution containing Cu2 and phenol.

194
2 Materiels and Methods

2.1 The adsorbents

Raw sugar beet pulp was provided by Lyven (France). Its preparation has been previously
described by GBrente et al. [ 101 and the fixation of several metals have been studied and
published in [11,16]. The granular activated carbon, Pica NC60 from Pica Co. (France),
presents a high specific surface area (1200 mz.g-') coupled with a large microporosity
(94.5% VOI.).

2.2 Sorption experiments

Experiments were performed in batch reactors at 21 f 1C with a continuous stirring at
500 rpm and some ratio solidsolution fixed at 2.26 g.L-' for dry pulp and 0.5 g.L-' for
activated carbon. A pre-hydration of 90 min of the pulp was necessary and the pH of the
solution was stabilized at 5.5. Equilibrium times were deduced from the kinetics. The
mixed metallic solutions had equimolar initial concentrations (8.1 O4 mol.L-'). The
influence of benzaldehyde, benzoic acid and phenol on the fixation of CU" onto the pulp
was conducted using 100 mg.L-' (expressed in TOC) of organic compounds. The
adsorption on the mixture of sorbents of phenol and Cu2' ions was carried out with 50
mg.L-' of each components.

3 Results and Discussions

3.1 Fixation of Cu", Ni2' and Pb2' in mono and multi-metallicsolutions

--
alone CuZ'
............. &!. ...... ......................................................

"
-12% - 5 0 % -42% -10%-40% -43% -56% -47% -61%
% of decrease

Figure 1. Ion competition experiments, conducted at initial equimolar ionic concentrations (8.1O4 mo1.L-I)

In the case of single metal ions adsorption, it has been seen (Fig.1) that the adsorption
capacities order following Pb" > Cu2' > Ni2' [ 161. Smith and Martell [171 and Makridou
et al. [181 have shown that Pb2'presents a higher stability constant with galacturonic acid
than Cu". Moreover, no value are featured for Ni2'and Makridou et al. [ 181 assert that a
complex between this metal and galacturonic acid does not exist. These results confirm

195
the order obtained above. Some experiments are then performed to study the competition
of adsorption of these metallic species. Experiments conducted with two different cations
show that Ni2' ion exhibits the lowest competitive effect: whereas its presence induces
decreases of adsorption capacities of 12 % for Pb" and 10% for Cu", these last two
cations prevent the Ni2'fixation at 47 and 56 % height respectively. Pb" seems to have a
higher effect than Cu". In a mixed solution of copper and lead, the influence of the two
metals is slightly different: Pb2' induces 40 % decrease in Cu2+fixation, and vice-versa
the percentage reaches 50 %. As Pb2' is supposed to have the main preference to the pulp,
this value should be abnormally great, especially as the addition of the third metal induces
a lower decrease (- 42 %). The clear sorption preference for Cu2' and Pb2+is always
marked in the three metal solution: the addition of Ni2' seems to have no effect on the
other metals fixation.

3.2 Influence of organic compounds on Cu" fwation onto raw sugar pulp
The selected organic compounds were simple aromatic molecules (phenol,
benzaldehyde, benzoic acid) which could simulate moieties present in natural organic
matters. Preliminary, removal kinetics of these organic compounds have been performed
using the same concentrations; they showed that no fixation occurred (data not presented).
Secondly, isotherms of Cu2' ions sorption have been performed with a high constant
organic charge and they are plotted on Figure 2. In comparison with the fixation of copper
alone in solution, the presence of benzoic acid, at a concentration of 180 mg.L-', induces a
reduction of about 30% on the fixation capacity. The two other compounds decrease it
lower but the experiment does not enable to separate their influence. An explanation about
this special decrease in the presence of benzoic acid could be a complexation in solution
between Cu2+ions and benzoic acid, which would prevent the cation to be fixed on the
pulp. The stability constant given by Smith and Martell [ 171 between these two species
has a value of 1.6, which is close to that obtained with the complex Cu2'-galacturonic acid
(1.8). The effect of competition is obvious and these results confirm a strong affinity of
the carboxylic functions towards Cu2+ ions. Taking into account that the organic
concentrations are relatively important, it can be supposed that less concentrations would
have little effect on copper elimination.

3.3 Association ofpulp and activated carbonfor the removal of Cu2"and phenol
Preliminary experiments were carried out on each sorbent. As a little part of the organic
content of sugar beet pulp was soluble in water (close to 35 mg.L-' expressed in TOC for a
ratio pulp/water of 2.26 g.L-'), experiments with activated carbon were carried out in a
liquid medium of pulp-water. It was verified that this specific soluble organic matter,
probably composed of great molecules, was not adsorbed on NC60. The results are
presented in Table 1. In a first approach, it was confirmed that phenol was not removed by
sugar beet pulp at this concentration. The removal of copper was efficient with and
without the presence of phenol since the removal percentages reached a value close to
60 % in both cases. This interesting result confirms those obtained above and shows that
copper ions could be treated with pulp, even with a moderate organic charge. As far as
activated carbon is concerned, on one hand the efficiency of this kind of material is
verified towards the organic molecule since NC 60 exhibits a high removal percentage for
phenol (73 %) and low for Cu" ions (14 %). On the other hand, the presence of metal

1%
decreases around 20 % the fixation of phenol and the presence of phenol slightly affects
the copper fixation. The mechanisms of adsorption must be different, favoring a
chemisorption based on an attraction with the surface functions for metal fixation and a
physisorption, highly influenced by steric hindrance when another pollutants, for example
hydrated cations, are present in solution.
0.3

0.25

ECD
3 0.2
3
2
20
0.15
-ca
-ild
E 0.1 %
Cu alone
0.05 0 Cu + 130 mg.L-1 of benzaldehyde
4 Cu + 180 mg.L-1 of benzoic acid
X Cu + 140 mg.L-I of phenol
0

Figure 2 Organic influence of high concentrationon Cu2+fixation on pulp

Table 1. Removal percentage of copper and phenol, on pulp, NC 60 and a mixture of the two adsorbents
(Initial concentration of copper and/or phenol 50 mg.L-'; pulp ratio 2.26 g.L'; NC 60 ratio 0.5 g.L").

Pulp NC60 Pulp+NC60

cu2+ 63 14 67
Cu2'(with presence of phenol) 60 11 66
Phenol 0 73 74
Phenol (with presence of Cu2+) 0 54 53

When these two different adsorbents are used together, the results show that their
respective properties are conserved. In other terms, the sugar beet pulp exhibits a high
affinity with copper ions (67 YO),even if phenol is present (66 %), and NC 60 keeps its
efficiency with phenol (74 %) and is greatly influenced by the presence of copper.

To conclude, these preliminary results have shown the important ability of a low-cost
sorbent, the sugar beet pulp, to remove metal ions from aqueous solution. When several
metals are present in solution, a selectivity can be highlighted. The polysaccharide
exhibits high affinities towards Cu" and Pb2+whereas organic molecules are not retained
on it. The influence of organic matter on metal fixation occurs at high concentration or if a
complexation between species is possible. Finally, if the effluent contains organic and
metallic pollutants, the association of two kinds of adsorbent, namely sugar beet pulp and
activated carbon, seems to be efficient since they would keep their respective properties in

197
terms of adsorption. Further investigations would confirm these results in a dynamic pilot
unit and with real industrial wastewaters.

References
,
1. Bailey S. E., O h T. J., Bricka R. M. and Adrian D. D., A review of potentially low-
cost sorbents for heavy metals, Wat. Res. 33 (1999) pp. 2469-2479.
2. Ricou P., Lecuyer I. and Le Clouec P., Removal of heavy metallic cations by fly ash
in aqueous solution, Environ. Technol. 19 (1998) pp. 1005-1016.
3. Guibal E., Milot C. and Tobin J. M., Metal-anion sorption by chitosan beads:
equilibrium and kinetic studies, Ind Eng. Chem. Res. 37 (1 998) pp. 1454-1463.
4. Gerente C., Andres Y. and Le Cloirec P., Uranium removal onto chitosan:
competition with organic substances. Environ. Technol.,20 (1 999) pp. 5 15-521.
5. Volesky B., Advances in biosorption of metals: selection of biomass types. FEMS
Microbiol. Rev., 14 (1994) pp. 291-302.
6. Texier A. C., An&& Y. and Le Cloirec P., Selective biosorption of Lanthanide (La,
Eu, Yb) ions by Pseudomonas aeruginosa. Environ. Sci. Technol. 33 (1999) pp. 489-
495.
7. Solari P., Zouboulis A. I., Matis K. A. and Stalidis G. A. Removal of toxic metals by
biosorption onto nonliving sewage sludge, Sep. Sci. Technol., 31 (1996) pp. 1075-
1092.
8. Ho Y. S. and McKay G., The sorption of lead(l1) ions on peat, Wat. Rex, 33 (1999)
pp. 578-584.
9. Dronnet V. M., Renard C. M. G. C., Axelos M. A. V. and Thibault J.-F., Binding of
divalent metal cations by sugar-beet pulp, Carbolydr. Polym., 37 (1997) pp. 73-82.
10. Gerente C., Couspel du Mesnil P., Andres Y., Thibault J.-F. and Le Cloirec P.,
Removal of metal ions fiom aqueous solution on low cost natural polyssacharides:
sorption mechanism approach, React. Funct. Polym., 46 (2000) pp. 135-144 .
11. Reddad Z., Gerente C., Andres Y. et Le Cloirec P., Ni(I1) and Cu(I1) binding
properties of native and modified sugar beet pulp, Carbohydrate Polymers, 49 (2002)
pp. 23-3 1.
12. Crittenden B., Thomas WJ., Adsorption technology and design (1998), Butterworth-
Heinemann, Boston
13. Economy J., Lin R.Y., Adsorption characteristics of activated carbon fibers, Applied
Polym. Symposium, 29 ( 1976) pp. 199-21 1.
14. Faur-Brasquet C., Metivier-Pignon H., Le Cloirec P., Activated carbon cloths in
water and wastewater treatments, Res. A h . in Water Res., 2 (2002) pp. I - 19.
15. Faur-Brasquet C, Reddad Z, Kadirvelu K, Le Cloirec P, Modelling the adsorption of
metal ions (Cu2',Ni2',Pb2+) onto activated carbon cloths using surface complexation
models, Applied Surface Science, 196 (2002) pp. 356-365.
16. Reddad Z., GBrente C., Andres Y. and Le Cloirec P., Adsorption of several metal ions
onto a low-cost biosorbent : kinetic and equilibrium studies, Environ. Sci. h Technol.,
36 (2002) pp. 2067-2073.
17. Smith R. M. and Martell A. E. Critical Stability Constants (Plenum Press, New York,
1989).
18. Makridou C., Cromer-Morin M. and Schatff J.-P., Complexation de quelques ions
metalliques par les acides galacturonique et glucuronique, Bull. Soc. Chim. Fr.,
(1977) pp. 59-63.

198
 ADSORPTION CHARACTERISTICS OF PROTEIN-BASED LIGAND FOR
HEAVY METALS

MASAAKI TERASHIMA, NORIYUKI OKA, TAKAMASA SEI, KAZUYA SHIBATA, AND

HIROYUKI YOSHIDA
Department of Chemical Engineering, Graduate School of Engineering, Osaka Preficture
Universi& 1-1, Gakuen-cho.S a k i CiQ, JAPAN

E-mail: terasimaachemeng. osakfu-u.ac.jp

A fusion protein was engineered from maltose binding protein (pmal) and human metallothionein
(MT). The recombinant protein (pmal-MT) expressed in E. coli was purified, and immobilized on
ChitopearlTMresin. As expected from a tertiary structure of metallothionein, the prnal-MT ligand
adsorbed 12.1 cadmium molecules per one molecule of the ligand at pH 5.2. We have found that
the prnal-MT ligand also bound 26.6 gallium molecules per one molecule of the ligand at pH 6.5.
Adsorption isotherms for the both ions were correlated by Langmuir-type equation. Two types of
binding sites have been elucidated based on HSAB (hard and soft acid and base) theory: gallium ion
specifically binds to amino acid residues containing oxygen and nitrogen atoms, while cadmium ion
binds to specific binding sites formed by multiple cysteine residues. The pmal-MT protein bound
these metals in the concentration range of 0.2 - 1.O mM, and the bound metal ions could be eluted
under relatively mild condition (pH 2.0). The pmal-MT ChitopearlTMresin was stable and could be
used repeatedly without loss of binding activity. Thus, this new protein-based ligand would be
useful for recovery of toxic heavy metals and/or valuable metal ions from various aqueous solutions.

Introduction

Recently, the recovery and reuse of valuable metal ions such as rare earth metals,
from process waste water of electronic industries and waste electronic devices, is strongly
desired for saving precious resources and for achieving sustainable development. While
synthetic ligands or chelators are widely studied, biosorbents prepared from biomass of
bacteria, fungi, and algae have several advantages over synthetic chemical ligands. The
biosorbents, for example, show high selectivity to various ions depending on their tertiary
structures, and require only relatively mild conditions for adsorption and desorption [ 11.
Peptides and proteins could be efficient metal binding ligands, because they have the
functional groups for metal binding in their amino acid residues, and they can be produced
at low cost by recombinant technologies. While many peptides and proteins are known
to work as metal transport proteins in biological systems, metallothioneins (cysteine rich
proteins with molecular weight of ca. 7 kDa) have attracted researchers attention for
decades because they bind heavy metals in vivo [2]. The metallothioneins are considered
to be involved in detoxication and metabolism of heavy metals.
In this work, a fusion protein has been engineered from maltose binding protein
(pmal) and human metallothionein (MT). The fusion protein (pmal-MT) has been
expressed in E. coli, and purified with an amylose column. The purified fusion protein
was immobilized on a solid matrix, and its characteristics as metal binding ligand have
been studied. We have found that the pmal-MT ligand efficiently binds gallium ion, one
of the valuable rare metals desired to be recovered from aqueous solution [3]. Different
binding mechanisms for two metal ions have been elucidated based on HSAB (hard and
soft acids and bases) theory [4].

199
Methods

Preparation of p-ma1 MT protein, and immobilization of the gmal MT protein on

ChitopearlTMresin were described in detail in the previous work [3]. Chitopearlm resin
inmobilizing p-ma1 MT protein (pmal-MT ChitopearlTM),ChitopearlTMimmobilizing
p-ma1 protein (pmal ChitopearlTM),and ChitopearlTMresins have been prepared. The
latter two resins are prepared as negative controls.
The ChitopearlTMresin was packed in a glass column (inner diameter 1. 4 cm, bed
height 4.3 cm). The column was first equilibrated with a 20 mM MES buffer containing
20 mM NaCl and 10 mM 2-mercaptoethanol. Then 60 ml of the MES buffer containing
metal ion was applied at the flow rate of 0.5 ml/min. Adsorption capabilities of the
ligands were examined for cadmium, gallium, cupric, zinc, or nickel ion. Afier the
column was washed with MES buffer, the adsorbed metal ion was eluted with the MES
buffer (pH of which was adjusted to pH 2.0). In order to examine effects of pH on the
adsorption, pH of the MES buffer was varied from pH 5 to pH 9. The eluted solution
was collected as several fractions of 10 ml each, and the metal concentration of each
fraction was determined with atomic adsorption analysis (SAS 7500A, Seiko Instruments,
Japan). Total amount of the eluted metal ion was defmed as the adsorbed metal ion on
the resin. The total amount of the adsorbed metal ion was divided by the total amount of
immobilized protein to calculate the number of metal molecules bound to one mole of the
protein. The adsorption experiments were carried out multiple times, and the maximum
experimental error was 25%.

Results

Amounts of the protein immobilized on the ChitopearlTMresin were 3.55 (mg/g-wet

resin) for the pmal and 1.51 (mg/g-wet resin) for the pmal-MT. The optimal pH for
cadmium binding was pH 5.2 (data not shown). Figure 1 shows an adsorption isotherm
at 298 K for cadmium adsorption on the pmal-MT ChitopearlTMresin at pH 5.2. Neither
the pmal ChitopearlTMresin nor the ChitopearlTMresin adsorbed cadmium ion under the
employed experimental condition. These results clearly show that cadmium ion binds to
the metallothionein moiety of the pmal-MT ligand. The adsorption equilibrium was
correlated by a Langumuir-type equation. The equilibrium constant K, and adsorption
capacity for cadmium binding Q were 15.74 [mM-'] and 3.76 x 1 0 ' [mol/g-wet resin],
respectively. The maximum amount of adsorbed cadmium ion per metallothionein
molecule is 12.1 (mol cadmium/mol metallothionein), which is relatively close to a
theoretical value 7 confirmed by NMR [ 5 ] . These results strongly suggest that the
methallothionein moiety of the h i o n protein bind cadmium as it works in vivo. Figure 2
shows effect of NaCl concentration in the metal solution applied to the column on the
cadmium adsorption. The amount of adsorbed cadmium ion drastically decreased at
NaCl concentration about 45 mM, suggesting that the tertiary structure of the metal
binding site probably change at this salt concentration, and thus the ligand lose its binding
ability for cadmium ion. The binding ability, however, was easily recovered by washing
the column with the MES buffer (pH5.2).

200
 0.2 a4 0.6 QI) 1 0 50 100 150 m
Cadmium eweem. fmMl NaCI ancen. [mMl
Figure 1 Adsorption isotherm Figure 2 Effect of NaCl on
for cadmium ion cadmium adsorption

We have found in this work that the pmal-MT ligand also binds a valuable rare metal
gallium ion. Adsorption characteristics of the pmal-MT for gallium ion, however, were
different from those for cadmium ion. An optimal pH for gallium ion adsorption was pH
6.5, which was different fiom that for cadmium ion (pH 5.2). This result suggests that
.the conformation of metallothionein suitable for binding of cadmium ion is not preferable
to gallium ion, and vise versa. An adsorption isotherm for gallium ion at 298 K is shown
in Figure 3. Unlike the case of cadmium ion, the pmal ChitopearlTMresin and the
ChitopearlTMresin adsorbed gallium ion about 4.0 x lom7(mol/g-wet resin) at pH 6.5.
Since these results suggest that gallium ion bind to the proteins non-specifically,
non-specific binding of gallium ion to the proteins was examined by comparing the
adsorption of gallium ion on BSA. The numbers of gallium ions adsorbed on BSA (mol
gallium iodmol BSA) were 0.0335 at pH 5.2 and 0.114 at pH 6.5, while those of
cadmium ions (mol cadmium iodmod BSA) were 0.392 at pH 5.2 and 0.293 at pH 6.5.
These results showed that BSA did not adsorb cadmium ion and gallium ion under the
employed experimental condition, and suggest that both ions do not bind to proteins by
simple ion-exchange effects. In order to evaluate the adsorption on metallothionein
moiety, the amount of gallium ion adsorbed on the base matrix was subtracted from the
experimental data for the pmal-MT ChitopearlTM resin. The corrected result is shown in
Figure 4, and adsorption equilibrium is correlated by Langmuir-type equation. The
equilibrium constant K, and adsorption capacity for cadmium binding Q were 5.2 1 [mM']
and 9.09 x [moVg-wet resin], respectively. It should be noted that the maximum
amount of adsorbed gallium ion per metallothionein molecule reached 26.6 (mol gallium
iodmol metallothionein), which is much higher than that of cadmium ion. Effect of
NaCl concentration in the metal solution on the gallium adsorption is shown in Figure 5.
The drastic decrease in metal binding at NaCl45 mM, which was seen for cadmium ion,
was not observed in the case of gallium ion. These results also strongly suggest that the
binding mechanism of gallium ion is different from that of cadmium ion. The amounts of
adsorbed metal ion on the pmal-MT ChitopearlTMfor various ions are summarized in
Table 1. These results show that the metal binding on metallothionein is highly
selective.

201
 I,

0 .

0 0.2 0.4 0.6 OS I 0 0.2 0.4 0.6 M 1

Gallium C O I I C ~ . [mMl
Gallium cancea. [mMl
Figure 3 Adsorption of gallium ion Figure 4 Adsorption isotherm
for gallium ion

Table 1 Amount of adsorbed metal ion for

various ions

Metals Adsorbed Ionic radius

amount [x 1P m]
Cia 26 62 Hard
Ni 0.3 69 Little hard
Zn 0.2 74 Little soil
Cd 12.1 91 Soil

0 50 100 150 200

NaCI canem. [mMl
Figure 5 Effect of NaCl on gallium
adsorption

Discussion

We have found that metallothionein, which selectively binds cadmium ion in vivo,
binds gallium ion. The number of the gallium ion bound to one molecule of
metallothionein, 26.6, was about twice as large as that of cadmium ion. The results of
adsorption experiments to BSA, and the adsorption of various metal ions show that
metallothionein selectively binds gallium ion. The specificity is very high because other
metals which as similar ionic radius did not bound to the metallothinein as shown in Table
1.
A binding mechanism of metallothionein for cadmium ion and gallium ion is
elucidated as follows. According to HSAB (hard and soft acid and base) theory, metal
ions are classified into hard acid (ion) and soft acid (ion). Cadmium ion is classified as
soft acid that has strong binding affinity to S and P atoms. On the other hand, gallium
ion classified as hard acid that has strong binding affinity to N and 0 atoms. Therefore,
cadmium ions bind to S atom of cysteine residues of the metallothionein as indicated by
NMR. The conformation of binding sites formed by multiple cysteine residues should be

202
strongly affected by the change in tertiary structure of the metallothinein. Drastic change
of the amount of adsorbed cadmium ion by the increase of NaCl concentration (Figure 2)
suggests that the large conformation change of the cadmium binding sites of
metallothionein. On the other hand, gallium ion should bind to N and 0 atoms of amino
acid residues of the metallothionein molecule. Human methallothionein posses 3 Asp, 1
Glu, 7 Lys, 8 Ser, 2 Thr,1 Gln,1 Asn residues per molecule. Some of these negatively
charged residues are probably located the outer surface of metallothinein might form
binding sites.
As elucidated fiom the adsorption experiment on BSA, adsorption of gallium ion by
simple ion-exchange effect is negligible. The characteristicsof metal ions such as ionic
radius and hardness/sobess, and the conformation of the metallothionein probably affect
the selectivity of metal adsorption. The understanding of the mutual interactions among
those factors would be a key factor in designing the protein-based ligand suitable for a
specific metal ion.

References

1. Gutnick, D. L., and Bach, H., Engineering bacterial biopolymers for the biosorption
of heavy metals; new products and novel formulations,Appl Microbiol Biotechnol. 54
-
(2000) pp. 45 1 460
2. Romero-Isart, N. and Vasak, M., Advances in the structure and chemistry of
metallothioneins,J. Inorg. Biochem. 88 (2002) pp.388-396
3. Terashima, M., Oka,N., Sei, T. and Yoshida, H., Adsorption of cadmium ion and
gallium ion to immobilized metallothionein fusion protein, Biotechnol. Progress, in
press (2002)
4. Pearson, R. G., Hard and soft acids and bases, J. Am. Chem. SOC.85 (1963) pp.
3533-3539
5. Messerle, B. A., Schaffer, A., Vasak, M., Kagi, J. H., and Wuthrich, K.
Three-dimensional structure of human [ 1 13Cd7]metallothionein-2 in solution
determined by nuclear magnetic resonance spectroscopy, J Mol Biol. 214 (1990) pp.
765-779

203
 PREPARATIVE CHROMATOGRAPHY AT SUPERCRITICAL CONDITIONS

ARVIND RAJENDRAN, MARC0 MAZZOTTI AND MASSIMO MORBIDELLI

ETH Swiss Federal Institute of Technology, CH 8092 Zurich, Switzerland
E-mail: [email protected]

The basic issues in packed bed supercritical fluid chromatography (SFC) are: i) characterizing
adsorption equilibrium; ii) charracterizing mass transfer dynamics; iii) predicting column dynamics
and iv) the effect of polar modifiers. The first three issues have been addressed through experiments.
Experiments have been performed on a preparative column with significantpressure drop using pure
COz as an eluent and phenanthrene dissolved in toluene as solute. Parameters relating to retention
behaviour and mass transfer characteristics were measured. The observations are reported and the
deviations with respect to HPLC and GC behaviour are highlighted and discussed.

1 Introduction

Preparative chromatography is a proven technology for the separation of specialty

chemicals mainly in food and pharmaceutical industries, particularly the
enantioseparation of chiral compounds on chiral stationary phases. The potential of
preparative chromatographic systems were further increased by the development of
continuous chromatographic processes like the simulated moving bed (SMB) process.
Compared to the batch column chromatography, the SMB process offers better
performance in terms of productivity and solvent consumption [2].
Supercritical fluids poses properties that lie in between those of liquids and gases.
These properties, which are functions of their density, can be tuned according to the
process requirements. Supercritical fluid chromatography (SFC) is a proven analytical
tool which allows one to exploit the advantages of GC (faster and better separation) for
non-volatile substances which are usually analysed using HPLC. The solvent power of the
supercritical fluid is a function of its density. Hence, when a solute dissolved in a
supercritical fluid is in contact with a stationary phase, i.e. a solid adsorbent, the affinity
of the solute to the stationary phase, which under linear conditions is characterized by the
Henrys constant Hi, depends on the density of the supercritical fluid. At higher densities,
the supercritical fluid has a higher solvent power and hence the solute has a lower Hi
value and vice versa. This property allows for the establishment of a solvent gradient in a
SMB unit, thereby enhancing separation performance. In fact it was shown that the
supercritical fluid simulated moving bed (SF-SMB) process operated under the pressure
gradient mode (where the pressure in each section is regulated by a back pressure
regulator) offers a productivity improvement by a factor of 3 compared to the isocratic
mode (without any pressure regulation) [l]. In this context, it is therefore important to
study and understand the fundamentals of SFC under conditions where the pressure drop
along the column is significant, because these are the conditions in preparative
applications, particularly in SF-SMB.
The pressure gradient in the column causes substantial density gradient, which in turn
leads to a gradient in velocity. Since the mass transfer properties are a function of density,
they change at every point along the column. Moreover, unlike HPLC and GC systems,
where the Henrys coefficient, Hi, does not depend on the pressure, in SFC systems the
retention characteristics are a function of pressure (density) [l]. Hence, the dynamics of
an injected pulse is governed by several factors. This establishes the need to develop a

204
model, which could be used to simulate the pulse response of a packed column under
supercritical conditions with substantial pressure drop. A dynamic model is especially
useful, since it can be easily extended to simulate the SF-SMB process.
In the present study, experiments have been performed on a preparative column using
pure COz as the mobile phase. Experiments have been performed at four different back
pressure levels. For each back pressure setting, runs were performed both at low flow
rates, i.e. where the pressure drop was negligible and at high flow rates. The main process
aspects, i.e. pressure drop, mass transfer and retention have been experimentally
evaluated and analysed.

2 Experimental setup and procedure

Carbon dioxide (99.995% pure, obtained from PanGas, Switzerland) was used as the
mobile phase. Phenathrene @urity>97%) dissolved in Toluene (purity>99.7%, both

100

80 - +- BP =I80 bar
n
L
+-BP =210 bar
lu
a
LI
60-

0
E!
a
2 40-
v)
v)
2
n
20 -

0 I I

I I I I
0 I 2 3 4x1Om2
Mass flow [g/s]

Figure 1. Pressure drop characteristicsof the SFC column at different back pressure levels.
obtained fiom Fluka, Switzerland) was used as a solute. A Lichrospher 100 RP-18
column (Merck, Darmstadt) 125x4 mm, with an average particle size of 5pm, was used
for the experiments.
The experimental set-up consists of a syringe pump (Isco 260D) capable of
producing a continuous flow of COz which flows through an injection valve (Valco

205
C14W) that has an internal loop volume of 60 nL. The mobile phase then enters the
column at the end of which is a UV detector (Jasco W-1570).The pressure in the system
is determined by a back pressure regulator (Jasco BP1580-81) which is located
downstream of the U V detector. The column and the injection valve are housed in a
temperature controlled water bath. Upstream and downstream pressures are measured
using pressure transducers (Trafag 8891).
The experiments are performed by setting the back pressure regulator at the desired
level and programming the syringe pump to operate at a given flow rate. The system is
then allowed to reach a steady state. Once the pressure profile in the system is established,
a mixture of phenanthrene in toluene (2% w/w) is injected into the column through the
injection valve and the data acquisition is started simultaneously. For each setting, the
experiment is repeated more than three times to ensure reproducibility.

3 Experimental Results

Both high flow rate and low flow rate experiments were performed at 4 different
back pressure levels, namely 130, 150, 180 and 210 bar. Some low flow rate experiments
were also perfomed at intermediate back pressure settings. The operating temperature for

0 BP=150 bar
A B P 4 8 0 bar
V BPe210bar

0 200 400 600 800

(mass flow ratej[(g/sj]

Figure 2. Retention time against the inverse of mass flow rate at different back pressure levels.

206
all the runs was 64C.The pressures upstream and downstream of the column and the UV
signal were measured.
From the UV detector readings, the retention time and HETP values were calculated.
The HETP was calculated using the formula
N = 5.45(t,/~)~
HETP = L/N
where t R is the retention time, w the width of the peak at half peak height, N the number of
plates and L the length of the column. These equations have been used to describe the
HETP behaviour though they have the limitation that they assume the velocity to be
constant along the column.
The pressure drop characteristicsfor the four different sets of experiments are plotted
against the mass flow-rate in Fig. 1. The mass flow rate is used as the independent
variable since it is the only variable which remains constant throughout the column. At
higher flow rates there is a deviation from linearity. For a given mass flow rate, the runs
at a low back pressure setting show a larger pressure drop.
The measured retention times are plotted against the inverse of mass flow rate in Fig.
2. It can be seen that the points corresponding to a particular back pressure setting fall on
a straight line and the slope of the line depends on the back pressure setting. The line
corresponding to a lower back pressure setting has a larger slope than the one at a higher
back pressure setting. This is in contrast to HPLC where for a given temperature, under
linear conditions, all points, irrespective of the back pressure setting, will fall on one
straight line whose slope is proportional to Hi. This shows that in the case of SFC, Hi,is a

6o
50 -

-
E3.
Y
40 -

n 30-
F
w
I
20 -

10 -

0
0 10 20 30 40
0
r
50x10"
Mass Flow [glsJ
Figure 3.HETP values at different back pressure levels.

207
function of density. Further at high flow rates, there is a velocity gradient and a density
gradient in the system and these affect the retention time. Hence, the observed, or the
apparent, Hi is a combined effect of these two gradients.
The mass transfer characteristics, which are described by the HETP, are shown in
Fig. 3, where the HETP is plotted against the mass flow rates for different back pressure
levels. In general, the HETP curve has the typical shape of the well-known van Deemter
plot. Though plotting the HETP in this fashion (i.e. grouping runs with the same back
pressure setting) does not offer the provision to extract the mass transfer parameters 6om
the van Deemter equation, it nevertheless offers a qualitative picture of the mass transfer
kinetics. In general, at low flow rates, the H E P values fall with increasing flow rates,
reach a minimum, and gradually rise. Let us focus on the part of the curve after the
respective minima. The curve is flat for the runs with a back pressure of 130 bar
compared to those corresponding to 150 and 180 bar. The slopes of the later part of the
curves increase with increasing back pressure. It can also be seen that the curves show
cross over at larger flow rates. For the curve corresponding to 210 bar, it was however,
not possible to perform high flow rate experiments as the upstream pressures rose beyond
the maximum allowable pressure of the syringe pump.

4 Conclusion

Experiments have been performed on a preparative SFC system using pure CO2 as
the mobile phase under significant pressure drop. The retention times, pressure drop
characteristics and the mass transfer behaviour were studied. The trends observed differ
6om the behaviour of HPLC systems. These trends also emphasize the complexity
involved in analyzing the data for SFC measurements, which imply in turn greater
complexity of the SFC model as compared to standard liquid chromatography model.

Reference

1. Denet, F., Hauck, W., Nicoud, R. M., Di Giovanni, O., Mazzotti, M., Jaubert, J. N.
and Morbidelli, M., Enantioseparation through supercritical fluid simulated moving
bed (SF-SMB) chromatography, Ind. Eng. Chem. Res. 40 (2001) pp. 4603-4609.
2. Juza, M., Mauotti, M. and Morbidelli, M., Simulated moving-bed chromatography
and its application to chirotechnology, Tren& Biorechnof.18 (2000) pp. 108- 1 18.

208
 ADSORPTIVE SEPARATIONOF OLIGOSACCHARIDES INFLUENCE OF
CROSSLINKING OF CATION EXCHANGE RESlNS

JOHANA. VENTE'.~,HANS BOSCH', ANDRE B. DE HAAN', PAUL J.T. BUSSMANN~

I Separation Technology Group, Faculty of Chemical Technology, Universiw of Twente, P. 0.Box

21 7, 7500 AE Enschede, The Netherlands

TNO-MEP, P.O. Box 342, 7300A H Apeldoorn, The Netherlands

The influence of crosslinking on the sorption propetties of poly(styrene-co-divinylbenzene) (PS-

DVB) strong acid ion exchanger in K ' or Ca2+form was studied. Isotherms of sugars were deter-
mined for resin containing 2, 4 and 8% DVB. Sorption was strongly influenced by the degree of
crosslinking. At increasing crosslinking the amount of sorption by (non-selective) distribution
decreases and complexation is the dominant sorption mechanism. Resin with 2% DVB was hardly
selective, but resin with 8% DVB was very selective. A selectivity of 7.0 for fructose/sucrose and
Ca" loaded resin was obtained due to a combination of size exclusion and complexation. However,
sorption capacity decreared with increasing crosslinking. Chromatograms of fructo- and
galactooligo-saccharides (0s)showed that the separation of the monosaccharidesfrom the 0s was
best for 8% DVB but the fractionationof the 0s was best for 2%DVB.

1 Iatroductioa

Oligosaccharides (0s)are applied as functional food and feed ingredients. Fructo-OS

(FOS) are made by partial hydrolysis of inulin and galacto-0s (GOS) by trans-
galactosylation of lactose. They are produced as mixtures of different types of carbohy-
drates and further separation is required for most applications. Examples of desired sepa-
rations are the removal of monosaccharides in order to decrease the amount of calories
and the sweet taste. A technique to perform the desired separation is chromatography. The
adsorbent used in a chromatographic separation has a large influence on the performance
of the separation process. Fructose/glucose is separated on process scale with Ca" loaded
crosslinked poly(styrene-co-divinylbenzene) (PS-DVB) strong acid ion exchange resins.
To investigate the influence of resin properties on the separation, we selected PS-DVB
resins. There are two chemical properties of the PS-DVB resin material to optimise: (1)
the type of cation and (2) the degree of crosslinking. Earlier work focussed on the choice
of the cation. K ' was selected as the cation for GOS separation and Ca" for FOS
separation [1,21. The work presented here focussed on the degree of crosslinking.
The degree of crosslinking of the PS-DVB resin can be varied by the amount of DVB
used during the synthesis of the resin. Crosslinking decreases the elasticity of the resin
and thereby the swelling and equilibrium water content [3]. Saccharides can be sorbed
into the resin, either by distribution of the saccharides between the liquid inside the resin
and the liquid outside the resin or complexation [4]. To be able to distribute, the
saccharide molecules have to be smaller than the size of the interstices of the resin. The
size of hydrated monosaccharides is in the same order of magnitude as the size of the pore
diameter of a resin [5]. The separation of monosaccharides from 0 s may be improved by
choosing a DVB content such that disaccharides and 0s are excluded, but the
monosaccharides are still able to be sorbed by the resin. Higher water content increases
the sorption of the saccharides by distribution. The amount of cations per volume unit
water increases with increasing DVB content. Therefore, complexation driven sorption

209
might improve relative to the sorption by distribution with increasing DVB content. The
DVB content not only influences the equilibrium sorption properties of the resin, but also
the sorption kinetics and mechanical properties such as elasticity and attrition resistance.
Published data illustrate the influence of the degree of crosslinking of PS-DVB resins
on the separation of sugar alcohols [6], monosaccharides [7],and malto-OS [8]. Resins
with DVB contents between 3% and 8% are suitable for the separation of two hexoses
[7]. For sugar alcohols the optimal DVB content is 7% DVB for Ca2' form resins [6].
The separation of glucose from maltose improves with increasing DVB content. For
malto-OS and DVB contents between 2% and 6%, the separation improves with
increasing DVB, but is completely lost for a DVB content of 8% due to size exclusion of
the larger 0s by the resin. Resin with 6% DVB was optimal for the separation of malto-
0s with a degree of polymerisation of 1 to 7.It can be concluded from these literature
data that crosslinking has a large effect on the sorption and separation properties of PS-
DVB resin and that the effects are dependent on the type of sugar. However, no
information could be found on the separation of FOS or GOS. The goal of the work
presented in this paper was to determine the influence of the crosslinking on sorption
properties of PS-DVB cation exchange resins for FOS and GOS separation. Improved
understanding of the interactions of saccharides with sorbents may lead to the
development of highly selective sorbents for low cost separations.

2 Materials and methods

Gel type strong acid PS-DVB cation exchange resin, Dowex 50W (particle diameter 38-
74 pm), was used with different DVB contents. The resin was ion exchanged into the
desired cation form. Table 1 summarizes the properties of the columns (internal diameter
0.160 m) packed with resin. Porosity was calculated from the retention time of Dextran
T2000 (Pharmacia, Sweden). Dry substance content was measured by air-drying of
filtrated resin at 105C until constant weight. Isotherms and chromatograms were
measured at 60C in a column set-up, as was described earlier [2]. Glucose (0-300 gll),
galactose (0-10 a),
fructose (0-300 a),
sucrose (0-300 a),
lactose (0-200 and a)
mixtures of FOS (Raftilose@60, Orafti, Belgium), containing fructose, sucrose and 0s
and GOS (Elixor@259, Borculo Domo Food Ingredients, The Netherlands), containing
glucose, galactose, lactose and 0s were used. The isotherms were correlated with
q=uz+bc, with q the saccharide concentration in the resin, a and b fitparamters and c the
concentration in the liquid phase. The chromatograms were plotted as a fimction of
dimensionlesstime, defined as: (r-r,mcer)/r,mce,., with rrmer the retention time of dextran and
r the time. The selectivity of component i relative to component j , was calculated as:
(qh)4@$

Table 1: Properties of columns packed with strong acid PS-DVl3 cation exchange resin (i.e.=ionexchange).

210
3 Results and discussion

3. I Isotherms of saccharides on cation exchange resins

Fig. 1 shows the single sugar isotherms for K+ and Ca2' loaded resins with different
degrees of crosslinking. High sugar concentrations were applied, because concentrated
sugar solutions are used in commercial processes. The isotherm data were correlated with
the equations in Table 2. A stronger crosslinked resin, resulted in less sorption of sugars.
This result can be explained by the decreased elasticity and swelling of the resin with an
increased crosslinking. This results in lower water content of the resin (see also Table 1)
and decreasing sorption of saccharides by distribution. Moreover, crosslinking had a
larger effect on sorption than the type of cation [I, 21. For K
' loaded resin, the observed
order of adsorption was galactose2fructose>glucose>lactose>sucrose. The sorption order
of the sugars was for Ca" loaded resin the same as for K" loaded resin, except that
fructose2galactose. Fructose forms a complex with a Cat+ ion [9], which explains that
fructose sorbed better than glucose or galactose in highly crosslinked Ca" loaded resins.
For low crosslinking however, the monosaccharides sorbed almost to the same extent.
The explanation for the loss of selectivity may be that at low crosslinking the resin
contains more water and the amount of cations per volume unit resin is lower.
Consequently, more distribution may occur and the amount of fructose sorbed due to
complexation may decrease relative to the amount sorbed by distribution. In that case, the
sorption of non-complexing monosaccharides such as glucose or galactose is favoured
relative to fructose, as was observed experimentally.
At equal degree of crosslinking, the disacharides lactose and sucrose sorbed less than
the monosaccharides due to their larger size and hence the restricted accessibility to the
resin interstices. The sorption of the disaccharides decreased strongly with increasing
crosslinking. For 8% DVB the sorption was almost vanished. At increasing crosslinking
less water is available in the resin for distribution. In addition, part of the water is
hydration water of the cations [lo]. Apparently, for 8% DVB almost no water was avail-
able for distribution of sugar. The effect of crosslinking was stronger for sucrose than for
lactose. Lactose (0-PD-galactopyranosyl-(1,4)-D-gIucopyranose)and sucrose (0-a-D-
glucopyranosyl-(1,2)-PD-fructofuranoside)differ in the constituent monosaccharides and
the bond between the monosaccharides. Although sucrose contains a fructose unit, it is
not able to complex with Ca" in the same way as the monosaccharide fructose, because
the glycosidic bond of sucrose occupies the complexation site of the fructose unit. How-
ever, the differences in sorption may be the result of differences in the structure of the
molecules. Lactose is able to convert via the open chain form to another anomeric form,
whereas sucrose does not. Sucrose exhibits two interresidue intermolecular hydrogen
bonds in aqueous solution [I 13. These structural differences may result in a larger effec-
tive size of sucrose compared to lactose, hence increased size exclusion and less sorption.

Table 2: Isotherm correlations for 2,4 and 8% crosslinking of PS-DVBresin loaded with K' or Ca*' at 60C.

211
 0 1w 200 300 0 100 m 300 400

Fig. 1: Isotherms of glucosc for 2%,4%and 8% DVB on PSDVB rcsin with K

' (left) or Caz+(right) at 60C

3.2 Selectivity of cation exchange resins

Table 3 lists selectivities for sugar concentrations at 200 g/l, which were calculated from
the single sugar isotherms in Table 2. Besides the fiuctose/glucoseselectivity, Table 3 in-
cludes the selectivities for glucose/lactose and fructose/sucrose, which are indicative for
the separation of glucose from GOS and fructose from FOS, respectively. All selectivities
increased with increasing crosslinking. The increase was already explained for fructose/
glucose in the previous section. The selectivity of glucose/lactose increased with increa-
sing crosslinking due to the larger size of lactose compared to glucose. The selectivity of
fructose/sucrosereflects a combination of complexation of fructose with Ca2' cations and
size exclusion of sucrose and resulted therefore in the highest selectivity. Selectivities as
high as up to 2.5 for glucosellactoseand up to 7.0 for fhctoselsucrose were obtained.

Table 3: Selectivityof resin with P?,

4%and 8% crosslinking at an individual sugar concentration of 200 gll.

1.1 1.3 1.5 1.1 1.3 2.5

1.2 1.7 3.8 1.4 2.4 7.0

3.3 Chromatograms
Fig. 2 presents chromatograms of GOS on K ' resin and FOS on Ca" resin. Due to
the use of a short column and a relatively large resin particle diameter, baseline separation
was not achieved. However, qualitative effects of different DVB content of resins could
be obtained from the chromatograms. It appeared that with decreasing DVB content of
the resin, saccharides eluted over a longer period and the peaks became wider. The
components in the mixtures that were not retained, eluted first and with increasing time,
more retained components eluted. Glucose was the last eluting peak in the chromatogram
of GOS and fructose the last peak for FOS. From injection of pure components it
appeared that the retention times of the saccharides, which were determined only by
equilibrium effects, were in agreement with the measured isotherms [I]. Fig. 2 clearly
shows that, due to increased sorption capacity, the elution times increased with decreasing
crossliiking. Separation of fructose was in particular good for 8% DVB, as was expected
from the high selectivity of fructose/sucrose. Furthermore, Fig. 2 shows that 4% and 2%
DVB improved the hctionation of OS, which resulted in more peaks on the
chromatogram. Although 0s eluted over a longer period, the peaks on the chromatogram

212
for 2% DVB were wider than the peaks on the chromatogram for 4% DVB. Especially in
the case of GOS this resulted in poor separation. The long elution times lead to excessive
product dilution, which is for large-scale applications an economical disadvantage.

GOS m 4% DVB
nrinmK+fan 1 GOS m 2% DVB
rcsin m K' fam

0.0 0.5 1.0 1.5 20

dk*nsM...nktbnk.(J
25 0.0 0.5 1.0
dlnm-
1.5
- 20
-13
25 0.0 0.5 1.0 1.5
dkmnimhu-mC)

Fig.2: Chromatograms of G O S (K' resins) and FOS (Ca resins) with different DVB content (8?4 4%, and 2%
*+
20 25 :

DVB), injection volume 1 ml, concentration 100 g synrp/l, flow rate 0.545 ml/min, temperature 60C.

4 Conclusions

The isotherms of sugars on PS-DVB resin showed that increased crosslinking resulted in
more selective sorption of sugars at the cost of sorption capacity. The effect of cross-
linking on capacity and selectivity is larger than the effect of cation type [2]. Non-
selective sorption of sugars decreases with increasing crosslinking, due to a decrease in
available space in the resin for distribution. Also, with increasing crosslinking, the
relative contribution of complexation to the sorbed mount increases and as a result the
selectivity increases. The chromatogramsof FOS and GOS showed that a DVB content of
8% is better than 4% DVB for monosaccharide removal. A PS-DVB resin with 8% DVB
is best for the removal of fructose from FOS.However, if the goal is to separate the FOS
in several fractions with different degree of crosslinking, then it is better to use 4% DVB.
For separations including molecules with different size, it is recommended to select first
the optimal degree of crosslinkingbefore optimising the cation type.

References

11 J. A. Vente, et al., to be submitted, 2002.

2!I J. A. Vente. et al. in: AIChEAnnual Meeting - Symposium on industria/App[icationsof
I4,dkorption and Ion Exchange. 2001. Reno (NV), USA: AIChE, p. 786.

3 J. Tiihonen, et al., Journal of Applied Polymer Science, 2001.82: p. 1256.

4 S. Adachi, et al., Bioscience Biotechnology and Biochemistry, 1997.61( 10): p. 1626.
1

5 M.Saska et al., Journal of Chromatography, 1992.590: p. 147.

6 H. Caruel, et al., Journal of Chromatography, 1992.594: p. 125.
7 S. Adachi, et al., Journal of Chemical Engineering of Japan, 1999.32(5): p. 678.
a S. Adachi, et al., Agricultural and Biological Chemistry, 1989.53(12): p. 3193.
J R. W. Goulding, Journal of Chromatography, 1975.103: p- 229.
1 9 R. S. D. Toteja et d.,Langmuir, 1997. 13(11): p. 2980.
1 I] S. Immel and F. W. Lichtenthaler, Liebigs Annalen, 1995(1 I): p. 1925.

213
 IDENTIFICATION AND PREDICTIVE CONTROL OF A SIMULATED
MOVING BED PROCESS

IN-HYOUP SONG AND HYUN-KU W E E

School of Chemical Engineering h Institute of Chemical Processes
Seoul National Universiw, Kwanak-ku, Seoul, 151 742, Korea -
E-mail: [email protected]

MARC0 MAZZOTTI
Institute of Process Engineering, ETH Zurich, CH-8092 Zurich, Switzerland
E-mail: Mazzotti@ivuk mavt.ethz.ch

Identification technique is applied to a simulated moving bed (SMB)process for chiral separation of
enantiomers of TrOgers base and an advanced predictive controller is designed on the basis of the
identified model. To obtain the identified model, an artificial continuous system is constructed by
keeping the discrete events such as the switching time and the number of columns to be switched
constant. The SMB process is identified as an inputloutput data-based prediction model, which is
then used to design a linear predictive controller. In this study the internal flow rate ratios are chosen
as the input variables whereas the pair of product purities are taken as the controlled outputs. It is
demonstrated by simulation studies that the designed predictive controller performs satisfactorily for
the disturbance rejection as well as for the setpoint tracking in the SMB process.

1 Introduction

In the chemical industry, chromatographic separation process is an emerging

technology for the separation of pharmaceutical products, food and fine chemicals. To
improve the economic viability, a continuous countercurrent operation is often desirable
but the actual movement of the solid leads to a serious operating problem. Therefore, the
simulated moving bed (SMB) process is an interesting alternative option.
In recent years, several researchers have applied some advanced control strategies to
simulated moving bed units to treat the dynamic operation of SMB processes, ranging
from the nonlinear control strategies such as the input-output linearizing control [4] to the
repetitive model predictive control [S]. One of the shortcomings of these control strategies
is that they use the first principles model. Although these controllers may be effective to
treat various control problems, it may be difficult to implement them to actual SMB
processes because of their heavy computational load and complex design procedure.
To overcome these drawbacks of the controller based on first principles model,
various identification techniques were applied. Neural network based model predictive
control was used for the dynamic control of SMB unit [ 11, however, its implementation to
actual process can be very difficult because of the complexity of identified neural net
model.
In this study we identify an SMB process using the subspace identification method.
The well-known inputloutput data-based prediction model is also used to obtain a
prediction equation which is indispensable for the design of a predictive controller. The
discrete variables such as the switching time are kept constant to construct the artificial
continuous input-output mapping. With the proposed predictive controller we perform
simulation studies for the control of the SMB process and demonstrate that the controller
performs quite satisfactorilyfor both the disturbance rejection and the setpoint tracking.

214
2 Brief description for SMB process

In this study, the SMB process is divided into four sections, each of which consists of
2 columns of chromatographyplaying a specific role in the separation. Ethanol solution of
the racemic Troger's base is taken as the feed stream and unsupported microcrystalline
cellulose triacetate(CTA) bead is used as the stationary phase. The separation is carried
out in the two central sections. For the reference conditions of simulation study, one may
refer to the previous work[7].
The principle of operation can be best described with reference to the equivalent true
countercurrently moving bed (TCC) configuration. Since the two configurations are
equivalent, i.e., they achieve the same separation performance provided the geometric and
kinematic conversion rules are fulfilled, the simpler model of the equivalent TCC unit can
be used to predict the steady state separation performances of SMB units, in particular, for
design purposes.
The first principles model of the SMB unit is constructed with reference to the
previous works[4,6] and considered to be the actual plant.

3 Subspace identification of SMB process

Figurel. Validation of identified model.

In order to solve the first principles model, finite difference method or finite element
method can be used but the number of states increases exponentially when these methods
are used to solve the problem. Lee et d [ 8 ] used the model reduction technique to reslove
the size problem. However, the information on the concentration distribution is scarce and
the physical meaning of the reduced state is hard to be interpreted. Therefore, we intend to
construct the input/output data mapping. Because the conventional linear identification
method cannot be applied to a hybrid SMB process, we construct the artificial continuous
inpudoutput mapping by keeping the discrete inputs such as the switching time constant.
The averaged concentrations of rich component in raffinate and extract are selected as the
output variables while the flow rate ratios in sections 2 and 3 are selected as the input
variables. Since these output variables are directly correlated with the product purities, the
control of product purities is also accomplished.

215
 We adopt the inputloutput data-based prediction model using the subspace
identification technique. To find the correlation between the inputs and outputs, we need
to obtain the input and output data. On the basis of the triangle theory[6], the optimal feed
flow rate ratios at steady state are calculated. Then, the pseudo random binary input signal
is generated on the basis of this optimal value. Figure 1 compares the output from the
identified model (dot) with that from the first principles model (solid curve). Clearly, we
observe that the identified model based on the subspace identification method shows an
excellent prediction performance. The variance accounted for (VAF) indices for both
outputs are higher than 99%. The detailed identification procedure can be founded in the
literature [3,5,9,10].

4 Predictive control for SMB process

The inputloutput data-based predictive controller based on the identified model is

designed and applied to a MlMO control problem for the SMB process. We use the
input/output data-based prediction model in the MPC algorithm. The QP method is used
to obtain the control input u,by minimizing the objective function defined as

minJ(u f ) = ( L p
"f
+ L,wp - r f ) T@Lu,u, + L , W ~- r,)+u;Ru,
where w p is defined as OPT u ~with ~ values of inputs up and output yp , and rj
~ the) past
denotes the set-point trajectory. Here, Q and R are the weighting matrices for the output
and input, respectively and the controller parameters L,, and L," are determined during the
identification procedure. It is to be noted that there is no need to explicitly calculate the
state estimate or the state space model. The complete separation region in the triangle
theory is considered as the input constraint, and the output constraint comes from the
requirement for the purity. One may refer to our previous work[9] for the details of the
controller design procedure.
Here we shall treat two typical control problems of practical interest; one is the
disturbance rejection and the other is the setpoint tracking. First, we assume that the feed
pump stops during 40 minutes after 40th switching. These may be considered as
unmeasured disturbances introduced to the process. Figure 2 shows that the controller
successfully rejects the unmeasured distu rbance.

216
 B
j
g .r
t
B

Figure 2. Disturbance rejection performance.

Figure 3. Tracking control performance.

In the second case the setpoints for the average concentrations of A at extract and
that of B at raffiate are changed simultaneously after 20 switching times as shown in
Figure 3. For the control purpose, the prediction and control horizons are set equal to 5
and 2 switching periods, respectively. The weighting matrices are tuned by the trial and
error method. Here it is noticed that the control inputs act predictively to bring the control
output to their new respective setpoints. It is clearly seen that the control performance is
quite satisfactorily.

217
5 Conclusions

An SMB process is identified by using the subspace identification method. The

inputloutput data-based prediction model is used to obtain the prediction model. The
identified model exhibits an excellent prediction performance. The inputloutput
data-based predictive controller based on the identified model is designed and applied to
MIMO control problems for the SMB process under the presence of the input and output
constraints. The simulation results demonstrate that the controller proposed in this study
shows an excellent control performance not only for the di'sturbance rejection but also for
the setpoint tracking.

References

1. Wang, C.; Engell, S.; Hanisch. F.; Triennial World Congress, Barcelona, Spain
2002,1145-1 150.
2. Dunnebier, G.; Fricke, J.; Klatt, Karsten-Ulrich.; Ind. Eng. Chem. Res. 2000,39,
2290-2304.
3. Favoreel, W.;De Moor, B.; Van Overschee, P.; Gevers, M., Proceedings of the
American Control Conference 1999,3372-3381.
4. Kloppenbwg, E.; Gills, E.D. Journal ofProcess Control 1999,9,41-50.
5 . K.-Y. Yoo; H.-K. Rhee, AZChE J. 2002,48(9), 1981-1990.
6. Migliorini, G.; Mazzotti, M.; Morbidelli, M., Journal of Chromatography 1998, A,
827, 161-173.
7. Pedefem, M.; Zenoni, G.; Mazzotti, M.; Morbidelli, M., Chem. Eng. Sci., 1999, 54,
3735-3747.
8. Natarajan, S.; J. H. Lee, Computers and Chemical Engineering 2000,24, I 127-1 133.
9. Song, I.-H.; K.-Y. Yoo; H.-K. Rhee; Ind. Eng. Chem. Res. 2001,40,4292-4301.
10. Verhaegen, M.; Dewilde, P. Inr. J. Control 1992,5, 1187-1210.

218
 QUICK AND COMPACT OZONATION USING SILICEOUS ZEOLITE

Hirotaka Fujita*, Akiyoshi S&da*, Taka0 Fujii* and Jun Inmi**

*Institute of Industrial Science, University of Tokyo
4-6-1 Komaba,Meguro-ku, Tokyo, Japan
**Nagasaki R&D Center, Mitubishi Heavy Industries, Ltd.
5-717-1 Fukahori-cho. Nagasaki, Japan
We developed a novel omnation using high silica mIites as an adsorptive concentrator of
ozone and processing organics, resulting in a significant increase in reaction rate. Throgh
TCE degradation test, it was found that the ozone reaction toward TCE was significantly
increased.

Key words: ozone. adsorption, high silica zeolite

Introduction
Drinking water resoufrces are inmasingly contmimted with chlorinated pollutants
such as hichloroethene (ICE), etc. ozonation is an oxidation process extensively applied to
water treatments for eliminating such pollutants. However, single omnation isn't always effective
in terms of time required for technical pcess. To enhance the omnation effectiveness, many
studies have focused on AOP ( a d v d oxidation pmcesm) such as omnation combined with
H&[ Beltran et al., 1998 I, etc, in which OH radicals, a much more reactive specie than omne
itself, play a main role. 'Ihe present study performs a new trial aiming at the enhancement of the
omne mction,designing a novel oxidationprocess in which the omne reaction rateis sigruficantly
increased through the use of an omne &orbent. We found in our pvious work [ Fujita et al,
ux)2] that high silica zeolitespcwxseda m n g abiity to adsorb omne. In addition to that,it has
been reposed that dissolved organics were also highly adsorbed onto these zeolites [ Giaya et al.,
20001. The above findings will raise the possibility that a considerable increase of reaction rate can
be achieved due to the adsorptive enrichment of omne and target Organics inside the zeolites. The
objectives of this work are to investigate the effectivenessof the novel omnation process proposed
and to elucidate the fundamental phenonena throughTCE degradation tests using a tubular flow
mtor..

Materials and methods lsMelIligb~zeolitesUsed

Adsorbents
employed are listed in Table
TCE was selected as a
model substance to be
decomposed and its adsorption property was examined by batch and brealahrough tests. In order to
investigate TCE degradation, the experimental appmlus shown in figure 1 was employed. Ozone
solution was produced by bubbling gaseous omne generated by an ozone generator (POX-10, Fuji
Electric Co., Ltd). A constant concenhation of TCE solution was prepad, miXing hued water
and high comntrationof TCE solution fed by a syringe type microfeeder (KDS-100, Kd Scientific
Inc.,). From different paths, Ozone solution and TCE solution were pumped into the mixing
vessel before the inlet to the fixed bed, using a dual plunger pump without pulsating c m n t

219
(NP-KX-120, Nihon Seinritsu Kagaku Co., Ltd), and this mixed solution was fed into the glass
column. TCE concentrations at the inlet and oudet of die column at a steady state were detected,
and die conversion of TCE, x, at a steady state was
given.
For die determination of TCE, samples
were withdrawn into an aqueous solution phase of
saturated SQs2 tor die destruction of residual
1
ozone, simultaneously extracting TCE in die
aqueous phase into hexane phase, and analyzed on
a gas chromatograph equipped with an electron
capture detector. For die analysis of chloride ion,
samples were mixed with a little amounts of
phenol solution for die destruction of residual
ozone, and analyzed using an ion chromatograph
(PIA-1000, Shimadzu GLC Ltd) equipped with a
Shim-pack IC-A3 column (Shimadzu GLC Ltd)
Fig. 1 Schematic diagram of
and a conductivity detector. The transfer phase was experimental apparatus employed for
the test of TCE deradation (J) Oxygen
an aqueous solution of 8mM p-Hydroxy benzoic tank @Ozonizer(DOzone stock solution
acid (Wako) and 3.2mM Bis (2-hydroxcyediyl) distilled water Mixing vessel Dual
plunger pump without pulsating
iminotris (hydroxymethyl) methane (Wako). The current syringe pump with high
concentration of TCE solution Packed
concentration of aqueous ozone was monitored column with an adsorbent Column
packed with wet activated carbon for the
with an UV spectrometer (UV-1600, Shimadzu degradation of exhausted ozone
Co., Ltd) at 258 run.

Result and Discussion

/. Adsorption Property of TCE
Adsorption isotherms of TCE are presented in Figure 2. The adsorption performances
were strongly influenced by SiQ/AljQ ratio and the pore structure of die zeolites. Amount
adsorbed increases in
loop
the following order
ZSM-5(SiOj/Al2O5ratio:3000)
Mordenite 100 ZSM-5(SiOj/Al2O3ratio:30)
ratio: 10) < A Mordenile(SiOj/Al2O3ratio:90)

Mordenite (SiQ/AlA
10 O Mordenitc(SiOj/Al2O3 ratio: 20)
D MordenitefSiOj/AljOjratio:10)
ratio: 20) < ZSM-5
ratio: 30)
Mordenite 0.1

ratio: 90) OrjP

0.01
3SM-5 (SiCVAlA 0.1 1 10

ratio: 3000). This order Equilibrium concentration [mg/L]

corresponded closely
Fig..2 Adsorption isotherms of TCE

220
with that of ozone [ Fujita et al., 2002 1.
2. Dep&&n of TCE
F d y , the ratio of
Chloride ion to TCE disappeared
was examined. Almost the
Stoichiornehic release of chloride
ion was found to occuc when
using ZSM-5(SiO#d2@ ratio
3ooo) , which pvided the
evidence that TCE was
decomposed in the column and
the steady state was achieved. D
TCE decomposition behaviors in Reaction time [s]
the column with and without the
Fig. 3 TCE degradation with and without siliceouszeoItte
&orbent were examined. As Wow rate:lO mL/mtn. Particle d i a m e k 5 0 0 - 5 9 0 ~ .
AqueousTCE concentratlon:1.27-3.0 mg/L)
shown in figure 3, much fastex
TCE decomposition was observed
in the column packed with ZSM-5 (SiWAl2Q ratio: 3000) in comparison to bulk reaction. The
d e w o n behavim when using different adsorbents were examined. TCE conversion increased
with higher adsorption performance of ozone and 'ICE, which indicates that the incmse of
reaction rate is likely due to the high enrichment of ozone and TCE inside the adsorbent With the
further increase of adsorption perfomance, TCE conversion converged to a certain value,
independent of high silica zeolite species. In other words, a maximum limit of the increase of
reaction rate existed. Apparent TCE demmposition behavior in the packed column would be
expressed in following Equation(l), which was based on simple mass balanceequation :
r,, = u -
d[TCE]
dz
[O, 1" [TCElp
=4 , (1)

For the determination of these unknown parameters, a , B , Kob,, the dependence of

decompsition upon ozone concentration and n=E concentration was examined using ZSM-5
(SiOJAlfi ratio : 3000) which gives the above-mentioned maximum limit. A semi-logarithmic
plot of fICEll4TCE]~vs z was found to be approximately b, which meatls a is
approximately equal to 1.0. As shown in f i p 4, ozone concentration has little effect on TCE
conversion, h m which we can determine pis approximately equal to 0. As a result, Equation(1)
is consequently changed into Equation(2).
x = 1-exp(- k,a,.r) (2)
Intereshgly, this results suggests that n=E conversion, x, was apparently i n f l u e d only by
reaction time r , not by ozone concentration, which differs h m the bulk reaction directly
influenced by ozone concenttation moigne et d, 1983 I. This suggests that there will be
conclusive factors that have a strongly effect on the apparent reaction behavior, independent of the
Charactenstl
' 'csof the adsorbent natu~. Thus,the effect of other factorssuch as particle size (biier

221
@cle size), flow rate
should be investigated in
order to elucidate ihese
phenomena with smallea
@cle size, hi*
conveasions~fwndto
-
be attained. 'Ihe effect of 0.2

the particle size on the

apparent reaction rate is ' 0 2 4 6 8 1 0

likely due to the Reaction time, r [sl

mass-transfer resistance. Fig. 4 El& of ozone concentrationon TCE degradation
Fkhemme,theincrease (Flow rate:loml/min.particle diameter: 5 0 0 - 5 9 0 ~ .
TCE concentratlon: 1.27-3.Omg/L)
of flow rate gave higher
TCE conversion, which
raised the possibility that film transfer resistance between the adsorbent surface and current w i l l
strongly influence the apparent d o n . The limitation of reaction enhancement was likely due to
the mass transfer limitation of tila Taking these into consideration, a kinetic TCE destruction
model was propod as follows. Apparent reaction rate r-hcan be described in Equation(3)
When ~ 1 is equal
, to 0, r-hcan give the maximum reaction rate, r h , , as described in
Equation(4)
r, = - h a , ([TCa - [TCEJ,) (3)
r,- =-k,a,[TCEI (4)
?his equation can be transformed kto the following Equation (5). k can be guessed by the
equation proposed by carbery et al[Caheny et aL, lW].and that proposed by W h n W h n et
al., 1%6]., while a,can be given by Fiquation(6).
x = 1-exp(- k,a,T) (5)

a, =- 3P,
%PP
We assumed that this equation would give the maximumvalue of TCE conversionobserved in the
experimental results when the adsorption performanas for TCE and ozone were high enough.
Close concordances were exjmxsed between the experimental results and the kinetic model. .As a
result, it can be deduced the adsohen& wexe kept almost in a virgin state. Aqueous ozone
mncentration has no effect on TCE conversion under our experimental condition as observed in
the experimental results due to the relatively fast reaction in the pores in comparison to TCE film
diffusion. However,such acase will be very specific. In the case that the substance to be degraded
is more Unreactive or is adwrbed less, or in the case that ozone CoIlCentration is much lower, other
states would occur.

222
CONCLUSION
Remarkable increase in ozone reaction rate toward TCE was observed in the presence
of high silica zeolite in comparison to bulk reaction. TCE conversion increased with the increase in
adsorption of ozone and TCE, which provided the evidence that this significant increase in reaction
rate is likely due to the high enrichment of ozone and TCE inside the adsorbent With further
increase of adsorption, TCE conversion converged to a certain value, which indicated a maximum
limit of reaction improvement exists. The kinetic model provides an accurate description of the
experimental results, which suggested the limit was due to the limitation of mass transfer in film.
This suggested that ZSM-5 (SiO/AlAratio: 3000) were kept almost in virgin state due to the fest
ozone reaction with TCE inside the adsorbent

NOMENCLATURE
w:Linear flow rate[ ms"1 ], L:Column lengthf m], r :Reaction time defined as L/u,[ s ], (TCE]:TCE
Concentration [ molm"3, or mgL"1 ], (TCE]i,:lhfluent TCE Concentration[ molm"3, or mg/L ],
[TCEL^Effluent TCE Concentration[ molm"3, or mg/L ], gLength[ m ], k^: apparent
Reaction rate constant[m3mor1s"1],[O3]:Ozone Concentration [ molm"3, or mg/L ], a,: Surface area
of adsorbent per unit volume[ nfari3 ], kL :Mass transfer coefficient between liquid and solidf ms"1 ],
rah,: Apparent reaction rate[ molL's"1 ], Dm :Difrusion coefficient of TCE in water[ mV ]

Reference
Beltran FJ., Encinar. J. M., Alonso M. A. (1998) A Kinetic Model for Advanced Oxidation
Processes of Aromatic Hydrocarbons in Water Application to Phenanthrene and Nitrobenzene, bid.
Eng. Chem. Res., 37,32-40
Carberry J. J,.(1960) A boundary-layer model of fluid-particle mass transfer in fixed bed. AIChE J.,
6,460-463
Fujita H., Sakoda A., Fujii T., Izumi J., (2002) Adsorption and decomposition of water-dissolved
ozone on high silica zeolite. Wat Res., submitted
Hoigne J., Bader H. Wat Res. (1983) Rate constants of reactions of ozone with organic and
inorganic compounds in water-1 non-dissociating organic compounds. Wat Res. 17,173-183
Wilke C. R., P. Chang. (1955) Correlation of diffusion coefficients in dilute solutioa AIChE J., 1,
264-270
Wilson E. J., Geankoplis CJ. (1966) Liquid mass transfer at very low Reynolds numbers in packed
beds. I&EC Fundamentals, 5(1), 9-14

223
 TIME RESOLVED MULTICOMPONENTSORPTION OF LINEAR AND
BRANCHED ALKANE ISOMERS ON ZEOLITES, USING NIR SPECTROSCOPY

ALEXANDRE F. P. FERREIRA, M. MITTELMEIJER, M. SCHENK, A. BLIEK AND B. SMIT.

University of Amsterdam, Department of Chemical Engineering, Nieuwe Achtergracht 166, 1018
WV Amsterdam, The Netherlad
E-mail: a.ferreira@cience. uva.nl

Snap shot oE

- butane on MFI - iso-butane on MFI

Pressure Swing Adsorption (PSA) unit is a dynamic separation process. In order to create a precise
model of the process and thus an accurate design, it is necessary to have a good knowledge of the
mixtures adsorption behaviour. Consequently, the diffusion rates in the adsorbent particles and the
mixture isotherms are extremely vital data. This article intends to present a new approach to study
the adsorption behaviour of isomer mixtures on zeolites. In a combined simulation and experimental
project we set out to assess the sorption properties of a series of zeolites. The simulations are based
on the configurational-bias Monte Carlo technique. The sorption data are measured in a volumetric
set-up coupled with an online Near Infta-Red (NIR) spectroscopy, to monitor the bulk composition.
Single component isotherms of butane and iso-butane were measured to validate the equipment, and
transient volumetric up-take experiments were also performed to access the adsorption kinetics.

1 Introduction

Branched hydrocarbons are preferred to linear hydrocarbons as ingredients in petrol

because they enhance the fuel octane number. By catalytic isomerisation linear
hydrocarbons are converted into mono and di branched hydrocarbons, and it becomes
necessary to separate the mixture. A variety of zeolites may be used for this purpose,
either on the basis of sorption thermodynamics or on the basis of sorption kinetics. Such
data are relevant to the development of sorption based separation methods, but also they
provide key information regarding the catalytic isomerisation over zeolites themselves.
Zeolites are crystalline materials with a well-defined system of micro-pores. Zeolitic
materials are used in a variety of applications, one of the majors is in the area of
separation processes because of their unique porous properties. The size and structure of
the pores as well as the molecular structure of the adsorbate determines the adsorption
capacity and selectivity [S].
Simulation data on mixture adsorption can be used to screen zeolites as adsorbents,
but experimental data are necessary to validate the simulations and to accurately design
the separation process. The first step of the process design is to obtain such data. However,
the experimental assessment of multi-component adsorption equilibria and kinetics is not
straightforward and is highly time-consuming. As a result, some theories have been
developed that predict adsorption behaviour for a mixture based on the pure component
equilibria [1,3]. The isotherm data have to be correlated before their use in a design model

224
for easier handling. Therefore, the experimental systems have to be measured accurately
over a wide range of pressures and temperatures.
Gravimetric or manometric techniques have been used to establish adsorption data of
gases on zeolites. Both techniques present problems, manometric equipment has an
accumulation of the error; and data obtained by the gravimetric method are influenced by
effects associated with flow patterns, bypassing, and buoyancy. In the mixtures
adsorption behaviour, isomers mixtures have the highest degree of difficulty to study.
Isomers can not be differentiated in standard commercial adsorption equipment. This
problem has been solved in this study by coupling a manometric apparatus with an NIR
spectrometer, which allows us to measure the gas phase composition (in time, if
necessary). In this paper we report this new approach to study the adsorption of mixtures
of butane and iso-butane.

2 ExperimentalSection

In a combined simulation and experimental project we set out to assess the adsorption
properties of a series of zeolites. In the present work the adsorption properties of n-butane
and iso-butane on MFI are being studied. The experimental part consists in the validation
of the molecular simulation model, by confirming its results. The experiments were
performed in a constructed in-house manometric apparatus coupled with a NIR
spectrometer (Perkin Elmer, FT-IR system, GX Spectrum). Figure 1 is a scheme of the
experimental set-up.
Liquidpump

Figure 1. Scheme of the manometric apparatus coupled with the Figure 2. SEM picture of MFI.
NIR spectrometer.
A sample of commercial MFI from ZeoIist was used, with a silicon aluminium ratio
(SYAl) of 100, the crystals do not present a regular shape. On figure 2 Scanning Electron
Micrograph (SEM) of MFI crystals is presented. The sample was calcined for 6h at 873K.
The adsorptive gases used were n-butane with a 99.5% purity, and i-butane with
99.95% purity from Praxair. No further treatment was performed on them before their
admittance into the experimental set-up.
For single component and mixtures calibration curves, the NIR spectra of pure gases
are recorded at several pressures and the data gathered are treated as explained in the
result section.
To measure a point of the single component isotherm, pure gas is admitted to the set-
up with valve 2 closed (figure 1). By closing valve 1 and opening valve 2 the pressure

225
starts to drop, and the final value is measured after equilibrium is reached, and the loading
can be then calculated.
Transient volumetric up-take experiments are performed following similar procedure
of measuring one point of the isotherm, but before opening valve 2, NIR spectra are
recorded in regular intervals. After a short period valve 2 is opened to start the adsorption.
Spectra are recorded until the equilibrium is reached.

3 Results and discussion

Experimental data on single component adsorption isotherms of normal-butane and iso-

butane, on MFI zeolite, at 373K, for a pressure range of OSkPa to 200kPa,were obtained.
1.8 1.o

=-
2?
1.6
1.4
1.2
-
0.8
s0.7
0.9

5
;
1.0
0.8
3 0.6
> 0.4
E
v
-
E 0.6
0.5
0.4
3
0.3
0.2
0.2 0.1
on . 0.0
1 .OE-05 1 .OE-03 1.OE-0 1 1 .OEM 1 1.OE+03 1 .OE-05 1 .OE-03 1 .OE-01 1.OEM1 1 .OEM3
P(kPa) P(kPa)

Figure 3. nButane and i-butane isotherms on MFl at 373K. (- Mol. Simulations, 0 Exp. Results, - - - Exp.
Results ref [ 6 ] )
In the figure 3 we present data on single component adsorption isotherms and
simulation results. Data obtained from the literature [6] are included for comparison. The
increase of the loading observed on the high pressures region can be explained by
capillary condensation in the exterior secondary pore system, in particular between the
crystals. This has been observed also by other authors [5].
Using the presents used manometric set-up coupled with a NIR spectrometer
produced results that are in agreement with literature data, and in agreement with
simulations ( simulations details are provided on ref [4]).
Near Infra-Red spectroscopy is a non-intrusive technique that allows to monitor the
composition of the gas phase (differentiate isomers) and its changes in a time resolved
manner. In the NIR spectra (figure 4) some small differences between iso-butane and n-
butane can be observed. We can also observe that mixtures exhibit a spectral behaviour
that is a linear composition of the pure component spectra. It is necessary to quantify the
spectral differences of the two isomers, so that the composition of the mixtures can be
determined by NIR spectroscopy.
Spectra of pure n-butane and pure i-butane are recorded at several pressures. The
single component calibration curves can be calculated by integrating the spectra in a
certain wave length interval obtaining the total spectral intensity (T.S.I.) for that same
interval (figure 5). After some preliminary studies we conclude that the interval of
[6700,7350]+[8200,8700] cm-' is the best to use for the calibration, since it presents the
lowest degree of non-linearity. We did correct all spectra for baseline drift and water
bands. In the figure 5 we present the single component curves that were obtained.

226
 0.8
h

v
0.6
0

0.4
e
w 0.2
9
0
5000 7000 9000
Wave lengh (cm-1)
Figure 4. n-Butane, i-butane and 50/50% mixture NIR spectra.

2 7 1
n-bu tnae i-butane
-1.5
:
- 1
-!
m
& 0.5 0 Experimental
-Calibration
-Cahbratio n
0
0 60 120 180 240 0 60 120 180 240
p Wa) P (kPa)

Figure 5. Single component calibration curve.

0.6 0.6
h n-butane h

j 0.5 4 0.5
f0.4
s
0.3
>
m 0.2
2 0.1
0
0 60 180 240 0 60 120 180 240
P (kPa)

Figure 6. Multi-componentcalibration curves - NAS value.

To calculate the amount of each component in one mixture we use the Net Analyte
Signal (NAS) theory [2]. In figure 6 we can see the NAS calibration curves for n and i-
butane.
In figure 7 we present the volumetric up-take experiment results. The time-base spectra
for n-butane are recorded. We can see a overlapping of spectra on the higher region, that
correspond to the initial part before opening valve 1, so the pressure is kept constant for
150s. On the lower part we see high number of spectra overlapping, that correspond to the
equilibrium. These two regions can also be seen on the total up-take curve, the line at
OmmoVg for the initial part between 0 and 15Os, and the flat part at 0.96moI/g for the
equilibrium, reached after 1000s.
Figure 8 shows adsorption data for n-butane and i-butane mixture. The total loading is
represented by the grey symbols, the black symbols represent the partial n-butane loading
and the open symbols the partial i-butane loading. We can observe that the separation
between the two isomers increases with the pressure, but we need to measure more points
in the high pressures region.

227
 0.21 1.2 ~

Total Uptake -final pressure 3.90kPa

c?.
I
t = loo to 700s ,
3 0.14
0 v

P 0.07
n
3> 0.4
< 0.2 - .
0.00 0 7
6500 7500 8500 0 500 1000
Wave Lenght (cm-1) time (s)
Figure 7. nButane time-based spectra and total volumetric uptake versus time.

h
1.6 0.6
I -
Fartial Uptake h a 1 p s u r e 10.6kpa 1
f
E 0.8
v
U
0
$ 0.4
p 0.2
>
0.0 0
l.E-03 l.E-01 l.E+01 l.E+03 0 400 time (s) 800 1200
P(kPa)
Figure 8. Mmnm equilibrium data and partial uptake for mixture components versus time.

4 Conclusions

We conclude that with a manometric set-up coupled to a NIR spectrometer it turns out to
be possible to measure equilibrium and kinetic data for single components and mixtures.

5 Acknowledgements

This research was carried out within the project CW/STW 349-5203. The authors thank
the Stichting Technische Wetenschappen for their financial support.

References

1. Krishna R., Diffusion of Binary Mixtures in Zeolites: Molecular Dynamics

Simulations versus Maxwell-StefanTheory. Chem. Phys. Lr. 326 (2000) pp 477-484
2. Lorber A., Faber K., Kowalski R., Net Analyte Signal Calculation in Multivariate
Calibration. Anal. Chem. 69 (1997) pp 1620-1626
3. Ruthven D. M., Past Progress and Future Challenges in Adsorption Research. I n d
Eng. Chem. Res. 39, (2000) pp2 127-2131
4. Schenk M. et al., Sep. of alkane isomers by exploiting entropy effects during adsorp-
tion on silicalite-1: a CBMC simulation study. Lungmuir 17 (2001) pp 1558-1570
5. Stach H., Lohse U., Thamm H. and Schirmer W., Adsorption Equilibria of
Hydrocarbons on Highly Dealuminated Zeolites. Zeolites 6 (1986) pp 74-90
6. Zhu W. et d.,Adsorption of Light Alkanes in Silicalite-1: Reconciliation of Exp.
Data and Mol. Simulations. Phys. Chem. Chen Phys., 2 (2000) pp 1989-1995

228
 PORE SIZE EFFECTS IN THE LIQUID PHASE ADSORPTION OF ALKANES
IN ZEOLITES

JOERI F.M. DENAYER', KURT DE MEYER', JOHAN A. MARTENS' AND GIN0 V. BARON'
'Departmentof Chemical Engineering, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussel,
Belgium. Joeri.deMverO.vub.ac.be Centerfor Surface Science and Catalysis, Katholieke
UniversiteitLeuven, KasteelparkArenberg 23, B-3001 Leuven, Belgium.

The liquid phase adsorption of C5-C22 linear alkanes on ZSM-5 was studied using a batch
adsorption technique. Saturation capacities of the alkanes depend strongly on the chain length. The
CHI packing density in the pores of ZSM-5 increases from pentane to heptane, then decreases steeply
to octane, and subsequently increases gradually to reach a plateau where the theoretically highest
packing density is observed. Binary adsorption experiments show pronounced selectivity effects
between alkanes with different chain length. For most binary mixtures, the longest alkane is adsorbed
preferentially, but for certain binary alkane mixtures, the adsorptionselectivity is inverted.

1 Introduction

The adsorption of the homologous alkane series on zeolites and other adsorbents has been
extensively studied by many research groups. These studies demonstrate the existence of
linear relationships between adsorption enthalpy and entropy, and the carbon number [ 1-
41. Contrarily, there are few studies dealing with the adsorption of alkanes in liquid phase.
This can be explained by the lack of selectivity effects that occur in the adsorption of
alkanes on adsorbents in liquid phase. Indeed, classical stationary phases for HPLC show
no separation of alkane mixtures, as a result of the rather weak interactions between the
molecules and the force field exerted by the surface of the amorphous material.
When the alkane molecules are trapped in the pores of a zeolite, much stronger
energetic interactions occur compared to those on amorphous surfaces. Pulse
chromatographic experiments in liquid phase showed a slight increase of the alkane
retention with carbon number on a column packed with a large pore Y zeolite [5]. When
ZSM-5 was used, a zeolite with much smaller pores than Y, very large differences in
adsorption between short and long alkanes were observed in liquid phase.
Among all zeolites, ZSM-5 is undoubtedly the most studied one, both from the
adsorption and catalysis point of view. Several experimental studies reported the
occurrence of a kink in the pure component adsorption isotherms of certain alkanes in gas
phase on ZSM-5 [6-81, and an inflection point in the adsorption enthalpy and entropy
versus zeolite loading curves [9,10]. Generally, the two-step behavior of these linear and
branched alkanes is interpreted in terms of the ZSM-5 channel geometry. For example, it
has been observed that at low partial pressures, isobutane is adsorbed preferentially in the
channel intersections of ZSM-5, while at higher pressures, these branched molecules are
"pushed" into the channel segments, resulting in a significant loss of entropy and a kink
in the adsorption isotherm [ 1 11. The packing and sitting of alkanes in the pore system of
ZSM-5 has been investigated by computational techniques, such as Configurational-Bias
Monte Carlo simulation [12-14]. These simulations all confirm the importance of entropy
effects in the adsorption of alkanes on ZSM-5.
In the present work, we have investigated the liquid phase adsorption behavior of n-
alkanes on a ZSM-5 zeolite. The saturation capacity of linear C5-C22 alkanes was
determined to study alkane packing effects. Adsorption isotherms of binary mixtures were

229
measured to determine the effect of alkane packing on the competition between short and
long alkanes.

2 Experimental

The H-ZSM-5 zeolite sample used in this study was obtained by deamoniating a NH4-
ZSM-5 from Zeolyst (CBV 8014, SiOZ/A1203 mole ratio of 80) at 673 K in a muffle oven
in presence of air for 48h.
Experimental adsorption isotherms were obtained by means of batch experiments.
Zeolite samples (-I g) were put in 10 ml glass vials and weighed with a balance. After
regeneration overnight at 673 K, the vials were immediately sealed with a cap with
septum in order to avoid water uptake from the air, and were weighed to determine the
regenerated zeolite mass. Mixtures of two linear alkanes or only a pure alkane in solvent
-
were added to the sample (each 20 ml). Iso-octane (99% purity, Acros) was used as a
solvent in the adsorption measurements. Immediately after sealing the cap and weighing
the sample, about 10 ml of the mixture was injected through the septum into the zeolite
containing vials, and another 10 ml was added to a vial without zeolite, to be used as
blank sample. Samples were kept at 277 K, to be sure no compounds could evaporate, and
stirred frequently.
Liquid samples were taken with 2 ml syringes after 24 h and 48 h, to verify if
equilibrium between adsorbed and bulk phase was achieved, and analyzed in a gas
chromatograph with flame ionization detector. For every binary mixture, a calibration line
was obtained by analysis of the blank samples, for which the concentration of the
components is exactly known. For each sample the amount adsorbed (qwhte) at
equilibrium was obtained by calculation of the mass balance.

3 Results and Discussion

Fig. 1 shows the uptake of pure hexane and decane from their mixture with iso-octane on
ZSM-5. The amounts adsorbed in the zeolite remain constant after 20 hours, indicating
that the experiments were performed under equilibrium conditions. In order to verify that
iso-octane can be used as inert solvent, a comparative experiment was performed in
which the binary adsorption isotherm of hexane and decane was determined using iso-
octane and 1,3,5-trimethylbenzene as respective solvents. The same adsorption isotherms
are obtained with both solvent (Fig. 1b), demonstrating that iso-octane does not interfere
with the adsorption of the linear alkanes, and that the relative adsorption of the short and
long n-alkanes is not influenced by the nature of the solvent, given that the solvent is not
able to enter the pore system.
Fig. 2a gives the saturation capacity of the C5-C22 n-alkanes, expressed in number of
molecules adsorbed per unit cell. Fig. 2b gives the total number of carbon atoms per unit
cell. For pentane and hexane, about 7.7 molecules are adsorbed per unit cell of ZSM-5.
This corresponds to 39 C atoms for pentane and 45 C atoms for hexane. Although the
number of adsorbed carbon atoms is the same for heptane as compared to hexane, only
about 6.4 heptane molecules are adsorbed per unit cell. A sudden drop is observed
between heptane and octane: only 4.5 octane molecules are adsorbed per unit cell. From
octane on, the number of C atoms adsorbed per unit cell increases steadily, to reach a
plateau of about 55 C atoms adsorbed unit cell.

230
 0 20 40 m w l W l 2 0 1 4 0
LO 02 04 LB 08 10
U r n lh) xw

Fig 1a: uptake of hexane and decane in Fig 1b: Binary adsorption isotherm of C6
ZSM-5 fiom their mixture with iso-octane and C10. Open symbols: mesitylene
solvent; closed symbols: iso-octane solvent

. ....................... ..........

t i 1............................................................. 1
a 4

e 2

5 6 7 8 9 10 11 12 13 14 15 18 20 P 5 8 7 0 9 10 11 12 13 14 15 18 20 P
C.rbmnumbm Urbnn m b r

Fig 2a: Saturation capacities of n-alkanes on Fig 2b: Saturation capacities of n-alkanes
ZSM-5 (molecules/UC) on ZSM-5 (C-atoms/UC)

Obviously, the saturation capacity depends both on the dimensions of the adsorbing
molecules, and the geometry of the pore system. The pore system of ZSM-5 is constituted
of linear channels, with a fiee pore diameter of 5.6 x 5.3 A, intersecting with sinusoidal
channels, with a free diameter of 5.5 x 5.1 A (see Fig. 5a). The length of the linear
channel segments between two intersections is equal to 4.5 A, while the length of the
sinusoidal channel segments between two intersections equals 6.65 A. Each unit cell
contains 4 intersections, 4 sinusoidal segments and 4 linear segments. An octane molecule
has a length of 11.1 A, exceeding the length of a linear channel segment and an
intersection, but is also longer than a sinusoidal channel segment. As a result, not all
channel segments can be occupied by C8 molecules, explaining the lower saturation
capacity compared to C5-C7. Even with these shorter molecules, no use is made of all the
available space for adsorption. This is shown in Fig. 2b, where the total number of C-
atoms adsorbed per unit cell is plotted as a function of the carbon number. Only from C 13
on, a plateau is reached. When the number of adsorbed molecules per unit cell is
multiplied by their molecular length, a plateau value of about 70 A is obtained. This is
longer than the total length of the intersections, linear and sinusoidal channels (66.2 A),
which seems contradictory at fmt sight. However, it should be considered that the kinetic
diameter of the n-alkanes is 4.3 A, leaving an additional 1.1 A at every intersection for
adsorption. This gives a maximal length available for adsorption of 66.2 A + 4* 1.1 A =
70.6 A, which corresponds very nicely to the plateau value. This 100 % occupancy of the
ZSM-5 pores is only possible if the long molecules are able to bend, and occupy both
linear and sinusoidal channels.

231
 1

0.8

0.6
E
0.4

02

0
0.0 0.2 0.4 0.6 0.0 1.O 0 02 0.4 0.8 0.8 1
XCO XCO

Fig 3a: Binary adsorption isotherm of CB Fig 3b: Selectivity diagram for the
and C 12 on Z S M J competitive adsorption of C8 and C 12
1.4
1

..
ao 02 0.4 0.8 08 10
0.0 0.2 0.4 0.0 0.8 1.0
XQ)
XQ)

Fig 4a: Binary adsorption isotherm of C6 Fig 4b: Selectivity diagram for the
and C8 on ZSM-5 competitive adsorption of C6 and C8

In Fig 3%the binary adsorption isotherms of a CB/C12 mixture is shown. C12 is adsorbed
in a very selective way from the mixture, as can be seen in the selectivity diagram (Fig
3b), in which x and y represent the molar firactions in the liquid and adsorbed phases
respectively. This selectivity for the longer chain can be explained by its higher
interaction with the zeolite.

Fig 5a: Pore structure of ZSM-5 Fig 5b: Packing of C6 and CS molecules
from their mixture in the pores of z5m-5

232
A peculiar behavior is observed with a C6K8 mixture (Fig 4), where, at low C6
concentrations (thus a high C8 concentration), the lightest component is adsorbed
preferentially over the heavier alkane. This adsorption selectivity reversal is again
explained by the packing of the molecules in the pores. Since octane can only adsorb in a
!+action of the channel segments, vacancies remain available in which C6 can adsorb, as
is shown in Fig. Sb, where an example is given how the octane molecules can pack in the
linear channels, leaving open space in the sinusoidal channels for hexane. Selectivity
inversion has been observed with a range of binary mixtures, but these data will be treated
elsewhere.

4 Conclusions

The adsorption of linear alkanes on ZSM-5 is governed by geometric and packing effects,
which result in pronounced selectivity effects, and even in an inversion of the normal
adsorption selectivity. These effects are certainly important with respect to catalytic and
separative applications, and will be studied in further work.

Acknowledgements
This research was financially supported by FWO Vlaanderen (G.0127.99). J. Denayer is
grateful to the F.W.0.-Vlaanderen, for a fellowship as postdoctoral researcher.

5 References

1. Bond, C.G., Keane, M.A., Kral, H., Lercher, J.A., Catal. Rev. - Sci. Eng., 42(3), 323-
383,2000.
2. Ruthven, D.M., Kaul, B.K., Adsorption, 4,269-273, 1998.
3. Eder, F.; Lercher, J.A. J Phys. Chem. B 1997,101, 1273-1278.
4. Hampson, J.A.; Jam, R.V.; Rees, L.V.C., Characterization of porous solids I1
Rodriguez, F. et al, Amsterdam 1991, 509,517
5. Denayer, J.F.M., Bouyermaouen, A., Baron, G.V., Ind. Eng. Chem. Res., 37 (9),
1998,3691.
6. Richards, R.E., Rees, L.V.C., Langmuir, 1987,3,335-340.
7. Zhu, W.,van der Graaf, J.M., van den Broeke, L.J.P., Kapteijn, F., Moulijn, J.A., Ind.
Eng. Chem. Res., 1998,37,1934-1942.
8. Sun,M. S., Talu, 0. Shah, D. B., J. Phys. Chem., 1996, 100(43), 17276-17280.
9. Yang, Y., Rees, L.V.C., Microporous materials, 12, 1997, 117-122.
10. Millot, B., Methivier, A., and Jobic, H., J. Phys. Chem. B, 1998, 102(17), 3210-
3215.
11. Zhu, W.,Kapteijn, F., Mouiijn, J.A., PCCP, 2000,2, 1989-1995.
12. Maginn, E.J., Bell, A.T., Theodorou, D.N., J. Phys. Chem., 1995,99,2057-2079.
13. Smit, B., Siepmann, J.I., J. Phys. Chem, 1994,98, 8442-8452.
14. Krishna, R, Paschek, D., PCCP, 2001,3,453-462.

233
 DETECTION OF FREEZING POINT ELEVATION IN SLIT NANOSPACE BY
ATOMIC FORCE MICROSCOPY

M. MIYAHARA, M. SAKAMOTO,H. KANDA AND K . HIGASHITANI

Department of Chemical Engineering, Kyoto University, Kyoto 606-8501, Japan
E-mail: [email protected]

An experimental trial for finding the freezing point elevation phenomena was conducted, employing
the so-called colloidal-probe Atomic Force Microscopy. The elevated freezing point had been
predicted in the earlier molecular simulation work by the authors, which is thought to be caused by
the attractive potential energy from pore wall. To make up a strongly attractive nanospace, a
carbon microparticle was attached to the top of the cantilever tip, and its interaction force with
cleaved graphite was measured within a liquid cell filled with organic liquid, controlled at a desired
temperature above the bulk freezing point of the liquid. The two surfaces will form a slit-shaped
nanospace because the radius of the particle is far larger than the separation distance concerned.
For two kinds of liquids, freezing behavior has been detected above the bulk freezing point.
Though the extent of the elevation itself was rather small, the finding of the definite existence of the
elevation, not only in the micropores but also in a nanospace with the size of a few nanometers,
would be of much importance in the research field of the phase behavior in nanopores.

1 Introduction

Solid-liquid phase transition (fieezing) in confined space, which is of importance not only
in adsorption but in the nanomaterial fabrication, nanotribology and in the pore size
determination, has recently been explored by several research groups including ours. In
contrast to the long-believed phenomena of freezing point depression in pores[ 11, some
molecular simulation studies have clarified that the freezing point in slit nanospace in
equilibrium with saturated vapor or liquid in bulk can be higher than the bulk, depending
on the potential energy of the confining walls [2,3]. Note that the pore shape other than
slit geometry, and the equilibrium vapor-phase pressure less than the saturated one would
bring depressing effect in freezing point [4,5]. The present understanding of the freezing
in nanospace is roughly reviewed below.
Our study on the first point [2] clarified the following. Depending on the strength of
the attractive potential energy from pore walls, fluid in a slit pore in equilibrium with
saturated vapor showed freezing point elevation as well as depression, and the critical
strength to divide these two cases was the potential energy exerted by the fluids solid state.
The excess attraction relative to the critical one was considered to bring the confined
liquid to a higher-density state that resembles a compressed state, which would result in
the elevated freezing point.
The above result is in accord with other recent studies. Dominguez et al. [6]
examined freezing of LJ fluid in slit pores of purely repulsive and weakly attractive walls,
employing thermodynamic integral technique to determine true equilibrium points. The
freezing points showed significant downward shift, relative to the bulk, in purely repulsive
walls, while the downward shift was much smaller in magnitude for weakly attractive
walls. Further, Radhakrishnan and Gubbins [3] used a different approach of
determination of freezing point in slit pores employing the Landau free energy calculation,
and simple fluid in strongly attractive slit pore was shown to exhibit elevated freezing
points.

234
 Now the research effort goes toward experimental verification of the elevation
phenomena in the simplest geometry, a slit. Our main interest is in the range of a few to
several nanomenters. Some experimental studies have already reported fieezing point
elevation in slit pores [7-91,but the materials used were activated carbon fibers (ACFs),
which have only micropores less than 2nm. In such small pores the first layer adjacent to
the attractive pore wall, which is known to form an ordered phase at a temperature well
above the bulk freezing point, will occupy most of the pore spaces, and the freezing
behavior in the interior of the pore space is difficult to be detected. Further, there may
still remain some controversy if a liquid confmed in a larger nanopore would exhibit
elevation unless an experimental verification is made for such a pore.
The employed technique for this purpose was the so-called colloidal-probe AFM
(Atomic Force Microscopy). The results clearly demonstrate that cyclohexane in the
nanoscale slit space between the carbonaceous solids freezes at a temperature above the
bulk freezing point. Though the extent of the elevation itself might look rather small, we
believe that the finding of the definite existence of the elevation would be of much
importance in the research field of the phase behavior in nanopores.

2 Experimental

An AFM apparatus, PicoSPM manufactured by Molecular Imaging (MI), which is

schematically shown in Fig.1, was used. The temperature of the liquid cell can easily be
controlled from the bottom side of the cell, onto which Peltier cooling device is equipped.
MI guarantees temperature control with 0.1 C accuracy.
A slit-like nanospace can be made up by applying so-called the colloidal probe
A F M technique, which was in its origin developed to measure the force between a solid
surface and a particle of micrometer size. The particle was glued on the cantilever tip,
and this colloidal probe is to be used instead of the usual cantilever. The vertical
scanning of the cantilever controls the distance between the particle surface and a solid
surface, and the usual manner of detecting the bending of cantilever gives the

AFy Piezo synner/Laser unit f Laser f

figure 1. Schm@ic drawing of AFM apparatus and image of the nanospace formed between surfaces.

235
forcedistance relation. If we look at the system with the separation distance of single
nanometers, the space between the two surfaces would be almost slit geometry because the
radius of the particle (order of 10' pn) is far larger than the separation distance.
A graphite plate (HOPG) and carbon particles, which were made from phenolic resin
by pyrolyzing above 2000"C, were used to form a nanospace with carbonaceous surfaces.
The graphite plate was freshly cleaved before measurement by using Scotch tape.
Cyclohexane, and octamethylcyclotetrasiloxane(OMCTS), whose bulk freezing point
were 6.4"C and 18C respectively, were used for the examination of freezing behavior.
The reasons of the choice were: i)aBity to carbonaceous surface, ii)freezing point near
ambient temperature, and iii)almost spherical molecular shape. Methanol, with bulk
6eezing point of -98"C, was also used as a reference.

3 Results and discussion

3. I Ultrahighsensitivity of vdWforce on temperature and its trick

Figure 2 illustrates the force curves between the carbon particle and graphite plate
immersed in cyclohexane at various temperatures above the bulk freezing point. (Note that
the distances shown in the figures are those detrmined from usual mannar of AFM
measurement, and is not the distances between the center of nuclei of surface atoms,
which are often used for simulation results.) As is expected the force at higher
temperature, e.g., at 18.6"C, exhibits typical van der Waals (vdW) force that acts mainly
in the single nanometer range. On the other hand what was NOT expected is its ultrahigh
sensitivity on the temperature. Only about 10 degree C of temperature change brought
multi-fold variation in the force. From theoretical point of view the effect of temperature
on the vdW force would be only from the variation in density of solids and liquid, which
at most would affect only a few percent in this small range of the temperature examined.

I '

0.3 0 17.0"C
c.l
E + ll.0"C
z 0.2
-s
A 10.0"C
E 9.0T
0.1 0 8.0"C

.
3

0-0.1
0

f.4 -0.2

-0.3I I I I I I I
0 10 20 0 10 20 30
Distance [nm] Distance [nm]
Figure 2. Force curves between carbon-graphite Figure 3. Force curves between carbon-graphite
surfaces immersed in cyclohexane for various surfaces immersed in methanol
temperatures above freezing point (6.4"C)
As a matter of fact, the same measurement in methanol gives the results shown in Fig.
3: Quite naturally the force curves do not show any detectable difference against the
temperature variation. We should, then, not take the abnormal temperature dependence

236
as is apparently shown, but should interpret it on the basis of the insensitivity of vdW
force on temperature.
The observed abnormality can be reasonably interpreted if we look at the trick in the
determination of the separation distance in the AFM measurement, which follows below.
Unlike the surface force apparatus, the AFM measurement does not have a direct
method to detect the distance between the surfaces. Instead, one will take the linear
signal of cantilever deflection against sample displacement as the origin of the surface
distance: The linear signal results because the two surfaces are in contact and move
together. This manner of wall detection usually works well. However, what would
result if the freezing point elevation is the case?
Suppose that the liquid between the two surfaces may freeze when the two surfaces
come close to a certain distance, at which the superposition of the potential energy from
each wall exceeds a critical strength that would be needed for the liquid to freeze. The
two surfaces, then, can never go any closer because of the steric repulsion of the frozen
phase (Fig. 4). The system replies with a linear signal as if it reached the situation of real
contact. The observed force curve would then represent a part of the real vdW force cut
out at this distance, taking this point as the origin of the surface distance, A weak
force thus appears.

0.3
CI

-
ul
a
o.2
0.1

.
3
0
8
0
-0.1
0
-0.2

-0.3
0 10 20
Distance [nm]

Figure 4. Schematic illustration of contact Figure 5. Shifted force curves and emerged
with frozen phase between surfaces. walls for cyclohexane.

3.2 Force curves against real separation distance: fieezing point vs. distance
relation
Now we try constructing the real force curves plotted against real separation distance.
The key assumption for this trial would be the insensitivity of the vdW force on the
temperature. The attractive force before the freezing or the apparent contact should stay
almost unchanged. Thus each force curve was slid laterally to exhibit a best fit in the
region of attractive force with long separation distance. The results are shown in Fig. 5.
Each force curve converges into a single curve in longer distance region, which is in
line with the insensitivity of the vdW force on temperature and stands for the validity of
the above assumed interpretation of the apparent phenomena. Then the walls standing
at certain distances for lower temperatures would be a direct reflection of the freezing at a
temperature above the bulk freezing point. Namely, the graphiticlcarbonaceous surfaces

237
in their nature exert strongly attractive potential energy to cyclohexane confined within,
and the superposition of the two potential energies grows with decreasing distance
between them. At the position where the wall stands in Fig. 5 , the potential energy in
the nanospace exceeds the critical value for the confined liquid to freeze. Thus the
distance of the wall gives the pore size needed for the freezing at each elevated
temperature. Quantitative discussion will further be presented in the conference.
We observed similar walls also for the case with OMCTS, though the page
limitation does not allow us to show the results here.
The superposition of the potential energy is typically the case for the micropores.
Note that, however, superposition of potential energy in some larger pores of a few
nanometers would still be sufficient to cause detectable elevation in the freezing point
because the potential energy of carbon surfaces itself shows quite a large value in the unit
of temperature when converted with Boltzmanns constant: It will be more than hundreds
of Kelvins in the vicinity of the surface and can easily be tens of Kelvins even with a
distance of a few nanometers.

4 Conclusion

An experimental trial for finding the freezing point elevation phenomena was conducted,
employing the so-called colloidal-probe Atomic Force Microscopy. A carbonaceous
nanospace with slit geometry was successfully made up by this technique.
The results demonstrated that cyclohexane in the nanoscale slit space between the
carbonaceous solids freezes when the distance comes down to 4 nm, even at 8.4C,which
is above the bulk freezing point of 6.4C. Measurements with octamethyl-
cyclotetrasiloxane, bulk freezing point being 18C, also detected the freezing behavior at
an elevated temperature of 22C. The detected elevation of a few degree C might be felt
as a matter of less significance, but we believe that the finding of the definite existence of
the elevation would be of much importance in the research field of the phase behavior in
nanospaces.

References

1. e.g., Patric W. A. and W. A. Kemper, J. Chem. Phys., 42 (1938) 369; Durn J. A., N.
J. Wilkinson, H. M. Fretwell, M. A. Alam and R. Evans, J. Phys. Cond. Matter, 7
(1995) L713.
2. Miyahara M. and K. E. Gubbins, J. Chem. Phys., 106 (1997) 2865.
3. Radhakrishnan R. and K. E. Gubbins, Mol. Phys., 96 (1999) 1249.
4. Kanda H., M. Miyahara and K. Higashitani, Langmuir, 16, (2000) 8529.
5. Miyahara M., H. Kanda, M. Shibao and K. Higashitani, J. Chem. Phys., 112 (2000)
9909.
6. Dominguez H., M. P. Allen, and R. Evans, Mol. Phys., 96 (1999) 209.
7. Kaneko K., A. Watanabe, T. Iiyama, R. Radhakrishnan and K. E. Gubbins, J. Phys.
Chem. B, 103 (1999) 7061.
8. Watanabe A. and K. Kaneko, Chem. Phys. Lett., 305 (1999) 71.
9. Sliwinska-Bartkowiak M., R. Radhakrishnan and K. E. Gubbins, Mol. Simul., 27
(2001) 323.

238
 MODELJNG OF HIGH-PRESSURE EQUILIBRIUM ADSORPTION OF
SUPERCRITICAL GASES ON ACTIVATED CARBONS. DETERMINATION OF
PORE SIZE DISTRIBUTION USING A COMBINED DlV AND EOS
E. A. USTINOV
St. Petersburg State Technological Institute (Technical University). 26 Moskovsky prospect,
St. Petersburg, 198013 Russia
E-mail: [email protected]

D.D.DO
Department of Chemical Engineering, University of QueensIand,St Lucia, QLD 4072, Australia
E-mail: [email protected]

The density functional theory modified by including the Bender equation of state is applied to the
analysis of high-pressure (up to 50 MPa) adsorption of argon, nitrogen, methane and helium in
activated carbon Norit R1 in temperature range from 298 to 343 K. The approach allows us to
determine the pore size distribution (PSD) and the adsorbent density without recourse to helium
experiments. The PSD obtained with nitrogen at 77.35 K and those obtained with the high-pressure
argon, nitrogen and methane are in fair agreement with each other. However in the latter case the
PSD shows small pores of about 0.3 nm (physical pore width), which do not appear in the 77.35 K-
PSD due to diffusional limitations. It was shown that the adsorption of helium is not negligible at
room temperature and it cannot be used as an inert gas to determine the adsorbent density. The
approach can be easily extended to near critical gases such as carbon dioxide and ethane.

1 Introduction

Adsorption of supercritical gases at high-pressure has some features, one of which is that
the adsorption excess vs pressure exhibits a maximum. There are some simplified theories
involving Langmuir, Langmuir - Freundlich, Toth, Dubinin - Astakhov equations for the
absolute adsorption isotherm and 2-Dequation of state (EOS) to the adsorbed phase.
Some approaches were based on Ono - Kondo equations for lattice gas and the simplified
local density theory that utilizes the van der Waals EOS and the Peng - Robinson EOS
modified for the case of confined slit pores. Brief review of these models was presented
earlier [ 13. There are some investigations of supercritical and near to critical temperature
adsorption of gases on the basis of the density functional theory (DFT) [2,3] and the
grand canonical Monte Carlo (GCMC)simulations [4-71. In both cases the PSD obtained
using high temperature and high-pressure gas adsorption and that determined with
nitrogen at 77 K for the same sample of activated carbon was shown to be markedly
different. A group of small pores of the PSD obtained with near critical carbon dioxide
and supercritical methane is not revealed with the low temperature nitrogen adsorption,
which is attributed to diffusion limitations or even blocking these pores by frozen
nitrogen 161. By this reason the PSDs obtained with C02are considered to be more robust
and reliable than those obtained with N2at 77 K. Thus it might suggest that supercritical
adsorption is a suitable means to derive PSD. But there exists one obstacle to this
approach. Since the capillary condensation does not occur in supercritical adsorption the
dependence of amount adsorbed expressed in terms of excess per unit surface area on
pressure is nearly the same for sufficientlywide pores (wider than 6 collision diameters in
the case of methane adsorption [4]). It does not allow us to reliably determine the PSD in
the region of large pore widths and even the total pore volume. This problem may be

239
overcome by increasing the pressure range for experimental isotherms and specifying the
total pore volume, which could be evaluated from N2 adsorption at 77 K.
In this paper we use a modified DFT approach to the data on Ar, N2, CH4 and He
adsorption on activated carbon Norit R1 at pressures up to 50 MPa and four temperatures:
298, 313, 328 and 343 K. Experiments were carried out at Institute of Non-classical
Chemistry of University of Leipzig, Germany, using a magnetic suspended gravimetric
system (Rubotherm PrtlzisionsmesstechnikGmbH, Bochum, Germany) [11. Our aim was
to develope a method of the PSD analysis from high-pressure adsorption data without
recourse to helium adsorption to determine the skeletal density of the adsorbent.

2 Model

The widely used non-local density functional theory [8-101 is known to involve a number
of assumptions. The Helmholtz free energy of a fluid confined in a pore is divided by two
parts, which account for separately attractive and repulsive forces. The former is usually
calculated by integrating the Weeks - Chandler - Andersen (WCA) fluid - fluid potential
over the confined pore on the basis the mean field approximation (MFA). The latter is
modeled by the equivalent hard sphere fluid using the Camahan - Starling equation of
state (CS EOS). The Helmholtz free energy of the hard sphere fluid is supposed to be a
function of a smoothed density, which is defined as a weighted average according to the
T a w n a prescription [el. The solution for a given pore at a specified temperature and a
bulk phase pressure is obtained by minimizing the grand thermodynamic potential with
respect to the density profile across the pore. Mathematically the problem is reduced to a
system of highly nonlinear equations with respect to the set of densities at different
distances from the pore wall, which can be solved by an iteration scheme. In the sub
critical region one may obtain two or more density profiles depending on initial
conditions. In such a case the true density distribution corresponds to the minimal value
of the grand potential. In the case of supercritical gas adsorption only one density profile
is obtained at each bulk phase pressure. The parameters for the fluid - fluid interaction
(the potential well depth m, the collision diameter aff and the equivalent hard sphere
diameter dHS)are chosen in such a manner that the agreement between the bulk phase
properties (dependence of the saturation pressure and densities of coexisting phases on
temperature) and the surface tension and their respective experimental values is
reasonably well. The solid - fluid parameters are calculated by the empirical Berthelot
mixing rule, which sometimes can be corrected using the data in the Henry law region.
For the solid - fluid potential the 10 - 4 - 3 Steele potential is usually used.
Application of the NLDFT to the case of high-pressure adsorption, being principally
the same as in the case of sub critical adsorption, is however accompanied by an
additional difficulty. The problem is that the bulk phase density is comparable with that of
the adsorbate phase and must be described as a function of pressure very precisely. The
Carnahan - Starling EOS does describe this dependence for Ar, Nz and CH4 reasonably
well but its accuracy is not enough to lead to reliable results, especially in the PSD
analysis, which is known to be very sensitive to small experimental errors and hence to
small changes in the governing equations. Moreover, the bulk phase properties near to
critical gases such as COz and C2& the CS EOS cannot fairly correlate at all. The
question is whether the bulk gas could be described by any more accurate equation than
the CS EOS. In the case of high-pressure adsorption the contribution of the central part of
a sufficiently wide slit pore to the excess must tend to zero due to the solid - fluid
interaction becomes very weak, Obviously, it is possible only in the case when the same

240
EOS is applied to the adsorbed phase and the bulk phase. It suggests the following
scheme of a modification of the NLDFT. The Carnahan - Starling EOS is replaced by the
much more accurate Bender EOS [ 111 and then the corresponding molar Helmholtz free
energy is split into three parts: (i) the ideal component, (ii) the excess fiee energy
associating with only repulsive forces and (iii) the component accounting for only
attractive forces. The latter is assumed to be described by the WCA potential in just the
same way done in the original NLDFT. Then the excess free energy is obtained by
subtracting the ideal part and the attractive part (defined for homogeneous fluid) from the
molar Helmholtz fiee energy corresponding to the Bender EOS. At a specified
temperature the excess free energy is considered to be the single-valued function of the
smoothed density, which is defined according to the Tarazona prescription in the case of
inhomogeneous fluid.
The Bender EOS is given by
p = pRT + Bp2 + Cp' + Dp4 + Ep' + Fp6 + (G + Hp2)p3exp(-a,,p2) (1)
where p and p are the pressure and density, respectively. The constants B, ..., H are
temperature-dependent. Employing this equation, after subtracting the ideal component
and that accounting for fluid - fluid attractive interactions for the molar excess Helmholtz
free energy one can obtain:
F,(p) = Bp + Cp212 + Dp3 /3 + Ep4 / 4 + FpS15
(2)
+(Z/2)[G +ZH-(G +ZH+HpZ)exp(-a,,p2)]+C,p
where Z = l/azo,CO= (2'"16/9)lre&, and NA is the Avogadro number. The term --Cop is
the contribution of attractive interactions in the case of homogeneous fluid. In order to
determine the potential well depth Q and the collision diameter uffthe CS EOS was used
as a first approximation:
p = RTdl +(4<-2G2)/(1 -i)3]-C,oiNAp2, 6 = 7tdiSp/6 (3)
We fitted two parameters in this equation, namely dHSand the product q,u$ to achieve
the best agreement with the Bender EOS in the pressure range up to 300 MPa, which
approximately corresponds to the maximal smoothed density in the pore of I nm width at
50 MPa in the bulk phase. In the case of supercritical fluids it is not possible to evaluate
w and urn separately because we cannot invoke additional information such as the surface
tension. By this reason the hard sphere diameter was taken equal to the collision diameter.
The parameters fitted for 298.15 K are listed in Table I .
TaMe 1. Parameters for the Camahan - Starling equation of state at 298.15 K

Nz Ar CH4 He
dHS, nm 0.34764 0.32124 0.36177 0.21592
OK q&, nm3 K 4.1203 3.7780 6.9560 0.048202
M B 7 K 98.07 113.97 146.91 4.79

3 Discussion

Figure 1 gives a representation on primary experimental data in the form of dependence

of measured weight change on pressure in the case of methane adsorption. Buoyancy
contribution is very high and strongly depends on the activated carbon density, which was
considered as a fitting parameter. The experimental data shown in Figure 1 was treated by

241
the modified NLDFT approach and the PSD was determined using Tikhonov
regularization method. All isotherms at different temperatures were treated
simultaneously (in order to diminish sensitivity of the PSD to experimental errors) using
databases for local isotherms generated for these temperatures. The total pore volume
(0.58 cm3/g)was specified and taken from low temperature nitrogen adsorption. Once the
PSD and the skeletal density had been determined the excess isotherms were plotted.
Figure 2 shows excess argon isotherms at four temperatures.

0 1 0 2 0 3 0 4 0 5 0 0 10 20 30 40 50
Pressure (MPa) Pressure (MPa)

Figure 1. Methane adsorption on Norit R1. Figure 2. Argon isotherms on Norit RI.
Temperature, 'C: ( 0 ) 25; (m) 40; (A)55. Temperature, 'C: ( 0 )25; (m) 40; (A)55; (+) 70.
Solid lines in Figures 1 and 2 are calculated for the same PSD. The fitting is excellent,
which means that having determined the PSD from the analysis of one isotherm one can
predict isotherms of other temperatures with high accuracy.
Figure 3 illustratesthe PSD obtained with Ar, N2 and CH4.
Figure 3. Pore size distribution of Norit RI obtained
with different gases: (-) Ar, (-) N1, ( - x - ) CH+ The
pore width is physical, that is H - A, where A is the
separation of adjacent carbon layers in graphite.

It is seen that there are three groups of

pores centered at 0.3, 0.7 and 1.6 nm. The
PSDs obtained with different gases are
reproduced fairly well. The PSD analysis
of low temperature nitrogen adsorption
confirms the existence of the second and
third pore groups and does not reveal the
1 10
population of small pores. This is most
likely due to the diffusional limitations.
Pore width (nm)
Determination of the PSD with
helium is not possible due to small
amount adsorbed and low curvature of the isotherm. By this reason we used the PSD
determined with argon to calculate helium adsorption by the modified NLDFT approach.
In this case the only fitting parameter was the activated carbon density. Figure 4 depicts
the helium excess versus the bulk phase pressure. One can see that the excess amount is
not negligible at 25'C.

242
 Figure 4. Helium excess isotherm on Norit RI at
25C. (Points) experiment; (line) calculation with
PSD determined with argon.
The contribution of helium adsorption
in the measured weight change is rather
small and at 50 MPa amounts to 4 percents
compared to that of Archimedes buoyancy
force, which makes the description of
helium adsorption isotherm not reliable.
Probably this is the reason of relatively
3 I/
o,o . I . 2oI . 3o . , , 5oI
I 1
large deviations between experimental
data and calculated curve. Anyway
Pressure (MPa) determination of the helium-solid
density is by no means correct at room
temperatures. The skeletal density
determined from adsorption measurements separately for each gas is 2.006,2.050, 2.1 57
and 2.017 g/cm for Ar,N1,CH4 and He, respectively. The mean value 2.058 g/cm3seems
to be very plausible.

4 Conclusion

A new approach based on a combination of the non-local density fbnctional theory and
the Bender equation of state was successfully applied to high-pressure (up to 50 MPa)
argon, nitrogen, methane and helium adsorption. The approach allows reliably
determining pore size distribution and adsorbent density, and most importantly it does not
require helium experiments to determine the skeletal density.

5 Acknowledgements

Support from the Australian Research Council is gratefully acknowledged. The authors
are gratefbl to Dr.Peter Harting for the experimental data.

6 References

1. Ustinov E. A. and Do D. D. J. Colloid Interface Sci. 250 (2002) pp. 49-62.

2. Quirke N. and Tennison S . R. R. Carbon 34 (1996) pp. 1281-1286.
3. Scaife S., Kluson P. and Quirke N. J. Phys. Chem. B. 104 (2000)pp. 3 13-3 18.
4. Gusev V. Yu., OBrian J. A. and Seaton N. A. Langmuir 13 (1997) pp- 2815-2821.
5. Heuchel M., Davies G. M., Buss E. and Seaton N. A. Langmuir 15 (1999) pp. 8695-
8705.
6. Sweatman M. B. and Quirke N. Langmuir 17 (2001) pp. 501 1-5020.
7. Sweatman M. B. and Quirke N. J. Phys. Chem. B 105 (2001) pp. 1403-141 1.
8. Tarazona P., Marconi U. M.B. and Evans R. Mol. Phys. 60 (1987) pp. 573-595.
9. Lastoskie C., Gubbins K. E. and Quirke N. J. Phys. Chem. 97 (1 993) pp. 4786-4796.
10. Olivier J. P.J. Porous Muter. 2 (1995) pp. 9-17.
1 1. Platzer B.; Maurer, G . Fluid Phase Equilibria 84 (1993) pp. 79-1 10.

243
ON THE PECULIARITY OF THE MINIMUM OF N-HEXANE PERMEABILITY
IN ACTIVATED CARBON

JUN-SEOK BAE AND DUONG D. DO

Department of Chemical Engineering, University of Queensland. St. Lucia, Qld 4072, Australia
E-mail: duongdacheque.uq.edu.au

For the tortuous and irregular capillaries of porous media, it has been reported theoretically and
experimentally that a minimum in the permeability of adsorbates at low pressures is not expected to
appear. In our study of n-hexane in activated carbon, however, a minimum was consistently observed
for n-hexane at a relative pressure of about 0.03, while benzene and CCIJ show a monotonically
increasing behavior of the permeability versus pressure. Such an observation suggests that the
existence of the minimum depends on the properties of permeating vapors as well as the porous
medium. In this paper a permeation model is presented to describe the minimum with an introduction
of a collision-reflection factor. Surface diffusion permeability is found to increase sharply at very
low pressure, then decrease modestly with an increase in pressure. As a result, the appearance of a
minimum in permeability was found to be controlled by the interplay between Knudsen diffusion and
surface diffusion for each adsorbate at low pressures.

1 Introduction

In diffusion and flow of adsorbing gases or vapors through capillaries, a minimum has
been reported in a plot of flow rate versus pressure when the mean free path (k) is
comparable to the capillary radius (r) [1,3,8]. For the tortuous and irregular capillaries of
porous media, on the other hand, the minimum is not expected to appear as reported
theoretically and experimentally [3,9,10]. In our permeation study of n-hexane through
activated carbon, however, a minimum in permeability was observed consistently at
relative pressure of about 0.03 [2], while for benzene and carbon tetrachloride no
evidence of the minimum appearance was found. Such an observation suggests that the
existence of the minimum depends on the properties of permeating vapors as well as the
porous medium. it has been known that the diffusion and flow of sub-critical vapors
through activated carbon is controlled by Knudsen diffusion, gaseous viscous flow and
surface diffusion [4,6]. We argue that the interplay of these three mechanisms may
determine the appearance of the minimum in total permeability, and present iqthis paper
a permeation model which is modified from the Do and Dos model [6].This permeation
model introduces a collision-reflection factor, which takes into account the adsorption
effect on Knudsen permeability at low pressures. In the model surface diffusion is treated
in such a way that the mobility of molecules on the surface increases with surface
concentration due to the reduction of energy barrier with the progress of adsorption. This
paper will present a better description of the three diffusion mechanisms and feasible
reasons for the minimum appearance in the total permeability of n-hexane through
activated carbon.

2 Problem Formulation on permeability

For a given pressure gradient across a porous medium, the mass balance equation can be
described as follows, provided that Knudsen diffusion, viscous flow and surface diffusion
are additive to the total flux.

244
where E is the porosity of the particle, P is the intraparticle pressure and C, is the
adsorbed phase concentration. At steady state the terms in the square bracket on the RHS
of Eq. 1 are corresponding to Knudsen diffusion permeability (Bd, viscous flow
permeability (By) and surface d i m i o n permeability (BJ,respectively.
The pore diffusivity (D,), which is assumed to follow Knudsen mechanism, takes Eq.
2 for non-overlapping cylindrical pore of radius r.

L(1
Dk 2r 8RgT 2 - f
D,=-=-zk 3rk - - f (2)
where f is the fraction of molecules undergoing collision to the solid surface over
reflection From the surface, which is usually taken to be unity in most work in the
literature. Since we have found that the factor f decreases with an increase of pressure
(Carman [3] observed similar behavior in the transient period), we propose the following
pressure dependence form for the factorf for strongly adsorbing vapors:
f = f, +(fb -f,>.exP[-a(P@o)I (3)
wherefo and fm are the values at the limit of zero pressure and at high pressure (close to
vapor pressure), respectively, which will be obtained directly from experiments. The
viscous flow parameter (Bo) for cylindrical pore is equal to ?/(8rv). The tortuosity factors,
'5k and r, are for diffusive and viscous flow, respectively, which will be calculated from
the permeability of non-adsorbing gases such as helium.
The surface diffisivity (D,) is known to be a function of adsorbed phase
concentration and is equal to the corrected diffisivity (D*$ multiplied by a
thermodynamic correction factor (8lnP/i?lnC,). Assuming that the driving force for
surface diffision is the gradient of the chemical potential and that the mobility constant is
a function of adsorbed phase concentration, Do and Do [6] have introduced the following
form for the corrected surface dihivity:

where Eo, is the activation energies for surface diffusion at the limit of zero pressure.
Dollmand Do, are the corrected surface diffisivities at the limit of zero pressure and at the
infmite temperature and at the permeation temperature, respectively. As the pressure
increases, adsorption occurs on progressively lower energy sites, resulting in the
reduction of energy barrier for diffusion. This is described by the term (I-&/( l+pC,)) in
Eq. 4. The larger the parameter p is, the faster E, decreases with adsorbed concentration.
When p=O, Eq. 4 reduces to the classical Darken equation [5].

3 Experimental

A differential permeation method was used to determine diffusion kinetics of strongly

adsorbing vapors through an Ajax activated carbon (type 976) (whose physical properties
[7]: particle density of 733 kg/m3, micropore porosity 0.40, macropore porosity 0.3 1 and
mean macropore radius 0.8 pm). An activated carbon pellet was carefully mounted in a
copper block, separating two reservoirs. One reservoir is much larger in volume than the

245
other, and is used as the supply reservoir. Thus the supply reservoir was maintained
practically at constant pressure during the c o m e of diffusion and adsorption.
The flux of each adsorbate through the porous solid was measured with respect to
time. Knowing the receiving reservoir volume (V), the cross-sectional area (A) and the
length (L) of the particle, we can determine the total permeability (BJ from Eq. 5 which
gives us a straight line passing through the origin with a slope of BdA$TNL) in a plot of
I~[(P,-P~O)/(P~-P~)] versus time.

where P1 is the pressure in the supply reservoir, and P2(t) and P; are the pressure in the
receiving reservoir at any t and at PO, respectively. The detailed experimental procedure
and the set-up of permeation measurement can be found elsewhere [2,6].

4 Results and Discussions

The total permeabilities of n-hexane, benzene and carbon tetrachloride through activated
carbon were obtained from Eq. 5 and shown in Figure la. At very low pressures a sharp
increase in total permeability was observed for all adsorbates studied. The adsorbed
concentration of these adsorbates [2] increases drastically over this very low range of
pressure, indicating that surface diffusion must play a significantrole in its contribution to
the total permeability since the contribution of gas phase diffusion is expected to be
relatively small. Interestingly, it can be clearly seen in Figure la that only n-hexane
exhibits a minimum in total permeability at reduced pressure of around 0.03. This
phenomenon was consistently observed at three different temperatures. In this range of
pressure, Knudsen diffusion and surface diffusion are expected to dominate the transport
of molecules since gaseous viscous flow is insignificant.
To elaborate the contribution of surface diffusion to total permeability with our
model, several parameters are required. Firstly, tortuosity factors for Knudsen diffusion
(~k) and viscous flow (T,) were found to be 5.66 and 3.10, respectively by comparing the
total permeability of adsorbates with that of inert gases whose flow is governed by
Knudsen d i h i o n and gaseous viscous flow. Secondly, the collision-reflectionfactorfo at
the limit of zero pressure has to be 1.79 for the theoretical Knudsen diffisivity (Eq. 2) to
agree with experimental data. Thirdly, the equilibrium isotherms of the three adsorbates
were well described by the Toth equation. With these parameters, the mass balance
equation (Eq. 1) was solved for receiving reservoir pressure with respect to time by the
combination of the orthogonal collocation method and a standard integration method. All
parameters in Eqs. 3 and 4 at 303K were obtained by matching the simulated results
against permeation results and listed in Table 1.
Table 1. Optimal parameters for n-hexane, benzene and carbon tetrachloride at 303K

n-Hexane Benzene Carbon tetrachloride

1.79 1.79 1.79
c
3

a
1.03
2.15 lo3
1.22
2.05 x 10'
1.15
3.16~10'
Dopo(m%) 5.86 x 7.36 1 0 - l ~ 1.10 x 1 0 - l ~
p (m3/mol) 2.30 x lo4 1.75 x lo4 2.55 x lo4
Eo,, (J/mol) 46900 50100 50000

246
 One would physically expect that as pressure increases the solid surface may get
smoother due to the filling of small pores and cavities with adsorbed molecules, and as a
result the reflection time of gas phase molecules from the surface may become shorter.
The values off, in Table 1 are close to unity as expected and they are in an increasing
order of n-hexane, carbon tetrachloride and benzene. On the other hand, the parameter a
for n-hexane is much higher than that of the others. Since the parameter a in Eq. 3
represents how fast the Knudsen diffusivity increases with pressure, one would expect a
substantial contribution of the Knudsen diffusion for n-hexane to the total permeability at
very low pressures. Also the parameter j3 is a measure of how fast the activation energy
for surface diffusion decreases with adsorbed concentration. As Table 1 indicates, the
surface diffusion permeabilities of n-hexane and carbon tetrachloride are expected to
increase more sharply than that of benzene.

0.00 OlDp 0.01 0.08 0- 0.10

w w
I Id)
Carbon Taaachkrida

P,.",. [P.]

Figure 1. The permeabilities of strongly adsorbing vapors in activated carbon

Based on the optimal parameters in Table 1, the three permeabilities (BbB, and B,,)
are plotted in Figure 1 (b, c and d for n-hexane, benzene and carbon tetrachloride,
respectively) at 303K.The theoretical permeabilities calculated from 4 s . 1-4 for the
three adsorbates show very good agreement with those obtained from experiments. As
shown in Figure 1, at very low pressures the contribution of surface diffusion to total
permeability is considerable for benzene and carbon tetrachloride while Knudsen
diffusion is more dominant for n-hexane. Additionally it is interesting to note that the
surface diffusion permeability for the adsorbates studied is found to increase drastically at
very low pressures and then decrease moderately with an increase in pressure, which is in
agreement with out earlier theoretical study [4]. The fractional contribution of surface
diffusion to total permeability is the same order as the maximum adsorbed concentration,
confirming the fact that surface diffusivity is a function of adsorbed concentration.

247
Consequently, it can be clearly seen in Figure 1 that the interplay between Knudsen
diffusion and surface diffusion play an essential role in the appearance of a minimum in
total permeability.

5 Conclusions

The Knudsen diffusion, viscous flow and surface diffusion for strongly adsorbing vapors
are well described at low range of pressures in this paper. The collision-reflection factor
for Knudsen diffusion is found to be not constant but exhibit a modest increase with an
increase in pressure. The dependence of the Knudsen diffision for n-hexane on pressure
is stronger than that of the other vapors. Moreover the activation energy for the surface
diffusion of n-hexane exhibits a faster decreasing behavior in comparison with the others.
Conclusively, the reason for the minimum appearance in the total permeability of n-
hexane can be attributed by the interplay between the Knudsen diffusion and surface
diffision.

6 Acknowledgements

Support from the Australian Research Council is gratehlly acknowledged.

References

I. Adzurni H.,Studies on the flow of gaseous mixtures through capillaries: 111. The flow
of gaseous mixtures at medium pressures. Bull. Chem. SOC.Jap. 12 (1937) pp. 292-
305.
2. Bae J . 4 . and Do D. D., Study on diffusion and flow of benzene, n-hexane and C C 4 in
activated carbon by a differential permeation method. Chem. Eng. Sci. in press (2002).
3. Carman P. C., Diffusion and flow of gases and vapours through micropores I. Slip
flow and molecular streaming. Proc.Roy.Soc.(London) A203 (I 950) pp. 55-74.
4. Do H. D. and Do D. D., A new diffusion and flow theory for activated carbon from
low pressure to capillary condensation range. Chemical Engineering Journal 84
(2001) pp. 295-308.
5. Do H. D., Do D. D. and Prasetyo I., On the surface diffision of hydrocarbons in
microporous activated carbon. Chemical Engineering Science 56 (200 1) pp. 435 1-
4368.
6. Do H. D., Do D. D. and Prasetyo I., Surface diffusion and adsorption of hydrocarbons
in activated carbon. AIChE Journal 47 (2001) pp. 2515-2525.
7. Gray P. G. and Do D. D., Dynamics of carbon dioxide sorption on activated-carbon
particles. AZChE J 37 (1991) pp. 1027-34.
8. Knudsen M., Die Gesetze der Molekularstr6mungund der inneren Reibungsstr6mung
der Gase durch R6hren. Annalen der Physik (Leipzig) 28 (1909) pp. 75.
9. Pollard W. G. and Present R. D., On gaseous self-diffusion in long capillary tubes.
Phy~.Rev. 73 (1948) pp. 762-774.
10. Wicke E. and Vollmer W.,Flow of gases through micropores. Chem. Eng. Sci. 1
(1952) pp. 282-291.

248
SIMPLIFIED EXPERIMENTAL METHOD TO ANALYSE INTRA-ACTIVATED
CARBON PARTICLE DIFFUSION BASED ON PARALLEL DIFFUSION
MODEL

Y. MIURA, Y. OTAKE,S.IWASAWA AND E. G. FURUYA

Dep. lnd. Chem.,Meqi Universiy, Tama-Ku, Kawasaki, 214-8571, Japan
E-mail: [email protected]

H. T. CHANG AND N. KHALILI,

Dep. Environ, Chem. Eng., Illinois Inst. Tech., Chicago, 6160-37936, IL, USA
E-mail: [email protected]

The rate of contaminant adsorption onto activated carbon particles is controlled by two parallel
diffusion mechanisms of pore and surface diffusion, which operate in different manners and extents
depending upon adsorption temperature and adsorbate concentration. The present study showed
that two mechanisms are separated successfully using a stepwise linearization technique
incorporated with adsorption diffusion model. Surface and pore diffisivities were obtained based
on kinetic data in two types of adsorben and isothermal data attained from batch bottle technique.
Furthermore, intraparticle difisivities onto activated carbon particles were estimated by traditional
breakthrough curve method and final results were compared with those obtained by more rigorous
stepwise linearization technique.

Introduction

Granular activated carbon (GAC) has been used widely in environmental pollution
control due to its capability to absorb a broad range of organic and inorganic compounds.
To properly design a GAC column, information regarding accurate kinetic data of
intraparticle diffusion rate parameters is required. Intraparticle diffusion is, however, a
complicated process that includes at least two parallel mechanisms: pore diffision and
surface diffusion [7].
In this paper, we will present a procedure to separate pore diffusion and surface
diffusion in GAC adsorption by use of a stepwise linearization. Furthermore, a simplified
method to estimate concentration dependency of apparent diffusivities from breakthrough
curves will be proposed.

Theoretical Considerations

An intraparticle diffusion model including surface and pore diffusion has

been presented by other researchers [2,3,4,61. The mathematical equation for
parallel diffusion model is expressed as

On the other hand, a model for apparent intraparticle diffusion 111 is

(2)

This model was developed for its simplicity to avoid the need to determine

249
surface and pore djffusivity separately.
The relationship between apparent diffusivity (D3,surface diffusivity
(Dh and pore diffusivity (D,)can be derived by comparing (1) and (2).

The limitation of Equation (3) is that the isotherm slope @/& is a h c t i o n of

solution concentration; therefore it may not be a constant over a range of concentration
normally encountered in kinetic tests. It is also noted that D,and D,are a function of solid
concentration q, while D,is independent of adsorbate concentration.
Neretnieks 151 studied the effects of temperature and concentration on
surface diffusion, using a pseudo-linear technique to approximate the slope of
isotherm as depicted in Figure 1. An equilibrium point (c-2 and q=qe2) was
connected to an origin (c=O and q=O)by a straight line in order to approximate
actual slope of the isotherm at ce2. I t is obvious that his pseudo-linear
technique offers poor approximation for highly nonlinear isotherms, which
are often found in industrial applications.

Y
Liquid Concentration, c

Figure 1. Comparison Neretnieks and Stepwise Isotherm Linealization Technique

Using a stepwise linearization technique developed in this study can minimize the
problem associated with the Neretnieks technique. This technique uses a straight-line
section by connecting two isotherm points [(c,,, q,,) and (C,~,qe2)]to approximate the
actual isotherm as presented in equation (4).

The slope obtained by the stepwise method is, therefore, more representative of the actual
slope than that by Neretnieks technique. As seen in Figure 1, the slope (f3) in the
adsorption isotherm curve becomes gentle in high concentration range. Consequently, DS
becomes negligible in comparison with the pore-diffision term Dp/Pppaccording to
Equation (3).

Materials and Experimental Methods

The adsorbent employed in this study was granular activated carbon CAL
(GAC-CAL) manufactured by Toyo Calgon Ltd. The adsorbates used in this study
were pnitrophenol (PNP) and pchlorophenol (PCP), both of which were obtained fiom
WAKO Chemical Co. The concentration of the adsorbate solution was determined by UV

250
absorbance using an WNIS spectrophotometer (Shimazu UV-240). Three types of
adsorbers were used in the present study. For rigorous discussion, Differential Reactor
(DR, shallow bed) and Completely Mixed Batch Reactor (CMBR) were employed. With
aid of curve-fitting procedure, apparent diffusivities (D,)were obtained from uptake
curves in DR and from concentration decay curves in CMBR. Long Column Bed reactor
(LCB), commonly used in industrial application, was also studied to determine the
diffusivitiesby use of Mass Transfer Zone (MTZ)method from breakthrough curves.

Results and Discussions

The correlationsbetween D, and 1@ for PNP and PCP obtained in DR and CMBR are
presented in Figure 2, indicating that the apparent diffisivity is a strong function of the
adsorbate molecule and temperature.

XlW' ~10-7
25 I I , I I , #

20 . :g
DRCMBR
.
i
. . . .I
i
.
!
...i .....-i .......i ....i
.
5,
8

i.
joeg<.O i
* .i* . i .

XI
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

'18 Wm3 1/B Wcm?

(a) PNP - CAL systems (b) PCP - CAL systems
Figure 2. Effective Diffusivity as Function of i@.

Figure 3 also includes the correlation between D, and l/P for PNP in of traditional
long column bed (LCB). It is noted that D, in LCB was determined by use of the MTZ
method.

0.30 I I I I

0.25 -................................................
oi
.-.-.W.i.-.-.- 1-.-.a '.-_-.-. +.-._..
Dflm = 0.235

0.20 _. Dflm 0.22 ( D W M B R ) i

................................................

0.15 .............-...;.--.
_. PNP -
i LCB 0
: DWMBR .---*-

0.10 I I I I
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

'4 Is/cm3 Temperature plcl

Figure 3. Effective Diffusivity method Figure 4. Ratio of Pore Diffusivity to for LCB
Molecular Diffisivity

251
The slopes in Figures 2 and 3 were used in order to calculate the pore diffusivities (D,),
which are independent of adsorbate concentration, by use of Equation (3).
The ratios of Dp to molecular difisivities Dm are plotted against adsorption
temperatures in Figure. 4. The Dflm values are expected to be constant for a particular
adsorbent, irrespective of types of adsorbates and adsorbers over a range of temperatures.
From Equation (3), the value of the surface diffisivity Ds were calculated by
subtracting the pore-diffusivity term Dp/Pppfrom the apparent difisivity D,. The value of
D@, is then plotted against D, for DR and CMBR methods (Figure 5-a) and for LCB
method (Figure 5-b). It can be pointed out from Figure 5-a that that (i) DR and CMBR
methods yield similar estimate of D, and that (ii) intraparticle diffusion of PCP into GAC
is dominated by surface diffusion compared to PNP.

..........
1 !\AA..... ;
....................

-
) ....................
g 0.4 ......,.....
5? I..

n
Len-. i py- OR CUBR ffiB
... ......... i.......
. i ....
. i...:...:..:..:
. . . .....
soa. 0
0.2
.* . . . .. .. .. .. m- .
. . A
a m
AO A
318% 0 : : i i i i i 318%eO
0.0 x10l8 0.0 I I I I l l 1 XlOd
3 4 5 6 8 1 0 M M 3 4 5 6 8 1 0 2 0 M
4 lcn+4 4 ldkl
(a) DR and CMBR Method (b) Traditional LCB Method
Figure 5. Contributionof Surface Diffisivity to Molecular Diffisivity

It is also seen fiom Figure 5-b that the extents of Ds contribution to D, evaluated on
LCB method almost coincides with those for DR and CMBR, implying that LCB method
can be applied for the determination of concentration dependency of D,.

Conclusions

The kinetic adsorption studies in different types of adsorbers were performed with
two phenolic compounds of PNP and PCP on activated carbon. A technique of isotherm
stepwise linearization has been proposed and applied to approximate nonlinear isotherms
for GAC adsorption. The results showed that pore and surface diffisivity are estimated
satisfactorily using this stepwise linearization technique. This study also showed that the
apparent difisivity (D,),which possesses concentration dependence, could be estimated
on LCB by applying the technique in high-adsorption region.

References

1. Bird, R. B., Stewart, W. E., and Lightfoot, E. N. (1960). Transport phenomenon.

John Wiley, New York, N.Y., 542-544.
2. Crittenden, J. C., et al. (1993). Water Res., 27,725.
3. Fleck, R. D., Kirwan, D. K., and Hall, K. R. (1973). Ind. Eng. Chem. Fundam.. 12,

252
 95.
4. Mansour, A. D., von Rosenberg, D. U., and Sylvester, N. D. (1982). AlChE J., 28,
765.
5. Neretnieks, I. (1976). Chem. Engrg. Sci.. 31, 1029.
6. Noll, K. E., Gounaris, V.. and Hou, W. S. (1992). Adsorption technology for air and
water pollution control. Lewis Publishers, Chelsea, Mich., 1 - 20.
7. Weber, W. J. Jr, ed. (1972). Physicochemical processes John Wiley, New York, N.Y.,
199-259.

253
 IN-SITU CHARACTERIZATIONOF ION ADSORPTION
AT BIOMIMETICAIRlWATER INTERFACES

T.Y.KIM,G. S.LEE,ANDD. J . A H "

Department of Chemical and Biological Engineering,
College of Engineering, Korea University,Seoul 136- 701, Korea
E-mail: [email protected]

Biomimetic fatty acid Langmuir monolayers at the aidwater interface were investigated both by an
in-situ near-normal external reflection and by an ex-situ transmission FTIR methods. The
monolayers consisted of reactive fatty acids and nonreactive fatty alcohols to cadmium ions, and the
relative amount of each component was varied. For the monolayers interacting with pure water
showed the composition same for both methods regardless of the amount of reactive acid moieties.
However, when the cadmium ions at a concentration of ImM in the aqueous subphase (pH=5.5)
adsorbed to the monolayers, the in-situ method revealed that ion adsorption was significantly less
than full coverage relative to available reactive moieties in case of the amount of nonreactive ones
exceeding 20%. On the contrary, the ex-situ method showed nearly full coverage of cadmium
adsorption regardless of the monolayers' composition. This discrepancy occurs most probably
because the ex-situ analyses requires the sampled monolayers to be neutralized upon drying by an
additional adsorption of cadmium ions, which does not reflect the state of the monolayer at the
&/water interface. In-situ analyses showed that the adsorption of cadmium ions requiring 2:l
covalent complexation became to be restricted as the nonreactive alcohol molecules neighbored with
the reactive sites.

1 Introduction

Adsorption of ions at surfhces has been one of the classical fields of interest in regard
of catalysis and separation. Among various surfaces, Langmuir monolayers mimic the
structure of cell membranes and hence these biomimetic surfaces are a model system that
enables one to investigate surface phenomena analogous to the membrane surfaces [1 21.
Adsorptivity of various ions to Langmuir monolayers has been known to be quite
different fiom that in the bulk phase after extensive studies done by theoretical models and
empirical techniques involving X P S , neutron scattering, FTIR, etc [3, 41. Most of the
empirical studies were done ex-situ because there exists difficulties in dealing with the
monolayers at the aidwater interface in-situ. Recently, much progress has been reported
especially on the in-situ FTIR spectroscopy for the aidwater interfaces [5,6].
In the present study, we report new results on in-situ near-normal external reflection
FTIR on the monolayers containiig both reactive and inactive moieties to adsorption of
ions. Unlike the conventional ex-situ analysis, it was capable of observing suppressing
effects of mixed monolayer surfaces to ion adsorption.

2 Experimental

Compounds used to form Langmuir monolayers were stearic acid (>99.9%) and
stearyl alcohol (>99.9%) purchased fiom Aldrich. Stearic acid known to bind with
cadmium ions (CdC12, >99.9%, Aldrich) were mixed with stearyl alcohol at varying
compositions fiom 1OO:O to 50:50. Each solution was at the concentration of 1mM in
chloroform (>99%, Fluka).

254
 Each mixed solution was spread using a micro syringe on a KSV minitrough
(Finland) containing deionized (DI) water or 1mM aqueous CdCl2 solution both at pH of
5.5. After evaporating chloroform for 30min., the molecules were compressed to form the
close-packed monolayer at the surface pressure of 2OmN/m. After waiting 3Omin to reach
the equiIibrium, the monolayer was deposited onto a clean CaFz substrate for the ex-situ
FTIR analyses by the skimming method to make sure the sampled film to represent the
original monolayer at the aidwater interface. The sample was blown dry by the nitrogen
gas stream, and then scanned 512 times with a FTIR spectrophometer (Perkin-Elmer
Spectrum GX) equipped with a liquid-nitrogen cooled MCT detector at the transmission
mode with an unpolarized i n h e d beam.
For the in-situ FTIR analyses, mixed solutions were spread by the same fashion onto
a house-made minitrough that fitted into the sample chamber of the FTIR instrument. For
every in-situ experiment, the close-packed monolayer was initially formed on DI water
that was then replaced with lmh4 CdC12 solution by careful circulation of the subphase
using a peristaltic pump. The background spectrum was taken for a bare water surface
when the sample was the monolayer on pure water, and was taken for a close-packed
monolayer on pure water when the sample was the one on the cadmium-containing
subphase. The latter technique enhanced the carboxylate peak relating to ion adsorption.
All the experiments were done at room temperature.

3 Results and Discussion

The ex-situ transmission FTIR results of the mixed monolayers spread on pure water
@H=5.5) are shown in Fig. la. The spectra are adjusted so that methylene stretching
bands (v,CHz, 2918cm-' and vSCH2, 285Ocm-') have the same intensity height. The
carbonyl band (vC=O, 1705-173Ocm~')representing the acidic state of the stearic acid is
not interfered with peaks arising fiom the stearyl alcohol. As the relative amount of the
alcoholic moiety increases the peak area decreases as expected fiom the monolayer
composition, as shown in Fig. lb. It is also noted that the carbonyl band shifts fiom
1705cm" to 173Ocm-', which indicates that lateral hydrogen bonding among carboxylic
head groups becomes to be weaken as the stearyl alcohol molecules are mixed and
inserted among the stearic acid molecules. On another set of experiments of surface
isotherms (of which results not shown), the surface Gibbs fiee energy implies the two
components to be well-mixed at the surface pressure of 2OmN/m at room temperature.

255
 S Fnction of steuyl llcohol b 8 c t i v e molecules)
0 i o m a o ~ 0 ~ 8 0 m m o o t m
Tbeoretic.1 c.1cUhtion
0 vc-0

0.2 -
iw m 80 70 80 50 H) so m 10 o
0 Fraction of siearic acid (reactive molecules)

(a) (b)
Figure 1. Transmission FTIR results on the mixed monolayers spread on pure water with varying monolayer
composition (stearic acid : strearyl alcohol) (a) and the peak area of the carbonyl band at various composition
(b). The peak areas were normalized with respect to that of pure straric acid case (1OO:O).

The ex-situ transmission results on the mixed monolayers spread on the subphase
containing 1mM CdClz (pH=5.5) are shown in Fig. 2a. At this high concentration of
cadmium ions, every acidic molecule is bound with cadmium ions as indicated by the
complete disappearance of the carbonyl band and the existence of the asymmetric
carboxylate band (v,COO, 1542cm-') at the same time. In general, the peak intensity of
the asymmetric carboxylate band decreases as the relative amount of the reactive stearic
acid decreases, as observed in the case of Fig. 1.
The in-situ near-normal external reflection results on adsorption of cadmium ions to
mixed monolayers are given in Fig. 2b. Since the background spectra are taken for the
close-packed mixed monolayers on pure water at each composition, the methylene
stretching bands are nearly cancelled out. By contrast, the asymmetric carboxylate band
appears with reasonable S/N ratios even without a polarizer-modulator set-up [7].It is
noted that the near-normal incidence of the i n h e d beam resulted in negative signals on
the contrary to that of the grazing-angle incidence.
It is interesting in the in-situ investigation (Fig. 2b) that as the relative amount of the
inactive stearyl alcohols to cadmium ion exceeds 20% the asymmetric carboxylate band
suddenly decreases in its intensity and remains about the same afterwards. In Fig. 2c, it is
clearly seen that the adsorbed amount of cadmium ions at 80% of stearic acid is reduced
by more than half of those at larger compositions of the acid. By contrast, the ex-situ
results showed a linear decrease in cadmium adsorption with the monolayer composition
as well as a certain deviation fiom the theoretical calculation that is different to the case
without adsorption of cadmium ions.
 (c)
Figure 2. Ex-situ transmission FTDR results (a), in-situ near-normal external reflection FTIR results (b) on the
mixed monolayers spread on aqueous cadmium solution ( I d , pH4.5) with varying monolayer composition
(stearic acid : streayl alcohol), and the peak area of the asymmetric carboxylate band at various composition (c).
The peak areas were normalized with respect to that of pure strearic acid case (1OO:O).

It is strongly suggested that introduction of more than 20% of inactive alcohol

molecules into acid monolayers hinders significantly the 2: 1 covalent complexation
among cadmium ions and carboxylic head groups. When the monolayer is sampled on the
solid substrate and dried, additional neutralization of carboxylic groups with cadmium
ions are enforced, which results in higher amount of adsorption complexation that is yet
less than the complete adsorption coverage as calculated.

4 Conclusions

We investigated the adsorption of cadmium ions on mixed monolayers at the aidwater

interface using both ex-situ and in-situ FTIR spectroscopic methods.
The ex-situ investigation with the sampled monolayers did not always reflect the
original monolayers at the aidwater interface in analyzing adsorption of cadmium ions. It
was newly observed by the in-situ method that the adsorptivity of cadmium ions to acidic
monolayers could be abruptly reduced when the inactive alcohols were mixed into at more
than 20% in composition.
The present results offer usehl information on further characterization of binding of
biologically active ions to cell membranes and also on designing sensors and catalysts at

257
the molecular level.

5 Acknowledgements

This work was supported financially by the Korea Science and Engineering Foundation.

References

1. Vijendra K.A.,Langmuir-Blodgettfilms, Phys. Today (1988) pp. 40-46.

2. Swalen J.D., Allara K.L., Andrade E.A., Chandross E. A., Garoff S., Israelachvili J.,
McCarthy T.J., Murray R., Pease R.F. and Rabolt J.F., Molecular Monolayers and
Films, Langmuir 3 (1987) pp. 932-950.
3. Ahn D.J. and Franses E.I., Ion adsorption and ion exchange in ultrathin films of fatty
acids, AIChE 40 (1994) pp. 1046-1054.
4. Kim T.Y., Lee G.S. and Ahn D.J., Ion separation of binary aqueous solutions at
acidic Langmuir monolayer surfaces, Korean J. Chem. Eng. 18 (2001) pp. 977-985.
5. Geriche A., Huhnerfm H., In situ investigation of saturated long-chain fatty acids at
the adwater interface by external infrared reflection-absorption spectrometry,J. Phys.
Chem. 97 (1993) pp. 12899-12908.
6. Ren Y., Hossain M.M., Iimura K.I. and Kato T., CH3(CH2)COOH/Cd2+system on the
aqueous cadmium acetate solution investigated in situ by polarization modulation
infiared spectroscopy,J. Ply. Chem. B 105 (2001) pp. 7723-7729.
7. Blaudez D., Buffeteau T., Cornut J.C., Desbat B., Escafre N., Pezolet M. and Turlet
J.M., Polarization-modulated FT-IR spectroscopy of a spread monolayer at the
aidwater Interface, Appi. Spectroscopy 47 ( 1993) pp. 869-874.

258
SINGLE AND MULTI COMPONENT ADSORPTION OF VOLATILE ORGANIC
COMPOUNDS ONTO HIGH SILICA ZEOLITES DISCUSSION OF ADSORBED -
SOLUTION THEORY

P. MONNEYRON, M.-H. MANERO AND J.-N. FOUSSARD

LIPE, Dpt GPI, Institut National des Sciences Appliquees, 135 avenue de Rangueil
31077 Toulouse. France
E-mail: [email protected],manero@ch. iut-tlse3.jL

Pure and binary component adsorption isotherms were determined for three volatile organic
compounds daily used as industrial solvents (methyl ethyl ketone (MEK), toluene (TOL),and 1,4-
dioxan (DIO))on two high silica zeolites, a desaluminated faujasite Y (Fau Y) and a silidite (Sil Z).
Apart from steric exclusion taking place with TOL-containing mixtures on Sil Z, adsorption
occurred via three phenomena At low filling rates, chemisorption on specific sites happened in
favor of polar or major compound, whereas in the high pressure saturation domain the adsorbent was
selective for the minor compound. In between, micropores-filling had a similar behaviour to
distillation since the component with the lower volatility adsorbed preferentially. Secondly, binary
equilibria predictions were performed using the IAST. On Sil Z, the difference between Langmuir
and Toth model-based calculations of isobaric selective diagrams was small but significant in the
MEK-DIO mixture at low pressures. On Fau Y, IAST calculations lead to an excellent agreement
with experimental data when the adsorbed phase was closed to ideality. RAST correlations
efficiency of fitting was good in some cases. For engineering purposes, Fau Y is to be considered as
a high adsorption capacity adsorbent, whose selectivity can be described qualitatively by the
distillation analogy, and predicted quantitatively with the IAST in case of quasi-ideal mixtures.

1 Introduction

Volatile Organic Compounds (VOC) are widely found as solvents in industrial processes
and domestic use. Considered as major air contaminants for inducing directly health
troubles, or for being precursors of tropospheric ozone, polluted air streams are mainly
treated by adsorption on a porous solid [6].
Activated carbon is commonly acknowledgedto be very efficient as a VOC adsorbent.
However, in the last decade, interests arose to develop new materials, such as alumni-
silicate molecular sieve, towards operating conditions for which activated carbon was
inappropriate due to its inflammability and adsorption capacity dependence on effluent
relative humidity. Apart from hydrophobicity, the advantages of High Silica Zeolites
(HSZ)are notably a thermal and chemical stability, a high steric selectivity and a
complete regeneration at low temperatures [ 1 11. Yet, for adsorption and separation
processes development, organic compounds properties impact on adsorption and
crystalline framework influence on selectivity are to be clarified and efficiently modeled.
This study firstly aims at understanding adsorption properties of two HSZ towards
three VOC (methyl ethyl ketone, toluene, and IP-dioxane), through single and binary
adsorption equilibrium experiments. Secondly, the Ideal Adsorbed Solution Theory
(IAST) established by Myers and Prausnitz [lo], is applied to predict adsorption
behaviour of binary systems on quasi homogeneous adsorbents, regarding the pure
component isothenns fitting models [ 5 ] . Finally, extension of adsorbed phase to real
behaviour is investigated [4].

259
2 Experimental section

In this study, two commercial HSZ,were used in pellets form, a desaluminated faujasite Y
and a silicalite. The supplier kept undisclosed the chemical nature of the clay binder
together with details of desalumination technique. The main characteristics of adsorbents
are given in table 1.
Table 1. Main characteristics of zeolites.

Type (Symbol) Faujasite (Fau Y) ZSM-5(Sil Z)

Crystalline framework a-cages Interconnected channels
Pore internal diameter (A) 13 (5.7 * 5.1) and 5.4
Si02/Al2O3(mol mol-') 360 1880
Active porous volume (cm3g-I) 0.24 0.18

The three volatile organic compounds studied (99%), having very different chemical
structure as summarized in table 2, were all used without further purification.
Table 2. Main properties of adsorbates,

2-butanone methyl 1,4-

property (methyl ethyl benzene dioxacyclohexane
ketone) (toluene) (p-dioxan)
Symbol MEK TOL DIO
Formula C4H8O C7H8 C4H802
Molar volume 25C (cm3mol-I) 90.14 107.51 85.79
Vapour pressure at 25OC @Pa) 12.60 3.79 4.95
Dipolar moment (D) 2.78 0.375 0
Kinetic diameter (A) 5.2 5.8 5.3

Batch experiments were performed at 298 K via a standard volumetric method, using
a 1.1 L glass reactor at atmospheric pressure, containing typically 0.5 g of adsorbent,
polluted by liquid VOC injection, leading to an initial concentration of about 0.5 mmol.L-'.
Equilibrium times were 1 and 2 hours respectively for Fau Y and Sil Z, after which the gas
phase was sampled and analysed by chromatography (HP 5890 11). Using the conventional
assumption that the clay binder does not take part in the adsorption mechanism, data were
reported for pure zeolite material. Reproducibility and repeatability of experimental data
were checked by three different manipulators.
In case of binary systems, co-adsorption was performed using three different initial
-
mixtures being equimolar (Z, = Zt= 0.5) and rich in component 1 and 2 (respectively 2,
= 0.85 and Z, = 0.15) - whose molar ratio were formerly checked. The selectivity of
adsorbent (qlQ)was either directly represented in an X-Y isobaric equilibrium diagram, or
determined using the following equation, where Xi and 5 are the adsorbed and gas phase
molar tiaction of thcomponent, and plotted versus total amount adsorbed (&).

Xl-Y2
91/2 =-
x2 -yl

260
3 Qualitative discussion of adsorption mechanism via equilibria experiments

3.1 Pure component isotherms

Adsorption isotherms were determined in the low partial relative pressure range, &/po 5
0.1). For all compounds, the isotherm curves obtained on Sil Z and Fau Y were
respectively of type I and V of IUPAC classification. Using the assumption that the
adsorbed phase was in the liquid state, micropore fillings were calculated, highlighting the
steric exclusion of TOL on silicalite (52% for TOL versus 87 % for MEK and DIO). This
maximum adsorption capacity of 4 molecules per unit cell, adsorbed at channels
intersections, was commonly observed on pentasil zeolites by other researchers for
aromatics [7], and chlorinated compounds [2]. Contrary to the second step occurring in
the isotherms that they observed, no significant upload in adsorbed quantity of TOL on Sil
Z was found here, as the partial pressure increased up to &bo= 0.35). This difference
was attributed to the initial adsorbent being either directly synthesized silicalite or ZSM-5
with a lower SQAI ratio, or in our case a HSZ where the desalumination step induced a
small decrease in pores dimension. As expected, no exclusion phenomenon was observed
on the large pore faujasite: the difference of pore filling being much smaller (respectively
72, 73, and 82% for TOL, DIO, and MEK) was attributed to differences in molecules
flexibility towards maximum space occupation. These results were consistent with
previous studies carried out on similar systems [8,12].

3.2 Experimental selectivities of HSZ toward binary mixtures

Apart from specific adsorption sites - being generally framework aluminum in partially
-
desaluminated hydrophobic zeolites inducing some polar affinity [3], selective co-
adsorption on HSZ was reported to depend on adsorbent crystalline framework,
component volatility, initial molar fraction of the gas phase and micropore filling [5,13].
However, considering the different desalumination techniques, various amounts of Extra
Framework Aluminum (EFAL) and silanol group (SiOH) sites can be formed, and may
induce a change of selectivity at low loading [8]. As well, the clay binder can exhibit some
specific adsorption sites fiom its framework hydroxyl groups however small its specific
surface area may be [I]. This co-influence of chemisorption reported previously for an
aqueous system [9], was noticed on Fau Y for {aa, < 0.5 mmol g-1, leading to a selective
adsorption of component (1) in TOL-containing mixture (fig l.b), and of the major
compound in MEK-DIO binary system.
On both zeolites, qualitative identical adsorption behaviour were observed between
MEK-TOL and DJO-TOL mixtures. On Sil Z (fig. l.a), a low selectivity for TOL
occurred du to the great difference in active volume. With increasing loading, this
tendency increased naturally. However it is strongly suspected that a competitive
adsorption on TOL adsorption sites remained hidden behind this steric exclusion
selectivity, as confirmed by the higher selectivity for MEK predicted by the IAST. Indeed,
in the MEK-DIO mixture at high loading, the Sil Z was found to be selective for the less
volatile compound (DIO). In the same way, this distillation analogy was found for Fau
Y in the micropore filling pressure range, (fig. 1 b). In addition to volatility influence, in
the high loading range, both HSZ show higher selectivities towards minor compound,
inducing for the {Z,= O.I5)-TOL containing mixtures a competitive influence of these
two parameters.

261
 45 7 5 7
a) z 1 = 0.15 Z1 = 0.15
40 -
35 - 0 z 1 = 0.50 0 z 1 = 0.50
h
x Z 1 =0.85 x Z1 = 0.85
E30 -
4 *.
. .
25 -
1 2 0 -
H
**

P 5 % o m oroclKrq!ax
I I 1

0 0.5 1 1.5 2 0 1 2 3
Q t o t (mmol.g-') Q t o t (mmol.g-')

Figure 1. Selectivity towards h4EK-TOL mixture, on Sil Z (a) and Fau Y (b).

4. Binary equilibria modeling with Adsorbed Solution Theory

4.1 IAST prediction calculations

Since IAST-based calculationsrequire a precise description of pure component adsorption
data, especially in the low coverage range, Freundlich, Langmuir, Toth and Dubinin-
Astakhov models were compared.
Had no steric exclusion happen, HSZ are regarded as homogeneous adsorbents that
are to have a relatively small influence on selectivity. Agreement between IAST
calculations and experimental data was very good for the quasi-ideal MEK-DIO mixture
on Fau Y,using Freundlich model on three segments isotherms, as shown in figure 2.a,
whereas unexpectedly the IAST did not predict the selective adsorption of DIO on Sil Z.
Moreover, at very low pressure, as can be seen in figure 2.b, the quasi-ideal MEK-DIO
mixture show an azeotxopic behaviour on Sil Z, and in this case the difference between
Langrnuir and Toth models-based IAST calculations was significant.
1
+Ex)
_ _ _ IAST
._._ Langrnuir
0.8 0.8

0.6 0.6
XMEK (P= 30 Pa)
0.4 0.4

0.2 0.2
YMEK
0 I

0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Figure 2. Isobaric selective X-Y diagrams towards MEK-DIOmixture, on Fau Y (a) and Sil Z (b).

262
4.2 RAST correlation calculations
RAST was applied using Wilson model for activity coefficients and, as shown on figure
2.b, it successfully predict the azeotropic behaviour for the system considered. However
in care of TOL-containingmixtures, experimentallydetermined activity coefficients could
not be fitted over the whole adsorbed phase composition range with only a two adjustable
parameters model.

References

1. Alpay E., Haq N., Kershenbaum L.S. and Kirkby N.F., Adsorption parameters for
strongly adsorbed hydrocarbon vapours on some commercial adsorbents, Gas
Separation and Purijkation 10 (1996) pp.25-33.
2. Bouvier F. and Weber G., Adsorption of a polar or a non-polar chloroalkene on a
ZSM-5 zeolite at 298 K, Journal of Thermal Analysis 54 (1998) pp. 881-889.
3. Brosillon S., Manero M.-H. and Foussard J.-N., Adsorption of acetoneheptane
gaseous mixtures on zeolite: co-adsorption equilibria and selectivities, Environmental
Technology 21 (2000) pp. 457-465.
4. Costa E., Sotelo J.L. Calleja G. and Marron C., Adsorption of binary and ternary
hydrocarbon gas mixtures on activated carbon: experimental determination and
theoretical prediction of the ternary equilibrium data, AIChE Journal 27(1) (1981)
pp.5-12.
5. Garrot B., Couderc G., Simonot-Grange M.-H., and Stoeckli F., Co-adsorption of
1,2-dichloroethaneand 1-bromo,2-chloroethaneon zeolite ZSM-5 from the liquid and
vapour phases, using the Myers-Prausnitz-Dubinin model, Microporous and
Mesoporous Materials 52 (2002) pp. 199-206.
6. Le Cloirec P. Les composes organiques volatiles (COV) dans Ienvironnement
(Lavoisier Tec&Doc, Pars, 1998)
7. Li J.-M. and Talu O., Effect of structural heterogeneity on multicomponent
adsorption: benzene and p-xylene mixture on silicalite, in: M. Suzuki (ed.) Proc.Ivh
Int. Con$ on Fundamentals of Adrorption, (Elsevier, Amsterdam, 1993) pp. 373-380.
8. Meininghaus C. K.W. and Prins R., Sorption of volatile organic compounds on
hydrophobic zeolites, Microporous and Mesoporous Materials 35-36(2000) pp. 349-
365.
9. Monneyron P., Faur-Brasquet C., Sakoda A., Suzuki M., and Le Cloirec P.,
Competitive adsorption of organic micropollutants in the aqueous phase onto
activated carbon cloth: comparison of the IAS model and the neural networks in
modeling data, Langmuir 18 (2002) pp. 5 163-5169.
10. Myers A.L. and Prausnitz J.M., Thermodynamics of mixed-gas adsorption, AlChE
Journal 11(1) (1965) pp.121-127.
11. Otten W., Gail E. and Frey T., Einsatzmoglichkeiten hydrophober zeolithe in der
adsorptionstechnik, Chem.-Znd-Tech.64(10) (1992) pp. 915-925.
12. Rochester C.H. and Strachan A., The adsorption of dioxan by porous silicas, Journal
of Colloid and Interface Science 177 (1996) pp. 456-462.
13. Takeuchi Y., lwsrmoto H.,Miyata N., Asmo S., and Harada M., Adsorption of 1-
butanol and p-xylene vapor and their mixtures with high silica zeolites, Separations
Technology5 (1 995) pp. 23-34.

263
 INFLUENCE OF VOCs MOLECULAR CHARACTERISTICSON
EXOTHERMICITY OF ADSORPTION ONTO ACTIVATED CARBON

P.PRE, C. FAUR-BRASQUET AND P. LE CLOIREC

Ecole des Mines de Nantes. GEPEA, UMR-CNRS 6144

4, rue Arfrd Kastler, BP 20722, 44307 NANTES Cedex 03, France

E-mail: P. Preaemn.fr

The objective of the study was to point out the influence of the molecular properties of volatile
organic compounds (VOCs)on the intensity of the energetic interactions with activated carbon. The
integral adsorption enthalpies of 44 organic species were first measured onto one type of granular
activated carbon. Experimental data showed that depending on the nature of the VOC, the
adsorption enthalpy may vary from 40 to 80 kJ.moF'. To account for the influence of the molecular
properties on the variations observed, quantitative structure propem relationships (QSPRs) were
investigated. QSPRs were set up through different statistical approaches which enabled to
discriminate the molecular characteristics which have a significant influence on the adsorption
energy. Physical data representative of both dimensional and electronic properties of the organic
molecules were retained to form the input variable set. As a simplest tool, a multiple linear
regression was first investigated. The best linear regression obtained involved 3 explicative
variables :the ionization potential, the polarisability and the molar mass. The linear model permits to
compute the adsorption energies by less than 15% in error. In a second approach, a non linear model
was also attempted, using neural networks. According to the size of the experimental database, the
number of neurons in the input layer was restricted to 3. The best neural network was selected after
training was achieved with 56 input variable triplets. It is noticeable that the same molecular
properties previously involved in the linear regression, also merge in the best neural network. The
predictive ability of the neural network is then proved to be comparable to that of the linear
regression.

1 Introduction

Solvent recovery from air by adsorption onto granular activated carbon (GAC) is an
highly exothermic process which may bring about ignition risks on industrial fixed bed
units. Indeed, the saturation of the bed being progressive from the inlet to the outlet, at
each time, the adsorption spreads over a small area. Due to heat accumulating, hot spots
then locally arise. When the temperature becomes sufficiently high to start the oxidation
reactions of the adsorbent or the organic compound, chain mechanisms set up, leading to
ignition. Thus, to insure the efficiency and the safety of the process, prediction of the
temperature maximum attained within the bed is required.
Our previous work mainly paid attention to experimental characterization and
modeling of the thermal behavior of a GAC bed during adsorption of VOCs [ 1-31. One of
our conclusions was that the integral adsorption enthalpy, which is a data depending both
on the nature of the VOC and on the type of activated carbon material employed, has to be
primarily determined in order to accurately predict temperature rises with operating
conditions. Figure 1, which reports the temperature rises observed during experimental
tests conducted with various pollutants at different inlet concentration levels, demonstrates
the strong influence of this data.
The aim of the present study was to quantify the changes in the integral adsorption
enthalpies induced by the nature of the organic adsorbate. For that purpose, energy
measurements were achieved for a wide panel of VOCs, representative of different

264
chemical groups, using one type of carbon adsorbent. A database was thus generated,
which was appropriate for fiuther statistical analysis.
120

- 100

g 80
c.
a 60

1
E
40

I 20
3
O O
0 20 40 60 80 100 120
- AH x C a (kJ.m")

Figure 1. Maximal temperature rises observed during dynamic adsorption tests of 15 VOCs onto one GAC bed,
the inlet concentration C M varying from 20 to 100 g.m".
Quantitative structure property relationships (QSPRs) were thus developed using
different statistical models : multiple linear regressions on one hand, and neural networks
on the other hand. The reliability of each of these models was determined from their
predictive ability.

2 Experimental Section .

2.1 Materials and Methods

One type of GAC (NC60) was mainly employed to set up the adsorption enthalpy
database, with a panel of 44 VOCs. For a few compounds, adsorption energies were also
measured on other GACs. The magnitude of the changes observed were then considered
as indicative of the influence of the characteristics of the GAC. The characteristics of the
GACs employed are listed in Table 1.

Table 1. Main characteristics of the GACs tested

Manufacturer Pica Chemviron Norit

Name NC60 NClOO BC120 BPL RB2
Raw material coconut coconut wood coal peat
Specific surface area
(m2.g-') 1063 1783 1675 1043 1013

A differential scanning calorimetry coupled to a thermogravimetric analysis (model

SETARAM TG-DSC 111) was used to measure the integral adsorption enthalpy. The
adsorption energy was determined at 20C, following the sample mass variations and heat
released until the complete saturation of the adsorbent. A gaseous mixture, composed of
helium loaded with the VOC at a concentration of 50 g.m-3, was continuously flowing
through the cell. The experimental error was shown to be less than 5 %.

265
2.2 Adsorption enthalpies data
The whole experimental data are presented in tables 2 and 3. Adsorption energies range
from 40 to 82 kJ.mol%nd are 1.5 to 2 times higher than condensation heats.

Table 2. Experimental adsorption energy database for NC60

VOC Adsorption heat VOC Adsorption heat
(k~.mo~-') (W.mol-')
Acetaldehyde 48.5 Ethyl alcohol 48.6
Ethyl acetate 61.2 isoiropylether 62.2
Iso-propyl acetate 68.5 Formaldehyde 42.3
Acetone 50.6 Ethyl-formate 51.4
Ethyl acrylate 16.2 Heptane 12.9
2-Bromopropane 54.2 Hexane 61.4
Butylamine 14.9 1-Hexene 63.0
Butyraldkhyde 53.2 1-Hexyne 56.9
Chloroform 50.5 Isobutylvinylether 64.1
2-Chloro-propane 49.4 Methyl-alcohol 40.8
Cyclo-hexane 55.4 Methylethylketone 56.2
Cyclo-hexene 63.5 Methylethyldioxolane 70.9
1.2-Dichloro6thane 51.2 Methylisobutylketone 67.6
Dichloromethane 48.6 2-Methyl-1 -propano1 56.6
Diethylether 56.9 Pentane 51.9
Diisopropylamine 14.4 Propyl alcohol 50.0
Dimtthoxymtthane 64.2 Propionaldehyde 49.4
Tetrachloroethylene 70.2 Benzene 59.4
Tetrahydrofurane 49.1 Fluorobenzene 61.1
Toluene 63.1 Acrylonitrrile 51.2
Trichloroethylene 65.6 Methylbutanone 64.2
Triethylamine 81.9 Mean 59.44
Trimethylpentane 75.3 Standard deviation 9.67

Table 3. Adsorption enthalpies measured for different GACs

-BPL RB2
Dichloromethane 46.3 48.5
Ethyl alcohol 52.1 61.4
Ethyl-formate 51.4 49.9 50.0 52.4 54.1
Methylethylketone 56.2 55.0 53.1 61.3 62.2
Benzene 59.4 52.8 40.6 51.2 49.2
Fluorobenzene 60.0 65.1
Acrylonitrile 46.4 51.6 52.2
Methylbutanone 64.0 62.8
Hexane 68.3 60.9 52.6 64.3 76.2
Mean
Standard deviation I 6.70 I 5.71 I 4.84 I 5.94 I 8.40

From the standard deviations reported in tables 1 and 2, we notice that the energetic
interactions are sensitive to both the adsorbent and the organic specie. Indeed, energy
deviations observed for a same compound on different carbon materials are in the same
order of magnitude as that measured with different VOCs on one type of GAC.
The energy database for various GACs has to be extended before any quantitative
analysis, accounting for the influence of the intrinsic properties of the adsorbent, could be
achieved. However, we notice that on average, adsorption enthalpies measured on NC 100
and BC120,which have the highest specific surface areas, are lower than on the other
GACs.

266
3 Quantitative Structure Property Relationships(QSPRs)

3. I Data selection
In order to calculate the adsorption energies as a function of the VOCs molecular
characteristics, QSPRs were investigated allowing for a sufficiently large energy database,
which was that given for 44 VOCs on only one GAC (table 1). These experimental data
were randomly divided into a working data set, which was used to fit the parameters of the
statistical models, and a testing data set, which allowed to assess their predictive ability
and generalisation capacity. So the statistical models were set up fiom data relative to 36
compounds, and their predictive capacity was tested for 8 VOCs .
The dimensional and electronic properties collected for each organic species
represented a set of 8 input variables : the molar mass M, the molar volume V,, the
ionization potential IP, the polarisability , the molar refraction R,,,, the dipolar moment
, the surface tension . and the dielectrical constant . These variables were selected
because different authors have shown their influence on sorption energies [4-61.

3.2 Statistical models

As a first attempt, a linear model was assumed and a multiple linear regression (MLR) was
investigated. Applying a stepwise procedure, the explicative input variables were
discriminated. The explicative variables were successively included in the model provided
that the coefficient of determination ? could be significantly improved. The best linear
regression was then derived based on the maximum ?criteria.
In a second approach, a non linear relationship was supposed and a neural network
(NN) was considered in order to recognise the pattern embodied in the energy database. A
neural network is a highly interconnected structure, composed of many simple elements,
called neurons. The most usual architecture retained is the Multi-Layer Perceptron,
constructed fiom three different layers [7]. The input layer contains the input variables
that are speculated to affect the output data, which form the output layer. The hidden layer
placed between the input and the output ones, enables to compute highly non linear and
complex functions. It contains a variable number of hidden neurons, which are connected
through connection weights that represent the relative strength of an input neuron in
contributing to the output neuron. The training process leads to the determination of the
connection weights values, given a number of nodes in the networks architecture. The
weights are updated by testing the network on validation data which were not used during
the training.
In this paper, the training was performed with the supervised learning feedforward
error backpropagation algorithm [8], accounting for 24 compounds in the working data set,
and 12 in the validation data set. Its stability was tested by random settings of the
connection weights and random splitting of the training and validation data sets. The
testing data set, composed of the same 8 compounds, was used separately, to check the
generalisation ability of the NN.

267
4 Results and discussion

4. I Multi-linear regressions (MLR)

Results of the best MLR are presented in figure 2, which compares experimental and
computed energies. The best MLR corresponds to the maximum coefficient of
determination ?which was 0.853 on the working data set and 0.879 on the testing data set.
This means that more than 85% of the variability in the data may be explained with the
linear model. Adsorption energies are then predicted with a precision around +/-lo%.
Following the stepwise procedure, 3 explicative variables were discriminated :

- Ha&= 89.1-5.I3 IP+I .55 +0.0724 M

The effect of the polarisability . and of the ionization potential IP may be directly
related to their impact on the energy of the dispersive forces (London, Debye and
Keesom) which govern physical adsorption onto activated carbon [4]. The lower positive
effect of the molar mass M may be related to the influence of the molecular overcrowding
which increases the surface contact with the solid, leading to more intensive interactions.

4.2 Neural Networks (NN)

The architecture(3,1,1) was primarily retained to investigate the best neural network. The
56 triplets of the input variable set were discriminated. For each triplet, the training was
repeated up to 10 times. Each triplet was removed as soon as the coefficient of
determination calculated on the working base was less than 0.8. At last, the only mplet
remaining was the same as that resulting from the linear regression i.e , ZP,M. The best
result obtained from the NN model is reported in figure 3. The coefficient of
determination ? computed for the working base was then close to 0.850, while that
computed for the testing data set may be up to 0.899.
The method of weight partitioning proposed by Garson [9] was used to determine the
sensitivity of the selected input variables on the output. The relative importance of IP,
and M was thus estimated (in percent) as 54.3, 32.6 and 14, respectively. The architecture
(3,2,1) was also tested but it was shown that one neuron more in the hidden layer did not
improve the prediction capacity of the NN. +lo%
+lo% ,. ideal line
80
'-10%
-
h

70
8

1
0
40 50 60 70 80 90 40 50 60 70 80 90
8
experimentai energies (kl.rnoF') experimental energies ( I C J . ~ ~ ~ ' )

Figure 2. Results of the best MLR Figure 3. Results of the best NN

268
 4.3 Discussion
It is noticeable that both linear and non linear modeling approaches explain changes in
adsorption enthalpy data with the same molecular properties. This means that the lack of
agreement observed is not due to the way of combining the explicative variables in the
relationship. A better representation would only be found provided that an other pertinent
molecular property is included among the input variables set.
This study demonstrated that statistical models may be successfblly applied to set up
QSPRs available for prediction of adsorption enthalpy of an organic specie on one type of
activated carbon. But for generalization, the specific influence of the properties of the
GAC have also to be taken into account.

References

1. Prd P ., Delage F. and Le Cloirec P., A model to predict the adsorber thermal
behavior during treatment of volatile organic compounds onto wet activated carbon,
Environ. Sci. Technol.,Environ. Sci. Technol, in press (2002).
2. Prd P., Delage F. and Le Cloirec P., Modeling the exothermal nature of V.0.C
adsorption to prevent activated carbon bed ignition, Fundamentals ofahorption 7, K.
Kaneto, H. Kanoh, Y.Hanzawa Editors, IK International, Chiba, Japon, (2001) pp. 700 -707.
3. Delage F., Prd P ., and Le Cloirec P., Mass transfer and warming during adsorption of
high concentrations of VOCs on an activated carbon bed : experimental and
theoretical analysis, Environ. Sci. Technol.,34 (22) (2000). pp. 4816-4821
4. Ruthwen M.D., Principles of adrorption and ahorption processes, John Wiley &
Sons, New-York, US (1984).
5. Mocho P., Adsorption de composds organiques volatils sur charbon actif -
rdgdndration in situ du charbon par chauffage par induction dldctromagnetique,PhD
Thesis, Universitede Pau et des pays de 1 Adour- France (1994).
6. Terzyk A.P, Rychlicki G., Empirical relationship describing energetics of adsorption
at low coverages of macroporous carbons, J. Thermal Anal., 55 (1 999) pp 101 1- 1020.
7. Bishop C., Neural networksfor pattern recognition,Clarendon, Oxford, U K , (1995)
8. Rumelhart D. E., Hinton G. E., and Williams R. J., Learning representations by back-
propagation emors., Nature, 323 (1986) pp. 533-536.
9. Garson G.D., Interpreting neural network connection weights, Al Expert, 6 (1991)
pp. 47-5 1.

269
 THE INFLUENCE OF Ar AND He ON THE RATE OF ADSORPTION AND ON
THE ADSORPTION EQUILIBRIUM OF ALKANES IN ZEOLITES

M.C.MITTELMEIJER-HAZELEGER, A.F.P. FERREIRA AND A: BLIEK

Department of Chemical Engineering, University of Amsterdam. Nieuwe Achtergracht 166.
1018 WVAmsterdam, The Netherlands
E-mail: marjom@cience .uva.nl

Diffision of alkanes in zeolites is widely studied. The experimentallyobtained diffisivities, however,

vary by orders of magnitude. In many cases noble gasses are used and it is generally believed that
they neither influence the adsorption rate nor the adsorption equilibrium. We studied the influence of
argon and helium on the adsorption rate and equilibrium of n- and iso-butane in the zeolites BEA,
MFI and FER at 383 K and found a pronounced effect on the adsorption equilibrium as well as on
the adsorption rate.

1 Introduction

Adsorption and diffusion of alkanes in zeolites and in well-structured porous materials

like MCM-41 materials are studied widely [1,5,8,9,1I]. The reported difusivities however
differ sometimes by orders of magnitude. These differences are sometimes attributed to
the use of microscopic techniques in stead of macroscopic techniques [ 121. We, however,
think that a major part of the found differences must be imputed to the use of a carrier gas.
Adsorption is often studied in diluted systems with methods as ZLC [3],
gaschromatography [4], inverse gas chromatography [ 101, gravimetry [ 121, while others
are not using camers gasses at all.
In this paper we show that the use of a carrier gas can influence the rate of adsorption and
the adsorption equilibrium. We studied the adsorption of n- and iso- butane in the
presence of helium or argon. We also show that helium or argon has no influence at all,
when introduced after butane is pre-adsorbed.

2 Experimental

The samples used to study the adsorption process are:

1. H-FER (ferrierite) with a Si to A1 ration of 55, Provided by the Shell Research and
Technology Centre Amsterdam (sample code CLA28401).
2. MFI (ZSM-5) with a SVAI ration of 100 from Zeolyst and
3. BEA (Beta) with a SVAI ratio of 300 also from Zeolyst.
All the zeolites were calcined in air at 873K for 6 hours.
Before each measurement the samples were outgassed at 573K to a vacuum better than
3.104 Pa.
The adsorptive used in this study are n-butane and iso-butane from Praxair with a
purity of 99.5 and 99.95%. Helium and argon both were 99.999% pure and were also
from Praxair.
The experimental set-up consists of a home-made manometric adsorption apparatus
coupled to a Rubotherm three-position magnetic suspension balance. The introduction
part of this home-made set-up is designed to work at 473 K and 10 MPa. During the

270
experiments the balance was kept at 383 K, while the remaining system was kept at a
higher temperature in order to prevent any condensation. A scheme of the set-up is given
in figure 1.
Buoyancy corrections can in the present set-up be made by the in-sifumeasurement of
the density of the gas by means of the weight change of a sinker with a known volume.
This balance is described extensively by Dreisbach ef al. [2].

8 gas in
@ = Pressuretransducer tanium sinker
ample holder

figure 1. Schematic representation of the experimental set-up for adsorption measurements.

In a first series of experiments adsorption was studied using undiluted n- or iso-

butane. In a second series helium or argon were admitted first, and subsequently (within 5
minutes), n- or iso-butane was admitted to the sample compartment. In both series exactly
the same amount of n- or iso-butane was introduced.

3 Results

3.1 BEA
In Figure 2a and b the results are given for the adsorption of n- and iso-butane at 383 K on
BEA. If there was no influence at all of helium and argon in each figure only two lines
should have been present. From these figures it can be observed that the pre-admission of
helium or argon slows down the rate of adsorption, this being the case to a stronger degree
for argon. Also the equilibrium loading is adversely effected by pre-admission of helium
or argon. The effect on the adsorption equilibrium results in a higher (partial) pressure of
butane in case argon or helium is present, although the same amount of butane was
introduced.
After introducing helium or argon no change in weight could be observed, indicating
that no adsorption of helium or argon is taking place (less than 0.01 pmol He/g or 0.1
pmol Ar/g) before the butane is admitted.

27 1
 1.00 Im
.-.
0.75 0.75

-c
6 O M O M

0.25 0.25

0.00 a I 0.00 y I
50 100 150 200 0 50 100 I50 m
lime (min) cim (min)
Rgure 2a and b. The amount adsorbed versus time for n- and iso-butane on BEA at 383K. The indicated
pressures represent the (partial -in 100 Wa He or Ar-) pressures of n- or iso-butane after 200 minutes
equilibrium time.

3.2 MFI
In figures 3a and b the results are shown for n- and iso-butane on MFI at 383K. Basically
the same effects observed for BEA can also be seen in case of MFI.
(iu (s)
1.00 1.00

3-
0.75 0.75

2 0.50 E- 0.50

1 0.25
1 0.25

0.00 0.00
0 50 100 Is0 200 0 50 100 I50 Mo
time (min) time (min)
Figure 3a and b. The amount adsorbed versus time for n- and iso-butane on MFI at 383K. The indicated
pressures represent the (partial -in 100 kPa He or Ar-) pressures of n- or iso-butane after 200 minutes
equilibrium time.

3.3 FER
For n- and iso-butane on FER at 383 K the results are given in figure 4a and b. also in
these figures an influence can be found of the presence of a noble gas. From figure 4b
clearly the conclusion can be drawn that no equilibrium is reached after 200 minutes.

I
0.00 r SO 100 I50 200
0.K
0 50 100 I50 m
time (min) time bin)
Figure 4a and b. The amount adsorbed versus time for n- and iso-butane on FEFt at 383K. The indicated
pressures represent the (partial -in 100 Wa He or Ar-) pressures of n- or iso-butane after 200 minutes
equilibrium time.

272
 The influence of noble gasses on the adsorption isotherm is demonstrated in figure 5
for n-butane on FER at 383K.As the influence on the amount adsorbed is dependent OR
the loading the used of noble gasses influences the shape of the isothenn.

2.00 1.00 ;

A
130 -
ij
1
1.00-

>
0.3 -

0.00
10 LO 100 10 10 10
0 40 80 120 160 200

Of D - b W e Orps)
@artid)plcss~rr lime (Inin)

figure 5. Adsorption isotherms of pure n-butane and figure 6. Effect of sequence of admitting helium and
n-butane in 100 kPa helium on FER at 383 K. n-butane to MFI at 383K.

Re-admitting of a noble gas influence the adsorption, while admitting a noble gas
subsequently to butane adsorption has no influence at all. This is demonstrated in figure 6
for n-butane on MFI at 383 K.

4 Discussion

As there is no circulation pump present in the experimental set-up, part of the described
phenomenas could be due to poor mixing. Therefore we performed a test in a
microbalance. A BEA sample was pre-treated in a helium flow at 573 K.The sample was
cooled down to 383 K in a helium flow. After cooling down the helium flow was switched
to an n-butanehelium mixture. Comparison with the experiment described before revealed
that the rate of adsorption was faster, but still much slower than without any noble gas
present and that the amount adsorbed was even less than in the described experiments with
a noble gas present.
The presented equilibria are no thermodynamic equilibria. Therefore another test was
performed. After reaching equilibrium during adsorption of n-butane and 100 Wa argon
in FER at 373 K we heated the sample to 573 K during one day and cooled the system
again down to 373 K.We found that the amount adsorbed increased. But also in this case
the amount adsorbed did not reach the value obtained without argon being present. This is
shown in table 1.

Table 1. Amount of n-butane adsorbed on FER at 373 K.

(partial) pressure n-C4 Amount adsorbed
(Pa) (mm0Vg)
No argon 10.6 1.056
100 Wa argon 17.5 .876
100 Wa argon after a temperature cycle. 12 .968

?he te.mperahue cycle was equilibrium at 373 K, heating to 573 K and cooling down to 373 K.

273
 The difference between helium and argon can not attributed to the external mass
transfer, i.e. differences in the binary coefficients. These are DHe,c4= 49.106 m2/s and
DAr,c4=42.106 m2Is.
This reported effect of influence a noble gas is not restricted to zeolites only. We
found the same effect for n-butane on a Carbon Molecular Sieve 5A of Takeda.
The reported influence of a carrier gas for n- and iso-butane was observed before. For
instance K&ger et ul. [7]observed in NMR based diffusivity studies that the self-diffusion
coefficient of cyclohexene in NaX is negatively affected by the presence of argon.
We think that carefully studying the reported literature data on diffusion coefficients
and taken into account whether a carrier gas was used or not could lead to data sets that do
not show the big differencesas thought to be the case.
The influence of a noble gas may be used to improve separation processes as in our
case for MFI the adsorption of n-butane is influenced to a different extent by helium or
argon than is iso-butane. This was reported also by Jwalin et ul. [6] who concluded that
propene and propane could be separated over 5A or 13X zeolites, provided that the
mixture was sufficiently diluted with nitrogen.

Conclusions

Helium and argon, often used as carrier gasses during adsorption studies, have a
negative influence on the rate of adsorption and also on the adsorption equilibrium as
at the same partial pressure less is adsorbed. This is proven for n- and iso-butane on
BEA, MFI and FER.
The influence of argon is more pronounced than that of helium.
The equilibrium is less influenced at higher (partial) pressures.
Re-admitting of a noble gas influence the adsorption, while admitting a noble gas
subsequently to butane adsorption has no influence at all. This suggests that the
observed phenomena is a kinetic effect. Likely, the noble gasses are present in the
zeolite channels but no adsorbed, seriously hamper the diffusion of both n- and iso-
butane.
The reported phenomenas are not restricted to zeolites; also CMS show the same
effects.

Acknowledgement

This research was carried out within the project CWISTW 349-5203.

References

1. Calvalante C.L. and Ruthven, D.M., Adsorption of branched and cyclic parafines in
silicalite. 2.Kinetics. Ind Chem. Res. 34 (1 995)pp. 185-191.
2. Dreisbach F. and Losch, H.W., Magnetic suspension balance for simultaneous
measurement of a sample and the density of the measuring fluid. J. Therm. Anal.
Chem. 62 (2000)pp. 5 15-521.
3. Eic M., Micke A., KoEirik M., Jama M. and ZikhnovB A. Diffusion and
immobilization mechanisms in zeolites studied by ZLC chromatography. Auksorption

274
 8 (2002) pp. 15-22.
4. Fomi, L. Viscardi, C.F., Sorption-difision in molecular sieves. J Catal. 97 (1986)
pp. 480-492.
5. Jama, M.A. Delmas, M.P.F., Ruthven, D.M., Diffusion of linear and branched C6
hydrocarbons in silicalite studied by the wail-coated capillary chromatographic
Method. Zeolites 18 (1997) pp. 200-204.
6. JWalin H. and Fair J.R., Adsorptive separation of propylene-propane mixtures.
Ind.Eng.Chem.Res. 32 (1993) pp. 220 1-2207.
7. Kiirger, J. Zikanova, Z. Kocirik, M., NMR study on the influence of a carrier gas on
zeolitic diffision. Z P@s. Chem. Leipig, 265 (1 984) pp. 587-592.
8. Keipert, O.P. Baers, M., Determination of the intracrystalline diffision coefficients of
alkanes in H-ZSM-5zeolite by a transient technique using the temporal-analysis-of-
products (TAP) reactor. Chem. Eng. Sci. 53 (1 998) pp. 3623-3634.
9. Millot, B. Mdthivier, A. Jobic, H. Moueddeb, H. Dalmon, J.A., Permeation of linear
and branched alkanes in ZSM-5 supported membranes. Micropor. Mesoporo. Mater.
38 (2000) 85-95.
10. Thielmann, F. Baumgarten, E., Characterization of microporous aluminas by inverse
gas chromatography.J COILIntetf Sci. 229 (2000) pp. 4 18-422.
1 1 . Xiao, J. Wei, J., Diffusion mechanism of hydrocarbons in zeolites-11. Analysis of
experimental observations. Chem. Eng. Sci. 47 (1992) pp. 1 143-1 159.
12. Zhu, W. Kapteijn, F. Moulijn, J.A., Diffusion of linear and branched C6 alkanes in
siiicalite-1 studied by the tapered element oscillating microbalance. Micropor.
Mesopor. Mater. 47 (200 1) pp. 157- 17 1.

275
 MODELING THE DISCHARGE BEHAVIOR OF METAL HYDRIDE
HYDROGEN STORAGE SYSTEMS

S.A. GADRE, A. D.EBNER, S. A. AL-MUHTASEB AND J. A. RIITER

Department of Chemical Engineering, Swearingen Engineering Center
Universi@of South Carolina, Columbia,SC 29205, USA
E-mail: [email protected]
A new approach is introduced to model the discharge behavior of a metal hydride hydrogen storage
bed. The reversiblereaction kinetics and the empirical Van? Hoffrelationship used in a typical reactor
model are replaced by a solid phase diffusion equation and a semi-empirical equilibrium P-C-T
relationship. Two new semi-empirical P-C-T models are also introduced based on modified virial and
composite Langmuir expressions. By varying the heat and mass transfer coeffcients, the model was
calibrated to experimentalpressure and temperatm histories obtained from a commercially viable bed
containing Lm1.M N~.%A~O.M metal hydride. Overall, the results of this study showed that a fairly
simple numerical model can do a reasonable job in predicting the discharge behavior of a fairly
complicated metal hydride hydrogen storage bed over a wide range of hydrogen demands.

1 Introduction

The Savannah River Technology Center (SRTC) recently developed an on-board hydrogen
storage bed for a hybrid electricbus, [I] based on the commerciallyviable Lml.MNi4.&lo,m
metal hydride. A schematic of this unique bed is shown in Fig. 1. Due to its inherent
complexity, a mathematical description of the SRTC system is highly desirable for
improving on the design and understanding its dynamic behavior.

I
MFW lgdridc panicles
upliag the wid rpre of
\ /
r
nrnna~iwlaiion
thc aluminum folm
Figure 1. Schematic of the SRTC hydrogen storage vessel.

Most of the mathematical models presented in the literature make use of reactor design
equations that include a reaction kinetics equation. However, since the reaction rate tends to
be very fast, the discharge process 60m a metal hydride bed can also be modeled by
assuming solid diffusion as the rate limiting step. In this new approach, the heat and mass
transfer coefficients are obtained by comparing the theoretical discharge curves with the
experimental ones. The extreme heat and mass transfer limits are also explored to gain an
appreciation for the limiting and actual behaviors of the SRTC system. Theoretical
Referring to Fig. 1,the U-tube heat exchanger is approximated as a single coaxial tube with
the diameter doubled. The outside of the column is assumed to be perfectly insulated, and
the water temperature inside the heat exchanger is assumed to be constant throughout the
discharge process. All heat transfer processes between the material and the heat exchanger
are lumped together and represented by an overall heat transfer coefficient. It is further
assumed that the flux variation inside the column is linear. All radial gradients are also

276
ignored. Based on these assumptionsand assuming ideal gas behavior, the 1-D mass balance
is given by
I a7t 7t ae a0 a/
----- +-+-=o (1)
8 a~ e2a t a{ a t
where $, R, 0 and o denote dimensionless loading, pressure, temperature and molar flux,
respectively. The corresponding energy balance, including compression but ignoring
convection effects, is written as

+-- +kh(B-6w)=0 (2)

where, k,, k,,, k, and k,, are defined as

The intraparticle mass transfer mechanism is based on the following linear driving force
expression (which assumes solid diffusion control):

where k,,, is the mass transfer coefficient, and $*, the dimensionless equilibrium loading, is
obtained directly from the isotherm relationship. The model parameters are defined as inner
radius (ri) 4 . 0 2 m, outer radius (ro) = 0.045 m, length (L)= 1.52 m, heat capacity of H2
(Cpg) = 14.42 kJkgK,heat capacity of solid (Cps) = 0.419 kJ/kgK,mass of metal hydride
(md) = 26.078 kg, mass of aluminum foam (mAl) = I .789 kg, density of metal hydride (pmH)
= 8700 kg/m3and density of aluminum foam (pA,)= 2700 kg/m3. With h and k,,, specified,
Eqs (1) to (3), along with a suitable P-C-T relation are input to FEMLAB and solved
simultaneouslywith the following initial conditions: at t = 0 ,Ic = 7ti , 8 = ei and
4 = 4i. This model utilizes two new P-C-T relationships, denoted modified virial (MV)
and composite Langmuir (CL) [2]. They are
correlated with Lm,.M N&,&tlo,04
0.014
metal hydride data; the results are
shown in Fig. 2. The two 0.012
experimental isotherms at 3 13 and
333 K, and the numerous P-Tdata 0.010
points at constant loading above and m
A

below the phase transition region are 3 0.008

correlated equally well by both P
models. The corresponding hears of 9 2 0.006
adsorption predicted by the two 0.004
models are also reasonably similar
and relatively constant in the phase 0.002
change region: 28 and 33 kJlmol for
the CL and M V models, 0.000
0.1 1 10 100
respectively. P (atm)

Figure 2. Metal hydride-Hz isotherms (symbols: experiment;

solid lines: MV correlation, dashed lines: CL correlation).

277
2 Experimental

The SRTC column is equipped with 16 k-type thermocouples mounted externally on the
stainless steel surface at 4 axial locations with each axial location having four
thermocouples set approximately 90"apart. The pressure inside the column is measured
just at the outlet; and the hydrogen discharge flow rate is controlled with a mass flow
controller. The heat exchanger water temperature and flow rate are fixed at 302 K and
0.32 Us.In a typical discharge run, the bed is filled with hydrogen at 100 SLPM to the
desired pressure (17 to 25 am). The bed is then allowed to cool to the ambient condition.
Finally, the bed is discharged at a constant molar flow rate (5-40 SLPM) and the pressure
and temperatures are recorded.

3 Results and Discussion

Six hydrogen discharge experiments were carried out at 5, 10, 15. 20, 25 and 40 SLPM
hydrogen demand. One of these runs was chosen arbitrarily (the 20 SLPM run) to
calibrate the numerical model by varying the h and k,,, until the pressure and temperature
histories from the model matched the experimental data. The performances of the other
five runs were predicted using these coefficients without further adjustment. The heat and
mass transfer coefficients obtained from the M V model were h = 13.24 x W/cmz/K
and k,,,= 0.1, and those from CL model were 7.57 x W/cmz/K and k,,,= 0.1. These
values were uniquely defined in each case, with both models resulting in the same k,,,, and
nearly the same h. The factor of two difference in h was most likely caused by the
difference in the plateau heat of adsorption for the two P-C-T models. The CL model had
the lower heat of adsorption (by 5 kJ/mol) and this was consistent with the CL model also
resulting in a smaller value of h.
The experimental pressure and temperature histories obtained during discharge for
three of the experimental flow rates viz. 5, 20 and 40 SLPM are shown in Fig. 3, along
with model predictions from both the CL and MV models using the fitted values of h and
k,,, (Non Ad, Non eq). Note that each experimental temperature history corresponds to the
average temperature associated with all 16 thermocouples. The predictions from both
nonadiabatic-nonequilibrium models utilizing the two different P-C-T relationships (CL
and MV) were quite satisfactory and similar; however, the one based on the MV model
gave slightly better results, especially for the temperature histories of the 20 and 40 SLPM
runs. For this reason, the MV model was chosen to show the extreme behaviors, under
isothermal equilibrium conditions (Iso, Eq) with h = k,,, = 00, adiabatic equilibrium
conditions (Ad, Eq) with h = 0 and k,,, = -, and adiabatic non-equilibrium conditions (Ad,
Noneq) with h =O and k, = k,,, (fitted); these results are also plotted in Fig. 3. In all cases,
the adiabatic predictions deviated markedly from the experimental results, indicating that
the heat exchanger water flow rate was sufficient enough to allow the system to operate
nearer to isothermal conditions. In fact, isothermal-equilibrium conditions prevailed at the
lower flow rate of 5 SLPM, suggesting that the reaction rate was very fast. However, this
was not the case at the higher flow rates, where mass and heat transfer effects dominate.
Even though the pressure histories were predicted quite well in all cases, the models using
the fitted values of h and k,,, failed to predict the leveling off of the temperature history of
the 40 SLPM run. This indicated that the overall heat transfer approach was not rigorous
enough to capture the behavior of the bed at high hydrogen flow rates; more rigorous
models are under development.

278
 25
293
20

15

10
273!
5
(d) 5 SLPM
0 253
0 4 a 12 0 4 8 12 16
251 . I I I I
-
20

--E 15 -
Y

Q
m
10
I- 273 -1
8
-
'I
5 263 -I -
I I
(e) 20 SLPM
I
-
0 253
0 1 2 3 0 1 2 3
. ~ . ~ , l , l *

(c) 40 SLPM

0 0.4 0.8 1.2 1.6 2 0 0.5 1 1.5 2

t (hr)
Figure 3. Comparison of various model predictions with the experimental pressure and temperature histories
for 5,20 and 40 SLPM HZflow rates.

4 Acknowledgement

Financial support was provided by the NRO,contract no. NRO-00-C-0134.

279
References

1. Heung L. K. On-board Hydrogen Storage System Using Metal Hydride. Hydrogen

Power: Theoretical and Engineering Solutions, Proceedings of the HYPOTHESIS
Symposium,2nd, Grimstad, Norway, Aug. 18-22,1997 (1998) pp. 251-256.
2. Gadre S. A., Ebner A. D., Al-Muhtaseb S. A. and Ritter J. A. Practical Modeling of
Metal Hydride Hydrogen Storage Systems, Znd. Eng. Chem Res. (2002) submitted.
 THE ADVANCED MODELING TECHNOLOGY FOR
PERIODIC ADSORPTION PROCESS:
DmECT DETERMINATIONOF CYCLIC STEADY STATE

JEONG-HO YUN, ANDREW C. STAWARZ AND FELIX 0. JEGEDE

Aspen Technology Ltd, Titan House, Castle Park, Cambridge
CB3 OA United Kingdom
E-mail: Jeong-Ho.Yun@pentech. com

We report on our modelling progress for the determination of the cyclic steady state condition for
periodic adsorption processes.The modelling technique is the result of the complete discretisationof
both space and time that reduces a periodic adsorption process to a set of algebraic equations. The
new model was implemented into Aspen ADSIMm and used to examine the simulation performance
of a number of isothermal, non-isothermaland singldmulti-layerprocesses such as nitrogen PSA [I],
oxygen VSA [2], and hydrogen PSA [3f. The results obtained indicate that this approach offers an
extremely efficient design tool that can be readily used for faster optimisation of adsorption
PrOCeSSeS.

1 Introduction

Cyclic Steady State (CSS) is a unique feature of periodic adsorption processes. It is

defined as a condition whereby the state at the end of each cycle is identical to that at its
beginning. Within an engineering context, the theoretical determination of CSS for a
given set of periodic conditions is a key step towards the optimisation of adsorption
processes. CSS determination by computer simulation, however, is still one of the most
challenging procedures for a process engineer to implement. Once achieved, the engineer
is more readily able to maximise profit and minimise costs of the adsorption process.
A periodic adsorption process is operated as a series of sequential steps on one or
more adsorbers packed with either single or multiple layers of adsorbent. Although the
operation of each bed is batch-wise, the whole system is continuous owing to the use of
multiple adsorbers operated in phased cycles. The traditional approach for CSS
determination is to execute a dynamic simulation of the process, starting with a specified
set of initial conditions and then simulated over a large number of cycles until a CSS
condition is eventually obtained with a pre-defined criteria, i.e., the cycle initial state at to
must be identical with the cycle end state at t N (see Figure 1).

Figure I Conceptual illustration for the typical dynamic simulation of a periodic adsorptiou proccss

281
 In reality, a practical adsorption process for the separation of gases is often complex
and involves a number of sequential but interacting unsteady state cycle steps.
Accordingly, a precise rigorous mathematical model used to describe such a process
becomes computationally expensive and the solution time-consuming thus making the
optimisation of adsorption processes through dynamic simulation a slow procedure.
Owing to this inherent limitation several researchers have examined improved numerical
techniques to accelerate CSS convergence of adsorption process simulators [ref.2 and
references therein]. Existence of a periodic time boundary, however, encourages the
replacement of the initial conditions by periodicity conditions such that a dynamic system
may be reduced to a steady state system within a confined time period, the time for one
cycle. An instructive illustration is given in Figure 2.

I..
sbt.(a) -
p.rlodkBwnb.ry
-ww
CyCllC S t M I y sat.

Figure 2 Conceptual illustrationsof the steady state modelling for direct CSS determination

As seen in this figure, the enforced periodicity condition provides continuity to the system
by linking the starting point towith the ending point tN.In a computational point of view,
this suggests a steady state simulation is feasible by a complete discretisation of space and
time within the confined time length. In fact, the complete discretisation method had been
discussed in an earlier study [ 5 ] and investigated extensively [6].However the application
has been limited to computationally simple problems and was considered an unrealistic
method for complex non-isothermal process simulation owing to computational barriers
121.
The purpose of this study is to examine the capability of this modelling methodology
using readily accessible computing resources such as a Pentium computer and a general-
purpose modellingkimulation environment. The investigation examined a number of
periodic adsorption processes of general complexity such as nitrogen PSA [l], oxygen
VSA [2] and hydrogen PSA [3,4]. From the extensive study, we realised that direct CSS
determination modelling using steady state simulation can provide a remarkably faster
solution than traditional dynamic simulation. Furthermore CSS modelling enables direct
movement from one CSS condition to another CSS condition. However for a dynamic
model further cycles must be simulated until the new CSS condition has been achieved, a
step that can take a significantly longer time than with the direct CSS determination
model (hereafter CSS model).

282
2 Cyclic Steady State Model

Adsorption beds are essentially transient, spatially distributed systems, where the
properties in the solid and gas phases varying over time in one or more spatial
dimensions. The mathematical description of adsorption beds is usually described by a
series of partial differential equations and algebraic equations. In this paper,
distinguishable features of the CSS model are briefly given.

2. 1 Discretisation Methods
In this work, discretisation of both space and time derivatives was executed, based on
either central finite difference (CFD) or orthogonal collocation on finite elements (OCFE)
discretisation in the spatial domain and backward finite difference (BFD) discretisation in
the time domain.

2.2 Periodicity Conditions

Cyclic Steady State is the condition whereby the state at the end of each cycle is identical
to that at its beginning. For a non-isothermal adsorption system this may be represented
by a mathematical model, comprising material, energy and momentum balances as well
as adsorption equilibrium and kinetic models, the CSS can be expressed by:

Vjce(0,l)

where C, and ?, are bulk and solid concentrations for an individual component and 7^, Ts,
and TW are temperatures at gas phase, solid phase and column wall, respectively. tN is a
cycle time and x denotes a distributed variable along spatial distance L.

3 Simulation of Literature Processes

Several gas adsorption process examples were studied to examine the simulation
performance of the CSS model. Examples include: (1) binary, 4 step N2 PSA using 2
isothermal beds each containing a single CMS layer [1]; (2) binary, 8 step O2 VSA using
2 non-isothermal, non-adiabatic beds each containing a single Zeolite 13X layer [2]; (3) 6
step binary-, ternary-, five component H2 PSA using 2 non-isothermal, non-adiabatic beds
each containing a double layer of Activated Carbon/Zeolite 5A [3, 4]. In this work, the
results obtained by CSS simulation are compared to those by traditional dynamic
simulation. The computing efficiency is assessed in terms of solution convergence and
computational time requirements (all computational results in this paper were obtained
using a 1.7 GHz Pentium 4 computer with 1 GB of physical memory).
In Table 1, a summary of the model assumptions for each example and a comparison
of results obtained using the dynamic simulation and the CSS simulation has been given.
The temporal domain for each example made use of 20 nodes for N2 PSA, 32 nodes for
O2 VSA, and 22 nodes for H2 PSA, together with adapted time element length depending
on each operation step.

283
 Process example N2 PSA [13 0 2 VSA [2] H2 PSA [3,4]
2 bed, 4 steps 2 bed, 8 steps 2 bed, 6 steps
Cycle info
150 s per cycle 60 s per cycle 420 s per cycle
Adsorbate N2.02 N2-02
H2, a,
co,
N29 CO2
Activated Carbon I
Adsorbent CMS Zeolite 13X
Zeolite 5A
Non-Isothermal I Non-Isothermal I
Energy balance Isothermal
Non-Adiabatic Non-Adiabatic
Momentum balance Ergun equation Ergun equation Ergun equation
Extended Dual-
Equilibrium model IAST (Langmuir) site Lan LRC
Kinetic model Solid phase LDFA Solid phase LDFA Solid phase LDFA
1
~

Dvnamic CPutime 180 s 392 s 1080 s

Purity I 95.0 % N2 94.9 % 0 2 91.3 % H2
~

css CPUtime 240 s 176 s 595 s

solution Purity 95.8 % N2 94.2 96 0 2 90.6 9% H2

cycle Time f Pn.i,inn m

Cycle Time s Position rn

Figure 3 Typical CSS solution of pnssure and gas composition bed prohles for the 9VSA process
A typical solution given by CSS simulation is shown in Figure 3 whilst Figure 4
shows the difference between both simulation methods with respect to pressure and
product composition for the H2 PSA and 4 VSA cases.
From Table 1 and Figure 4 it is apparent that both solutions provide similar results,
although not exactly identical. Since the same spatial discretisation was used in both
forms of simulation, the difference in product purity between corresponding results arises
from the numerical error incurred through time discretisation. The errors resulting from
temporal discretisationbecome considerablewhen the temporal variation of bed condition
is steep. In CSS simulation, improved accuracy can be obtained if finer temporal
discretisation is used. However, by using more temporal nodes, the system size is
increased thus resulting in increased computing resources. This is considered the primary

284
limitation of this approach and therefore, careful consideration is required in deciding the
number and size of temporal elements.

14?{ -
12 0.9
L
0
2- 10 0.8'3
zi
2 8
3
$ 6 0.7 8
i?
e
n 4
0.6 I"
2 Hz PSA
0 I 1 0.5
0 105 210 315 420 0 20 40 60
Cycle Time, s Cycle Time, s

Figure 4 Typical CSS solution of pressure and gas composition bed profiles for the Q VSA process
-
(symbols CSS, lines - Dynamic)
In spite of the system size generated by temporal discretisation, it is has been found
that the method can provide a faster solution with acceptable accuracy for highly complex
examples such as oxygen VSA and hydrogen PSA. Furthermore, this modelling technique
becomes much more advantageous if several runs are required to be carried executed. For
example, when executing an optimisation study, the solution of each simulation run
serves as the initial guess for the next simulation run.This allows for rapid transition from
one CSS condition to another CSS condition with the minimum of iterations whereas for
a dynamic simulation fiuther time expensive cycles are required to achieve the new CSS
condition. This is a key benefit for the process engineer, as the technique can offer an
extremely efficient design tool that can be more readily used to determine optimal design
and operating conditions of an adsorption process.

References
1. Hassan, M. M., Ruthven, D. M.and Raghavan, N. S., Air Separation by Pressure
Swing Adsorption on a Carbon Molecular Sieve, Chem, Eng. Sci. 41 (1986) pp.
1333-1 343.
2. Todd, R. S., He, J., Webley, P. A., Beh, C., Wilson, S. and Lloyd, M. A., Fast Finite-
Volume Method for PSANSA Cycle Simulation-ExperimentalValidation, Ind. Eng.
Chem. Res. 40 (2001) pp. 3217-3224.
3. Yang, J., Lee, C. -H. and Chang, J. -W., Separation of Hydrogen Mixtures by a Two-
Bed Pressure Swing Adsorption Process Using Zeolite 5A, Ind Eng. Chem. Res. 36
(1 997) pp. 2789-2798.
4. Jee, J. -G., Kim, M. -B. and Lee, C. -H., Adsorption Characteristics of Hydrogen
Mixtures in a Layered Bed Binary, Ternary, and Five-Component Mixtures, Ind.
Eng. Chem. Res. 40 (2001) pp. 868-878.
5. Alpay, E., Kenney, C. N. and Scott, D. M., Simulation of Rapid Pressure Swing
Adsorption and Reaction Processes, Chem. Eng. Sci. 48 (1 993) pp. 3 173-3 186.
6. Nilchan, S . and Pantelides, C. C., On the Optimisation of Periodic Adsorption
Processes, Adrorption 4 (198) pp. 1 13-147.
 ADSORPTION AND DESORPTION CHARACTERISTICSOF ZEOLITE
IMPREGNATED CERAMIC HONEYCOMB FOR VOC ABATEMENT

H. S. KIM, Y.J. YOO, Y.S.AHN,M.K.PARK, K.T.CHUE AND M.H. HAN

Korea Institute of EnergV Research, 71-2, Jang-dong, Yuseong-gu, Taejon, 305-343. Republic of
Korea
E-mail: [email protected]

VOC such as toluene, benzene, MEK,MIBK, cyclohexanone is one of main causes for air pollution.
When the concentration of VOC is around 1% LEL (Low Explosive Limit), the operation cost of the
rotor concentrator with thermal oxidizer was known to be most inexpensive. The rotor concentrator
is made of ceramic paper and adsorbates like zeolite and active carbon. Ceramic paper, which is
prepared by casting inorganic fiber dispersed slurry, was corrugated and rolled to form honeycomb
shaped cylinder. Zeolite and active carbon was impregnated on the surface of the ceramic
honeycomb cylinder.
In order to evaluate the VOC adsorbing capacity, several small ceramic rotor ( I Ocm diameter, 40cm
length) were prepared, and the adsorption and desorption characterstics were measured using static
adsorption I desorption test equipment. In the experiment, VOC laden gas was artificially made and
provided by bubbling air into VOC liquid. The concentration of the VOC was adjusted between 150
ad 420 ppm, and its flow rate was from 150 to 600 liter/min.
With this experiment, the proper type of adsorbent and the amount of impregnation was determined,
and the operation parameter such as adsorption / desorption / cooling area ratio, rotational speed,
required size of VOC adsorbing honeycomb rotor was designed

1 Introduction

Zeolite or active carbon has micorpores into which VOC (volatile organic compound) is
adsorbed, resulting in removal or concentration of VOCs. They are usually used as
powder or pellet in packed bed but high energy cost due to pressure drop makes the
application hesitant. Honeycomb formation of the materials is one of the promising steps
to avoid the drawback. However the extrusion cost is also very high and large size
honeycomb is extremely difficult to extrude. Zeolite or active carbon impregnation into
ceramic paper made of ceramic fiber and glass fiber as main components is one of the
developed technologies showing low pressure drop. [ 1,2,3,4,5]
Although several thermal swing data using ceramic rotor were reported,
manufacturing process of the rotor was not described in detail. In this paper ceramic paper
making process from slurry containing ceramic fiber and glass fiber, corrugation,
honeycomb formation, and impregnation of zeolites were described step by step.
For evaluation of the rotor, thermal swing adsorption was tested in a static adsorption
/ desorption test equipment and proper type of adsorbent for target VOCs was
determined.

2 Experimentation

2. I Ceramic paper making ly casting

Ceramic paper was cast on a papermaking machine into a dimension of 0.25mm thickness,

286
50cm width from finely dispersed sluny of mineral fibers in water. The slurry cast on
porous endless wire mesh made of polyester and polyimide was dewatered using
vacuum, pressure rolling and hot air blowing, successively. The final water content of the
formed paper was around 65%. The obtained wet paper was dried in dry oven over 24
hours.

2.2 Corrugation of ceramicpaper

Two sheets of ceramic paper were supplied to a corrugation machine for forming
corrugated sheet having v-shape (2mm height, 4mm pitch) parallel flute, through which
VOC laden air could flow freely. The corrugated sheet was rolled into honeycomb shape,
which had many triangular prismatic parallel flutes and the flowing air did not penetrate
from one flute to another. The ceramic rotor was rolled until its diameter reached lOcm,
before both ends were cut in order to give determined length.

2.3 Impregnation and heat treatment

Organic binder in ceramic honeycomb rotor was removed by heating it for 5 hrs in 600C
before impregnation of zeolite. Heat treated rotor was soaked in zeolite (UOP HISIV
1000 and HISIV 3000) dispersed slurry with silica sol (Nissan Chemical ST-30) or
alumina sol (Nissan Chemical AS-520) as a binder. The amount of binder addition was
varied to 3, 5,7 wt?h.BET surface area and SEM micrograph was analyzed with respect
to type of binder and its amount.

2.4 Static a&orption / desorption test

Magnetic suspension balance (Rubotherm-30G500P) was used to measure equilibrium
adsorption of VOC by measuring weight change of the impregnated sample, where VOC
at constant pressure was adsorbed into the sample on a boat until no weight gain was
reported.
Zeolite impregnated ceramic rotor was put into the sample chamber of static
adsorption 1 desorption tester, the diameter and length of which are lOqm and 40cm,
respectively. Toluene, MEK, cyclohexanone were aerated by compressed air and mixed
with air in a line mixer, where a required concentration of a VOC of the experiment was
adjusted. As the produced VOC was supplied to the sample chamber, the concentration of
the VOC after passing through adsobent was measured with total hydrocarbon analyzer
(model Horiba THCJ 10). Afier the concentration of the VOC reached 40 ppm,, supply of
VOC laden air was stopped, and hot air through electric heater for desorption was
supplied to the adsorbent in a reverse direction compared to adsorption step, and the
concentration of the VOC at the outlet of the sample chamber was measured.

287
 (a) ceramic paper

(c) honeycomb mtor (tqiview) (a)honeycombmtci (sideview)

Figure 1. Ceramic honeycomb rotor formed with ceramic paper

3 Results and Discussion

3. I Equilibrium aakorption and BET sujbce area

BET surface areas of the HISIV 1000 and HISIV 3000 zeolites were 585 and 419 m2/g,
respectively. In order to determine proper type of inorganil: binder, equilibrium loading
amount of toluene into zeolite impregnated paper was measured. HISIV 1000 zeolite
showed that zeolite adhered by silica sol adsorbed more toluene than zeolite adhered by
alumina sol at low partial pressure as shown in Figure 2. Thus silica sol was used as a
binder for binding zeolite between fibers of the ceramic paper.
The BET surface areas of the HISIV 1000 impregnated with 3,5, 7 % silica sol as a
binder were 147, 159, 165 m2/g,respectively. Those of HISIV 3000 impregnated with 3,5,
7 % silica sol were 138, 137, 129 m2/g, respectively. HISIV 1000 zeolite impregnated
with 3, 5, 7 wt% silica sol had loading capacities of 5.2, 5.7, 5.7 wt% of toluene at 1 1.5
mmHg, respectively. HISIV 3000 zeolite impregnated with 3, 5, 7 wt% silica sol had
loading capacities of 4.7, 4.4, 4.5 wt% of toluene at 11.5 mmHg, respectively. Using 5
wt% of silica sol was the best selection of the binder system.

288
 10

0 1 2 3 4 5 6 7 8 9 10 11
Umrcb)

Figure 2. Equilibrium adsorption curve of toluene on HISN lo00 impregnated ceramic paper

3.2 Adsorption properties

-
The adsorption desorption properties of 3 types of zeolite impregnated honeycombs (10
cm diameter, 4Ocm length each) were tested in static adsorption 1 desorption tester. The
honeycombs were coated with HISIV 1O00, HISIV 3000 and (70% HISIV 1000 + 30%
HISIV 3000) before measurement, and their weights of (adsorbent + binder) were 177,
191, and 163g. respectively. The ratios of (adsorbent + binder) with regard to honeycomb
rotor were 30.6,32.3, and 31.2%,respectively.
The adsorbent with HISIV 1O00 adsorbed more toluene and cyclohexanone than that
with HISIV 3000. HISIV 3000 was efficient to adsorb MEK than HISIV 1000. In order to
adsorb all 3 types of target VOC's efficiently, honeycomb adsorbent with mixed zeolites
was made and tested. As expected, 9.72% toluene, 11.0%MEK, 13.0% cyclohexanone
were adsorbed to the adsorbent until the concentration of VOC in effluent gas reached 40
ppm,. VOC removal efficiency of the mixed zeolite impregnated honycomb with repect to
toluene, MEK,cyclohexanonewere 98.2,99,1,96.7 %, repectively.
The maximum concentrations of toluene and MEK were about 3500 ppm,, and that of
cyclohexanone was 1200 ppm. which could perform catalytic oxidation without
supplementaryheat source.

289
 Table 1. Summary of adsorption properties of honeycomb adsorbents.

4 Acknowledgements

We thank Ministry of Science and Technology and Greenhouse Gas Research Center for
financial support.

References

1. Kodama A., Goto M., Hirose T. and Kuma T., Temperature profile and optimal
rotation speed of a honeycomb rotor adsorber operated with thermal swing, J. Chem.
Eng. Japan 27 (1 994) pp. 644-49.
2. Mitsuma Y., Yamauchi H., Hirose T., Analysis of VOC reversing adsorption and
desorption characteristics for actual efficiency prediction for ceramic honeycomb
adsorbent, J. Chem. Eng. Japan 31 (1998) pp. 253-57.
3. Kodama A., Goto M., Hirose T. and Kuma T., Experimental study of optimal
operation for a honeycomb adsorber operated with thermal swing. J. Chem.. Eng.
Japan 26 (1993) pp. 530-35.
4. Krishna S. M., Murthy S. S., Experiments on a silica gei rotary dehumidifier, Hear
Receovey Systems & CHP. 9 (1989) pp. 467-73.
5. Kodama, A, Goto M, Hirose T. and Kuma, T., Performance evaluation for a thermal
swing honeycomb rotor adsorber using a humidity chart, J. Chem. Eng. Japan. 28
(1 995) pp. 19-24.

290
 REVERSE FLOW ADSORPTION TECHNOLOGY FOR THE RECYCLING OF
HOMOGENEOUS CATALYSTS: SELECTION OF SUITABLE ADSORBENTS

J. DUNNEWIJK, H. BOSCH AND A.B. DE HAAN

University of Twente, Faculty of Chemical Engineering, Separatim Technology Group
P.O. Box 21 7, 7500AE Enschede, The Netherlands
Telephone: +3153 489 4288, Fa: +3I 53 489 4821, Email: JDunnewijk@ct,utwente.nl
A promising concept for the recovery of homogeneous catalysts is Reverse Flow Adsorption. In
actual homogeneous catalyzed processes, a homogeneous transition-metal catalyst is at equilibrium
with its individual components: the fiee transition-metal center and ligands. Therefore, to apply
Reverse Flow Adsorption, a combination of two adsorbents has to be used to reversibly adsorb: the
transition-metal center and its ligands. The transition-metal center can be adsorbed by a suitable
ligand immobilized onto a solid carrier, while the ligand is adsorbed by an immobilized transition-
metal. Dichlorobis(triphenyIphosphine)cobaIt(lI), dissolved in 1-butan01 has been selected as a
homogeneous model catalyst. The adsorption of Co(I1) and PPh3 has been determined for two
groups of functionalized adsorbents: 1) nitrogen (Amberlyst ,421) and phosphor (polymerbounded
PPh3) bctionalized adsorbents and 2) transition-metal (A$ and Co") functionalized Amberlyst 15.
The Co(I1) adsorption decreased, as predicted by the HSAB theory, according to: N > P. Transition-
metal fbnctionalized adsorbents proved the adsorption of the PPh3according to: Ag' > Co".

1 Introduction

Homogeneous transition-metal catalysts offer a number of advantages [I] when compared

to heterogeneous catalysts. Higher selectivities are achieved due to the well-defined and
adaptable ligand structures of the homogeneous catalyst. Mass transfer resistances are
negligible because of the high degrees of dispersion of the reactants, products and the
homogeneous catalyst in one single reaction phase. And consequently, homogeneous
catalyzed processes are performed at relative mild reaction conditions in comparison to
heterogeneous catalyzed processes.
In spite of these advantages, homogeneous catalysis is still not as common in use as
heterogeneous catalysis due to the various draw-backs during the usual methods for the
recovery and recycling of homogeneous catalysts. Various processes have been proposed
for the recovery of homogeneous catalysts [I]: decomposition, distillation, extraction,
membrane filtration and - more recently - phase transition by using fluorous media [2].
However, these separations include additional solvents and/or are operated at process
conditions that negatively influence the stability of the homogeneous catalyst.
Recovery of homogeneous catalysts by adsorption excludes the need for additional
solvents. The combination of a reversible adsorption with the reverse flow technology [3]
- Reverse Flow Adsorption - is a potential method for the integrated recovery and
recycling of homogeneous catalysts. With the right choice of adsorbent, the stability of
the homogeneous catalyst is preserved as the adsorption can be carried out within the
stability region of the homogeneous catalyst, for instance at reaction conditions. The
catalyst is separated from the product flow by adsorption downstream the reactor. In the
subsequent step, the catalyst is recycled by desorption from the saturated adsorbent by
reversal of the process flow (figure I).
 Figure 1. Homogeneous catalyst recycling by Reverse Flow Adsorption (feed alternatesi between A, and
I

In this paper, the Hard and Soft Acids and Bases (HSAB) theory [4] is applied to
select potential adsorbents for the reversible adsorption of transition-metal complexes.

2 Approach

In actual homogeneous catalyzed processes, a homogeneous transition-metalcatalyst is in

equilibrium with its free transition-metal center and ligands. An excess of ligands is
normally added [l] to decrease the amount of free transition-metal which negatively
influences the selectivity of the reaction. Therefore, to apply Reverse Flow Adsorption, a
combination of two adsorbents has to be used to reversibly adsorb: the transition-metal
center and the excess of ligands.
The above mentioned equilibrium - o-bondn-backbond - also exist between the free
transition-metal- or ligand - and its immobilized counterpart if one of the components is
bound to a solid carrier (figure 2). The HSAB theory gives a first-order prediction for the
strength of the interaction between a transition-metal and its ligand. For Co(II), a
transition-metal with borderline acid strength, it is expected that the interaction with a
group V element containing ligand decreases according to: N > P > As > Sb. For a given
soft ligand, the trend is predicted to decrease according to: Ag' > Co' > Fez+> Co2+.

Ligand +Metal - Ligand Ligand - Metal + Ligand

Metal + Ligand Metal - Ligand

(b)
Figure 2. (a) Transition-metal adsorption by an immobilized ligand and (b) ligand adsorption by an
immobilized transition-metal.

Cobalt and tipbenylphosphine (PPh3) ligands are commonly encountered in

homogeneous catalyzed processes. Therefore, dichlorobis(triphenylphosphine)cobalt(II)
has been selected as a homogeneous model catalyst. In 1-butanol, this complex is in
equilibrium with the flee Co(II)CI2 and PPh3 ligands.

292
 We studied two groups of adsorbents, based on their interactions with the Co(I1)
transition-metal center or PPh3 ligands:

0 Nitrogen and phosphor functionalized adsorbents for the adsorption of Co(I1).

Hereby, Amberlyst A21 was selected for its nitrogen functionality. It is a
macroreticular polystyrene - crosslinked by divinylbenzene - anion exchange resin
fimctionalized with an alkylamine group. As a phosphorous functionalizedadsorbent,
polymerbounded PPh3has been selected. It is a gel-type polystyrene - crosslinked by
2 [%I divinylbenzene - resin functionalized with a PPh2group.

Ag' and Co2+functionalized adsorbents for the PPh3 adsorption. These transition-
metal functionalizedadsorbents were prepared by immobilizing Ag' and Co2' onto a
solid carrier, for which Amberlyst 15 has been selected. Amberlyst 15, a
macroreticular polystyrene - crosslinked by divinylbenzene - sulfonated cation
exchange resin, has been selected as carrier because of its large pore diameter of
approximately 100 [nm]. These macropores ensure the accessibility for the relatively
large PPh3 ligands.

3 Experimental

The three selected carriers were: Amberlyst A21 (4.8 [mmol N/g dry], Sigma-AIdrich),
polymerbounded PPh3 (3.0 [mmol P/g dry], Sigma-Aldrich) and Amberlyst 15 (4.7
[mmol H/g dry], Sigma-Aldrich). The Amberlyst 15 was firstly washed with de-ionized
water (Millipore) in a column. The functionalization of the Amberlyst 15 was done by
contacting the resin with either 0.1 [mM] CoCl2 (98 [%I, Sigma-Aldrich) or 0.1 [mM]
AgN03 (extra pure, Sigma-Aldrich) aqueous solutions. During the ion-exchange, the
hydrogen of the Amberlyst 15 was exchanged for the transition-metal. The exchange was
done until maximum loading was reached. All four adsorbents were pre-rinsed with de-
ionized water. Then, the remaining water was rinsed out of the resin with methanol (p.a.,
Merck). The remaining methanol was rinsed with 1-butanol (p.a., Merck). The adsorbents
thus prepared were taken fiom the column and used in the adsorption experiments.
The adsorption characterizations of the various adsorbents were done via batch
adsorption experiments. The series of nitrogen and phosphorous functionalized
adsorbents = 0.3 [gr]) were contacted in erlenmeyer flasks with 10 [ml] of
dichlorobis(triphenyIphosphine)cobalt(II) (98 [%I, Sigma-Aldrich) at various
concentrations of 1, 2, 4 and 8 [mM]. The Ag' and Co2+functionalized adsorbents were
conctacted with 10 [ml] PPh3 solutions of 2, 4, 8 and 16 [mM]. The erlenmeyer flasks
were then equilibrated at 90 ["C] in a thennostated shaking water bath for 15-16 [hr]
(approximately 5 times the real equilibration time). The liquid phases were decanted and
analyzed. UVNis spectroscopy (Shimadzu 2501) was used for the determination of the
PPh3 concentrations at 265 [nm]. The Co(I1) concentrations have been analyzed by AAS
(Varian Specrtaa 1 10). After adsorption, all equilibrated samples were contacted with 10
[ml] of fresh I-butanol for 15-16 [hr] at 90 ["C] to investigate the reversibility of the
adsorption. The equilibrium concentrations after desorption of the relevant components
were measured as described above. The amounts adsorbed were calculated from the
differences in initial and equilibrium amounts. The loading of the adsorbents after
adsorption and desorption are expressed with respect to the number of functional sites - N,
P, Ag' or Co2' - in the adsorbents.

293
4 Experimental results

The results of the Co(II) adsorption and desorption experiments over the nitrogen and
phosphor functionaiized adsorbents are presented in figure 3. The Co(II) loading onto
these two selected adsorbents is shown as a function of the equilibrium concentration of
Co(I1) in the liquid phase.

0.0 2.0 4.0 6.0

[&I
Go@)
Figure 3. Co(1I) adsorption (closed squares) and desorption (open squares) onto nitrogen functionalized
Amberlyst A21 and Co(II) adsorption (closed triangles) and desorption (open triangles) onto phosphorous
functionalized polymerboundedPPh,.

It can be concluded from figure 3, that the Co(II) adsorption onto Amberlyst A21
(closed squares) is strong. This adsorption is reversible, as the desorption results (open
squares) are located on the adsorption isotherms. The poIymerbounded PPh3 adsorbent
shows a less strong, but reversible Co(II) adsorption (closed triangles). The stronger
Co(II) adsorption by the nitrogen functionalized adsorbent was expected from the HSAB
theory. For both adsorbents, no PPh3 adsorption has been observed.
The experimental results of the PPh3 adsorption and desorption on both transition-
metal - Ag+ and Co" - functionalized adsorbents are shown in figure 4 as a function of
the equilibrium PPh3 concentration.
Because the immobilized Ag+ can be exchanged for Co2+from the homogeneous
model catalyst, the transition-metal functionalized adsorbents have only been contacted
with PPh3 solutions. Thus, to avoid the exchange of an immobilized transition-metal,
Reverse Flow Adsorption requires two separate adsorption beds. The transition-metal
center has to be recovered before the adsorption of the ligands. Figure 4 demonstrates that
the PPh3 adsorption (closed triangles) onto Co2+functionalized Amberlyst 15 is weak.
The low degree of adsorption is caused by two effects: 1) the interactions between Co2+
and PPh3 - as predicted by the HSAB theory - are small and 2) the steric effects of the
SOigroups with the relatively large PPh3. One Co2+is immobilized onto two SOi groups.
As predicted by the HSAB theory, the Ag+ functionalized Amberlyst 15 shows a
larger PPh3 adsorption (closed squares) compared to the immobilized Co2+. The
experimental results - the open symbols - indicate that the PPh3 desorbes from the
transition-metal functionalized adsorbents. However, no complete desorption was
observed, indicating that the desorption time has been taken to short.

294
 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0
CPPh3 [d]
Figure 4. PPh3 adsorption onto (closed squares) Ag' and (closedtriangles) Co2' and desorption from (open
squares) Ag' and (open triangles) Co2' functionalized Ambedyst 15.

5 Conclusion

To apply Reverse Flow Adsorption, a combination of two adsorbents has to be used for
the reversible adsorption of a homogeneoustransition-metal catalyst. The transition-metal
center can be adsorbed by a suitable ligand immobilized onto a solid carrier, while the
ligand is adsorbed by an immobilized transition-metal. Two groups of adsorbents have
been studied, based on the HSAB predictions on the interactions with the Co(I1)
transition-metal center or PPh3 ligands:

0 Nitrogen and phosphor - group V elements - functionalized adsorbents showed to

reversibly adsorb Co(I1) according to the HSAB theory: N > P.
0 The PPh3 ligands were adsorbed - as predicted by the HSAB theory - by Ag+ and
Co2+functionalized adsorbents according to: Ag+ > Co".

To avoid the exchange of the immobilized transition-metal, for the transition-metal

of the homogenous catalyst, Reverse Flow Adsorption requires two separate adsorption
beds. The first bed for the recovery of the transition-metal center and the second bed for
the ligand adsorption.

References

1. S. Bhaduri, et al, Homogeneous Catalysis; Mechanisms and Industrial Applications,

Wiley-Interscience,200 1
2. A. Behr, et al, Temperature Depended Solvent Systems; An Alternative Method for
Recycling Homogeneous Catalysts, proc. ECCE3,200 1
3. J. Dunnewijk, H. Bosch, A.B. de Haan, Reverse Flow Adsorption Technologyfor
Homogeneow Catalyst Recovery, proc. ISMR-2,200 1
4. R. G. Pearson, Absolute Electronegativity and Hardness: Application to Inorganic
Chemistry, Inorg. Chem., 27,734-740, 1988

295
 MOLECULAR SIMULATION OF GAS SEPARATION
BY ADSORPTION PROCESSES

J. P.B. MOTA
Departamento de Quimica. Cenm de Quimica Fina e Biotecnologia, F a l a h i e de Ciincias e
Tecnologia. Universidade Nova de Lisboa, 2829-51 6 Caparica,Portugal
A new molecular simulation te-chnique is developed to solve the pernubation equations for a
multicomponent, isothermal stirred-tank adsorber under equilibrium controlled conditions. The
method is a hybrid between the Gibbs ensemble and Grand Canonical Monte Carlo methods,
coupled to macroscopic material balances. The bulk and adsorbed phases are simulated as two
separate boxes, but the former is not actually modelled at the atomistic level. To the best of our
knowledge, this is the first attempt to predict the macroscopic behavior of an adsorption process
from knowledge of the intermolecular forces by combining atomistic and continuum modelling
into a single computational tool.

1 Introduction
Process modelling is a key enabling technology for the development, design and optimiza-
tion of every adsorption process. However, its success is critically dependent upon the
accurate description of adsorption equilibriumand kinetics. Molecular simulation has now
developed to the point where it can be useful for quantitative prediction of those proper-
ties. Although there are several molecular simulation methodologies currently available,
bridging techniques, i.e. computational methods used to bridge the range of spatial and
temporal scales, are still largely underdeveloped. Here, we present a new molecular sim-
ulation method that bridges the range of spatial scales, from atomistic to macroscale, and
apply it to solve the perturbation equations for a multicomponent, isothermal stirred-tank
adsorber under equilibrium controlledconditions.

2 Problem formulation
Consider an isothermal stirred-tank adsorber under equilibrium-controlledconditions. q
is the bulk porosity (volumetric fraction of the adsorber filled with fluid phase), qp is the
porosity of the adsorbent, Fi 2 0 is the amount of component i added to the adsorber in
the inlet stream, and Wi 2 0 is the correspondingamount removed in the outlet stream;
both fi and Wi represent amounts scaled with respect to the adsorber volume.
The differential material balance to the ith component of an m-component mixture in
the adsorber yields

where ci and qi are the concentrationsin the fluid and adsorbed phases, respectively. Since
the fluid phase is assumed to be perfectly mixed,
dWi =yidW = c i d C , (2)
where yi is the mole fraction of component i in the fluid phase and dG is the differential
volume of fluid (at the conditions prevailing in the adsorber)removed in the outlet stream,
scaled by the adsorber volume. Substitution of Eq. (2) into Eq. (1) gives
q dci + (1 - q)qp dqi = d E - Ci dG. (3)

296
When Eq. (3) is integrated from state n - 1 to state n, the following material balance is
obtained:

In Eq.(4) the superscript denotes the state at which the variable is evaluated and

is the average concentration of component i in the volume AG(") of fluid removed in

the outlet stream. If AG(")is small enough, then a first-order implicit approximation for
Eq. ( 5 ) holds,

and Eq. (4) can be approximated as

Given that the inlet value A FYI is an input parameter, the terms on the r.-h.-s. of Eq. (7)
are known quantities. To simplify the notation, the r.-h.-s. of Eq. (7)is condensed into a
single parameter denoted by wi and the superscripts are dropped. Eq. (7)can be written in
this shorthand notation as

(q + AG)ci + (1 - q)qpqi = Wi. (8)

This equation requires a closure condition which consists of fixing the value of either AG
or the pressure P at the new state.
Here we show that Eq. (8), together with the conditions of thermodynamic equilib-
rium for an isothermal adsorption system (equality of chemical potentials between the two
phases), can be solved using the Gibbs ensemble Monte Car10 (GEMC) method in the
modified form presented in the next section.

3 Simulation method
In the GEMC method' the two phases are simulated as two separate boxes, thereby avoid-
ing the problems with the direct simulation of the interface between the two phases. The
system temperature is specified in advance and the number of molecules of each species
i in the adsorbed phase, Nip, and in the bulk, N ~ Bmay , vary according to the constraint
+
NjB Nip = Ni, where Nj is fixed.
If Eq. (8) is rewritten in terms of N ~ B
and Nip, the following expression is obtained:

where NA" is avogadro's number and Vp is the volume of the box simulating the adsorbed
phase. The value of Cj has been expressed as a function of Vp instead of the volume VB of
the box simulating the bulk fluid. The reason for this is that Vp is always fixed, whereas,
as we shall show below, VB must be allowed to fluctuate during the simulation when the

297
pressure is an input parameter. Obviously, for Eq. (9) to be valid the values of VB and Vp
must be in accordance with the relative dimensions of the physical problem, i.e.

Since the GEMC method inherently conserves the total number of molecules of each
species, Eq. (9) is automatically satisfied by every sampled configuration provided that
each Ci turns out to be an integer number. This feature makes the Gibbs ensemble the
natural ensemble to use when solving Eq.(9). Unfortunately,in general it is not possible to
size VB and Vp according to Eq. (10) and Eq. (9) so that each Ci is an integer number. To
overcome this problem, Eq. (9) is satisfied statistically by allowing Ni to fluctuate around
the target value Ci so that the ensemble average gives

(Ni) = Ci. (11)

This approach is different from that employed in a conventional GEMC simulation in
which Ni is fixed. When AG is an input parameter, the sizes of the two simulation boxes
are fixed and their volumes are related by Eq. (10). On the other hand, when the pressure
of the bulk fluid is imposed, the volume VB must be allowed to fluctuate during the simu-
lation so that on average the fluid contained within it is at the desired pressure. Once the
ensemble average ( VB) is determined, the value of AG follows from Eq. (10):

It is shown in detail elsewhere2 that if an equation of state for the fluid phase is known,
the bulk box does not have to be explicitly modelled computations on the bulk box amount
to just updating the value the NiB as the configuration changes. Thermodynamic equilib-
rium between the two boxes is achieved by allowing them to exchange particles and by
changing the internal configurational of volume Vp. The probability of acceptance of the
latter moves (molecule displacement, rotation, or conformational change) is the same as
for a conventional canonical simulation:

min{1, exp(-pAU)], (13)

where B = l / k ~ Twith, kg the Boltzmann's constant, and AU is the internal energy
change resulting from the configurational move.
The transfer of particles between the two boxes forces equality of chemical potentials.
The probability of accepting a trial move in which a molecule of type i is transferred to or
from volume Vp is, respectively,

298
where U ( S ~ ~ Pis+ the ~ ) internal energy of configuration s ~ ~ P in+ volume
~ Vp, NB =
.
[ N I B ,. ., N ~ B ]and
, ~ ( N Bk ), is the fugacity of species i in a gas mixture at tempera-
ture ?' and mole-fraction composition
Yl =- NIB ... , yi = - NiB k + ... , ym = NNmB
-
NB+k' N B + k ' B+k'
How the equation of state is actually employed to compute depends on the type of
problem being solved. If AG is an input parameter, Vg is fixed throughout the simulation
and the gas mixture is further specified by its number density p ~ ~ +=k (NB k)/ VB. +
If, on the other hand, the pressure is fixed, its value defines the state of the mixture. The
statistical mechanical basis for Eqs. (14) and (1 5) is discussed elsewhere.2
All that remains to completethe simulation procedure is to generatetrial configurations
+
whose statistical average obeys Eq. (1 I). To get the best statistics Ni = NiB Nip must
fluctuate with the smallest amplitude around the target value Ci, which is the case when
+
Ni takes only the two integer values int(Ci) or int(Ci) 1. It is straightforward to derive
that for Eq. (1 1) to hold, the probability densities of finding the system in one of the two
configurationsmust be
nl(N + int(Ci)) O( 1 - S i , N ( N i + int(Ci) + I ) oc S i , (17)
where 6i = Ci - int(Ci). In order to sample this probability distribution, a new type
of trial move must be performed which consists of an attempt to change the system to
+
a configuration with int(Ci) or int(Ci) 1 particles. It is highly recommended that the
box for insertionhemoval of the molecule always be the bulk box (except for the rare
cases that NiB becomes zero). This choice is most suited for adsorption From the gas phase
where, in general,the bulk phase is much less dense than the adsorbed phase and, therefore,
more permeable to particle insertions. Furthermore, given that the bulk box is not actually
modelled the molecule insertionhemovalmove amounts to just updating the value of N ~ B .

4 Application example
Due to lack of space, the few results presented here are primarily intended to demonstrate
the validity of the proposed method. The pore space of the adsorbent is assumed to consist
of slit-shaped pores of width 15 A, with parameters chosen to model activated carbon. The
porosity values are fixed at q = 0.45 and q p = 0.6. The feed stream is a ternary gas mixture
of H ~ / C H ~ / C ZThe
H ~ .vapor-phase fugacities were computed from the virial equation to
second order, using coefficients taken from Reid et aL3
Methane and ethane were modelled using the TraPPESunited-atom potential in which
CH4 and the CH3 group are considered as single Lennard-Jones interaction centers. The
LJ parameters for hydrogen, ~ / k g= 38.0 K and 0 = 2.915 A, were taken from Turner
et aL6 The potential cutoff was set at 14 A, with no long-range corrections applied. The
interactions with the carbon walls were accounted for using the 10-4-3 Steele p~tential.~
The cross-term LJ contributionsbetween all molecules were calculated using the Lorentz-
Berthelot mixing rules. The simulations were equilibrated for 1O4 Monte Carlo cycles,
where each cycle consists of N attempts to change the internal configurations of volume
VP (equally partitioned between translations, rotations and conformational changes) and
N / 3 attempts to transfer molecules between boxes. Each particle molecule attempt was
followed by a trial move to adjust the total number of molecules of that type.

299
 Table 1. Results of simulations.The subscripts indicate the estimated error in the last digit of the value.

The simulations reported here consisted of pressurizing an initially evacuated adsor-

ber with four mixtures of different compositions. These simulations are very much like
the traditional flash calculations of chemical engineering thermodynamics applied to an
adsorption system. The first set of runs, which we refer to as set G-NVT,is equivalent
to solving Eq. (4) with c!') = 0, 41') = 0, AF!') = yj"AF(') 2 0, closure condition
A G ( ' ) = 0, and the equilibriumpressure as output of the simulation. Then, a second set of
runs (G-NPT) was performed with the same feed mixtures and the pressure fixed at the val-
ues obtained from the G-NVT runs. In this case ( VB)is an output of the simulation. Finally,
a third set of runs (GCMC) was camed out to check the validity of our simulation tech-
nique. These runs consisted of standard multicomponent GCMC simulationswith mixture
fugacities calculated from the virial equation of state using the pressures and gas-phase
compositionsobtained in the G-NVT runs.
The results obtained are listed in Table 1. The three sets of runs give the same results
to within the statistical uncertainty of the simulations, which attests to the viability of the
proposed method. The total number of molecules employed in the G-NVT and G-NPT runs
is given in the 8th column of the table. They were purposively set to noninteger values to
test the efficiencyof the method in generatingtrial configurationswhose statistical average
obeys Eq. (1 1). As can be easily verified, there is very good agreement between (NB) +
(Np) and the imposed N value for every run. The theoretical approach presented here
represents a successful attempt to develop an ab-initio or first-principles computational
methodology to predict the macroscopicbehavior of an adsorption process from knowledge
of the intermolecularforces and structural characteristicsof the adsorbent.

References

1. A.Z. Panagiotopoulos,Molec. Phys. 61,8 13 (I 987).

2. J.P.B. Mota, J. Chem. Phys, submitted (2002).
3. R.C. Reid, J.M. Prausnitz, and B.E. Poling, The Properties of Gases and Liquids, 4th
ed. (McGraw-Hill, Singapore, 1987).
4. W.A. Steele, f i e Interaction of Gases wih Solid Surfaces (Pergamon,Oxford, 1974).
5. M.G. Martin and J.1. Siepmann,J. Phys. Chem. B 102,2569 (1998).
6. C.H. Turner, J.K. Johnson,andK.E. Gubbins,J. Chem. Phys. 114,1851 (2001).
 METAL-DOPED SODIUM ALUMINIUM HYDRIDE AS A REVERSIBLE
HYDROGEN STORAGE MATERIAL

JUN WANG'. ARMIN D. EBNER~,KEITH R. EDISON',

JAMES A. RIT"ER' AND RAGAN ZIDAN'
'Department of Chemical Engineering, Swearingen Engineering Center
University of South Carolina, Columbia, SC 29208, USA,E-mail: [email protected]
zWestinghouse Savannah River Company
Savannah River Technology Center, Aiken, SC 29804. USA

In an ongoing effort to reduce the kinetic limitation of the dehydrogenation of NaAIH4, while
maintaining sufficient HZcapacity, the effect of different transition metal catalysts (Ti, Zr, Fe) in
various combinations have been investigated using thennopvimhic analyses. The Ti doped
systems, in all cases, exhibited the lowest Hz desorption temperam, with the HZdesorption kinetics
improving with an increase in the Ti loading, but at the expense of decreasing the Hz capacity. In all
samples doped with 4 mole% combinations of Ti, Zr and Fe, the Ti played the most important role;
however, an interesting synergistic behavior was revealed when doping NaAIb with 1 mole% Fe
and 3 mole% Ti. Overall, these results continue to prove that doping NaAW with transition metals,
especially Ti, improves the Hz dehydrogenationkinetics, but much more research needs to be done.

1 Introduction

Metal-doped N a A l b is becoming a very promising material for H2 storage because it

contains a high concentration of useful hydrogen (5.6 wt%). At standard conditions, the
dehydrogenation of N a A l b is thermodynamically favorable, but it is kinetically slow and
takes place at temperatures well above 200C in a two-step process involving the
following reactions:I4

3NaAlh + Na3A1H6+ 2A1+ 3H2 (1)

The first work on the doping of NaAlK with Ti used solution chemistry techniques,
whereby nonaqueous solutions of N a A l b and either TiC13 or Ti(0Bu"k catalyst
precursors were decomposed to solid Ti-doped NaAlb.' Zidan et aL2 and other
investigators3-' discovered later that a further lowering of the dehydrogenation
temperature was highly dependent on the doping and homogenization procedures. They
also found that Zr when mixed with Ti improved the dehydrogenation reversibility of
N a A l b over Ti alone. These favorable effects of using mixed metals as the dopant
generated interest in trying other combinations of mixed metal catalysts. The objective of
this study is to show the effects of Ti, Fe, Zr and their combinations on the H2 desorption
kinetics of NaAlh.

2 Experimental

TiC13 (Aldrich), FeC13 (Aldrich, 99.99%. anhydrous) and ZrCb (Aldrich, 99.9%) were used
as received as the catalyst precursors. Crystalline NaAl& (Fluka) was purified fiom a THF
(Aldrich, 99.9%,anhydrous) solution and vacuum dried. The dried N a A l b was mixed with

301
a predetermined amount of catalyst in THF to produce a doped sample in the desired
concentration up to 4 mole96 total metal. Samples containing a single catalyst or a
combination of them were all prepared in this manner. The THF was evaporated while
the NaAlh and the catalyst were mixed manually for about 30minutes using a mortar and
pestle, or until the samples were completely dry. These mixtures were then ball-milled for
2 h, using a high-energy SPEX 8O00 mill. The above procedures were carried out in a N r
laden glove box free of oxygen and moisture.
A Perkin-Elmer thermogravimetric analyzer (TGA) was used to determine the
hydrogen desorption kinetics at atmospheric pressure. This instrument was located in
another glove box under nitrogen atmosphere to prevent any exposure of the samples to
air and moisture. Samples were heated to 250C at a ramping rate of 5"Umin under 1 atm
of He, using an initial 1 minute delay to ensure an environment of pure He. Approximately
10 mg of sample were used in the TGA.

3 Resultsanddiscussion

Figure la shows the TGA results for catalyzed NaAl& with 1 to 4 mole% TiC13. The 4
mole% Ti sample exhibits the best behavior with respect to the H2 desorption kinetics,
while the 1 mole% Ti sample has the highest H2 capacity. In the recent study by Sandrock
et al,6 they found that the TiC13 was completely reduced by Na in the NaAl& to form
NaCl and most likely zero-valent Ti. This solid state reaction can be written as:

(1-x)NaAl&+xTiC13+( 1-4x)N~+3xNaCl+xTi+3xAl+6xH~ (3)

where x is the mole fraction of TiC13 in the NaAl&. This reaction shows that the H2
capacity depends on the amount of TiC13 in the sample. Theoretically, after doping with 4
mole% Ti, the H2 capacity decreases to 4.6 wt%; the experimental value obtained here is
very close to this value at 4.5 wt%. Clearly, the Tic13 loading has a negative effect on the
H2 capacity. In contrast, the TiC13 loading has a positive effect on the H2 desorption
kinetics, which increases with increasing TiC13loading.
Figure 1b shows the TGA analyses for NaAl& doped with 4 mole% each of the three
different metal chlorides. The 4 mole% Ti sample exhibits the best behavior with respect
to the H2 desorption kinetics, followed by 4 mole% Zr and then 4 mole% Fe. This result
confirms that Ti by itself is the best catalyst with respect to the kinetic behavior. Figure 2a
shows the TGA analyses for NaAlH., catalyzed with 4 mole% metal, but in different
combinations and with each containing with 1 mole% Fe. The 1 mole% Fe-3 mole% Ti
sample exhibits the best behavior with regard to the Hz desorption kinetics and again the
kinetics increase with increasing Ti loading. Figures 2b, 3a and 3b compare the 1 to 3

302
 5.0

f 4.0
[ 3.0
s
1 1.o
2.0

0.0
50 100 150 200
cc)
Figure 1. TGA analyses of NaAIH4doped with a) 1 to 4 mole% Ti; and b)
varying amounts of the three pure metal chloride catalysts.

mole% Ti samples with different amounts of Fe and Zr and Ti itself. All the mixed metal
samples with the same Ti loading have similar profiles, i.e., the samples with 1 or 2
mole% Ti exhibit similar kinetics and Hzcapacity. However, by comparing with Ti alone
at the same loading, the 1 or 2 mole% Ti mixed with different metals show an improved
kinetic profile, while losing some Hzcapacity. Surprisingly, the 1 mole% Fe-3 mole% Ti
sample is better than the 4 mole% Ti sample with respect to HZdesorption kinetics, but it
does nothing for improving the kinetics of the second reaction depict in eq 2. This
synergistic behavior with the Fe-Ti mixed catalyst system is very interesting and needs to
be explored in more detail. In general, however, all the samples containing Ti exhibited
the best behavior.

303
 f 4.0

1.o

0.0

5.0

4.0

1
1
3.0

i 20

1.o

0.0

Figure 2. TGA analyses of NaAI& doped with 4 mole% metal in different combinations
with each containing a) 1 mole% Fe; and b) 1 mole% Ti.

4 Conclusions

These results continue to prove that doping NaAlh with transition metals, especially Ti,
improves the Hz dehydrogenation kinetics. However, more research needs to be done to
lower the dehydrogenation temperature even further, especially for the second reaction
depict in eq 2. In this study, the effect of the different transition metals played an
insignificant role in reducing the temperature or increasing the kinetics of the second
reaction. Other metals and metal combinations are currently being explored for this
reason, and to further reduce the temperature (increase the kinetics) of the first reaction.

304
 5.0

f 4.0
2KTklxzr+l%F.

I ao

1::
0.0

1
4.0

3.0

g 2.0

1 1.0

0.0

Figure 3. TGA analyses of NaAlH4doped with a) 4 mole% metal with at least 2 mole%
Ti and varying amounts of Fe and Zr; and b) 1 mole% Fe-3 mole% Ti, and 4 mole% Ti.

5 Acknowledgements

Financial support was provided by SCURJZF/WSRC/DOE under contract WEST052,

KGO9725-0. and the NRO under contract NRO-00-C-0134.

References

1 B. Bogdanovic, M. Schwickardi, J. Alloys Comp. 253 (1997) 1.

2 B. Bogdanovic, R. A. Brand, A. Marjanovic, M. Schwickardi, I. TolIe, J. Alloys
Comp. 302 (2000) 36.
3 R. A. Zidan, S. Takara, A. G. Hee C. M. Jensen, J Alloys Comp. 285 (1999) 119.
4 C. M. Jensen, R. Zidan, N. Mariels, A. Hee, C. Hagen, Inter. J. Hydrogen Energy 24
(1999) 461.
5 C. M. Jensen, K. J. Gross, Appl. Phys. A Mat. Sci. Proc. 72 (2001) 213.
6 G. Sandrock, K. J. Gross, G. Thomas, J Alloys Comp. 339 (2002) 299.

305
 SYNTHESIS AND DEHUMIDIFICATION BEHAVIORS OF MONODISPERSE
SPHERICAL SILICA GELS WITH DIFFERENT PORE AND CHEMICAL
STRUCTURES

C.H.CHO, Y.J. YOO,J. S.KIM, H.S.KIM, Y.S.AHN AND M.H.HAN

Centerfor Functional Materials Research, Korea Institute of Energy Research, 71-2 Jang-dong,
Yusong-gu, Taejon 305-343, Korea
E-mail: [email protected]

Monodisperse spherical silica gels were prepared by aging a gel precursor in different basic
conditions. The precursor was synthesized by the SFB process and was monodisperse, spherical and
200nm sized in diameter. The specific surface area of the precursor was so small ( I 6m2/g). In the
precursor, there was no pore except the micropores of which average diameter was around 16A. On
the aging process, there was no spherical morphology change. As the basic condition strengthened,
the specific surface area minutely increased to 25m2/g. In addition to the pore structure, the chemical
structure of the precursor didn't change in the aging process. All the silica gels prepared in the
present study showed the same Si-NMR spectrum in which main Q2 and Q3 peaks and minor Q'
peak appeared. Clearly the reaction to form anhydrous silica has not gone to completion as
evidenced by the complex distribution of Q' through Q' species. In the present study, effect of
synthesis and aging conditions on the pore and chemical structures of silica gels will be investigated
and then the relationship between pore and chemical structures and dehumidification behavior will
be discussed.

1 Introduction

Colloidal silica gels have been applied to numerous industrial fields such as thermal
insulation, catalyst supports, filters and sorbents[11. Their industrial performance is
dependent not only on their chemical structures but also on their physical pore structure,
which comprises pore sue, pore size distribution, pore volume and surface area.
Therefore, it is important to elucidate the relationship between the pore and chemical
structures and the performance efficiency.
From the viewpoint of materials processing, the pore and chemical structures are
affected by three kinds of processing conditions: one is concerned to the parameters of
sol-gel based synthesis (hydrolysis and condensation) route[2-5], another is related to the
aging and drying conditions of the wet gels[5-10], and the third is in close relationship to
the calcination conditions of the dried gels[ 1 11.
C. J. Brinker et al. showed that the pore size of silica gel increased by the simple
immersion in a basic solution[4]. R. K. ller has summarized the dependence of solubility
of amorphous silica in the aqueous solvent as a hnction of pH[1]. The solubility gently
decreases at the low pH values, reaches a minimum at pH 7-8, and shows a steep
increasing behavior above pH 8, as pH increases. It is expected that the aging of silica gel
in basic solutions increases the pore size due to the activated particle growth by the
Ostwald ripening process. Therefore, it is important to investigate the structural and
chemical change of silica gels during the aging in basic conditions.
In the present study, monodisperse spherical silica gels were prepared by aging a
monodisperse spherical silica gel precursor in different basic conditions, and then the
effect of the basic strength in aging process on the pore and chemical structures were
investigated.

306
2 Methods

2. I Synthesis of Silica Gels

Silica gels with different particle size were synthesized by the SFB process[ 121, a
base-catalyzed synthesis route. In the present study, the particle size was controlled by
changing the volume ratio of H20 to EtOH and the addition amount of NH3, a
base-catalyst. In the synthesis process, tetraethoxysilane(TE0S) was used as a silica
source material.
Detailed preparation procedures of silica gels are as follows. At first, a mixed
solvent of ethanol and water was prepared by mixing the anhydrous ethanol(EtOH, 99.9%,
Hayman Chemical Co., England) with the lab.-made DI water. The volume ratio of H20to
EtOH was controlled to be from 0 to 1.5, and the total volume of mixed solvents was
constant to be 100ml. In the mixed solvent, H20 simultaneously plays both roles of a
solvent and a reactant. As a synthesis catalyst, NH3 aqueous solution (30%, Junsei
Chemical Co., Japan) was added in the mixed solvent and then homogenized by a
vigorous stirring process. The addition amount of NH3was changed from 2 to 12ml.
After the homogenization process, a 2ml of TEOS (99.9%, Aldrich Chemical Co.,
USA) was added and mixed into the catalyst-included mixed solvents. As reaction time
goes on after the TEOS addition, the hydrolysis and condensation reactions initiated, so
that the transparent solution became white and white. The reaction rate highly depended
on the reaction conditions such as the volume ratio of H20 to EtOH and the addition
amount of NH3. The synthesis temperature and time were room temperature and 4hrs,
respectively. After the synthesis reaction had proceeded for 4 hours, the synthesized silica
gel was washed with water three times by repeated centrifuging and dispersion in water,
and then dried at 110C for 72hrs. All the chemicals used in the present study were used
without any fkthermore purification.

2.2 Aging of Silica Get%h Basic Conditions

A silica gel with 400nm in diameter was used as a precursor for the as-following
aging process in basic conditions.
The 0.4g of dried silica gels were dispersed in lOOml NH3 aqueous solutions with
different contents of NH3 aqueous solutions (0, 0.01, 0.1, 0.5, 1, 2, 4, 6 and lOml), and
then aged in slow stirring mode for 24hrs. The aged gels were water washed at three times
by repeated centrifuging and stirring process, and then dried again at 110C for 72hrs.

2.3 Characterizationof Silica Gels

Particle morpholoa, size and distribution, phase, pore structure and chemical
structure of the synthesized silica gels were characterized by SEM (XL30, Phillips,
Holland), Laser Scattering(ELS-8000, Otsuka, Japan), XRD(DIMAX2000, Rigaku,
Japan), BET(ASAP2400, Micrometrics, U.S.A.1 Autosorb I, Quanta chrone Instrument,
USA.) and Si-NMR (DSX-300,Bruker, Germany) analysis, respectively. Before the
BET analysis, the silica gels were degassed at 210C for 4 hrs.

3 Results and Discussion

Generally, it is known that in the SFB process, particle size of silica gels is affected
by the processing parameters such as temperature, time, [NH3], [TEOS] and volume ratio
of H20 to EtOH. In the present study, the particle size was controlled to be 50 to 50Onm

307
by changing the volume ratio of H2O to EtOH and mH3], the addition amount of the base
catalyst.
Figure 1 represents the average particle size of the synthesized silica gels as a
function of the volume ratio of H20 to EtOH and mH3]. As the volume ratio of H20 to
EtOH increased, the particle size increased to maxima at about 0.1 of volume ratio of H 2 0
to EtOH, and then again decreased. In addition to the volume ratio, the content of NH3
affected the particle size. The particle size increased as the content of NH3 increased.

VokM. ntio of H,O Lo EIOH

Fig. 1 . Average particle diameter of silica gels as a function of the

volume ratio of H20 to EtOH and the addition amount of
NH3, a base-catalyst.

Figures 2(a) to (c) represent SEM images of the silica gels synthesized in the mixed
solvents with (a) 0, (b) 0.25 and (c) 0.67 volume ratio of H20to EtOH, respectively. In all
cases, the addition amount of NH3 in the synthesis process was 6ml. Without relation to
the volume ratio of H20 to EtOH, all the synthesized silica gels were monodisperse and
spherical. Also it was clearly shown that the particle size evaluated by the SEM analysis
was well in accord with the particle size results characterized by the laser scattering
method as shown in Fig. 1.

Fig. 2. SEM images of the silica gels synthesized in the mixed

solvents with (a) 0, (b) 0.25 and (c) 0.67 volume ratio of H 2 0
to EtOH.

G. H. Bogush ef al. have suggested that the growth of silica gel in the SFB process
should be governed by an aggregative growth mode1[13-14]. This model states that the
particle growth occurs due to an aggregation of primary particles that are nucleated in a

308
supersaturation of silica, and nowadays their model is generally accepted for the growth of
silica gel in a sol-gel process.
If the growth of the synthesized spherical gel is governed by the agglomeration
model, it is expected that the silica gels have high specific surface area, because the
spherical gels are composed of primary nanogels.
R. K. Iler has systematically summarized the dependence of solubility of amorphous
silica on pH[ 11. As pH increases, the solubility of amorphous silica gently decreases at the
lower pH, reaches a minimum at pH 7-8, and shows a steep increasing behavior above pH
8. The growth of primary particles in Ostwald ripening process will be activated by the
increment of NH3 content in the aging process. It is expected that the pore size can be
systematically controlled by aging the gels in different basic conditions, in other words, by
controlling the growth rate of primary nanogels. From the SEM analysis, it was known
that there was no spherical morphology change before and after the aging process in basic
conditions.
To confirm that the spherical silica gel comprises of primary nanogels, the pore
structure was analyzed by the BET analysis. Contrary to the expectations, the precursor
and aged silica gels have relatively very low specific surface area. In Table I, the specific
surface area of the precursor and aged silica gels was presented. As the addition amount of
NH3 increased, the specific surface area increased.

Table I Specific surface area of the precursor and aged gels

Addition amount of NH3
fml\ 0 0.01 0.1 1 10

Specific surface area

16.6 16.6 17.6 19.9 24.1
(m2/g)

The fact that the silica gels had low specific surface area means that the synthesized
silica gels were so dense. Therefore, the aggregative growth model suggested by G. H.
Bogush et al. might be impertinent for the silica gel prepared in the SFB process.

Fig. 3. Cumulative pore volume curves of precursor and aged silica gels as
a function of average pore diameter.

The cumulative pore volume curves as a hction of average pore diameter were
calculated to elucidate that the spherical silica gels are dense, and represented in Fig. 3. It

309
was clearly shown that all the silica gels have micropores with the pore diameter less than
20A and diffuse mesopores with the 10 to 20Onm of pore diameter. The diffuse mesopores
originated in the randomly loose packing of the spherical silica gels.
To investigate the detailed size distribution of the micropores, the BET analysis was
minutely conducted at the very low pressure of N2.Fig. 4 represents the detailed
incremental pore volume curves as a function of average pore diameter. It was clearly
shown that the pore diameter of the micropores was about I5.k irrespective of the aging
conditions.

-
Fig. 4. Incremental pore volume pore diameter curves for silica gels aged
with (a) 0.5. (b) 2 and (c) 4 ml of N H 3 .

It is interesting to answer the question where did the micropores originate in?. Fig. 5
represents the XRD patterns of the silica gels before and after the aging process. It is
obvious that there was no trace for the progress of crystallization during the aging process.
Therefore, it was concluded that the micropores didn't originate in the lattice pore in
crystalline phase.

,r I.0 I. I, (0 I. I0
I

Fig. 5. XRD patterns of silica gels aged in basic conditions.

310
 As mentioned in the introduction part, gas adsorption and desorption behaviors of
the porous solids depends not only on the pore structure but also on the chemical
structure. Therefore, it is interesting to investigate if the aging process can affect the
chemical structure of silica gels or not. Fig. 6 represents the Si NMR spectra of the
precursor and aged silica gelss. All the synthesized silica gels were mainly composed of
' and Q
the Q ' component, and also contained small amounts of Q'. The aging process
didn't affect the chemical structure of silica gels. Clearly, in the synthesis and aging
processes, the reaction to form anhydrous silica has not gone to completion as evidenced
by the complex distribution of Q' through Q3species.

Fig. 6. Si-NMR spectra of the silica gels (a) before and (b) after the
aging process with 6 ml of N H 3 aqueous solution.

4 Conclusion

Under the consideration that silica gels prepared by SFB process were so dense,
the growth of silica gels in SFB process was governed by Ostwald ripening process
rather than the agglomeration model. Silica gels prepared by base-catalyzed routes
such as the SFB process are not suitable for the preparation of adsorbents due to the
small specific surface area. Also, the aging process in basic conditions is not suitable
to control the pore and chemical structures of silica gel.
In the presentation, the effect of synthesis and aging conditions on the pore and
chemical structures of silica gels will be introduced and then the relationship between
pore and chemical structures and dehumidificationbehavior will be discussed.

5 Acknowledgements

This work was financially supported by National Research Laboratory Program (Korea
Ministry of Science and Technology).

31 1
References

1. R K. Iler, The Chemistry of Silica: Solubility, Polymerization, Colloid and Surface

Properties, and Biochemistry (John Wiley & Sons, New York, 1979)
2. C. J. Brinker and G. W. Scherrer, Ultrastructure Processing of Ceramics, Glasses, and
Composites (Wiley, New York, 1984).
3. C. J. Brinker and G. W. Scherrer, Sol-gel Science (Academic Press, New York, 1990).
4. A. Yasumori, M. Anma and M. Yamane, Phys. Chem. Glasses 30 (1989) 193.
5. D. C. L. Vasconcelos, W. R. Campos, V. Vasconcelos, W. L. Vasconcelos, Muter. Sci.
& Eng. A334(2002) 53.
6. M. Yamane and S . Okano, Yogyo-Kyokai-Shi87(8) (1979) 56.
7. R. Takahashi, K. Nakanishi and N. Soga, J. Non-Cryst. Solids 189( 1995) 66.
8. P. J. Davis, C.J. Brinker and D. M.Smith, J. Non-Ctyst. Solids 142(1992) 189.
9. P.J. Davis, C.J.Brinker, D. M.Smith and R. A. Assink, J. Non-Cryst. Solids 142(1992)
197.
10. J. H. Harreld, T. Ebina, N. Tsubo and G. Stucky, J. Non-Cryst. Solids 298(2002) 24 1.
11. Z. J. Li, C. R. Liu and Q. S. Zhao, J. Non-Ctyst. Solids 265(2000) 189.
12. W. Stiiber, A. Fink, E. Bohn, J. Colloid & Inter. Sci. 26(1968) 62.
13. G. H. Bogush and C. F. Zukoski IV,J. Colloid& Inter. Sci. 142(1991) 1.
14. G. H. Bogush and C. F. Zukoski IV, J. Colloid & Inter. Sci. 142(1991) 19.

312
 PRODUCTION OF HARD CARBONS FOR LITHIUM ION STORAGE BY THE
COXARBONIZATION OF
PHENOLIC RESIN PRECURSORS

S. R. M U M , T. TANIGAWA, T. HARADA AND H. TAMON

Dept. of Chem. Eng., Grad. School of Eng., Kyoto University Yoshida-Honmachi,Sahyo-ku, Kyoto
606-8501,Japan
E-mail: [email protected]~p

T. MASUDA
.
Div. of Materials Sci. and Eng. Grad. School of Eng..,Hokkuido UniversityNI3W8 Kita-ku,
Sapporo 060-8628,Japan
E-mail: takaeeng.hokudai.ac.jp

Hard carbons were synthesized by carbonizingvarious combinations of phenolic resin precursors in

order to obtain a material with a structure suitable to be used as the anode material in lithium ion
battery systems, i.e. a carbon with a large pore volume and small pore openings. The lithium ion
capacities of thus obtained carbons were also measured. From the obtained results, strategies to
obtain hard carbons with large reversible capacities and small irreversible capacities are proposed.

1 Introduction

Lithium ion batteries have dominated the market of portable secondary batteries, due to
their higher energy densities and longer shelf lives. In this battery system, carbon
materials that can store a significant amount of lithium ions within their structure are used
as the anode material. There are numerous types of carbon materials, but in commercial
cells, graphitic carbons are mostly used due to their high stability. However, as there is a
limit in the lithium ion capacity in this type of carbon (372 mAhg-), and materials with
capacities close to this limit have already been developed, worldwide scale research is in
progress to find alternative anode materials that possess higher capacities.
Among various types of carbon materials, hard carbons, which are predominantly
formed by graphene sheets stacked like a house of cards [4], is a potential alternative.
However, although most hard carbons possess large lithium ion reversible capacities, their
irreversible capacities are also rather large, making them difficult to be used in
commercial batteries. It is widely recognized that a large part of this irreversible
capacity arises from the formation of a passivation layer on the outer surface of the carbon
[2]. In hard carbons, such layers are also likely to be formed within its pores, which
leads to an increase in irreversible capacity. Previously, we showed that the irreversible
capacities of hard carbons highly depends on their pore structures, and hard carbons
which pore openings are small enough so that C02cannot penetrate into them tend to have
smaller irreversible capacities [ 5 ] . However, the pore volume of such hard carbons tends
to be small, and it is hard to expect large reversible capacities from such materials.
As there are a wide variety of hard carbon precursors, there is a high possibility to
obtain a hard carbon with a large pore volume and small pore openings by combining
precursors of different natures. In this work, hard carbons were synthesized by
carbonizing combinations of various phenolic resin precursors. The lithium ion
capacities of thus obtained carbons were also measured. From the obtained results,

313
strategies to obtain hard carbons with large reversible capacities and small irreversible
capacities are proposed.

2 Experimental

One feature of phenolic resins is that they are usually synthesized via several stages.
First phenols react with formaldehyde by the catalysis of an acid catalyst and novolac
resin is formed. These novolacs are usually cured with agents such as
hexatnethylenetetramine and a thermosetting resin is obtained. Hard carbons can be
obtained by carbonizing this thermosettingresin.
A wide variety of hard carbons can be obtained by carbonizing mixtures of phenolic
resins derived fiom different phenols that are at different synthesis stages. In this work,
first various phenolic resins were synthesized from combinations of different phenols and
formaldehyde. The phenols employed were pure phenol, o-cresol and 3,5-xylenol.
Note that the relative reactivities of these phenols with formaldehyde are 1:0.26:7.55,
respectively. Next mixtures of phenolic resins derived from different phenols that are at
different synthesis stages were combined and carbonized at 1273 K for 1 h, yielding
various types of hard carbons.
The pore volumes of the obtained hard carbons were measured using the molecular
probe method [3]. Adsorption isotherms of the probe molecules were measured at 298 K
using an adsorption apparatus (Be1 Japan, Belsorp 28). The employed probe molecules
were C02, C2H6, n-C4Hlo and i'C4H10 (minimum molecular dimensions: 0.33, 0.40, 0.43
and 0.50 nm, respectively). By applying the Dubinin-Astakhov equation (n-2) [11 to the
measured isotherms, the limiting micropore volumes corresponding to the minimum size
of the adsorbed molecules were determined.
Measurements of the lithium ion reversible and irreversible capacities of the samples
were conducted using a two-electrode cell at a constant current of 25 mAg-'. Cut off
voltages were set to 0 and 2.5 V. Lithium metal was used as the counter (reference)
electrode. The carbon electrodes were constructed by supporting ball-milled carbon to
copper foil using PVDF. The electrolyte used was a 1 M LiC104-EC/DEC(1: 1) solution
(Mitsubishi Chemicals). Celgard 2400 (Hoechst Celanese) was used as the separator.

3 Results and Discussion

Through preliminary experiments, it was found that the pore structures of carbonized
phenolic resins differ significantly when different phenols are used for synthesis. When
the reactivity of the phenol with formaldehyde is high, the resulting carbon tends to have
small pore openings, and if low, the pore volume tends to become large. Therefore the
carbonization of a combination of phenolic resins, one derived from phenols with high
reactivity and the other fiom phenols with low reactivity, is expected to give hard carbons
with large pore volume and small pore openings.

314
 By testing various resin combinations, it was found that the carbonization of a
mixture of 0-cresol derived phenolic resin at the thermosetting stage (OCN-R)with the
3,5-xylenol derived phenolic resin at the novolac stage (3,5XN) gave a hard carbon which
has a large pore volume and small pore openings.

0
0.30 0.35 0.40 0.45 0.50 0.55
Minimum Dimensionof Probe Molecule [nm]
Figure 1 Accumulated micropore volume distributionsof the obtained samples
Figure 1 shows typical micropore volume distributions of the carbonized mixtures
along with those of carbons obtained from the carbonization of 0-cresol derived phenolic
resin (Sample A) and 3,5-xylenol derived phenolic resin (Sample D). Sample A has a
large pore volume but the sizes of the pore openings are rather large. On the other hand,
the pore openings of Sample D are extremely small, and it is natural to assume that its
pore volume is also small. These structures reflect the reactivity of the phenols used for
resin synthesis. By carbonizing a mixture of 10 wt?? 3,5XN and 90 wt?? OCN-R, the
pore openings of the resulting carbon (Sample B) becomes smaller, but it maintains a pore
volume close to that of Sample A. When the amount of 3,5XN is increased to 20wt?h,
the pore openings of the resulting carbon (Sample C) also become smaller, with a slight
sacrifice of its pore volume.

315
 500
IReversibleCapacity1 [IrreversibleCapacity 1
Theoretical 400
Capacity ---- --
372 mAh g-l
300

200

100

0
Sample A B C D A B C D
Figure 2 Lithium ion capacities of the obtained samples

Figure 2 summarizes the results of the electrochemical measurements of the obtained

carbons. Sample A has a large lithium ion capacity, but the proportion of its irreversible
capacity is also large due to its large pore openings. The lithium ion capacity of Sample
D is not so large, but the proportion of the irreversible capacity is small which also
reflects its pore structure. When compared with Sample A, a large reversible capacity
increase, and a slight irreversible capacity decrease was observed in Sample B. When
the amount of 3,5XN was increased to 20 wt'?! (Sample C), the reversible capacity
decreased but there was a significant decrease in irreversible capacity. These results are
consistent with the pore structures of the tested carbons.
It is obvious that the narrowing of the pore openings is an effective way to minimize
the irreversible capacities of hard carbons. We believe that it is much more feasible to
directly synthesize a hard carbon with small pore openings rather than narrowing the pore
openings of a synthesized hard carbon. However, the pore volumes of hard carbons
which pore openings are small tend to be small when the carbons are synthesized using
typical methods. The carbonizing of mixtures of phenolic resins derived from different
phenols that are at different synthesis stages gives a wide variety of hard carbons with
various pore structures. This method is thought to be a promising method to obtain hard
carbons which pore structures lead to large reversible and small irreversible capacities for
lithium ion insertion.

4 Acknowledgements

This research was partially supported by Industrial Technology Research Grant Program
in '01 from New Energy and Industrial Technology Development Organization of Japan.

316
References

1. Dubinin M. M. and Astakhov V. A., Description of adsorption equilibria of vapors on

zeolites over wide ranges of temperature and pressure. A&. Chem. Series 102 (1 97 1)
pp. 69-85.
2. Fong R., Sacken U. and Dahn J. R., Studies of lithium intercalation into carbons using
nonaqueous electrochemical cells. J. Electrochem. Soc. 137 (1990) pp. 2009-13
3. Lamond T.G.,Metcalfe J. E. 111 and Walker Jr. P. L., 6A molecular sieve properties
of saran type carbons. Carbon 3 (1965) pp. 59-63.
4. Liu Y., Xue J. S., Zheng T. and Dahn J. R., Mechanism of lithium insertion in hard
carbons prepared by pyrolysis of epoxy resins. Carbon 34 (1996) pp. 193-200
5. Mukai S. R., Masuda T., Tanigawa T., Harada T., and Hashimoto K., Structure of
Hard Carbons which Leads to Small Irreversible Capacities for Li Insertion.
International Symposium on Carbon Science and Technology for New Carbons
(Tokyo) (1998) pp. 305-306.

317
 NOVEL BIOACTIVITE CARBOMINERALSORBENTS, INCLUDING
CLUSTER AND CARBON NANOTUBES FOR SUPERSELECTIVE
PURIFICATION OF BIODIESEL FUEL LIQUID HYDROCARBONS AND-
CARBONHYDRATE FROM SULFUR CONTAINING IMPURITIES

DMITRY I. SHVETS

Institute for sorption and problems of endoecology NAS of Ukraine.13. General Nawnov Str...
Kiev-I64,03680, Ukraine
E-mail: [email protected]: [email protected]
The properties of bioactive carboncontaining sorbents are consided during clearing liquid
hydrocarbonaceouspropellant from sulfurcontaining substances. It was revealed, that the superselectivity
is stipulated by availability in bioactive carboncontaining sorbents of clusters and carbon nanotubes. It
was shown, that a type of clusters, their concentration, natm and the sizes of nanotubes are defining at
sorption of sulfurcontaining substances. It was found, that the purified solar oil contains minimum
amount of admixtures and conforms the requirements of the world standads. The mechanism of
superselective sorption with allowance of nanoclaster structuresis discussed.

1. Introduction
The diesel fuel recently involves the increasing notice both technologists, and consumers.
For making solar oil use alcohols, carbohydrates etc. Despite of the successes, achieved in
this direction, rather acute there is a problem of quality of propellant, namely contents of
toxic admixtures, first of all elementcontaining. It is stipulated by that element (Cl, 0, N,
S) - containing liquid carbohydrates and the hydrocarbons (i.e. propellant with
admixtures) at combustion will derivate toxic products of the special danger, which one
harm as the person, and ozone layer of the Earth. Therefore, problem of purification of
liquid carbohydrates and hydrocarbons from toxic admixtures is one of most significant
for today for a propellant industry.
The technologies applied for clearing of propellant from toxic admixtures are rather
diverse; however degree of clearing remains insufficiently high and does not correspond
to increasing requirements. Not subjecting criticism any of existing methods, in operation
is considered an opportunity of clearing of components of solar oil - liquid carbohydrates
and hydrocarbons from sulfurcontaining substances with usage new of bioactive
carboncontainingsorbents with nanocluster structures in porous space.

2. Methods
As objects of study there were used new modifications of carbon materials vegetative
(carbon material with heterostructures of a radical type), organocarbon (nanotube), carbon
(puffed up graphite) nature and also combined carboncontaining composite. Modification
of sorbents conducted during their synthesis (heat treatment at fixed temperature) or by
processing carbon materials by special (ecologically clean) reagents at ambient
temperature. As toxic matters used petroleum, oil products, and also various fuels
containing toxic substances of a various type - aromatic substances, asphaltenes, resin,
sulfurcontaining substances, dispersible sulfur etc. Analysis of properties of carbon
materials and sorbed products carried out with usage of physicochemical methods,
atomic-adsorptive analyzer, and also with applying of methods ESR-,IR-, UV-, X-ray
spectroscopy, Zpotentiometry etc.
As the object of investigation diesel fuel @F) was used. DF- is the product of coke-
chemical production, containing asphaltenes, soots, paraffins, sulfur-, nitrogen- and
oxygen-containing compounds. Purification of DF was done by method of their passing
through sorption column, which containing material with sorption-catalyticproperties on

318
the base of natural type's aluminosilicatesof common formula (SiO~)m(Al~O~),,(Meox),, .
y(0H). z(H20). Estimation of purification degree was done by visual method by changing
of DF color in comparison with DF, corresponded to technical specifications. Efficiency
of the process of purification was estimated by quantity of volumes of DF, purified by
one volume of compositional material.

3. Results and Discussion

To obtain the composite meeting our requirements we varied its composition, taking
different amounts of the components and modifying their properties.
Experimental results showed that carbonmineral composites are much better than
others adsorbents, for example, the mineral one. A good selectivity of carbon materials
made us to assume that it is a carbon substance is responsible both for selectivity and
synergistic effect of adsorption too. From our point of view one of the reason of such a
behavior could be specially organized carbon structures such as intermediate complexes
(clusters), which possess peculiar electron properties only to them. Probably similarly
toxic substances are adsorbed, such as phenols, cresols, quaiacol, aldehydes,
polyatomic alcohols, ethers etc. (Table 1).

Table 1. Efficiency of decontaminationof some aqueous solutions from organic pollutants by &on modified
andcarbonrmn ' eral* materials

I POUUtPDt

Phenol
concentration,m@l
Initial
2.1
I Final
Decontamination level, %

100
Cresol 4.8 lo4 100
Quaiacol 4.0. lo4 100
Ethers* 279.2 13.2 95
Aldehydes* 0.4 0.036 92
Polyatomic alcohols* 3,7 100

High efficiency of developed composite carbonmineral materials was demonstrated in the

processes of purification of water-alcohol mixtures. The results of experiment
demonstrated more effective extraction of toxic impurities from solution by our material
in comparison with known technology (Table 2).
TaMe 2. Efficiency of purification in relation to kind of sorption material

The results of comparative study of sorption properties of carbomineral sorbents

(initial and modified form) on the purification of technological solution are demonstrated
on Fig. 1-2. It is seen that carboncontainingsorbents provide efficient purification of iquid
-
hydrocarbons and diesel oil from toxic impurities chlorophyll and carotene. The absence
of toxic impurities in liquid carbohydrate - vegetable oil (from rape) provides, as it was
stated by us experimentally, its high heating capacity as a component of biodiesel fuel. As

319
can be seen from fig.3, the most cleaning degree is reached at use sorbent of mixed type
only, where synergic effect is maximal.
Efficiency of refining different mixtures with the usage of the combined sorbents was
established to be significantly higher than in the case of initial unmodified sorbents. The
observed synergism proposes new approaches to the selection of components of the
carbomineral sorbents.
A. X

4% '9
100
80
1 -NSP
60
2.Nsz
40 3-Nso
20 &NST
S-WW
0
S-NYNM 1 2 3 4 5 6 7 8 9
1 2 3 4 5 6

Fig.1 Influence of natun of dispersed material on Rg. 2. Influence of sorbents type and contents of
purificationdegree of oil (rape) from impurities compositional mixtures on type of DF purification
from toxic admixtures: 1.2.5.6.8 - natural sorbents;
-
3 . 4 - two-component mixtures; 7 fourcomponent
mkm; 9 -
two-stage process of purification

Modification of carbomineral sorbents causes not only change of sorption properties

resulted in increasing the quantity of sorbate, but also significant improving of selectivity.
It should be noted that increasing of the adsorption ability of carbomineral sorbents
towards toxic pollutants of different nature, such as ions of heavy metals, organic
compounds, was attained by thorough selection of complexing agents.
High efficiency of developed composite carbomineral materials was demonstrated in
the processes of purification of water-alcohol mixtures. The results of experiment
demonstrated more effective extraction of toxic impurities from solution by our material
in comparison with known technology.
It was established that the use of diffusion vortex affecting on heterogenous system,
composed from a liquid medium and composite sorbing materials allows to reach
nonadditive effect: it lies in abnormal raise of an extent of hydrocarbon fuel purification
from admixtures, as in case of a stream passage through fixed sorbent layers, as in case as
sorbent of composition mixture each component separately.
Nonadditive effects are characteristic for the processes of water purification from oil
(Fig 3) and benzene [5], also hydrocarbon fuel from toxic admixtures [6](Fig.4 ).
The perspective alternative by technology of purification of liquid mediums from
toxic impurities represents by usage of systems in a superextreme state [7,8].The used
term %systemin an extreme state D is similar to the term "stress" offered prof. E. Thukin
[9] or model of disastrous considered by us in [lo]. Specificity of such state is that the
system passes in a superextreme state at usual stresses and temperatures, and the
supercriticality of a state is caused by applying (in our case) combination of dispersible
materials with different significances of a power charge (first of all of opposite sign). At
such situation there is such system condition, when the separation of components of a
liquid phase (or extraction of one component from an admixture) proceeds practically
instantly through a stage "of stress", i.e. through a stage of instantaneousrough "reacting"
by addition in a liquid phase of dispersible materials. Confirmation of a developed
hypothesis are indicated below experimental data received by us recently. However, as
our investigationshave shown, not only mixed sorbents capable to show superselectivity
and provide high purifying degree of fuel from impurities. As it is seen from fig.5, similar

320
 1

1 2 3 4 5 6 7 8 9
Fig. 3. Influencecondition of heat treatmenton the type of DF purification by sorbents of different type (2.5.8 -
heat treatment of s o h t by stages;3.6,9 -heat treatment of sorbents mixture; 1.4.7 - sorbent without heat
treatment)

+petroleum
-)Ioil

1 2 3 4 5 6

Material type
Fig.4. The purification variation of liquid Wg.5. The influence of CadJoncontaining
hydrocarbons from toxic admixture depending on nanotubes n m (1 - 3 cadJon nature. 4 - 6
sorbent nature: 1 - carbon. 2 - mineral. 3 - organic nature)on the efficiency of the oilproducts
adsorption from water surface
carbonmineral

effect of superselectivity is observed for carbon nanotubes. The reasons of the anomalous
effect of such type are yet to be explained, however even today we can say about it value
for practical purposes for obtaining of high pure ecologically harmless kinds of biofuel.

Surfaces nature of aluminosilicates sorbents

On the base of data analyses Fig.3 and Fig.4 it can be suggested, that surfaces nature of
alumosilicates sorbents plays certain role in purification process. Its significant, effect of
purification can be observed in two situations, on the one hand-sorbents, containing acidic
centers, on the other hand-sorbents with basics properties. Accept acidic-basic properties
of alumosilicates surfaces, another factor influence on effectivity of purification.
Confirmation of rightness of such suggestion is the fact, at thermo-handling of natural
alumo-silicates loosing of water and also dehydroxilation of surface take place. But, as we
see from Fig.3, while increasing of water loosing purification degree of DF is achieving
maximum, and after that is falling down, that connected with releasing of such called
structural or zeolite water from pores space. Simultaneously, the character of
dependence of purification degree on thermo-handled samples is opposite to data on Fig.2.
More detail analyses show, that among natural surface (acidic and basic centers) as a
sorption, such a structural water play important role. Exactly zeolite water is maximally
releasing from alumosilicatessorbents at thermo-handling in the range of 350 C , in other
wards, while going away of water releasing from pores space take place. In these pores

321
space additional sorption of toxic substances can be happened. Its possible that impact of
such factor is significant, but its role in process of purification (as the role of OH-groups
and acidic centers) is not dominant.

Nature of surfacesions also play significant role in purification process

Confmnation for benefit of further mentioned argument is the fact, that exactly after the
thermo-handling of natural materials degree of purification of DF is increasing, but type of
natural sorbents is greatly influencing ). And we can suggest, that with structural-sorption
parameters nature of surfaces ions also play significant role in purification process. In
sorption process one materials saturated by toxic admixtures of DF and changes their color;
another one doesnt change their color. This fact confirms possible role of surfaces ions
nature. This fact takes a great attention, because it allows to suggest, that surfaces centers
are not only the sorption centers, but in the same time they are catalytically many-
functional centers of clusters type, Independent confirmation of these suggestion is a point,
that quantity of purified fuel in counting of it per lgramm (>20 g/g) exceeds any variants
(practical and theoretical), from volume of sorption pores Vs = 0,6-0,7 cm3/g and density
of admixtures d=0,8 g/cm3for sorption of 20 g of admixtures 25 cm3 of volume is needed,
and really system has 1,8-2,0 cm3, i. e. only sorption scheme of purification is unreal.
Relatively to the possibility of proceeding of the sorption-catalyticprocess, it is shown by
the fact , that functional-depended thermo-handling strictly increasing purification degree,
as in absolute values, such in effectivity.

4. Conclusions
So. ?resented experimental results shows, that at first time weve found the effect of over-
selectivity, based on fact, that in oxides systems of mixed type, passed through the
functional-depended thermo-handling, it is formed, clusters center and carboncontaining
nanotubes nature , providing the possibility of proceeding of over selective sorption-
catalytical process of purification of liquid hydrocarbons from toxic admixtures

References
1. Shvets D. I., Adsorption Science and Technology 17 (1999) pp. 709-714
2. Shvets D. I., Carboncontaining sorbents of mixed type: properties and applying in
extreme situations, Curbon-02 In Prossiding International Conference on Carbon,
September 15-20,2002 Beijing, China. ISBN 7-900352-03-7/16-03.
3. Shvets D. I., Kravchenko O.V., Urvant O.S. ,The applying of mixed carboncontaining
sorbents for removal of oil products from the water surface, Curbon-02 In Prossiding
International Conference on Carbon, September 15-20, 2002 Beijing, China. ISBN 7-
900362-03-7/IG-03.
4. Shvets D.I., Lapko V.V., Urvant O.S., Sorption-catalytic purification of diesel fuel
from toxic admixtures by oxides systems of mixed type, Adsorption Science and
Technology 21 (2003) (in press).
5. Shvets D. Biocatalysys on the basis of carbon - and carbon-mineral sorbents and their
property.Ros. 5 European Congressjn Catalysis. Limeriick, Iceland. 2001. p.129.
6. Baiker A. Chem. Rev., 99 (1999), pp. 453-454.
7. Sagave P.E.Chem. Rev., 99 (1999), pp. 603-605.
8. Shchukin E, Amelina E, Izmailova V. In Roc. NATO Advanced Research Workshop
Role of interface in Environmental Protection, Miskolc - Hungary, 2002, p. 6.

322
9. Shvets D In Pmc. NATO Advanced Research Workshop Role of interface in
Environmental Protection, Miskolc - Hungaru, 2002, p. 161.
10. Shvets D.I., Chochlova L.I., Kravchenko O.V. et al.The physico-chemicalaspects
of oil sorptiol by the carbon sorbents from water surface. Chemical and Technology
Water, 24 (2002), pp.22-31 (Rus).

323
 TITANOSILICATEETS-10: SYNTHESIS,CHARACTERIZATIONAND
ADSORPTION FOR HEAVY METAL IONS

[GEORGE)X.S . ZHAO*, J. L. LEE AND P.A. CHIA

Department of Chemical and Environmental Engineering, National University of Singapore, I0 Kent
Ridge Crescent, Singapore I 1926; E-mail: cliens0,nus.edu.sg

Microporous titanosilicate ETS-I0 was synthesized in the absence of organic template and
characterized using XRD, FTIR, Raman, and nitrogen adsorption. The adsorption properties of
heavy metal ion PbZ+on ETS-I0 were studied by measuring the adsorption kinetics and equilibria
using a batch-type method. Highly pure ETS-I0 was obtained without the presence of ETS-4. The
adsorption rate of heavy metal ions on ETS-I0 is extremely rapid, less than 5 seconds is required to
attain maximum adsorption capacity in a 10 moYL solution with a batch factor of 200 m u g . The
kinetic data can be fitted by pseudo-second-order model whereas the equilibrium data is better fitted
to Langmuir isotherm than to Freundlich isotherm. The maximum adsorption capacity of Pb2+and
Cu as predicted by the Langmuir equation was 1.12 and 0.578 mmol/g, respectively. The
remarkable adsorption rate coupled with the high adsorption capacity promise potential applications
of ETS-I0 for the removal of heavy metals present in drinking water and wastewater.

1 Introduction

Heavy metals such as lead (Pb) are common groundwater contaminants that must be
controlled to an acceptable level according to the increasingly stringent environmental
regulations. The heavy metals, especially Pb present in drinking water are extremely
detrimental to human beings. Depending on the existing form of the metals, they can be
removed by different technologies such as chemical precipitation, membrane filtration, ion
exchange, and adsorption [l]. Unfortunately, none of them affords reducing the heavy
metals to an acceptable low level at a minimal contact time, which is of significance in the
treatment of waters, especially in purification of drinking water.
ETS-10, a microporous titanosilicate ETS-10 discovered by Engelhard in 1989 [2] is
zeolite material with a pore-opening size of 0.8 nm [2-4]. The basic anhydrous formula of
as-synthesizedETS- 10 is Na1.5&.5TiSi5013. Unlike conventional zeolites, the framework of
ETS-I0 is constituted fiom SiO, tetrahedra and TiOs octahedra by corner-sharing oxygen
atoms [3]. The presence of each tetravalent Ti atom in an octahedrum generates two
negative charges, which are balanced by exchangeable cations Na+ and K+.Such a unique
framework property manifests itself a promising and potential ion exchanger for many
cationic metal ions that are present in waters such as Pb2+,Cd, Cu2+,Zn2+,etc. However,
adsorption data of heavy metals on ETS-10 have been hardly available [5,6]. Al-Attar and
Blackbum compared the uptake properties of uranium on ETS-10 materials synthesized
with different Ti sources and noted that the method of ETS-10 preparation has a
considerable effect on the uptakes of uranium [5]. Kunicki and Thrush [6] observed that
ETS- 10 and ETAS- 10 (Al-containing ETS- 10) displayed an extraordinarily rapid
adsorption rate towards Pb2+.The concentration of Pb2 was reduced to a negligible amount
fiom 2000 ppm in a very short contact time at a liquid to solid ratio of 100:2.4 (g:g).
Unfortunately, adsorption equilibrium data were not available.

324
 Motivated by the work of Kunicki and Thrush [6], we have carried out a systematic
study on the adsorption equilibria and kinetics of several heavy metal ions including PbZ+,
Cd2+,Cu, Zn2+and Ni2+on ETS-10 using a batch-type technique. Our observations have
not only confirmed that ETS- 10 does exhibit a remarkable adsorption rate towards heavy
metal ions but also demonstratedthat the maximum adsorption capacity of Pb on ETS- 10
is as high as 1.12mmoVg according to the prediction of Langmuir model. This is the highest
uptake that has been observed on zeolite materials [ 11.
In this paper, we present the unusual adsorption properties of ETS- 10 towards heavy
metal ions Pb2+and Cu2+.Adsorption equilibrium and kinetic data are reported. Fitting of
the experimental equilibrium results to both Langmuir and Freundlich isotherms and the
kinetic data to both pseudo-fist- and pseudo-second-orderkinetic models is described.

2 Methods

The method of synthesis of ETS-10 was similar to that reported by Yang and co-workers [7].
The synthesis recipe was 8NaOH:2KOH:TiF.,:5.7Si02:350H~0. Sodium silicate solution
(Merck) and TiF4(Aldrich) were used as the Si and Ti source, respectively. All chemicals
were used as received.
Samples were characterized by using X-ray diffraction (XRD) on a Shimadzu
XRD-6000 diffractometer (CuKa radiation), physical adsorption of nitrogen on a
Quantachrome NOVA 1000, Fourier transform infrared (FTIR) spectroscopy on a Biorad
spectrometer using the KBr method, Raman spectroscopy on a Bruker FRA 106/S
FT-Raman spectrometer, and scanning electron microscopy (SEM) on a Joel JSM-5600LV.
Adsorption of heavy metal ions on the ETS-10 sample was conducted using a
batch-type method at room temperature (23 OC). For kinetic measurement, 1 g of air-dried
ETS-10 was added to 100 ml of solution pre-acidified by nitric acid under shaking so as to
generate a solution of pH = 5.8. Then, 100 ml of 20 mmoVL Pb(NO& (or Cu(N03)z)
solution was added to obtain a mixture with an initial Pbz+(or Cu2> concentration of
approximately 10 mmoVL, a final pH value of about 5.0 and a batch factor (ratio of liquid
volume to solid mass) of about 0.2 L/g. 5 ml of the mixture was withdrawn at an appropriate
time interval by using a 5 ml syringe and rapidly filtered through a 0.2 pm nylon membrane
filter. The filtrate was collected in a sample valve and analyzed for Pb (or Cu), Na and K
concentrations using a spectrometer (Perkin-Elmer Analyst 300). The amount of metal
adsorbed at time t (s), qc(mmovg), was deduced from the mass balance between the initial
concentration (Co)and concentration at time t (CJ. The experimental data were fitted to
pseudo-second-order equation (t/q, =l/vo +t/q,) [8], where k (g/mmoVs) is the adsorption
rate constant, qc(mmoVg) is the amount of metal adsorbed at equilibrium, and vo (mmol/g/s)
is the initial adsorption rate which is kq;.
Adsorption equilibrium data were collected in a similar way as the kinetic
measurement. The equilibrium time was 10 min, which, according to the kinetics data, was
found to be sufficiently long to attain adsorption equilibrium. The experimental data were
fitted to both the Langmuir isotherm ( q , = q,,,bC,/(1+ bC,) ), where qm (mmol/g) is the
maximum adsorption capacity, C, (mmol/L) is the equilibrium concentration of the heavy
metal ion in solution, and b (L/mmol) is the Langmuir constant, and the Freundlich isotherm
(4,= KC:),where K and n are constants.

325
3 Results and discussion

3. I Characterization of Em-I0 sample

Fig. 1 depicts the XRD pattern, Raman and ETIR spectra of the ETS-I0 sample used in this
study. The XRD pattern is identical to that of ETS-10 materials reported previously [2,7],
showing that the sample is a pure ETS-I0 phase without the presence of ETS-4 impurity (it
has been shown that ETS-4 is a thermodynamicallymore stable phase than ETS-10 and it is
normally present in an ETS-I0 product [9]). The Raman spectrum (left-hand insert) further
confirms the purity of the sample. A most intense peak at 728 cm", assigned to Ti-0-Ti
stretching in comer-shared Ti06 chains [lo], and a small band at 305 cm-', attributed to
Ti-0-Si bending [1I] can be seen. The absence of any peak above 800 cm-' on the Raman
spectrum further confirms the inexistence of ETS-4 in this sample [10,111. The observation
of a main band at about 1024 cm-' due to Si-0 stretchingand a few small bands at 668,570
and 434 cm-' because of Ti-0 stretching, Si-0 rocking and 0-Ti-0 bending, and 0-Si-0
and 0 - T i 4 bending and Ti-0 rocking, respectively, on the FTIR spectrum (right-hand
insert) is consistent with the literature data of ETS-10 [121.
SEM image (not shown) displays cuboid-shape crystals of about 5 pm, but incomplete
crystal growth was also observed. The specific surface area of this sample calculated by
using BET model was about 258 m2/g.
c
I 728 I
a
E

g
Ba
r
.-
200 400 600 800 1000 1200 400 aoo 1200
Raman shift (cm-l) Frequency (cm-l )

l l I I I I I I

5 10 15 20 25 30 35 40
Two theta
Figure 1. XRD pattern, Raman spectrum (insert,left) and FTIR spectrum (insert, right) of
as-synthesized ETS-10

3.2 Adsorption kinetics of h e w metal ions on E n - I 0

Fig. 2 shows the adsorption kinetics of Pb2+ on ETS-10 together with the
pseudo-second-order kinetic curve. It is seen that the adsorption rate is extremely fast.
Under the experimental conditions, less than 10 s was required to attain saturation
adsorption. When the concentration of Pb was about 2.5 mmoVL, Pb2+was not detected

326
after 5 s. This rapid-adsorption behavior of ETS-10 towards Pb is of interest and
significance in terms of purification of drinking water as Kuznicki and Thrush suggested [6].
This unusual adsorption behavior of ETS-10 towards heavy metal ions is currently being
investigated. It is believed that the rapid adsorption rate must be related to the unique
structure of ETS-10. The experimental data fit well to the pseudo-second-order equation.
The experimental data were fitted to pseudo-second-order equation (t/q,=\/v0+t/qe),7
where k (g/mmol/s) is the adsorption rate constant, qe (mmol/g) is the amount of metal
adsorbed at equilibrium, and v0 (mmol/g/s) is the initial adsorption rate which is kq*.

O Experimetnal data Experimetnal data

Pseudo-second-ordvr model Langmuir isotherm

40 60 BO 100 0.00 0.05 0.10 0.15

ce (m mol/L)
Figure 2. Adsorption kinetics of Pb2+ on Figure 3. Adsorption isotherms of Pb2+
ETS-10 (C0 = 10 mmol/L, V/m = 0.2 L/g) on ETS-10 (23 C)

3.3 Adsorption isotherms of heavy metal ions on ETS-10

Fig. 3 shows the adsorption isotherm of Pb2+ on ETS-10. The experimental data were fitted
to both Langmuir and Freundlich isotherms and the results are included in Figure 2 as well.
The parameters derived from the two models are presented in Table 1. As can be seen, the
Langmuir isotherm predicts the experimental data much better than the Freundlich isotherm.
The maximum adsorption capacity of Pb2+ on ETS-10 as predicted by the Langmuir
isotherm is 1.12 mmol/g or about 232 mg/g. Such a high adsorption capacity of Pb2+ on
zeolite materials had never been observed [1]. The adsorption of Pb2+ on a commercial
zeolite NaY sample (Si/Al = 2.45) was measured as well and the results showed that its
maximum adsorption capacity towards Pb2+ was about 56.3 mg/g, much less than ETS-10.
The adsorption of other heavy metals including Cd2+, Cu2+ and Zn2+ on ETS-10 and zeolite
NaY was also studied and compared. A similar adsorption behavior was observed, namely,
the adsorption rate of these metal ions on ETS-10 was extremely fast and the adsorption
capacity was much higher on ETS-10 than on zeolite NaY. The maximum adsorption
capacity of Cd, Cu and Zn on ETS-10 was found to be all around 0.5 mmol/g while it
became about 0.2 mmol/g on zeolite NaY.
Table 1. Langmuir and Freundlich Parameters for Pb2* Adsorption on ETS-10
Freundlich model Langmuir model
A:(mmol L 1/n g 1 )
M/n n /v qm (mmol g"1) b (L mmol"1) I?
1.50 0.102 0.932 1.12 480 0.992

327
3.4 Adsorption mechanism of heavy metal ions on Em-10
During the measurementsof adsorptionkinetics, the concentrations of both Na+and K+were
also monitored. It was observed that the decrease in Pb2+was at the expense of the increase
in Na+and K+. In addition, the concentration of K' in the solution was always one third of
that of Na+, indicating the equal opportunity of ion exchange of Pb2' with the two alkali
metal ions. Furthermore, the concentration sum of Na' and K' in the solution exactly
doubled the concentration of Pb2+at any time, suggesting that each Pb2+ion replace the pair
of (1.5Na++0.5K?. These results suggest an ion exchange mechanism.

4 Conclusion

In conclusion, we have demonstrated the unusual adsorption behaviors of microporous

titanosilicate ETS-I0 towards heavy metal ions with an extremely fast rate and in a large
adsorption capacity, showingthe application potentials of ETS- 10 for water and wastewater
treatment. The adsorption is most likely via ion exchange. It is believed that the unique
compositional framework together with the large pore size of ETS-10 play a vital role in
determining its remarkable adsorption properties towards heavy metal ions.

References

1. Bailey S. E., Olin T. J., Bricka R. M. and Adrian D. D., A review of potentially low-cost
sorbents for heavy metals, Wat. Res. 33 (1992) pp. 2469-2479.
2. Kunicki S. M., Large-pored crystalline titanium molecular sieve zeolites, US Patent
4853202 (1989).
3. Anderson M. W., Terasaki O., Ohsuna T., Philippou A. MacKay S. P., Ferreira A.,
Rocha J. and Lidin, S., Structureof the microporous titanosilicate ETS-10, Nature 367
( I 994) pp. 347-351.
4. Rocha J. and Anderson M. W., Microporous titanosilicates and other novel mixed
octahedral-tetrahedral framework oxides, Eur. J. Inor. Chem. (2000) pp. 80 1-818.
5. Al-Attar, L., Dyer, A. and Blackburn, R., Uptake of uranium on ETS-I0 microporous
titanosilicate, J. Radioanal. Nucl. Chem. 246 (2000), pp. 45 1-455.
6. Kunicki S. M. and Thrush K. A., Removal of heavy metals, especially lead, ffom
aqueous systems containing competing ions utilizing wide-pored molecules of the
ETS-I0 type, US Patent 4994191 (1991).
7. Ho Y. S.and McKay G., The kinetics of sorption of divalent metal ions onto sphagnum
moss peat, Water Res. 34 (2000), pp. 735-742
8. Yang X.,Paillaud J.-L., van Breukelen H. F. W. J., Kessler H. and Duprey E., Synthesis
of microporous titanosilicate ETS- 10 with TiF4and Ti02, Micropor. Mesopor. Mater.
46(2001)pp. 1-11.
9. Xu H., B a n g Y., Navrotslq A., Enthalpies of formation of titanosilicates ETS-4 and
ETS-10, Micropor. Mesopor. Mater. 47 (2001), pp. 285-291.
10. Kim W. H.,Lee M. C.,Yo0 J. C. and Hayhurst D. T., Study on rapid crystallization of
ETS-4 and ETS-10, Micropor. Mesopor. Mater. 41 (2000), pp. 79-88.
11. Su Y., Balmer M. L. and Bunker B. C., Raman spectroscopic studies of silicotitanates,
J. Phys. Chem. B 104 (2000), pp. 8160-8169.
12. Mihailova B., Valtchev V., Mintova S. and Konstantinov L., Vibrational spectra of
ETS-4 and ETS-10, Zeolites 16 (1996), pp. 22-24.
 ORDERED MACROPOROUS MATERIALS STRUCTURALLYTEMPLATED
BY COLLOIDAL MICROSPHERES

Z. Zhoy W.C. Ong, (George) X. S. Zhao*

Department of Chemicaland Environmental Engineering, National University of Singapore,
Singapore 119260: E-mail: [email protected]

Ordered macroporous materials (OMMs)are a new family of porous materials that cm be

synthesized by using colloidal microspheres as the template. The most unique characteristicsof
OMMs are their uniformly sized macropores arranged at micrometer length scale in three
dimensions. Colloidal microspheres (latex polymer or silica) can self assemble into ordered arrays
(synthetic opals) with a threedimensional crystalline structure. The interstices in the colloidal
crystals are infiltrated with a precunor material such as metal alkoxide. Upon removal of the
template, a skeleton of the infiltrated material with a three-dimensionally ordered macroporous
structure (inverse opals) is obtained. Because of the 3D periodicity of the materials, these
structures have been extensively studied for photonic applications. In this paper, the synthesis
and characterization of highly ordered macroporous materials with variouS compositions and
functionalities (silica, organosilica, titana, titanosilicate, alumina) are presented. The application
potential of OMMS in adsorptionlseparation is analyzed and discussed.

1 Introduction

Porous materials can be classified, according to the pore size, into microporous materials
with ore size smaller than 20 A, mesoporous materials with pore size between 20 8, and
500 1, and macroporous materials with pore size larger than 500 A. Microporous and
mesoporous materials have been studied extensively and have found wide applications in
many areas, such as adsorption, separation and catalysis. One of the frequently used
methods of synthesizing porous materials is template strategy. A well-known example is
the Mobils liquid-crystal templating mechanism by which many mesoporous materials
can be made. This discovery provides an opportunity of treating and processing relatively
large molecules. However, when macromolecules such as enzymes are dealt with,
macroporous materials with an ordered pore structure and uniform pore size are desired.
OMMs were first synthesized for the purpose of photonic a lications because of their
3D spatial structure with periodically varied reflective index!Since then, OMMS with
various chemical compositions have been prepared and they have been demonstrated to
find wide applications in the fields other than photonics? With the availability of porous
inorganic-organic materials: and the successful synthesis of surfactant-mediated highly
ordered mesoporous organosilica materials, organic-inorganic macroporous composite
materials would afford high application potentials in adsorptiodmembrane separation. In
addition, OMMs are a suitable material for processing macromolecules such as enzymes.
In this paper, the synthesis and characterization of highly ordered macroporous materials
with various chemical compositions and functionalities (silica, organosilica, titana,
titanosilicate, aliumina) are presented. The application potential of the OMMs in
adsorptiodseparationare analyzed and discussed.

329
2 Methods

2. I Synthesis of OMMS by using self-assemblystrategy

The synthesis strategy of OMMs is similar to the conventional template method (see Fig.
1). The template used is self-assembled microspheres instead of single molecule or
surfactant micelles. Colloidal microspheres with uniform size and morphology are
induced to spontaneously organize into a crystalline lattice (artificial opals), which could
have a face-centered cubic (fcc) or a hexagonal-closed-pack (hcp) structure or the
combination of them depending upon experimental conditions. The artificial opals are
slightly annealed to improve the stability of the crystal structure and to form connecting
necks between the adjacent spheres. Then the voids among the opals are tilled with
another material such as oxides, polymers, and hybrid materials, etc. Finally, the artificial
opals which act as the template are removed, leaving behind an OMM.

li Self Assembly

a Witration

removai

Figure 1. Illustration of preparation procedures of OMMs using self-assembled template

2.2 Synthesis of uniform-size microspheres

Polystyrene (PS) spheres were prepared with emulsifier free emulsion polymerization.
Silica spheres were prepared following the modified Sttiber method! All chemicals were
used without further purification.

330
2.3 Ctystallization of the microspheres
The formation of artificial opals was achieved by using a number of methods including
sedimentation, filtration, evaporation, and drip method.
2.4 Opal annealing
After the opal forming, annealing was used. The PS opal was heated in oven at 110 '
C
(slightly higher than the glass temperature of PS) for 5 to 10 min.
2.5 Opal injiltration
Infiltration of the artificial opals was carried out by either filtration, or chemical vapor
deposition, or soaking method.
2.6 Template removal
The template was removed either by calcination at 550 OC for the PS opal or by chemical
etching in HF solution for silica spheres.
2.7 Characterization
Samples were characterized by using scanning electron microscope (SEM) (JEOL JSM-
5600LV), transmission electron microscope (TEM), physical adsorption, FTIR, Raman.

3 Results and discussion

3.1 Opalformation
Highly crystalline opals can be obtained by sedimentation, filtration, evaporation, and
drip methods as demonstrated by the SEM images shown in Fig. 2.
Thermodynamically, atoms or molecules tend to adopt the structure with the lowest
Gibbs free energy. When colloidal microspheres are allowed to self assemble at a close-
to-equilibrium state, they tend to form closely packed crystalline structures, such as face
cubic center (fcc) and hexagonal closed packed (hcp) lattices. As a result, maintenance of
equilibrium plays a vital role in obtaining a highly crystalline lattice. In addition, because
the fcc structure has the lowest Gibbs free energy it is always the observed structure
during self assembly as can be seen from Fig. 2. Another important factor having
influence on the self assembly process is the size distribution of the spheres. It was
observed that a size derivation of larger than 5% destroyed the long-range order.
3.2 Opal annealing
The opals obtained by self-assembly are mechanically unstable because there is only Van
der Waals force between spheres. The subsequent infiltration process could easily destroy
the ordered colloid arrays. So we annealed the opals of polymer sphere to increase their
stability. As a result, there would form interconnections between spheres, which come
from the slight melting of the sphere surfaces. These necks can provide the opal with
necessary mechanical stability. In addition, they are important for producing inverse opal
structure. After infiltration, when the samples are treated with calcinations, these necks
can act as channels for the transport of the products formed during calcination like COz.

331
Figure 2. Self-assembled artificial opals fabricated with different methods: (A) sedimentation, (B)
filtration, ( C ) evaporation, and (D) flow-controlled evaporation.

3.3 Opal infdtration and template removal

Figure 3. SEM images of (A) PS opal infiltrated with SiOz, (B) after removal of template of (A), (C)
macroporous Ti02 prepared via core-shell method, and (D) macroporous organosilica materials

332
Macroporous materials with an ordered structure of various framework compositions
were prepared by using both PS and silica spheres as the template. Shown in Fig 3 (A) is
the SEM image of PS opal infiltrated with Si02. It is seen that the self-assembled opal
was essentially completely inverted with SiOz. After removal of the PS spheres by
calcinations, polystyrene decomposed into C02 and a reverse opal structure was left
behind (Fig. 3 B). Fig. 3 (C) shows the SEM image of macroporous Ti02prepared by self
assembly of core-shell spheres. PS spheres were first hctionalized, coated with a layer
of TiOz on the surface to obtain PS-Ti02 composite spheres, followed by self assembly to
obtain an order structure. Upon removal of the PS spheres, a macroporous shell was
obtained. Using similar approach, macroporous organosilica materials was also obtained
(Fig. 3 D) by using silica spheres as the template.

4 Conclusion

In conclusion, we have successfully fabricated OMMs with different pore size by

using opal templated method. The pores are in micron scale and have narrow size
distribution. We also studied the synthesis conditions during each step and found a
relative feasible route to prepare OMMs.

References

1. Imhof A. and Pine D. J., Ordered macroporous materials by emulsion templating,

Nature 389 (1997), pp. 948-951.
2. Vlsov Y.A., Xiang Z. B., Sturm J. C. and Norris D. J., On-chip natural assembly of
silicon photonic bandgap crystals, Nature 414 (2001), pp. 289-293.
3. Xia Y.,Gates B., Yin Y.and Lu Y., Monodispersed colloidal spheres: old materials
with new applications,A h . Muter. 12 (2000), pp. 693-713.
4. Subramania G.,Constant K., Biswas R., Sigalas M. M. and Ho K.-M., Inverse face-
centered cubic thin film photonic crystals, A h . Muter. 13 (2001), pp. 443-446.
5. Stein A., Sphere templating methods for periodic porous solids, Micropor. Mesopor.
Muter. 44-45 (2001), pp. 227-239.
6.Loy D. A. and Shea K. J., Bridged polysilsesquioxanes. Highly porous hybrid
organic-inorganic materials, Chem. Rev. 96 (1995), pp. 1431-1442.
7. Inagaki S., Guan S., Ohsuna T. and Terasaki O., An ordered mesoporous organosilica
hybrid material with a crystal-like wall structure, Nature 416 (2002), pp. 304-307.
8. T. Okubo, T. Miyamoto, K. Umemura and K. Kobayashi, Colloid Polym. Sci., 2001,
279,1236.
9.Okubo T., Miyamoto T., Umemura K. and Kobayashi K., Seed polymerization of
tetraethyl orthosilicate in the presence of colloidal silica spheres, Colloid Polym. Sci.
279 (2001), pp. 1236-1240.

333
 ADSORPTION OF NITROGEN, OXYGEN AND ARGON IN TRANSITION AND
RARE EARTH ION EXCHANGED ZEOLITES A AND X

RAKSH VIR JASRA, JINCE SEBASTIAN AND CHINTANSINH D. CHUDASAMA

Discipline of Silicates & Catalysis, Central Salt & Marine Chemicals Research Institute, G.BMarg,
Bhavnagar 364 002 INDIA
E-mail: [email protected]

Adsorption ofN2,O~and Ar in zeolite A and X exchanged with silver, cerium and europium at 15 and 30C
temperatures has been studied. Silver exchanged zeolite A show higher adsorption capacity and selectivity
for nitrogen compared to Zeolite CaA.Enhanced Nitrogen interactions of nitrogen molecules with AgA are
also observed in very high heat of adsorption (38kJmor'). AgA also exhibits argon selectivity over oxygen
atypical of zeolites. Both cerium and europium exchanged Zeolite X show oxygen selectivity over nitrogen
in Henry region which has been attributedto interaction of oxygen with non-stoichiometric oxides of these
cations apparentlyformed inside zeolites cavities.

I Introduction

Zeolites are of immense interest in gas and chemical industries for purification and
separation due to their unique adsorption properties. The extra framework cations
invariably present in zeolites play significant role in determining their adsorptive properties
[l]. In particular, if coordinately unsaturated metal ions can be incorporated inside the
zeolite cavities, novel adsorption behavior may be fashioned on the basis of coordination of
guest molecules. Exchangeable transition metal ions in activated zeolites are generally
coordinately unsaturated and readily form complexes with a variety of guest molecules [2].
This is due to the van der Waals and Coulombic interactions between the extra framework
cations and the guest molecules. Synthetic zeolites of type A, X and mordenite having alkali
and alkaline earth metals as the extra framework cations have been extensively studied and
are mainly used as the nitrogen selective adsorbents for the adsorptive separation of oxygen
from air [3,4,5,6,7,8]. However, there are few studies on the adsorption behavior of
transition or rare earth cation exchanged zeolites. In the present paper, we report the
adsorption of nitrogen, oxygen and argon in some transition and rare earth metal ion
exchanged zeolite A and X.

2 Methods

2. I Cation Exchange
Commercially available zeolite A and X from Zeolites and Allied Products Mumbai
India was used as the starting material without any further purification. For exchangingwith
transition and rare earth metal ions, the zeolite was mixed with 0.1 M aqueous solution of the
specific cation and refluxed at 80C for 4 hours. The zeolite was filtered and washed with
distilled water until the washings were free fiom ions and used for the adsorption
measurements after drying. The percentage of ion exchange was determined by Atomic
Adsorption Spectroscopy after acid digesting the sample.

334
2.2 X-ray Powder Difiaction
Structuralchanges due to the cation exchange, if any, was determinedby measuring the
X-ray powder diffraction with Philips X'Pert M P D system using Cu Kctl(l= 1.54056) in
the 20 range 5 to 65.

2.3 Adsorption Isotherms

Oxygen, nitrogen and argon adsorption at 15C and 3OoCwas measured using a static
volumetric system (Micromeritics ASAP 20 10, USA), after activating the sample at 350C
under vacuum for 8 hours. Dosing of the adsorbate gas was done at volumes required to
achieve a targeted set of pressures ranging from 0.5 to 850mmHg. The adsorption and
desorption were completely reversible, hence it was possible to remove the adsorbed gases
by simple evacuation.

2.4 Heat of Adsorption

Isosteric heat of adsorption was calculated from the adsorption data collected at
different temperatures using Clausius - Clapeyron equation.
&dH" = R [alnpl/[a( 1IT)] 1 0
where R is the universal gas constant, 8 is the fraction of the adsorbed sites at a pressure p
and temperature T. A plot of Inp against 11T gives a straight line with slope of A,,+H"/R.

3. Results

3. I The Silver exchanged Zeolites A & X

Silver exchanged zeolite A shows anomalousadsorption behavior towards nitrogen and
argon. The adsorption isotherms measured at 303K on AgA are compared with those
measured in NaA and CaA at the same conditions in figure 1. The nitrogen adsorption
capacity for AgA is higher than (1.5 times at about 1 atmosphere pressure) that of CaA,
which has been reported to have highest adsorption capacity for nitrogen among Zeolite A
based adsorbents. The completely silver exchanged form shows an adsorption capacity of
22.3cclg for nitrogen, 4.36 cclg for oxygen and 6.25 cclg for argon at 30C and 765mmHg.
AgA shows argon selectivity (around 2) over oxygen, which is not generally observed in
other zeolite adsorbents. The high heat of adsorption value (-38 klmol-') for AgA compared
to 25 klmol-' for CaA shows that the nitrogen molecules undergo chemisorption-assisted
physorption with the silver species inside the zeolite cavities. However the silver exchange
on zeolite X increases the nitrogen adsorption capacity only by 20% compared to sodium
form. The adsorption isotherms on A@, NaX and CaX is given in figure -2. But the N2
selectivity over O2 is as high as 20 in the Henry's region. The heat of adsorption value for
nitrogen is high in the low-pressure region.

335
 Nitrogen Adsorption Isotherms

0 200 400 600 800

Pressure in mmHg

Figure -1 Adsorption Isotherms measured at 30C

~

A%r 1
Nitrogen Adsorption Isotherm
M
30
2 12
25

20

15

10

Pressureinnun&
Figure - 2 Adsorption isotherms measured at 30C

3.2 Cerium & Europium Zeolite X

Cerium and Europium exchanged zeolite X having different cerium and europium
loading was prepared and N2,O2and Ar adsorption measurementswere carried out at 15C
and 3OOC. The heat of adsorptionvalue for oxygen is found as high as -69 klmol-' in cerium
exchanged zeolite X. Samples having specific cerium loading displays oxygen adsorption
selectivity over argon and nitrogen. However, form our measurements extending up to
850mmHg equilibrium pressure, oxygen selectivity over nitrogen is limited to lower
pressure only ( 6 0 0 mmHg), i.e., Henry's region as which is shown in Fig.3.

336
 Europium exchanged zeolite X also shows higher heat of adsorption (-47 klmof')
value for oxygen. Samples having specific Europium loading displays oxygen adsorption
selectivity over argon and nitrogen in Henry' s region only which is shown in Fig.3.

CeX II Eux
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 200400600800 II It 0 20 40 60 80
Pressure in mmEIg
Figure -3 Adsorption Isotherm measured at 3OoCon CeX & EuX

The isosteric heat of adsorption on various zeolite adsorbents are given in table 1

Table-I Heat of Ahorption in kJ mot'K' at O.04mmoUg Coverage

Sample Nitrogen Oxygen I Argon
25.01
20.51
AgA 38.77F
NaX 22.45 15.62 13.98
CaX 29.15 15.23 14.12
Agx 37.86 15.84 14.76
CeX 17.57 69.48
EUX 16.10 46.90 I 15.60

4 Discussion

The various interactions contributing towards total energy of physical adsorption

include dispersion, polarization, field-quadmpole and close range repulsion interactions.
Framework oxygens and extra M e w o r k cations are the principal sites for interactionswith

337
the adsorbates molecules for zeolitic adsorbents. However, in case of zeolite adsorbents
used for the air separation, the main factors influencingthe nitrogen adsorptioncapacity and
selectivity are the difference in the quadruple moment of the adsorbate molecules and
accessibilityand the charge density of the cations [4,5].
Silver ions are reported to form neutral and charged silver clusters in the zeolite cavities
on vacuum dehydration at higher temperature [9].These coordinately unsaturated species
interact very strongly with nitrogen molecule [5].The other major factor contributing
towards the interaction of silver zeolite with nitrogen is enhanced polarization interactions
between AgA and nitrogen molecules. Ag+ ion with higher polarizing power compared to
Na'/Ca'2 ions and nitrogen molecules with relatively high polarizability interacts strongly
with each other resulting into enhanced adsorption for nitrogen.
The argon selectivity showed by the silver zeolite A can be explained in terms of the
difference in polarizability values. Argon having polarizability higher than that of oxygen
shows relatively high adsorption capacity and selectivity.
Higher interaction of oxygen molecules with CeX may be explined in terms of
interaction of oxygen molecules with non-stochiometricoxides of cerium probably formed
inside zeolite cavities due to interaction of cerium with zeolitic framework oxygens. Cerium
is known for its non-stochiometric oxides, which are oxygen selective.

5 Acknowledgements

We grateful to Dr.P. K. Ghosh, Director, CSMCRI and Department of Science and

Technology for the financial assist and support.

References

1. R.M. Barrer, Zeolites and Clay Minerals as Sorbents and Molecular Sieves; Academic
Press: London, (1978).
2. E.Y.Choi, Y. Kim and K.Seff, Microporous and Mesoporous Materials, 41 (2000),
61-68
3. S. Sircar, Ind Eng. Chem. Res., 41 (2002) 1389-.I392
4. R.V.Jasra, N. V. Choudary and S.G.T.Bhatt, Separation Science and Technology,26
(1991)885-.930
5. R.T.Yang, Y,D.Chen, J.D.Peck and N.Chen , Ind Eng. Chem. Res., 35 (1996)
3093-3099.
6. S.Sucar, R. R. Conrad and W. J. Ambs, US Patent 4,557,736,1985.
7. C. C. Chao US Patent 5,454,857,1995
8. J. Sebastian and R. V.Jasra PCTApplication (2002)
9. T.Sun and K Seff Chemical Reviews 94 (1994)857 - 870.
10. J.G.Nery, Y.P.Mascarenhas, Bonagamba and N.C.Mello, Zeolite, 18 (1997)44-49.
11. E.F.T.Lee and L.V.C.Rees ,Zeolite,7 (1987)446-450.

338
 ADSORPTION OF METHYLENE BLUE FROM WATER ONTO
ACTIVATED CARBON PREPARED FROM COIR PITH,
AN AGRICULTURAL SOLID WASTE

C. Namasivayam * and D. Kavitha

Environmental Chemistv Lab, Department of Environmental Sciences,

Bharathiar University, Coimbatore -64 I 046, iNDIA

*Comespon&ng author: Tek 91422422222; F m +92422425706
E-mail: [email protected] (C. Namasivayam)

The adsorption of methylene blue by coir pith carbon was carried out by varying the
parameters such as agitation time, dye concentration, adsorbent dose, pH and temperature.
Equilibrium adsorption data obeyed Langmuir isotherm. Adsorption kinetics followed a
second order rate kinetic model. The adsorption capacity was found to be 5.87 mg dye per g
of the adsorbent. There was no significant change in the per cent removal with pH. The pH
effect and desorption studies suggest that chemisorption might be the major mode of the
adsorption process.

Key wordr: Methylem blue, adrorption, coir pith carbon. isotherm, pH efeci,
desorption studies

1. Introduction
The conventional methods for removal of dyes using alum, femc chloride, coconut shell
based activated carbon etc., are not economical in the Indian context. Pollard el QL, [l]
and Namasivayam [2] have reviewed non-conventional adsorbents used for the removal of
dyes and heavy metals. Adsorption of acidic and basic dyes by activated carbon and bone
char [3] acid dye on bagasse pith [4] and orange peel [5], acidic and basic dye by biogas
residual sluny [6] and banana pith [7] have been reported. Industrial solid wastes like
Fe(III)/Cr(III) hydroxide for the removal of congo red [8] and red mud for the removal of
procion orange [9] have been investigated. Dyeing wastewater treatment using organic
coagulant and Fentons reagent [lo] and ozone [ l l ] have also been examined. The
photocatalytic oxidation and photoelectrocatalytic oxidation of rhodamine B using the
Ti/TiOz have been investigated and compared [12]. Biosorbents like chitosan beads for
reactive dye [13] and mixed culture of bacillus species and pseudomonas stutzeri for
methyl red [14] have been studied. Coir pith is a waste byproduct of coconut coir
industries in southern India. It is a soft biomass separated from the coconut husk during
the preparation of coir fiber. The purpose of this work was to investigate the removal of
methylene blue by coir pith carbon as a model study.

2. Experimental
2.1. Materials and Methods
Carbonized coir pith was prepared fkom dried coir pith powder (250-500pm) using a
muffle furnace at 7OO0C for 1 h. Adsorption experiments were carried out by agitating
300 mg of carbon with 50 nil of dye solution of desired concentration and pH at 200 rpm,
35OC in a thennostated rotary shaker (ORBITEK, Chennai, India). Dye concentration
was estimated spectrophotometrically by monitoring the absorbance at 660 nm using
UV-Vis spectrophotometer (Hitachi, model U-32 10, Tokyo). pH was measured using pH

339
meter(Elic0, model LI-107, Hyderabad, India). The dye solution was separated fiom the
adsorbent by centrihgation at 20,000 rpm for 20 min and its absorbance was measured.
Effect of pH was studied by adjusting the pH of dye solutions using dilute HCl and
NaOH solutions. Effect of adsorbent dosage was studied with different adsorbent doses
(25-600 mg) and 50 ml of 10,20,30,40 mg& dye solutions. For desorption studies, the
adsorbent that was used for the adsorption of 10,20 mg/L of dye solution was separated
from the dye solution by centrifugation. Then the spent adsorbent was agitated with 50
ml of distilled water, adjusted to different pH values for 60, 80 min. The desorbed dye
was estimated as before. For temperature studies, adsorption of 10 mg/L of methylene
blue by 200 mg of adsorbent was carried out at 35,40, 50 and 6OoC in the thermostated
rotary shaker.

3. Results and discussion

3.1 Effects of agitation time and concentration of dye on adsorption
The amount of dye adsorbed (rng/g) increased with increase in agitation time and reached
equilibrium. The equilibrium time was 40 and 60 rnin for 10 and 20 mg/L dye
concentration, respectively and 120 min for both 30 and 40 mg/L dye concentration..
The amount of dye removal at equilibrium increased from 1.4 to 5.4 mg/g with the
increase in dye concentration fiom 10 to 40 mg/L. It is clear that the removal of dyes
depends on the concentration of the dye.
3.2. Adsorption dynamics
3.2.1 Adsorbent rate constant
The rate constant of adsorption is determined from the first order rate
expression given by Lagergren [6]: log (qe -9) = logq, - kl t / 2.303 (1)
where qe and q are the amounts ofdye adsorbed (mg/g) at equilibrium and at time t (min),
respectively and kl is the rate constant of adsorption (Ymin).The disagreement between
the calculated and experimental qe shows that the adsorption of dye onto coir pith
carbon is not a first order reaction.
The second-order kinetic model [151 is expressed as t/q=l/kzq:+t/qe (2)
where kz (g/mg/min) is the rate constant of second order adsorption. The calculated qe
values agree very well with the experimental data. This indicates that the adsorption
system belongs to the second order kinetic model. The second order rate constants were
in the range 0.61-0.07g/mg/min. Similar phenomena have been observed in the
adsorption of Congo red on coir pith carbon [ 161.
3.3. Effect of adsorbent dosage
Increase in adsorbent dosage increased the per cent removal of dye, which is due to the
increase in absorbent surface area. The per cent removal was quantitatively at
300,400,500 and 600 mg/50ml adsorbent dose for 10,20,30 and 40 mg/L dye
concentration, respectively.
3.4. Adsorption isotherms
Langmuir isotherm is represented by the following equation [171:
CJqe = 1/Qob + C e IQo (3)
where C, is the concentration of dye solution (mg/L) at equilibrium. The constant Qo
signifies the adsorption capacity (mg/g) and b is related to the energy of adsorption
(L/mg). Qoand b were found to be 5.87 mg/g and 0.93 L/mg, respectively. Adsorption
equilibrium data do not follow the Freundlich isotherm [ 181.

340
3.5. pH effect.
The per cent removal was >90% in the pH range 2-1 1. At pH 2, though positively
charged surface sites on the adsorbent do not favor the adsorption of dye cations due to
the electrostatic repulsion, dye removal was still high (>90%). As the pH increased, the
removal increased slightly. This suggests that the chemisorption might play a major role
in the adsorption process. A similar trend was observed for the adsorption of methylene
blue by clay [19] and rhodamine B by biogas residual slurry [20]. The per cent desorption
was <lo% in the entire pH range 2-10 though there was slight increase from 0.5 to 9.0.
Insignificant desorption at different pH values confirms that chemisorption is the major
mode of adsorption process.
3.6. Effect of temperature
Increase of temperature increased the per cent removal. The change in standard free
energy, enthalpy and entropy of adsorption were calculated using the following
equations:
AGO = -RTlnK, (4)
where R is gas constant and K, is the equilibrium constant and T is the temperature in k.
Values of AGO were found to be in the range -9.239 to -13.57 (kJ/mol).
According to van't Hoff equation,

logloK, = AS0/2.303R - AHo/ 2.303RT (5)

Values of ASoand &were found to be 185.45(J/kfmol) and 48.02(kJ/mol), respectively.
Positive values of & show the endothermic nature of adsorption. Similar results have
been reported for the adsorption of methylene blue by clay [19].The negative values of
AG'indi~te the spontaneous nature of adsorption for methylene blue at 35, 40, 50 and
60' C. The positive values of ASosuggest the increased randomness at the solid /solution
interface during the adsorption of dye on coir pith carbon. Equilibrium data at different
temperatures for the adsorption of dye on coir pith carbon do not follow the first order
kinetic model but follows the second order kinetic model. The second order rate constants
were in the range 0.357-0.879g/mg/min.
4. Conclusion

The present study shows that the coir pith carbon is an effective adsorbent for the removal
of methylene blue from aqueous solution. Adsorption follows Langmuir isotherm. Kinetic
data follow second order kinetic model. The adsorption capacity was found to be
5.87mg/g. The results would be useful for the fabrication and designing of wastewater
treatment plants for the removal of dye. As the raw material, coir pith is discarded as
waste in coir industries, the treatment method using coir pith carbon is expected to be
economical.

341
REFERENCE:

1 S. J. T. Pollard, G. D. Fowler, C.J. Sollars and R. Perry, Low cost adsorbents

for water and wastewater treatment; a review, The Sci. Tot.Environ., 1 16,
31-52, 1992.
2. C. Namasivayam, Adsorbents for the treatment of wastewaters, In: Trivedy
R.K Editor, Encyclopedia of Environmental Pollution and Control. Karad,
Maharastra, India. Enviro-media, Vol. 1,3049, 1995.
3. G. .M. Walker and L. R. Weatherley, Adsorption of dyes fiom aqueous
solution - the effect of adsorbent pore size distribution and dye aggregation,
Chem. Eng. J.,83,201-206,2001.
4. B. Chem, C. W. Hui and G. Mckay, Film-pore diffusion modeling and contact
time optimization for the adsorption of dyestuffs on pith, Chem. Eng. J., 84,
77-94,2001.
5. C. Namasivayam, N. Muniasamy, K. Gayatri, M. Rani, K. Ranganathan,
Removal of dyes from aqueous solutions by cellulosic waste orange peel,
Bioresour. Technol., 57,3743, 1996.
6. C. Namasivayam and R. T. Yamuna, Utilizing Biogas Residual Slurry for dye
adsorption, Am. Dyestug Rep., 83,22-28, 1994.
7. C. Namasivayam and N. Kanchana, Waste banana pith as an adsorbent for
colour removal fiom wastewaters, Chemosphere, 25,169 1- 1705,1992.
8. C. Namasivayam, R. Jayakumar, R.T. Yamuna. Dye removal fiom waste
water by adsorption on waste Fe(III)/Cr(III)hydroxide. Waste Manag., 14,
709-7 16, 1994.
9. C. Namasivayam, R. T. Yamuna and D. J. S. E. Arasi, Removal of procion
orange fiom wastewater by adsorption on waste red mu4 Sep. Sci. Technol.,
37,2421-2431,2002.
10. S.El-Kadri ,'0.Dabbit and H. Kakhia, Treatment of wastewater containing
dyes used in the Syrian textile industry, J. of Chem. Tech. Biotech., 77,
11. 437443,2002.
M. M. Hassan and C. J Hawkyard, Reuse of spent dyebath following
12. decolorisation with ozone, Color. Technol., 1 18, 104- 1 11,2002.
X.Z. Li, H. L. Liu, F. B. Li and C. L. Mak,Photoelectrocatalytic oxidation
of rhodamine B in aqueous solution using Ti/Ti02 mesh photoelectrodes, J.
13. Environ. Sci. Health, 37,55-69,2002.
M. S. Chiou and H. Y. Li, Equilibrium and kinetic modeling of adsorption
of reactive dye on cross-linked chitosan beads, J. Haz. Mat., 93, 233-248,
2002.
14. K. Itoh, Y. Kitade, M. Nakanishi and C. Yatome, Decolorization of methyl red
by a mixed culture of bacillus sp. And pseudomonas stutzeri, J. Environ. Sci.
Health, 37,415421,2002.
15. G. McKay and Y.S. Ho, Pseudo-second order model for sorption processes.
Process Biochem., 34,451465, 1999.
16. C. Namasivayam and D. Kavitha, Removal of Congo red fiom water by
adsorption onto activated carbon prepared from coir pith, an agricultural solid
waste. Dyes Pigments, 54,47-58,2002.
17. Langmuir, The adsorption of gases on plane surfaces of glass, mica and
platinum, J. Amer. Chem. Soc,, 40, 1361, 1918.

342
18. F. Slejko, Adsorption Technology: A Step by Step Approach to Process
Evaluation and Application, Marcel and Dekker, New York, 1985.
19. D. Ghosh and K.G. Bhattacharyya, Removing colour fiom Aqueous Medium
by sorption on Natural Clay: A Study With Methylene Blue, Indian J. Enviro.
Prot., 21,903-910,2001.
20. C. Namasivayam and R. T. Yamuna, Removal of rhodamine B by biogas
residual sluny from aqueous solution, Wat.Air Soil Poflut.,65, 101-109, 1992.

343
 SEPARATION OF OXYGEN-ARGON MIXTURE BY
PRESSURE SWING ADSORPTION

XU JIN AND S.FAROOQ

Department of Chemical & Environmental Engineering, National University of Singapore
4 Engineering Drive 4, Singapore I I 7576
E-mail: chesAanus.edu.sg (S.Farooq).

Kinetic selectivities between 02 and Ar in four adsorbents were compared based on equilibria and
kinetics measured at low pressure. The equilibrium and kinetic parameters of Ar on the chosen sample,
a Carbon Molecular Sieve, were then measured over a wide pressure range. These equilibrium and
kinetic parameters were used to theoretically investigate separation of Oz-Ar mixture by Pressure
Swing Adsorption.

1 Introduction
The purity of oxygen product from air separation by equilibium controlled Pressure Swing
Adsorption (PSA) is limited to 95% by the presence of argon, which has the same
adsorption characteristics as 02.To meet the need of high purity 0 2 or Ar, cryogenic
distillation and catalytic hydrogenation are normally applied. Attempts to separate 02-Ar
mixture using kinetically controlled PSA have been reported in the literature [ 1,2]. The
present work is a part of the study aimed at choosing a suitable adsorbent and developing
feasible PSA cycles to separate 02-Ar mixture. Three CMS samples, Bergbau-Forschung
(BF) and two Takeda CMS (Takeda 1, Takeda 2) and a modified 4A zeolite (Union Carbide
RS-10)were investigatedfor this purpose. Although PSA processes may operate at elevated
pressure, the equilibrium and kinetics in the linear range are often used to give an
approximation for the initial screening of adsorbents. A proper kinetic model is crucial to
any dynamic simulation of PSA processes. While it is generally agreed that in CMS
adsorbents, the dominant resistance for gas diffusion is in micropores, three different
mechanisms have been suggested for transport in the micropores: distributed pore interior
resistance (pore model), barrier resistance confined at the pore mouth (barrier model) and a
combination of both (dual model) [3]. The dual model seems to best capture the uptake data
measured in this study. The bidispersed pore model developed earlier in this laboratory [4]
for kinetically controlled air separation has been modified to include dual transport
resistance in the micropores.

2 Adsorption equilibrium and kinetics in the linear range

Adsorption equilibrium and kinetics of Ar on the 4 adsorbents were measured by volumetric
method at three temperatures. The constant volume apparatus and the calculation method
have been described in detail elsewhere [3]. Small pressure steps (-0.1 bar) were given to
make sure that the measurementswere in the linear range (q* = Kc, where K is Henry law
constant, q* and c, in mmol/cc, are concentrations in adsorbed phase and gas phase). As
such, constant, limiting kinetic parameters were extracted.
Kinetics was determined by fitting the experimental uptake curves with a bidisperded
model. Only molecular diffusion was considered for the transport in macropore. The
micropore diffusion in CMS samples was expressed by a dual model of the following form
[3,51:

344
where DJr: (s-) is time constant accounting for distributed pore resistance and kb (s-) is
transport coefficient accounting for banier resistance. On RS-10, micropore diffusion
followed the pore model. Pore model was simulated by assigning a large value to in eq
(1).
F i g I. Uptake of A r r t 293K Fig 1. B r e a k t h r o u g h o f A r in Takeda 2
1 1
0.8 0.8
3
t0.4
0.6
- 0.6
0 0.4 0 cxpat 293 K
0.2 0.2 -model

0 0
60 0 200 400 600 800 1000
0 *ot,.,(s ..bj(O
D the t(s)

linear range) together with Henrys constants K are listed in Table 1. Representativematch
between experimental and model uptake curves is shown in Figure 1. It is obvious that the
uptake curve (plotted vs. t) for Takeda2 is of shape
9 and cannot be predicted by the
pore model. The same observation was also true for the other CMS samples.
Table 1. Equilibramand kinetics of Arand o2 adsorption on the 4 samples in the linear range

To validate the results from the volumetric method, breakthrough experiments were
also conducted using helium as carrier gas in columns packed with CMS samples. The
volumetric data predicted the breakthrough performance very well, as may be seen from
Figure 2.

3 Comparison of kinetic selectivities

Referring to O2 data [3] listed in Table 1, Ar diffuses much slower in all adsorbents. The
-
kinetic selectivity is defined by: a, - q~/ C A ,where qi is the average concentration in the
q B /cB
adsorbed phase.
Considering only micropore transport with constant parameters, the single component
uptake by pore diffusionin RS-10, following a step change in gas concentration, is given by

In CMS samples (dual model), the uptake is given by [5,6]:

$ cotpn+ L -1 = 0; L = -
kb
30, :1.

345
 Based on the assumption that the two species Fig 3. Kinetic i e l e e t i v i t y
diffie independently, which is twe in the linear 30
range, the calculation of UK as a function of time is
straightforward and are shown for all four adsorbents 20
Y
in Figure3. e
All the values approach the limiting value of equil 10
Among the 4 samples, Takeda 2 CMS gives the highest
for kinetically controlled PSA for 02-Arseparation. 0
0 2o t(s) 40 60
4 PSA simulation
PSA simulations were performed using a modified two-bed Skarstonn cycle without
external purge for the production of Ar [7,8]. The bidispersed pore model for kinetically
controlled air separation developed by Gupta and Farooq [6] has been extended to
incorporate dual resistance observed for transport of oxygen and argon in the CMS samples.
In view of space limitation, only some important features are highlighted here.
Adsorption equilibrium was represented by extended Langmuir isotherm with the same
saturation capacity (qs) for both adsorbates and constants (bi) following Arrhenius
temperature dependence:

where i denotes component, 8 is fractional coverage of the adsorbent, AU (kcal/mol) is the

change in internal energy, & (kcaVmoVK) is gas constant and T (K) is temperature. The
isotherm parameters are listed in Table 2 and the match to experimental results is shown in
Figure 4. The value of Henry's constant obtained from Langmuir fit is consistent with that
obtained ffom the linear range experiments.
Extended Langmuir model Dual model at 293 K
Adsorbate box1 O3 AU 9s D&*/rc2 kl*
(cdmmol) (kcdmol) (mmoVcc) (lo-%-') (IO-~S-')
pb
Oxygen 3.585 3.950 85.3 98.4 5.56
4.475
Argon 1.630 4.451 0.844 2.07 6.08

Micropore kinetics was described by a dual model with strongly concentration

dependent transport parameters. In a binary Lumgmuir system, the analogue of equation (1)
is:

The above equations are written for componentA. Similar equations apply for component B.
The concentration dependence of pore diffisivity followed Darken's equation. However,
the concentration dependence of the barrier coefficient was much stronger than that
expected from the analog of Darken's equation, which following other studies in this
laboratory [3,9], was represented by:
( h o = K"),
)I

346
(kio)Aris the limiting value obtained in the previous section. Representative results are
shown in Figure 5 and the extracted pb values are given in Table 2. Some simulation results
are summarized in Table 3.
Table 3: PSA simulation results

1. PR: pressurization;BD; blowdown; HP: high pressure adsorption;SP: self purge.

. -
2. Productivity: in units of cc(latm)/hr/cc adsorbent.
Feed composition: 95% 02 and 5% Ar on molar basis; temperature: 293 K; number of cycles: 25.
Bed length: 40 cm, bed radius: 1.9 cm; bed voidage: 0.35; particle radius: 0.159 cm; particle voidage: 0.33.

The purity of Ar in the high pressure cycles was limited to 50%. The purity and
recovery were improved in subatmosphericcycles with a consequent loss of productivity. In
blowdown steps, oxygen product was also concentratedto 96-98% with recovery up to 90%.
It is noted that 3 and 1 atm are the pressures of O2product in the two main industrial PSA air
separation methods, Lindox process and VSA process [2], respectively.
Studies are underway to further improve the purity and recovery of the argon product.
The results will be presented at the conference.
Fig 4. Iaotherms o f 0, and Ar Fig 5. Concentration dependence of
o n Takeda 2 D, and k , f o r A r inTakeda 2
3.5

g
5
CI 2
2.53

-
: 2 -model, : p 1
E 1.5 ij 10
a
-E 1
0.5
9" 5
no
0 0
0 0.1 0 2 0.4
C(mmbl/cc) 0.3 0 0.2 q,q. 0.4 0.6

References
1 . S. Hayashi; M. Kawai; T. Kaneko. Dynamics of high purity oxygen PSA. Gus. Sep.
Furif: 1996, 10, 19.
2. Salil U. Rege; Ralph T. Yang. Kinetic separation of oxygen and argon using molecular
sieve carbon. Adsorption. 2000,6, I5
3. Huang Qinglin; S M Sundaram; S Farooq. Revisiting transport of gases in the
micropores of carbon molecular sieves. Submitted for publication.
4. Gupta, R. and Farooq, S. Numerical Simulation of a Kinetically Controlled Bulk PSA
Separation Process based on a Bidisperse Pore Diffusion Model. Proceedings of the 81h
APCChE Congress, Vol. 3, p. 1753-1756, August 16-19, 1999, Soul, Korea.
5. Loughlin, K. F; Hassan, M. M.; Fatehi, A. I.; Zahur, M. Rate and equilibrium sorption
parameters for nitrogen and methane on carbon molecular scieve. Gas Sep. fur$ 1993,

347
 7,264.
6. J. Crank. The mthematics of difision. Clarendon Press, Oxyford, 1975.
7. Farooq, S; D. M.Ruthven. Numerical simulation of a kinetically controlled pressure
swing adsorption bulk separation process based on a diffusion model. Chem. Eng. Sci.
1991,46,2213.
8. S. Farooq; M. N .Rathor; K. Hidajat. A predictive model for a kinetically controlled
pressure swing adsorption separation process. Chem. Eng. Sci. 1993,48,4 129.
9. S Farooq, Huang Qinglin and S M Sundaram. Diffusion of Oxygen and Nitrogen in
CMS Micropores at High Coverage and its Impact on Kinetically Controlled PSA
Separation. In AICHE Annual Meeting, Reno, Nevada, Nov. 2001.

348
DUAL REFLUX PRESSURE SWING ADSORPTION CYCLE FOR
GAS SEPARATION AND PURIFICATION

ARMIN D. EBNER AND JAMES A. RITTER

Department of Chemical Engineering, Swearingen Engineering Center
University of South Carolina, Columbia,SC 29208, USA,
E-mail: ritterl3ener.sc.edu

A new dual reflux (DR) PSA cycle that combines the essential features of a conventional (stripping
reflux) PSA cycle with those of a relatively new enrichingreflux PSA cycle, is analyzed to show its
potential for separating gas mixtures with a minimal expenditure of energy. Based on isothermal
equilibrium theory applied to a binary linear isotherm system, the ultimate separation is carried out
where the binary feed is separated into two pure components with 100% recovery of each. This
idealized analysis reveals that such an ultimate separation is indeed possible over a wide range of
conditions and most surprisingly with pressure ratios as low as 1.1 to 1.5! In reality, it may be
entirely feasible to separate binary gas mixtures into two relatively pure components at very high
recoveriesusing only two DR PSA columns operatingin tandem with a very small pressure ratio.

1 Introduction

Pressure swing adsorption (PSA) has gained considerable commercial

acceptance over the last three decades [I]. Today, PSA is being used in a
wide variety of applications for bulk gas separation and purification.
Because of its modus operandi, however, state of the art PSA has been
limited mostly to the purification of the lightest component from bulk gas
streams. This is a direct consequence of the four basic steps associated
with the Skarstrom cycle, which is utilized by most PSA processes. These
steps include a high-pressure adsorption or feed step, a countercurrent
blowdown step, a countercurrent low-pressure purge step, and a light
product or feed pressurization step. This cycle allows for the complete
purification of the light component during the high pressure feed step.
However, during the low-pressure purge step, the best enrichment of the
heavy component in the exhaust stream is the pressure ratio, which is a
thermodynamic limit that is rarely achieved. This limitation vanishes,
however, if the pressures corresponding to the feed and purge steps of the
Skarstrom cycle are switched. Hirose and co-workers [2] demonstrated
that in such a PSA process the enrichment of the heavy component is no
longer limited by the pressure ratio, and more recently Ebner and Ritter 131
showed that complete purification of the heavy component is possible.
The consequence of these works is the redefinition of PSA cycles as
stripping and rectifying, by drawing a direct analogy with distillation.
Hirose and coworkers [2] and Mclntyre, et al. [4] showed further that it is
possible to combine both configurations into a dual reflux (DR) PSA

349
system, with light and heavy component enrichments only constrained by
the mass balance. The extreme limit of a DR PSA system is the complete
separation of a binary feed into two pure components; this concept has
never been demonstrated experimentally or theoretically. Therefore, the
objective here is to show that this perfect separation is entirely feasible
according to isothermal equilibrium theory based on linear isotherms [ 5 ] .

2 Description of the process

A 4-stage, twin-bed DR PSA system is depicted in Fig. 1. The process

consists of two stages carried out simultaneously at constant pressure (Fig.
la) and two stages carried simultaneously at varying pressure (Fig. Ib).
The feed is located along the column, and a recycle loop is located at each
end. Each column undergoes the same four steps: a) a constant low
pressure step at PL,where a feed NF is supplied to the column and a
stripping recycle NFLis entering the top (LHS column in Fig. la); b) a
pressurization step (LHS column in Fig. lb) from PL to PH,where the top
of the column is closed and then pressurized from the bottom with gas
from the other column (i.e., Npr = NBI);c) a constant high pressure purge
step at PH (RHS column in Fig. la), where the top of the column is opened
to produce pure light product NVL(= NFL+NPL) while a rectifying reflux
NVEis admitted to the bottom, and pure heavy product is produced (NPE)
from the other column; and d) a blowdown step from PH to PL,where the
top of the column is closed and blowdown (&I) occurs through the bottom
(RHS column in Fig. 1b) and pressurizes the other column.

3 Model

It is assumed that the DR PSA unit is able to separate and purifjl

Figure 1. Schematic of the
initial and final concentration
states (in gray) of the a) two
constant pressure steps and b)
two pressure varying steps that
constitute the DR PSA cycle.
The feed and pressurization
steps are depicted in the LHS
column, and the purge and
blowdown steps are depicted in
the RHS columns. The two
horizontal lines indicate the
1 possible location of the feed to
keep YAP=YAI during the feed
step.
completely the components of a binary feed. Thus, the mole fractions of
the heavy component (A) in all the enriched @A,E in NFE,NVEand N ~ E )
L NFL,NVL
and lean flows ~ A , in and NpL) are equal to 1 and 0, respectively.
Figure 1 shows the existence of constant profiles in the middle and upper
sections of the columns. During the constant pressure steps, these are YA,I
and YAJ (feed and purge), respectively, and are related according to [5]:

where 'I[;= PH/PL.~ A , Iis assumed to be equal to YA,F to avoid diluting during
blending. For a system with linear isotherms [6], all the gas that is
required to pressurize the column that just underwent the feed step comes
fiom the blowdown of the other column, i.e., Nw=Na. By defining
S=NFE/NF,the resulting relationship follows for the purge to feed ratio (y),
REand RL fkom local equilibrium theory for linear isotherms [6]:
(2)

(3)

(4)

Prepresents the relative affinity of the bed for the heavy component over
the light one (B) [5] and is equal to PA/& with fi given by
&
Pi = E + (1 - &)ki i=A,B (5)

where ki the dimensionless Henry's law constant and E is the bed void
fraction. The dark rectangle located on the LHS of the column undergoing
feed in Fig. 1 indicates that the location of the feed is not necessarily
restricted to a fixed position; it can be anywhere in between ZF,,,,~" and
ZF,ma. The former ensures no breakthrough takes place at the bottom of
the bed, while the latter ensures the feed concentration does not change
when mixing with the upstream flow. The expressions for ZF,min and ZF,ma
are given by [6]:

351
AT is related to length L, feed velocity UF and constant P step time At as
AT - 'F
= -pAAt
L

4 Discussion

Table 1 shows results for different values of p, 'II and S, for YA,P0.2, L=l
m, b 2 0 s and a p ~ 4 . 0 5 y,
. REand RL are obtained directly from Eqs (2)
to (4), and yA,2 is obtained from Eq. (1) with YA,I=YA,F. ATBTis the
dimensionless feed time defined according Eq. (8), for which ZF,min in
Eq.(6) is zero, i.e., it is the time above which breakthrough of the pure
heavy component always takes place [6].AZ~RBT is half this time. ZF,min,
ZF,- and UF are obtained from Eqs. (6) to (8), respectively, using ATII~BT.
Recall that perfect separation was assumed, i.e., y ~ , ~ and=0 y~,~1.0;
this necessarily results in 100% recovery for both components. However,
for this DR PSA process to be feasible, R ~ l l . 0 which
, was always the
case. What is even more remarkable is that the separations were achieved
at relatively small z (I 1S)! These results suggest that in reality it may be
possible to separate a binary mixture into nearly pure components at high
recoveries with a minimal expenditure of energy using the DR PSA
concept. It is also interesting to note the strong influence of 6and /3 on
While the first result is obvious, the second is not and due to the fact that
as the relative affinity for the heavy component increases (i.e., smaller p )
the amount of solute that desorbs during the feed step also increases,
which directly impacts the total gas leaving the bed. Notice the minuscule
impact of the three parameters on RE. Equation (3), as with Eq (4) for RL,
shows no dependence on 'II, a result that comes directly form the
application of equilibrium theory, and the fact that complete separation is
being achieved. RL is also independent of p: no matter the species in the
gas mixture one value of RLleads to the same result in all cases. The effect
of /3 on ZF,min, ZF,-, and UF is also strong, revealing that longer recti@ing
sections are needed for smaller /3. For the same reason, UF has to be
smaller. Finally, for small /?,only the rectifying section was required for
the separation.

352
 Table 1. Theoretical prededictions showing the feasibilityof the DR PSA concept for
different values of and .
6 x 6 Y & & Y&Z AfBT &I/ZBT q,min(@ uF(cmls)
0.2 1.2 0.01 3.38 0.95 0.01 0.174 0.047 0.024 0.211 0.000 2.353
0.2 1.2 0.1 3.75 0.96 0.11 0.174 0.043 0.021 0.211 0.003 2.128
0.2 1.2 1 7.50 0.98 0.56 0.174 0.022 0.011 0.211 0.015 1.087
0.2 1.2 10 45.00 1.00 0.93 0.174 0.004 0.002 0.211 0.026 0.185
0.2 1.1 1 8.18 0.98 0.56 0.186 0.022 0.011 0.319 0.015 1.087
0.2 1.3 1 6.92 0.98 0.56 0.164 0.022 0.011 0.145 0.015 1.087
0.2 1.5 1 6.00 0.98 0.56 0.146 0.022 0.011 0.073 0.015 1.087
0.8 1.2 0.01 0.84 0.84 0.01 0.194 0.660 0.330 0.466 0.004 32.990
0.8 1.2 0.1 0.94 0.85 0.11 0.194 0.604 0.302 0.466 0.033 30.189
0.8 1.2 1 1.88 0.92 0.56 0.194 0.327 0.163 0.466 0.177 16.327
0.8 1.2 10 11.25 0.99 0.93 0.194 0.058 0.029 0.466 0.317 2.920
0.8 1.1 1 2.05 0.92 0.56 0.197 0.327 0.163 0.482 0.177 16.327
0.8 1.2 1 1.88 0.92 0.56 0.194 0.327 0.163 0.466 0.177 16.327
0.8 1.3 1 1.73 0.92 0.56 0.191 0.327 0.163 0.452 0.177 18.327
0.8 1.5 1 1.50 0.92 0.56 0.187 0.327 0.163 0.428 0.177 16.327

5 Acknowledgements

Financial support provided by Westvaco and the Separations Research

Program at the University of Texas at Austin is greatly appreciated.

References
1 Ruthven, D. M., Farooq, S. and Knaebel, K. S., Pressure Swing
Adsorption, VCH, New York, 1994.
2 Diagne, D., Goto, M. and Hirose, T., Numerical Analysis of a Dual
Refluxed PSA Process during Simultaneous Removal and
Concentration of Carbon Dioxide Dilute Gas from Air, J. Chem. Tech.
Biotechnol., 65 (1996) pp. 29-38.
3 Ebner A. D., and Ritter, J. A., Equilibrium Theory Analysis of
Rectifying PSA for Heavy Component Production, AIChE J. 48 (2002)
pp. 1679-1691.
4 Mclntyre J. A., Holland C. E., and Ritter J. A., High Enrichment and
Recovery of Dilute Hydrocarbons by Dual-Reflux Pressure-Swing
Adsorption, Ind. Eng. Chem. Res. 41 (2002) pp. 3499-3504.
5 Knaebel K. S., Hill, F. B., Pressure Swing Adsorption: Development of
an Equilibrium Theory for Gas Separations, Chem. Eng. Sci., 40
( 1985) pp. 235 1-2360.
6 Ebner A. E. and Ritter J. A., Dual Reflux PSA for the Complete
Separation of Binary Gas Streams, AIChE J., submitted (2002).

353
 SIMULATION OF A COUPLED MEMBRANE / PSA PROCESS
FOR GAS SEPARATION

I. A. A. C . ESTEVES, J. P.B . MOTA*

Departamento de Quimica, Centro de Quimica Fina e Biotecnologia, FaculaMe de CiZncias e
Tecnologia, Universihde Nova de Lisboa, 2829-516 Caparica, Portugal
E-mail: iaesteves@dq. fct.unl.pt,[email protected]

An integrated model that successfully predicts the process characteristics of a novel hybrid gas
separation process combining membrane permeation and pressure swing adsorption (PSA) is
presented. The membrane performs most of the bulk separation operating in counter-current flow
mode to maximize the average driving force and therefore providing the most efficient arrangement.
Permeate and residue streams are fed to the PSA at different steps of the cycle for higher purity and
enhanced recovery.
Our simulation work shows that the efficiency of the pressurization and high-pressure adsorption
steps increase, thereby improving the separation performance as compared to a standalonePSA. The
process has been successfully applied to the bulk separation of a mixture of 50/50 HdCI-4 and
preliminary results have been obtained for COZ/CHJand Hz/COz/C& mixtures. The H~/CHJmixture
was selected because it is a challenging separation to show the benefits of the hybrid process since
the selectivity between CH& on activated carbon is high, making the PSA alone a very efficient
separation process for this mixture. Nevertheless, even in this unfavorable case the membrane
enhances the separation performance. It is expected that using the operating pressure of the PSA bed
as the driving force for membrane permeation minimizes recompression work and enhances
productivity.

1 Introduction

Membrane permeation and PSA are frequently considered as alternatives

to the conventional cryogenic gas separation processes. In a PSA process
the adsorbent is regenerated by reducing the partial pressure of the
adsorbate, which is accomplished by lowering the total pressure or by
using a purge gas. With the present generation of membrane modules, the
PSA process gains an advantage in the high-purity region while the
membrane system is more advantageous when product purity requirements
are less severe'. It is therefore it is reasonable to expect that an optimized
gas separation process integrating membrane and PSA improve product
purity and/or recovery as compared to the two-standalone systems.

2 Process Description

Fig. 1 shows the integrated cyclic process developed for bulk separation of
an A+B mixture, using a membrane and a dual-bed PSA unit. The process,
as presented here, operates under the assumption that the least adsorbed
component A permeates faster. The cycle starts with an incomplete
pressurization (PRI) of one bed with the permeate, which is stored in the

354
intermediate tank and is enriched in A. This stream was obtained during
the previous high-pressure adsorption step (HPA) operating on the other
bed. During PRI, valve VI is kept opened and V2 and V3 stay closed until
pressure equalization between tankhed is established. Then the tank outlet
is closed by shutting Vl. To complete the pressurization (PRz), V3 is
opened and the bed is pressurized with regular feed gas, which is less rich
in component A than the permeate employed in PR1. During PR2 the
membrane behaves as an empty tube, since both permeate and residue
sides are at feed pressure PH.HPA is initiated by opening V2 and feeding
the PSA with the residue from the membrane at a prescribed flow rate. The
residue is enriched in the strongly adsorbed component B, while the
permeate is stored in the tank to be employed in the next cycle. During
HPA the residue pressure is kept constant at the high-pressure value PH,
whereas the permeate pressure increases with time due to gas build-up in
the tank. The PSA cycle proceeds with a co-current blowdown (CD)to
recover the residual amount of A, which was pushed to the end of the bed
during the HPA; then a counter-current blowdown (BD) and finally a
purge (PG) to recover B and to regenerate the bed for the next ~ y c l e ~ . ~ .
During these steps the membrane is operating with other PSA bed in order
to provide continuity of flow. Although the operation of each bed is
batchwise, the system as a whole is a continuous one that is operated in
cyclic steady state (CSS).

0 M) I20 180 240 300 360 420

lime,.

Bed 1

Bed2
Munbruc

Figure 1 Flowsheet of hybrid membranaPSA process. Steps: pressurization (PR = PRI + PRz), high-pressure
adsorption (HPA), co- and counter-curred blowdown (CD, BD), Purge (PG);A : stream enriched in less
adsorbed component, B: stream enriched in strongly adsorbed component. Pressure histories of the integrated
cycle at CSS.

355
3 Theoretical Modeling

3.I Membrane Unit

The individual (i = 1,...,iVc- I ) and global material balances in the residue
(k= 1) and permeate (k= 2) sides of the membrane are

where 0 < z < Lm,Ak and A m are the cross-section and permeation areas,
respectively. If Qiis the permeance of species i, then its molar flux can be
Written as e.,l+2 -P2yi.?). For a binary mixture (A+B) the
= Qi(P,yiql
permeation flux of the less permeable species (B) can be expressed in an
alternative form as F ~= ~, ~ /~a ( p~-, ~P ~,~.J, I ~where
, ~ ) , a = Q ~ Q is
A the
ideal membrane selectivity.

3.2 Pressure Swing Ahsorption Unit

The individual material balances (i = 1,...,Nc-
1) in the adsorbent bed are

where 0 < z < L , A , = (1 - E)E, / E , and Rp= p p R g A p .The global material

balance and the energy equation read, respectively,
(1 + A ,)- (-)
a~ + P d (+I
at T az T
R, z:
)?& = 0
at
-

For a binary mixture (A+B)the pore-diffusion model6 can be written as

where 0,is the intraparticle effective diffusivity and rp the particle radius.

356
3.3 Multicomponent Adsorption Equilibrium
Multicornponent adsorption equilibrium is modeled using the Langmuir-
Freundlich (LF) isotherm model with the literature data given in [5].
The model equations were discretized in the spatial coordinate z using
a control-volume formulation with high-order flux correction applied to
the convective terms. The resulting large-scale stiff system of differential
algebric equations (DAEs) and process scheduling were solved using the
gPROMS simulation package.

4 Results and Discussion

The first case-study selected for assessment of process performance is the

bulk separation of a 50/50 H2(A)/CH@) gas mixture using a PSA with
activated carbon as the adsorbent, coupled to a membrane with selectivity2
CXAS = 35. The values of the parameters of the PSA model are given in and
are not reproduced here to save The standalone models for the
membrane and PSA unit were validated by comparing numerical results
with experimental data reported in the literat~re~~~. For the separation
under study the integrated system attained the CSS after the I 1th cycle from
startup for all runs. Unless otherwise stated, the operating conditions
considered are: PH/PL= 3Y1.2, Vc = 2500cm3, F(total feedlcycle) =
48L(STP), QA = IOOGPU, F m = 17. Parameter F m is a dimensionless
permeation flow defined as F,,,= QA A,,,(PH- PL)/ F, , where Ff is the total
feed amount in HPA. The process performance was study in terms of
products purities and recoveries, analyzing the effects of different cycle
variables on the separation results.
For the same separation performance, the integrated system has a
100 94
0
96
90 9
h E
5 92 * D A
8
*s
P
88 e
86 0 f
A
f
S 82
0
84 A

80 78
I 0 I
44 50 56 62 68 74 80 44 50 56 62 68 74 80
Feed, L(STP) Feed, L(STP)

Figure 2 Impact of feed flow rate on H2 product purity and recovcry for standalonc PSA (A, f H = 35 bar; A,
f H = 35 bar; m,
= 25 bar) and for integrated system (0, = 25 bar).

357
 100 0.8 1

$
95

90
tt! s
y
0.6

0.4
85
0.2
80

75 I I 0
0 0.05 0.1 0.15 0.2 0.25 0 0.2 0.4 0.6 0.8 1
P/F dL
Figure 3 Impact of P/F ratio on product purity and recovery for standalone PSA (Hz:A, purity; 0,recovery,
Ch: 0, purity; *,
recovery) and for integrated system (Hz:A, purity; m, recovery, Ch: *, purity;
recovery). CH4 concentration profiles in the PSA bed for a single and for the integrated system. PSA alone: 0
,

end of PR1(9s), 0 end of PRt(3Os), A end of HPA. Integrated process: H end of PRl(9s). + end of PRz, A end
of HPA.

higher feed throughput than the standalone PSA (Fig.2). Increasing the
purge to feed ratio increases H2 purity and C& recovery steadily, with loss
of H2 recovery and dilution of the CH4 product stream. The H2 recovery
and CI& purity for the hybrid system are higher than those for the
standalone PSA (Fig.3). The concentration profiles for the hybrid process
are sharper than those for the PSA alone, which enhances the separation
performance of the combined process (Fig.3). The reason for this is that,
unlike in a conventional PSA, the beds in the hybrid system are fed with a
varying-composition gas stream, initially rich in the more permeable but
less adsorbed component, which is progressively enriched in the other
component having opposite behavior. The results obtained show that the
coupled process increases the efEciency of the PR and HPA steps, thereby
improving the separation performance as compared to a standalone PSA.

5 Acknowledgements

I.A.A.C. Esteves thanks FCTMCT (Portugal) for a doctoral fellowship.

References
1. D.M.Ruthven et al., Pressure Swing Adsorption, VCH Pub.( 1994).
2. D.T.Coker, B.D.Freeman, GKFleming, AIChE J 44 (1998) pp. 1289-1301.
3. E.Drioli, M.Romano, I..Eng.Chem.Res. 40 (2001) pp.1277-1300.
4. R.T.Yang, Series on Chem. Eng., Imperial College Press Vol. l(1997).
5. R.T.Yang, S.J.Doong, AIChE J 31 (1985) pp.1829-1842.
6. R.T.Yang, S.J.Doong, AIChEJ 32 (1986) pp.397-410.
7. R.W.Baker, Ind.Eng.Chem.Res.41(6) (2002) pp.1393-1411.

358
 "CO AND ''CO SEPARATION ON Na-LSX USING PRESSURE-SWING
ADSORPTION AT LOW TEMPERATURES

J. IZUMI, N. FUKUDA, N. TOMONAGA, H. TSUTAYA, and A. YASUTAKE,

Mitsubishi Heavy Industries, Ltd., 5-71 7-1 Fukahori-mach, Nagasaki 851-0392,Japan
E-mail: [email protected]

A. KINUGASA and H. SAlKI

Mitsubishi Heavy Industries, Ltd., 1-1-1 Wadamisaki, Koube. Japan

Abstract. Economical I3CO separation is a very important process, not only for the production of isotopic
labels for nuclear magnetic resonance chemical analysis and medical diagnosis, but d S 0 to provide a means of
removing and fixing radioactive carbon (carbon-14) in the nuclear power industry. "C is normally separated
from 13C/12C CO or from 13C/12C CI% by cryogenic processes. As the separation factor is small, the cost of
separation is very high, so that the application of I3C is limited to use in the laboratory as a chemical reagent. It
is not viable to use it for industrial purposes. A marked difference was observed between the adsorption
equilibrium coefficients of "CO and I2CO on a low-Si02/Al203-ratio Na-X type zeolite (Na-LSX) at low
temperatures, so this phenomenon could be used to separate "CO from I2CO.

Keywords: Na-X type zeolite, I3CO selectivity, low temperature, equilibrium adsorbent, pressure-swing
adsorption

Introduction

Economical 13Cseparation is very important, not only for its use as an isotopic label in
nuclear magnetic resonance chemical analysis and medical diagnosis, but also to provide a
means for the removal and fixation of radioactive carbon (carbon-14) in the nuclear power
industry.
I3C is normally separated from l3C/''C CO or 60m I3CI1'C CH4 by cryogenic
processes. As the separation factor is quite low (the 13CO/12C0 separation factor is 1.007,
and the '3CH4/12CH,separation factor is 1.003), the cost of separation is very high. The
applications of I3C are therefore limited to use in the laboratory as a chemical reagent: its
use for industrial purposes is not viable. However, it is well known that CO is strongly
adsorbed on Na-X type zeolites'). Studies of the adsorption of CO on zeolites suggested
that some types of zeolite showed selectivity towards I3CO in the binary '3CO/12C0
system'). We therefore evaluated the relationships between the selective adsorption of
13
CO and a) the type of zeolite crystal, b) the Si02/A1203 ratios, and c) the chemical
modifications.

Experimental

A gaseous mixture of CO (13C/12C = 0.01 1, natural abundance) and helium as a carrier gas
was introduced into a small column of zeolite pellets for a selected time, then desorbed by
evacuation of the column.
Apparatus
A small-column adsorption apparatus was used to determine the adsorption

359
equilibrium and the rate characteristics of a bicomponent mixture of "CO and I3CO
(Figure-I). The column was loaded with 10-15 g of zeolite sample pellets, prepared by
extrusion as 1.dmm-diameter cylindrical samples. The samples were first heated to 400 K
in air for 2 hours to remove surface water, and then at 723 K, so that the adsorbents were
regenerated and became thoroughly free of pre-adsorbed water vapor. The column was
then placed in a refrigerator at a constant temperature of 2 13 K.
The inlet flow rate and pressure
were controlled by a mass-flow
controller. The mass flowmeter at the
outlet of the column detected changes
in the flow rate, and a sampling line
leading to the quadra-pole mass
spectrometer was placed at the exit,
for measuring the I2CO, I3CO, and
helium concentrations in the exiting
gas. The column was evacuated by a ~,
vacuum pump, with v-3 Opened and Figure-l. Small c-1- apparatus for '3CO/12CO
V-1 and V-2 closed. When the separation evaluationusing PSA.
column pressure reached the
desorption pressure, V-4 was opened, and helium was supplied for a countercurrent purge.
The amount of gas desorbed by evacuation was measured by a rotary flowmeter connected
to the outlet of the vacuum pump. "CO, "CO, and helium concentrations at the desorption
stage were not measured, but these values were determined from the mass balance
between the inlet and outlet at the adsorption stage.

Procedure
The adsorption pressure was adjusted by means of a pressure gauge installed on the
inlet gas line. The programmed sequence of the adsorption experiment was as follows: (1)
Valve V-1 was opened, the inlet gas of CO concentration cOl (I = I; I2co,I = 2; I3c0,I =
3; He) was introduced into the column, and the column pressure reached the adsorption
pressure Pa. This process took about 5 seconds. Valve V-2 was then opened, and
adsorption in the column fiom the inlet gas took place (adsorption step). During this
period, the flow rate and concentration of CO, GI,and C,,, were measured. (2) Valves V-1
and V-2 were closed, V-4 was opened, and the column was evacuated (desorption step).
At the end of the desorption step, the pressure was below 13 Pa after 600 seconds. (3) As
V-3 was also opened under evacuation, helium was supplied as a countercurrent purge to
remove CO thoroughly (countercurrent purge step). (4) With V-3 closed and V-4 still
open, the column was re-pressurized to the adsorption pressure with helium. The
measurement conditions are summarized in Table 1, and the samples that were screened
are listed in Table 2. The adsorption temperature was one parameter examined in this
study, and a sequence controller was programmed for each set of conditions, so that steps
( 1 x 4 ) were repeated over more than 4 hours. The total amount of desorbed gas G2 was
determined from the gas collected at the exit of the rotary flowmeter.

360
Table 1. Conditions for determination of the Table 2. Candidate adsorbent.

Adsorbent See Table 2

Adsorbent shspe and wi&t 1/16". 2 gram
Inlet gas rate 6W RI N/ninute
Inlet gas c a m s i t i o n CO 10 vol K. Ha 90 vol K
( W F ratio
:Natural abundance 0.011)
Adsorption tcrosrature 213-298K
Adsorption pressure 120 kPa
Dssorption pressure 1 kPa
Adsorption time 180 seconds

Calculrrtion of amount ahorbed

The amounts of I2CO and I3CO adsorbed during the pressurization and adsorption
periods were calculated by using the following equation:
4COi2 = {G2 ' c,,, - Gdead,ca)fw (1)
where qcoi denotes the amount of I2CO and "CO adsorbed on a unit mass of zeolite, and
GdeadCOi represents the amount of "CO and I3CO in the dead volume of the apparatus.
The amount of adsorbent packed in the column is denoted by w. Gdead,COi is estimated
from the dead volume of the small column.

Results and Discussion

The results of the screening test

The concentrations of I2CO and I3CO were measured at the outlet of the column
during runs with various samples (Figures-2a and -2b). The adsorption pressures, the
inlet oxygen concentrations, and the flow rates for all the runs were about the same; Pa =
120 kPa, Pd = 1 kPa, Co1= 0.1, Co2= 0.001 1, Co3= 0.9, and G I = 600 mL/min. The
adsorption temperature was 2 13 K. The adsorption period was 180 seconds.
As the inlet gas rate (mL Nhatch) increased, the mass transfer zone (MTZ) of '%O
reached the outlet, and the concentration of "CO became higher. Because the time taken
by the sample to reach the outlet increases with increasing amounts of CO adsorbed on the
sample, the sequence of the amounts of CO adsorbed by the samples can by determined
from the outlet I2CO concentration profiles.
The sequence of the amounts of CO adsorbed under these conditions is as follows:
NaX > Ca-A > Ca-X > Na-mordenite > Na-Y = Silicalite > Na-A.
The I3CO outlet concentration profiles of these samples at this time were also
determined (Figure-2b). Although these profiles are similar to the I2CO profiles (Figure-
2a), a slight difference was found on Na-X. In the case of Na-X, because the outlet
concentration of "CO was 8500 ppm (the effluent ratio 8.5%) at an inlet gas flow rate of
1500 mL Nhatch, I3CO could not be detected; this result strongly suggested that Na-X
could adsorb 13C0selectively, compared with I2C0.
The outlet 13CO-'2C0 isotope ratios ("C/"C) for the various samples were also
examined (Figure2c). The natural abundance of 13C is 1.1%. When a sample showed a
ratio lower than 1.1%, it adsorbed I3CO selectivity. In this screening test, Na-X
(SiO2/AI2O3 ratio 2.0, Na-LSX) and Na-Y (Si02/A1203ratio 5 ) belonging to FAU showed
selectivity towards I3CO. Na-LSX showed a particularly strong selectivity towards "CO.

361
 I d a qas m e ( d W b r c N

Figure-Za). Figure2b).
Outlet "CO concentration. Outlet 13c0concentration.

. :
$ 0
4
<
210 220 230 240 250 260 270 280
Adsorption t-rature (K)

- Figure-3.).
a Ida (PI m e ( drVbrtd3 Temperature dependency of
Figure-2c). "CO and "CO equilibrium
Outlet ~ 3 c 0 / 1 z cisotope
0 ratio. constants of Na-LSX.

I3COselectivity of Na-X
As a strong selectivity of Na-LSX towards I3CO was suggested by the screening test,
the adsorption behavior of Na-LSX towards I3CO and I2COwas evaluated. The adsorption
equilibrium constants of 13C0 and l2C0 ( B 13 and B 12, respectively) on Na-LSX were
determined at 2 13-273K (Figure-3a).
The lower the temperature, the greater were B 13 and 12. B 13 was greater than B
12 at all the temperatures studied. In this case, the separation factor cx in the binary
system '3CO/'2C0 is defined by Equation 3.
a = B 1318 12 (3)
The temperature dependence of the 13C0 separation factor of Na-LSX was studied
(Figure-3b). The separation factor increased with increasing temperature and reached a
value of 1.1 at 213 K. This separation factor for gas-phase adsorption on Na-LSX is
extremely high compared with the cryogenic separation factors for 13CH4/12CH4 (1.003)
and for '3CO/12C0(1.007). The mechanism for the separation is illustrated in Figure4 it
can be assumed that the interaction energy (mainly vibrational energy) between 13C0and
Na at the SIU site, which forms the active adsorption site, is greater than that between
l2C0 and Na at this site, and the statistical residence time of 13C0at Na at the SIU site is
greater than that of 1 2 ~ ~ 3 ) .

362
 -
v
s1.20, , I I ! , , ,
. .

0
Y
1 1 00
&? 210 220 230 240 250 260 270 280
Adsorption temperature (K)

Figure 3b). Figure 4.

Temperature dependence of 3CO/zC0 3CO/2C0separation mechanism.
separation factor.

13
CO/CO separation performance using pressure-swing ahorption
We are studying a system for the enrichment of a desorbed gas, by using a parallel
flow purge. When the adsorption is completed, the desorbed gas is supplied as a parallel
flow to the tower to substitute l2C0with 13C0for a three-tower VPSA unit (Figure 5).
We assumed that the enrichment performance of this VPSA unit can be calculated by
using a PSA simulation involving the parallel purge ratio R. The parallel purge ratio R is
defined in Equation 4.
R = Gp/Gd (4) Gp; Gas rate of parallel purge, Gd; Gas rate of desorption
The relationship between the separation factor and the enrichment ratio was
calculated (Figure-@. As the CO separation factor of Na-LSX at 213K is 1.1, it is
predicted that the enrichment ratio for a single stage of this VPSA will be 2.5.
3.

%D
icw Ipr

(1w 1.

Figure-5. PSA-3CO/2C0separation unit (single stage) Separation factor (-1

Figure 6. Separation factor and enrichment ratio

Summary

Isotope separation by means of gas-phase adsorption is quite a new area of study. In

the screening test for selective adsorption of 13C0 on zeolite samples from the binary
system 3CO/12C0,it was found that Na-LSX (Si02/A1203ratio 2.0) adsorbed 13C0
selectively at low temperatures, and that its 13C0separation factor reached 1.1 at 213 K.
Compared with cryogenic separation, this separation factor is extremely high.

363
 References

1) Japan Patent 3-072007, CO adsorption separation, Shirakawa, S., I m i , I., Tsutaya,

H., Kawamura, T.(1991)
2) I3COand "CO separation on Na-LSX using adsorption at low temperatures, Izumi, J.,
Tsutaya, H., Yasutake, A., Tomonaga, N., Saiki, H., Kinugasa, A., Abstract of 16th
Japan Adsorption Society Conference, p. 33(2002)
3) Interaction between cations inside FAU zeolite crystal and I2CO and I3CO, Yamazaki,
T., Yamada, H., Ohta, S., Inuni, J., Abstract of 16th Japan Adsorption Society
Conference, page 32(2002)

364
 High Purity Oxygen Generation PSA Process by Using Carbon
Molecular Sieve
Jeong-Geun Jee, Tae-Hoon Kwon, and Chang-Ha Lee'
Department of Chemical Engineering, Yonsei University, Seoul,Korea
Tel.: +82-2-2123-2762, Fax: +82-2-312-6401, E-mail: leech9vonsei.ac.k

Four different PSA cycles using CMS were compand for producing high purity oxygen over 99%
as well as the high productivity. The cyclic performances such as purity and recovery of four different
PSA processes were investigated experimentally under the n o n - i s o t h e d condition. The binary
(wh, 955 ~01.46)mixture was used as a feed gas for PSA experiments. As a result, the process I with
purge (BD2) after product (BDI) step produced the oxygen with about 98% purity and 66% recovery
while processes ll and Ill with product (BD2) after purge (BD1) step produced about 99.2 to 99.9%
purity and 56% recovey. Also,the process IV replacing the purge step in processes I to III with PE step
produced the 9 product with very high recovery about 78% as well as high purity about 99.7%.
Because the cyclic performances of the suggested processes were significantly affected by the diffusion
rate, the modified LDF model incorporating concentrationdependent rate parameter that was evaluated
from the Darken equation was proposed and successfully applied to the process simulation.
Key words: Q?PSA, CMS. concentrationdependent rate, nitrogen impurity

1. Introduction
High purity oxygen gases such as those having high purity of 99% or more are greatly
demanded industrially such as a welding gas, medical use, oxygen inhalation, combustion
process or a cylinder filling materials. Needless of this increasing demand on the high
purity oxygen, only a cryogenic process has been recognized as an effective method for
producing high purity oxygen product (99+ vol.%). However, its economical efficiency is
established only by the large-scale production of several tens of tons per day. Hence, when
a small amount of oxygen is used, the costs become extremely high because the above
high purity oxygen gas obtained from cryogenic plant is transported in the form of liquid
by using a tank lorry or packed in gas cylinders to supply it dividedly [ 11.
In that reason, many researchers suggested and have developed the PSA or VSA
processes for producing the high purity oxygen (99+ vol.%) since the mid 1980's. And the
most of these studies adopted a two-stage process consisted of zeolite bed for the first-
stage and CMS bed for the second. However, the additional equipment costs to set up the
second adsorption unit and the great product loss in CMS bed are a hindrance for
producing a high purity oxygen gas at a low cost. Therefore, the optimization of CMS
PSA and the development of new process to maximize the productivity need to stay
predominant over the cryogenic process.
In this study, the binary (Oz/Ar,9 5 5 vol.%) mixture assumed as the composition of the
products of equilibrium-controlled PSA process was used as a feed gas and CMS as
adsorbent. Also, four different PSA processes which have two-stage countercurrent
blowdown and one-stage cocurrent purge (or PE) steps with a different cycle sequence are
compared to c o n f m and develop both the high purity and high recovery oxygen
producing adsorption processes.
In a kinetically controlled separation system using CMS or zeolite 4A as adsorbents, it
is necessary to use more accurate rate model. Therefore, concentration dependent
diffusivity model based on Darken equation combined with Langmuir-Freundlich isotherm
was applied and each result was compared with the experimental data.

365
2. Mathematical models
All of the governing equations containing mass, energy, and momentum balances were
presented in our previous studies [2]. And, the boundary and initial conditions of
governing equations are presented below.
<Boundary conditions for feed pressurization, adsorption, cocurrent purge, and
pressurizing pressure equalization steps>

Where, y , I means feed composition for component i.

==0

<Boundary conditions for idle, blowdown 1, blowdown 2, and depressurizing pressure

equalization steps>

<Initial condition for fluid flow>

-
ci (:,O) =o ; qi (z,O) = 0 (4)

The multi-component adsorption equilibrium was predicted by the LRC model and the
sorption rate into an adsorbent pellet was described by the following modified LDF
model.
<Adsorption rate model: Darken equation + L F isotherm>
Farooq et al. [3] introduced a variable di hi vi t y model to a kinetically controlled PSA
separation process. They pointed out that the Darken equation with the Langmuir
isotherms predicted the experimental data better than the constant difisivity assumption.
Based on those results, following concentration dependent diffisivity model was
presented as the adsorption rate models.

Though the ideal gas assumption would cause some error in predicting result, the
reasonableness of the above suggested models can be explained by HI0 (Higashi, Ito,
Oishi) model (Higashi et al., 1963) which is based on the random walk of molecules. The
HI0 model was same with the model 4 in this paper when the single layer adsorption was
assumed.

3. Experimental
The schematic diagram of PSA apparatus was shown in Figure 1. The adsorption beds
were made of stainless steel pipe with a length of 100cm, ID of 2.2cm, and wall thickness
of 0.175cm. The beds were packed with CMS from the Takeda chemical company.
Calibrated three resistance temperature detectors (RTD, Pt 1000) were installed at the
positions of 10, 50, and 80cm fiom the feed end in order to measure the temperature
variations inside the bed. The two pressure transducers were located at the feed and
product ends in order to measure the bed pressure variation. The feed flow rate was

366
controlled by a mass flow controller
(Hastings, 202D-799). The total amount of
feed flow and the flow rate of each step
were measured by a wet gas meter
(Sinagawa Co. W-NK-1B). In order to
keep the pressure in the adsorption bed
constant, an electric back pressure
regulator was installed between the
adsorption bed and the product bed. The
concentration variations of the effluents at
adsorption and desorption steps were
analyzed by two portable oxygen analyzers
(Teledyne Analytical Instruments,
320BRC-D) calibrated by a mass
Rgure 1. Schematic diagram of 9 PSA
spectrometer (Balzers, QME 200). The apparatus
system was fully automated by a personal
computer with a developed control program and all measurements including pressure,
temperature, and 0 2 purity were saved on the personal computer through the AD
converter. The binary mixture (Oz/Ar;9 5 5 vol.%) was used as a feed gas for the PSA
experiments. The adsorption pressure and feed flow rate was fixed at 5atm and
5.5LSTP/min.

4. Description of the PSA cycles

Each process was consisted of conventional PSA steps such as pressurization (PR),
adsorption (AD), blowdown (BD), purge (PU), and pressure equalization (PE) steps.
However, the flow direction of all the steps except blowdown step was cocurrent.
Furthermore, the idle time was applied to keep the cyclic symmetry.
As can be seen in Table 1, the first blowdown step produces oxygen in the process I.
Then, the second blowdown gas is used as a purge gas for the other bed. Also, the one bed
is operated from the PR step to BDl step during the other bed inevitably undergoes idle
step. The desorption pressure at the end of BD1 step was 1.5atm at process I while 3.2atm
at processes 11 to Tv. However, the desorption pressure at the end of BD2 step was
ambient pressure at all the processes. Based on the cyclic sequence of process I, the
following modification was executed through process I1 and IV. In the process 11, the
purposes of two blowdown steps are opposite to the process I. Moreover, the idle step
(ID) between two blowdown steps is applied. And one bed is operated from @e ID1 step
to BD2 step during PR step in the other bed. The process I11 is nearly same as the process
I1 except non-idle step between two blowdown steps. Furthermore, only one idle step
exists and the PR step is hanged from the BD2 step to the middle of ID step. In the
process IV, the purge step in process I11 was replaced by PE step to maximize the
productivity and recovery with maintaining the high purity.

Table 1. Cyclic sequence of

BED1 PRf I A D T I BD13.

( 0 2 produced)
BD2 3.
(Purge)
m PUT

BED ll ID PUT PRf AD? BD14 BD2L

367
4. Results and discussion
The adsorption isotherms of Ar, N2,and 0 2 on P -
2.5 . ~~

C M S are shown in Figure 2(a). Although the rE 2.0

adsorption amount was 4, Ar, and N2in sequence,
the adsorption amounts were nearly same at low
1.5
.
f
adsorption pressure up to 5atm. As a result, the p 1.0
equilibrium selectivity among these adsorbents was eI!
not prominent in the range of typical operating 2 0.5
pressure of 0 2 PSA. However, as can be seen in 0.0
Figure 2(b), the kinetic selectivity between Ar and
O2 was very noticeable over all the experimental
pressure range. Therefore, the CMS bed for Ar u 10 -
removal from the O2 rich gas produced through s
L
zeolite bed PSA was expected to be successfully 2 10-
applied to this system. Nz,which exist as a minor
impurity in the product of equilibrium based 0 2
104
PSA, was also expected to be successfully removed
in the CMS PSA process because of the similar 0 2 4 0 0 10 12 14 18 18
Pnriura [am]
diffusion rate to Ar.
The performances of the suggested four different Figure 2. (a) Equilibrium isotherms and
(b) Diffusion time constants of three
PSA processes are shown in Figure 3 and Table 2. pum gases at 303K.
As a result, the process I showed relatively low 02
purity about 98%.It is because the large amount of
Ar in gas phase at BDl step was contained in the
product stream. However, the recovery was
relatively high about 66% because the effluent of
high-pressure region (5-1Jatm) was obtained as the
product. The processes 11 and 111 produced the high
purity oxygen in the range of 99.2 - 99.8%while the
low recovery about 56% because the high purity
oxygen in adsorbed phase of low-pressure region
(1.1-3.2am) was mainly obtained as product.
Furthermore, the process II that has the idle time
between BDl (purge step) and BD2 (product step) 90.5 (C) P m c a u ill
101 ,
produced the slightly lower purity oxygen than that
of the process 111. That is because the idle step 100 0/0aObMOOOWmoQ9
between purge and product step in process 11 caused
the desorption of Ar in adgorbed phase and this (d) P m w u iv
minor impurity was contained in following product 0 5 10 15 20 25
step. The process N with PE step showed Nunb.r ofeycba [ - 1

noticeably high recovery about 78% with Figure 3. Cyclic variation of 9purity of
maintaining high purity about 99.7% because the four different PSA cycles at Sam and 5.5
effluent gas of BD1 step used in both pressurizing LSTP/min
and cleaning the other bed.
As shown in Figure 3, the O2 purity drastically increased in first two cycles and the
cyclic steady state was reached at early cycles in the range of 3-5 at all the suggested
processes. In the PSA simulation, the modified LDF model introducing variable
diffusivity concept showed good agreement with experimental data.

368
 Table
Table 2. Ex 2. Experimental and predicted cyclic
A processes.

The predicted 0 2 concentration profiles in the

gas phase at the end of each step of two different
processes are presented in Figure 4(a) for process I
and (b) for process II. In Figure 4(a), the 0 2
concentration at the end of a pressurization step
showed a favorable decrease near to the product 0.70
- ".
end because the Ar in gas phase was increased by 0.60
the oxygen adsorption. However, the end position 1.10
of the bed kept clean with high O2 concentration @)-a
because the oxygen purged from earlier step
proceeded to the product end. At the end of an
adsorption step, the O2 concentration linearly but 0.80 4 I
slightly decreased to the product end. It means that
the adsorbed phase nearly reached the saturation 0.60 o.70
with high purity oxygen and only a small amount 0.50
of oxygen was being adsorbed at the vicinities of
1 -
the product end. At the end of the first blowdown Bed lerrslh [an1
step, the high purity oxygen over 99.5% was Figure 4. The predicted 9 concentration
produced as a bottom product through feed end. profiles in the gas phase at the end of each
However, the Ar rich stream was exhausted at early step of (a) process I, and (b) process II.
BD1 step. Also, at BD2 step, the high purity oxygen stream with similar composition to
that at the end of BDl step was purged to the next bed and the whole range of purge
receiving bed was kept clean with high purity oxygen. The O2MTZ of PR, AD, and BD2
steps at processes 11to IV was very similar to those at process I. However, as contrary to
the results of process I, the purge gas contained the Ar rich stream even at the end of BD1
step because pressure was decreased from 5atm to 3.2atm. Therefore, the feed end slightly
contaminated with Ar because the stream of BD1 step was used as purge gas.

5. Acknowledgement
The financial support of the Carbon Dioxide Reduction and Sequestration R & D Center
(COO2-0103-001-1-0-0) is gratefully acknowledged.

6. References
1. Haruna K. Process for Producing High Purity Oxygen Gas from Air. US Patent
4,985,052, 1991.
2. Yang J., Lee C.-H., Adsorption Dynamics of a Layered Bed PSA for Hz Recovery from
Coke Oven Gas.AZChE J. 1998,44,1325.
3. Farooq, S.; Rathor, M.N.;Hidajat, K. A Predictive Model for a Kinetically Controlled
Pressure Swing Adsorption Separation Process. Chem. Eng. Sci.. 1993,48,4129.
"."
4. Higashi, K., Ito, H., and J. Oishi, J. Aromic Energy SOC.,5, 846 (1963).

369
A STUDY ON THE PREPARATION OF DEODORIZING FIBERS BY COATING
TI01

S U N WHA OH, HYO JEONG KIM AND SO0 MIN PARK

Department of Textile Engineering, Pusan National UniversivvKeumjeong-Ku,
Pusan 609-735, Korea

E-mail: [email protected]

The fibers were coated with Ti02 for the purpose of deodorizing by dipping fibers into Ti02 sol
solutions and calcinated at 450 "C, 500 "C and 750 "C for 1 hr after coating. The surface structure
of the coated fibers were studied with XRD and SEM on the different preparation conditions of
coated Ti02 sol solutions and calcination temperatures. The deodorizing function of the prepared
fibers was studied by the determination of the decomposing capability of NH3, CH$H and
CH3CHO. The deodorizing effect of the prepared fibers showed the similar trend of decreasing
order as 5 wt% Ti02 sol solution calcinated at 450 "C < 3 wt?? Ti02 sol solution calcinated at 450
OC < 5 W? Ti02 sol solution calcinated at 500 "C < 5 W ?Ti01 sol solution calcinated at 750 "C.

1 Introduction

The photocatalytic minerlization of organic pollutants using active TiOz

photocatalysts has been established as an effective method for water and air purification
[ 11. However, the use of powdered photocatalysts in technological applications may
produce difficulties during processing. Consequently, recent investigations have focused
on the preparation of immobilized, active TiOz layers on suitable supporting materials.
Ti02 layers have been successfully prepared by the sol-gel process [2]. Their structures
and photochemical activity are strongly dependent on the temperature processing during
preparation [3]. The study on the heterogeneous photocatalyst for total oxidation of
organic and inorganic water pollutants has been carried out using suspensions of
powdered TiOz. However, the filteration of the slurries is very difficult for large
quantities. Recently, a few studies have been concerned with the preparation of
photocatalyst on different supports such as glass plate or glass bead [4]. The most
popular and efficient technique to immobilize TiOz on the support is the sol-gel technique
because it can modify the experimental parameters of this technique easily. The
photocatalytic degradation of organic materials in aqueous solution was also carried out
by the preparation of TiOz supported photocatalytic substrates on glass fiber using sol-gel
technique [5-61.
In the present study, we prepared Ti02 supported glass fibers by the dip-coating
using sol-gel technique in the different weight percent concentration of titanium
isopropoxide sol solution, calcination temperature and duration time. The prepared glass
fibers were investigated to determine the parameters for specific surface, crystallinity and
photocatalytic activity of the TiOz for the photodegradation of ammonia, methylmercaptan
and acetaldehyde.

370
2 Experimentals

2. I Materials
Titanium(IV) isopropoxide (Ti(OCH(CH3)2)4)and isopropanol were purchased fiom
Aldrich and used without any M e r purification. Ammonia water(28%), methyl
mercaptan standard solution(1pg/mL in benzene solution, Wako) and acetaldehyde
aqueous solution(3%) were applied as odors. All of these chemicals were used without
further purification.

2.2 Preparations of Ti02 sol and Ti02 supported on the glass fibers
To a solution of isopropanol (1.29 mL, 16.9 mmol) and titanium(IV)
isopropoxide(3wt%: 5 mL, 16.9 mmol, 5wt%: 8.33 mL, 28.2 mmol), 0.1 M HN03 (30
mL) was added by dropwise for 1 h at 25 "c . The reaction mixture was stirred for 8 h at
80 "C and then cooled to room temperature.
After removal of an adhesive dress from the technical product by firing for 10 h at
400 "C, the fabric was dipped into the prepared TiOz sol for 1 min. Then, the fabric was
removed from the solution at a constant rate of 30.00 cdmin. The samples were dried at
the ambient condition for 24 hr, and then calcinated in a furnace by increasing calcination
temperature from room temperature to 450 OC ,500 "C or 750 "C by 100 "C for every 30
min. and then each dip-coated glass fiber was continuously calcinated for 1 hr 450 "C,
500 C and 750 C,respectively.

2.3 Evaluation of deodorant activiy

The deodorant activity of the deodorizing fiber was determined using chromogenic
gas detector tubes (Gastec, Japan) to determine the degree of photodegradation of the
odor molecule with Ti02 photocatalysts incorporated glass fibers in the gas phase during
illumination with UV lamp. The gas phase solution was prepared by the evaporation of
inserted 0.1 mL each solution (ammonia, methyl mercaptan and acetaldehyde). These
experiments were carried out using a reservoir chamber of a Pyrex flask (1 L) illuminated
by a 40W UV lamp (40 W,Ushio) with 20 cm distance between the UV lamp and sample
flask. For all experiments, the glass fiber supported with Ti02 was introduced into the
flask. Each 0.1 mL odor molecule solution was added to the sealed chamber flask.
Then, the flask was incubated for 10 min at 45 'C to make phase transition from liquid
phase to gas phase. For each chamber flask, the initial concentration of odor molecules
was about 500 ppm and UV light irradiation was started from the initial concentration.

371
 I I I --' r ' I ' ~ -1
20 30 40 50 60 70
2Theta (deg.)
Figure 1. Low angle XRD panems of Ti& supported on glass fiber calcinated at 450 OC,500 "C and 750 "C
afIer coating with 5 wt?hTi02 sol solution.

3 Results and Discussion

The representative low angle XRD difiactograms for TiOz coated glass fibers
calcinated at 450 "C,500 OC and 750 "C for 1 hr after coating with 5 wt?! Ti02 sol
solution are shown in Figure 1.
This shows that the main phases at 450 OC and 500 "C are pure anatase, but the
calcinations temperature higher than 500 OC,the rutile phase begined to appear and major
phase of rutile appeared on the calcinations temperature at 750 "C. The increase of firing
temperature above 500 OC results in the phase transition fiom anatase to rutile. This is
already reported in previous literature [3, 51. The anatase phase is the more
photocatalytically active allotropic form of Ti02.
Most of the previous studies have been tried with photocatalytic mineralization of
organic pollutants using active TiOz photocatalysts as an effective method for water and
air purification. However, the use of powdered photocatlysts in technological applications
has many manipulation problems during processing. Such technological problems have
been tried to be solved by incorporation of Ti02 photocatalysts on the solid substrate,
glass fibers or fibers. The incorporation of TiOz photocatalysts on the glass fibers or
polymer fibers can increase the active surface area of Ti02 photocatalysts by
immobilization on the surface. In the present study, the TiOz photocatalysts has been
tried to incorporate on the surface of glass fibers by sol-gel process. This kind of study
has not been reported so far.

372
Figure 2. Disappearance of NH3 with 5wt% Ti02 Figure 3. Disappearance of NH3 with supported on glass
5wt% Ti% supportedon glass fiber. fiber.

The overall photocatalytic activity was chosen as the rate of decomposition of odor
molecules such as NH3, CH3SH and CH3CH0. Figure 2 shows that the degree of
photodegradation of NH3 molecule on the uncoated glass fiber, T i 4 photocatalysts
coated glass fibers twice and five times with 5 wt?? TiOz sol and calcinated at 450 'c.
The photodegradation rate of NH3 increased with increasing Ti02 sol coating number. The
same results were obtained for the photodegradation of CH3SH and CH3CHO.
Figure 3 shows the degree of photodegradation of NH3 during with Ti02
photocatalyst uncoated glass fiber, TiOz photocatalyst coated glass fiber with 3 wt?hTiOz
sol and calcinated at 450 "C,TiOz photocatalyst coated glass fiber with 5 W h Ti02 sol
and calcinated at 450 "C, TiOz photocatalyst coated glass fiber with 5 wt?! Ti02 sol and
calcinated at 500 "C and TiOz photocatalyst coated glass fiber with 5 wt?? Ti02 sol and
calcinated at 750 OC under the system of UV light illumination. The photodegradation
efficiency of NH3 molecule was greatly increased as increasing concentration of Ti02 sol
for the coating of glass fiber up to 5 wt?? at the cacination temperature of 450 "C. This is
interpreted as the increased concentration of TiOz sol solution results in the higher number
of coated TiOzphotocatalytic particle on the glass fiber surface. This enables to increase
the photodegradation efficiency of NH3 molecule. The photodegradation efficiency of
NH3 molecule was abruptly decreased with increasing calcinations temperature to 500 OC
and then 750 "C. This is possibly caused by the phase transition of Ti02 photocatalyst
coated on the surface of glass fiber from anatase to rutile phase. As it is already known,
the anatase phase has the higher photocatalytic effect than the rutile phase. So increased
calcinations temperature fkom 450 "C to 500 "C and furthermore to 750 "C decreases the
photodegradation efficiency of NH3 molecule due to phase transition of Ti02
photocatalyst coated on the surface of glass fiber. The same results were obtained for the
photodegradation of CHsSH and CH3CH0.

373
4 Acknowledgements

This study was supported by research grants from the Brain Korea 2 1.

References

1. Ollis D.F. and Al-Ekabi H., Photocatalytic Purification and Treatment of Water and
Air; Elsevier, Amsterdam (1993).
2. Aguado M.A. and Anderson M.A., Solution factors affecting the photocatalytic and
photoelectrocatalytic degradation of formic acid using supported Ti02 thin films J.
Photochem. Photobiol. A: Chem. 94 (1 996) pp 22 1-229.
3. Ding X.Z.and He Y.Z., Study of the room temperature ageing effect on structural
evolution of gel-drived nanocrystalline titania powders. J. Muter. Sci. Left. 15 (1996)
pp 320-322.
4. Hermann J.M., Tahiri H., Ait-Ichou Y., Lassaletta G. and Gonzales-Elipe A.R.,
Characterization and photocatalytic activity in aqueous medium of Ti02and Ag-Ti02
coatings on quartz. Appl. Catal. B 13 (1 997) pp 2 19-228.
5. Robert D., Piscopo A., Heintz 0. and Weber J.V., Photocatalytic detoxification with
TiO, supported on glass-fibre by using artificial and natural light. Catalysis Today 54
(1999) pp 29 1-296.
6. Brezova V., Blazkova A., Karpinsky L., Groskova J., Havlinova B., Jorik V. and
Ceppan M., Phenol decomposition using M"'/Ti02 photocatalysts supported by the
sol-gel technique on glass fibres. J. Photochem. Photobiol. A: Chem. 109 (1997) pp
177-183.

374
 COMPOSITE ADSORBENTS FOR THE REMOVAL OF CS AND SR IONS IN ACIDIC
SOLUTIONS

J. K.MOON, C.H.JUNG, S. H. LEE, E. H.LEE

Korea Atomic Energy Research Institute,P.O.BOX lOS, Youseong, Daejon, Korea
E-mail:njkmoon@ re.kr

H.T.KIM, Y.G. SHUL

Dept. of Chemical Engineering, Yonsei University. Seoul, Korea.
E-mail:[email protected]. kr

PAN-KCoFC and PAN4A composite adsorbents were prepared for the removal of cesium and strontium ions
from acidic nuclear waste solutions. High porous spherical composite adsorbents could be prepared using a dual
nozzle technique. The acid and radiation stability tests showed that the both composite adsorbents were stable
against acid solutions higher than pH = 2 and radiation dose less than 1.89 MGy, respectively. Adsorption tests
showed that the PAN-KCoFC was selective for cesium ion and the PAN-zeolite 4A was for strontium ions,
respectively. The ion exchange equilibrium isotherms were obtained and evaluated for the binary systems.

1 Introduction

Many inorganic ion exchangers have been studied for the removal of strontium and cesium ions, due
-
to their high selectivity for specific cations, thermal and radiation stabilities [1 51. In spite of those
advantages, they have been liited in their extensive applications because of the pressure drop
problem in column operation when they are used in microcrystallineor powdered form. To solve this
problem, in the case of synthetic zeolites, clay minerals are often used for pelletization, but it still
causes dissolution of clay minerals in aqueous solution.
For these reasons, the studies on the preparation of composite adsorbents using organic binder such
as polymer have been vigorously performed[6-12]. PAN(polyacrylonitri1e) has been viewed as one
of the most favorable binding materials, due to its physico-chemical properties such as excellent
bead formation, strong adhesive force with inorganic materials, good solubility for organic solvents
and chemical stability[lO].
Many researchers have studied the preparation and evaluation of the physico-chemical properties of
PAN-based composite adsorbents. In spite of some successful results on the preparation and
evaluation of the PAN-based composite ion exchangers, further studies on the preparation and
characterizationof various kinds of composite ion exchangers are still required.
Therefore this study focused on the preparation of PAN-4A and PAN-KCoFC composite ion
exchangers and characterization of their physico-chemical properties for the removal of Sr and Cs
ions in acid solution.

2 EXPERIMENTAL

To prepare PAN-KCoFC and PAN4A composite beads in which the inorganic powder content is
80 percent, the given amounts of inorganic powders were dispersed in DMSO(dimethylsu1furoxide)
solvent and then mixed with PAN(polyacrylonitri1e) polymer and a few drops of TWEEN-80

375
sufactant to make homogeneous composite dope. The detail procedures for preparation of
composite beads were the same as described in our previous study[ 1I]. Pore size distribution and
porosity of the prepared composite beads were measured by mercury porosimeter(Micrometrics,
AutoPoreIII). The pore structure and the distribution of the powders in the composite beads were
observed by scanning electron microscope(JE0L Co., JSM 5200).
Ion exchange tests for strontium and cesium ions were carried out to determine ion exchange
equilibrium isotherms in single component system. Ion exchange equilibrium isotherms were
obtained in batch system where the dosage of adsorbent and the pH were fixed constant at
1omWO.Ig and 2, respectivly, while the solution concentrations were varied between 0.0002 0.2 -
N.

3 RESULTS AND DISCUSSION

1. Preparation of PAN Based Composite Ion Exchangers

Spherical composite ion exchanger beads, PAN-KCoFC and PAN-4A, were prepared using a dual
nozzle technique. The composite bead sizes were found to be more or less controlled by air pressure,
with the fixed nozzle size, as in Figs. 1 and 3. In case of PANAA, the size variation is shown to be
larger than PAN-KCoFC, and it might be due to the differences in composite dope viscosities
depending on the kind of inorganic materials. The SEM images of the composite beads showed that
the inside pores were well developed as in Figs 2 and 4. The pore size distribution and porosity data
were represented in Figs. 5-6. They show that the porosity of the PAN-KCoFC beads is about 75%
with the average pore size of 0.08 m, and about 74% with the average pore size of 0.14 m for
PAN-zeolite 4A beads, which are obviously high porosity compared with other inorganic granules
such as zeolites.
2. Chemical and Radiation Stability of the Composite Ion Exchangers
The structural stability of the PAN-KCoFC and PAN-zeolite 4A composite beads in acid solution
was proven by observing no weight loss for 5 day contact with the nitric acid solution of pH = 2.
Also the results of ion exchange performance tests confirmed that the composite ion exchanger does
not deteriorate by contacting with the acid solutions of pH = 2 for 5 days.
Evaluation of radiation stability of the composite ion exchanger beads was carried out by measuring
weight loss and variation in ion exchange property together with observing morphology change for
the radiation dose ranging from 1.00~10'to 1 . 8 9 ~ 1 Gy.
0 ~ We found no radiation effect on the weight
loss, ion exchange property changes and any external deformation.

3. Evaluation of Ion Exchange Performances

Ion exchange isotherms for PAN-KCoFC/Cs, KCoFC/Cs systems and PAN-zeolite 4A/Sr, zeolite
4NSr systems were obtained to evaluate the equilibrium parameters such as ion exchange capacity
and equilibrium constant for kinetic calculations. The experimental data were modeled by Langmuir
equation given by
q,bc
1+bc

376
where q (meq/g) and c (meq/mL) are the equilibrium concentrationsin the solid and liquid phases,
respectively, and q,(meq/g) is the saturation concentration in solid phase, and b, the coefficients
fitted to the experimental data.
The results in Fig. 7 indicate that the Langmuir model fits the data. The ion exchange capacities of
the PAN-KCoFC and PAN-4A composite ion exchangersobtained by modeling the experimental
data using Langmuir model were 1.07 meq (Cs)/g and 2.90 meq (S?)/g, respectively. These values
are almost 80% of those for KCoFC and 4A powders, respectively. This fact is obviously due to the
difference in inorganic contents between them. It also indicates, taking into account the 80%
inorganic content in the composite ion exchanger, that the PAN binder provides a smooth pathway
for ion transfer without blocking the adsorption sites.

4 CONCLUSION

PAN-KCoFC and PAN-4A composite ion exchanger beads containing about 80Y0inorganic powders
were prepared to evaluate ion exchange behavior for cesium and strontium in acid solution. The acid
and radiation stability tests confirmed that the composite ion exchanger beads were stable in acid
solutions of more than pH = 2 and against radiation dose less than I .89 MGy, respectively. Also, ion
exchange tests proved that the prepared composite ion exchangerscould effectively be used for the
removal of cesium and strontium ions in acid solution.

5 ACKNOWLEDGEMENT

This project has been carried out under the Nuclear R&D program by MOST

REFERENCES

1. Jung K.T., Shul Y.G., Moon J.K., OH W.J.,Waste treatment and immobilization
technologies involving inorganic sorbents, IAEA-TECDOC-947( 1992-1996) 193.
2. Marageh M.G., Husain S.W., and Khanchi A.R., Selective Sorption of Radioactive Cesium and
Strontium on Stannic Molybdophosphate Ion Exchanger, Applied Radiation and Isotopes 50
(1999) 459.
3. Mimura H., Letho J., Harjula R., Ion Exchange of Cesium on Potassium Nickel
Hexacyanoferrate(II)s, J. of Nucl. Chem. Technol. 34(5) (1997) 484.
4. Moon, J. K , Kim, H. T., Shul, Y. G., Lee, E. H. and Yoo, J. H., Ion Exchange Behavior for
Mixed Solution of Sr and Cs Ions with Potassium Titanate, HWAHAK KONGHAK 38(6)
(2000) 847.
5. Liang T., Hsu C., Sorption of Cesium and Strontium on Natural Mordenite, Radiochimica Acta.
61(1993) 105.
6. Mardan A., Rumana, A., Mehmood, A., RAza, S. M. and Ghaffar, A., Preparation of Silica
Cobalt Hexacyanoferrate Composite Ion Exchanger and Its Uptake Behavior for Cesium, Sep.
Puri. Technol. 16 (1999) 147.
7. Mariamichel A., Kishnamoorthy S., Synthesis and Characterization of New Composite Ion
Exchangers, Scientific & Industrial Research 56 (1 997) 680.

377
8. John J., Sebesta F., Motl A., Composite Absorbers Consisting of Inorganic Ion-Exchangers and
Polyacrylonitrile Binding Matrix. Leaching of Cs-137 fkom Cemented NiFC-PAN Absorber,
Radiochimica Acta. 78 (1997) 131.
9. Sebesta F. and John J., An Overview of the Development, Testing, and Application of
Composite Absorbers, LA-12875-MS 1995.
lo. Sebesta F., John, J., Motl, A. and Watson, J., Proc. 5 Int. Conf. on on Radioactive Waste
Management and Environmental Remediation 1995, Berlin, p.36 1.
11. Moon J.K., Kim, K. W., Jung, J. H., Shul, Y. G. and Lee, E. H., Preparation of
Organic-inorganic Composite Adsorbent beads for Removal of Radionuclides and Heavy Metal
Ions, Radioanalyticaland Nuclear Chemistry, 246(2) (2000) 299.
12. Xue T.J., McKinney A., Wilkie C.A., The Thermal Degradation of Polyacrylonitrile, Polymer
Degradation and Stability 58 (1997) 193.

Figure 1. Effect of air pressure on the PAN-KCoFC composite bead sizes.

(a) Atmospheric pressure (b) 2 psi (c) 4 psi (d) 6 psi

(a) @)
Figure. 2. SEM images of the fracture(a) and magnification of macro pores(b) of PAN-KCoFC
composite ion exchanger beads.

378
 Figure 3. Effect of air pressure on the P M 4 A composite bead sizes.
(a) Atmospheric pressure (b) 2 psi (c) 4 psi (d) 6 psi

(a) ( b)
Figure 4. SEM images of the fiacture (a) and magnification of macro pores(b) of PAN-4A
composite ion exchanger beads.

Average Pore
Diameter
0.08 Frn

Porosi 75.70

Figure 5. Pore size distribution and porosity data of the PAN-KCoFC composite ion exchanger

379
 Average Pore
Diameter 0.14 pm
I
Porositv 73.67

Figure 6. Pore size distribution and porosity data of the PAN-4A composite ion exchanger

0.w 0.05 0.10

'f
1.o

0.0
0.w 0.05 0.10 0.15
I
0.20
Concentration in solution, CpeqlmL) Concentrationin solution. C.(meqlmL)

Figure 7 Ion exchange isotherms of the composite ion exchangers and the pure inorganic powders.
(a) PAN-KCoFC and KCoFC, (b) PAN-zeolite 4A and zeolite 4A.

380
 DEHUMIDIFICATION BEHAVIOR OF METALPI, AL, MG) SILICATES
IMPREGNATED CERAMIC FIBER SHEETS

Y. S. AHN, C. H. CHO, Y. J. YOO, J. S. KIM, H. S. KIM AND M. H. HAN

Centerfor Functional Materials Research, Korea Institute of Energy Research, 71-2 Jang-dong,
Yusong-gu, Taejon 305-343, Korea
E-mail:[email protected]

Ceramic sheets with magnesium silicate, aluminum silicate, titanium silicate adsorbents were
prepared by sol-gel based incorporation process of Mg, Al and Ti ions into silica-impregnated
ceramic sheets. In all the adsorbent-impregnated ceramic sheets, there were two kinds of pores:
one was the micropore with diameter less than 3nm and the other was the diffuse mesopore with
diameter of 3 to IOnm. In cases of the pure silica and magnesium silicate-impregnated ceramic
sheets, the average diameter of the mesopores was 4nm. On the other hand, the aluminum and
titanium silicate-impregnated ceramic sheets have the mesopores with average diameter of 5nm
and with more wide size distribution. In the aluminum and titanium silicate-impregnated ceramic
sheets, the micropores were well developed, compared with the pure silica and magnesium
silicate-impregnated ceramic sheets. The aluminum and titanium silicate-impregnated ceramic
sheets showed be@ dehumidification behavior than the silica and magnesium
silicate-impregnated ceramic sheets. The superior dehumidification efficiency originates from the
well developed micropres.

1 Introduction

Recently, rotor-type adsorption systems attract much attention, because the systems
can effectively control humidity and air pollutants such as volatile organic compounds
(VOCs), nitrogen oxide ("Ox) and sulfuric oxide (SOX). Gas remediation efficiency is
principally concerned to the gas capturing ability of the adsorbents impregnated into the
ceramic honeycomb rotor such as zeolite, silica, activated carbon, etc. Therefore, it is one
of the most important works to develop the adsorbent with good absorption-desorption
behaviors.
Nowadays, rotor-type dehumidification and air-conditioning systems are spotlighted
and begin to be industrialized, because the systems don't use CFCs which are known to be
representativechemicals destroying an ozone layer in earth. Generally, silica gel is used as
a dehumidifier in the dehumidification and air-cooling rotor. Therefore, it is important to
improve the dehumidification efficiency of silica desiccant in this field.
The dehumidification efficiency of silica desiccants very depends on the synthesis
and aging conditions such as pH[1,2]. But, there seems to be a limit to improve the
dehumidification efficiency by using pure silica desiccants.
Recently, Kuma et uf. have reported that the M-rotor treated with aluminum sulfate
showed better performance for dehumidificationthan the S-rotor treated sulfuric acid[3,4].
Dinnage et uf. have reported that the titanium silicate-impregnated ceramic sheet shows
better dehumidificationbehavior than the aluminum silicate-impregnated sheet[5].
In the present study, the silica ceramic sheets with Ti, A1 and Mg ions modified were
prepared and then their dehumidificationbehaviors were investigated. Finally the origin of
the superior dehumidification efficiency of Ti or Al-modified silica desiccants will be
discussed.

381
2 Methods

Ceramic sheets were prepared by a modified paper-making process with using AI2O3
and Si02 fibers. The detailed preparation conditions of the ceramic fiber sheet will be
reported, elsewhere.
Metal(Ti, Al, Mg) silicates were impregnated into the ceramic sheets by as-following
two-step process. The first step was the process to prepare silica-embedded ceramic sheets
and the second step was to incorporate Ti, Al and Mg ions into the silica-impregnated
ceramic sheets.
At first, the ceramic sheet with thickness of 0.2mm was tailored to the rectangular of
which width and length was 50 and 60mm, respectively. Then, the rectangular ceramic
sheets were dipped into a sodium silicate aqueous solution at 25C for Smin. The sodium
silicate aqueous solution was prepared by mixing 80wt?h of sodium silicate glass (No.3,
Shinheungjpsan, Korea) and 20wt?h of water. The dipped ceramic sheet was dried at
80C for 5min. The ceramic sheet was dipped and dried at the same conditions, one more
time. The dried ceramic sheet reacted with a sulfuric acid aqueous solution for 6hrs to
form silica gels on the ceramic sheet. In the sulfuric acid aqueous solution, the content of
the sulfiuic acid (Aldrich Chemical Co., USA) was 5wt%. After the silica gel formation
process, the silica-coated ceramic sheets were aged in the sulfuric acid aqueous solution at
pH 3 for 17hrs, and then washed in pure water at 25C for 5min to remove some
impurities such as sodium and sulfate ions from the silica-impregnated ceramic sheet.
The silica-impregnated ceramic sheets were dipped in three kinds of metal sulfate
aqueous solutions at 45OC for 3hrs. In the metal sulfate aqueous solutions, 1Owto/o of
Ti(S04)2 (Junsei Chemical Co., Japan), A12(S04),(Showa Chemical Co., Japan) and
MgS04 (Junsei Chemical Co., Japan) were dissolved, respectively. The dipped
silica-impregnated ceramic sheets were washed in pure water at 45C for lhr to remove
sulfate ions. After the washing process, the metal silicate-impregnated ceramic sheets
were dried at 200C for lhr and then used for the evaluation process of dehumidification
behavior.
Microstructure and pore structure of the prepared ceramic sheets were characterized
by SEM(XL30, Phillips, Holland) and BET(ASAP2400, Micrometrics, USA) analysis.
The dehumidification efficiency was evaluated by measuring the weight increment of the
adsorbent-impregnated ceramic sheets before and after the exposure to the 50% of relative
humid condition at 35OC for 3hrs. Also, the H20 adsorption isotherms were obtained by a
lab.-made mass equilibrium adsorption system on which the suspension balance
(Rubertherm magnetic suspension balance) is attached on.

3 Results and Discussion

Figures l(a) to (c) represent SEM images of (a) bare, (b) silica-impregnated and (c)
titanium silicate-impregnated ceramic sheets. In the bare ceramic sheet, there is large open
volume between ceramic fibers and the adhesion between ceramic fibers was induced by
the silica sol, an inorganic binder. On the other hand, in the silica or titanium
silicate-impregnated ceramic sheets, there was no free volume between ceramic fibers. In
the two cases, the open volume was fulfilled by the impregnated silica or titanium silicate
adsorbents. It was interesting that many microcracks were shown in the titanium silicate

382
phase located between ceramic fibers, compared with the silica-impregnated ceramic
sheet.

Fig. 1 SEM images of (a) bare, (b) silica-impregnated and (c) titanium
silicate-impregnated ceramic sheets.

Table 1 represents the weight variation in each experimental step and

dehumidification efficiencies of the adsorbent-impregnated ceramic sheets. The
interesting thing was that the dehumidificationefficiency increased in a sequence of silica,
magnesium silicate, alumhum silicate and titanium silicate-impregnated ceramic sheets. It
is evident that the incorporation of aluminum and titanium ions improved the
dehumidification efficiency of silica gel-based adsorbent, but the incorporation of Mg ions
doesn't affect the dehumidification efficiency of silica gel-based adsorbent.

Table 1 Weight variation in each experimental step and dehumidification efficiency of ceramic sheets
Specimen
w1 wz ws Dehumidification
(9) (9) (9) efficiencY (%1
Silica 0.4140 0.8043 0.9308 32.41
~

Magnesium Silicate 0.4262 ~ 0.7601 0.8687 32.52

Aluminum Silicate 0.4140 0.7903 0.9222 35.05
Titanium Silicate 0.41 50 0.8036 0.9528 38.39
W, : weight of ceramic sheet Mom impregnation. Wz : weight of ceramic sheet after impmgnafhm. WS : weight of
adrorb.nt-impregMted ceramic shed after dehumidification teat at 35C and 50% of relative humidity, Dehumidification
efficiency (K) = IWs-W,) I (Wz-W,) x 100

Figure 2 represents the H 2 0 adsorption isotherms of the adsorbent-impregnated

ceramic sheets. The aluminum silicate and titanium silicate-impregnated ceramic sheets
showed better adsorption behavior than the pure silica-impregnatedceramic sheets, as like
to the results for the dehumidification efficiencies obtained by the simple weight
increment test.
The same results were reported by few researchers. Kuma et ul. have reported that the
M-rotor treated with aluminum sulfate showed better performance for dehumidification
than the S-rotor treated sulfiuic acid. Dinnage et uf. have reported that the titanium
silicate-impregnated ceramic sheet shows better dehumidification behavior than the
aluminum silicate-impregnated sheet. They didn't suggest why the titanium silicate and
aluminum silicate show better dehumidificationbehavior than the silica gel desiccant.
If so, from where is the phenomenon that the dehumidification efficiency increases by
the incorporation of Ti and A1 ions in silica-based desiccants?

383
 Fig. 2. HzO adsorption isotherms of silica, aluminum silicate and titanium
silicate-impregnated ceramic sheets.

Figure 3 represents the N2adsorption-desorption isotherms of the silica, magnesium

silicate, aluminum silicate, titanium silicate-impregnated ceramic sheets. In all the
adsorbent-embedded ceramic sheets, there were the well-developed micropores which
gave the isotherms to the hysteresis phenomena at the relative pressures from 0.4 to 0.8.
Also, it was shown that the incorporation of Ti and A1 ions have the considerable volume
of adsorbed N2,compared with the pure silica and magnesium silicate-impregnated
ceramic sheets.
Figure 4 represents the incremental pore volume curves as a fimction of average pore
diameter in the adsorbent-impregnated ceramic sheets. In all the adsorbent-impregnated
ceramic sheets, there were two kinds of pores: one was the micropore with diameter less
than 3nm and the other was the diffuse mesopore with diameter of 3 to 10nm.In cases of
the pure silica and magnesium silicate-impregnated ceramic sheets, the average diameter
of the mesopores was 4nm. On the other hand, the aluminum and titanium
silicate-impregnated ceramic sheets have the mesopores with average diameter of 5nm and
with more wide size distribution. Relatively, the micropores were well developed in the
aluminum and titanium silicate-impregnated ceramic sheets, compared with the pure silica
and magnesium silicate-impregnatedceramic sheets.
It is interesting to indicate that the dehumidification efficiency increased in the same
order of the development of the micropore and mesopores. Therefore, it was concluded
that the superior dehumidification behavior of the aluminum silicate and titanium
silicate-impregnated ceramic sheets should mainly originate from the well developed
micropores and mesopores, especially the micropores with diameter less than 3nm.

384
 P-
88 IW

3 100

-B
0
> 5 0

0
0.0 0.2 0.4 0.8 0.8 1.0

Relative Pressure

Fig. 3. N2 dsorption-desorption isotherms of silica, magnesium silicate,

aluminum silicate and titanium silicate-impregnated ceramic sheets.

.....
10 100 lo00

Average Pore Diameter (Anstroms)

Fig. 4. Pore structures of silica, magnesium silicate, aluminum silicate and

titanium silicate-impregnatedceramic sheets.

4 Conclusion

The ceramic sheets with aluminum silicate or titanium silicate adsorbents showed
better dehumidification behavior than the silica and magnesium silicate-impregnated
ceramic sheets. The superior dehumidification efficiency originates om the well
developed micropores.

385
5 Acknowledgements

This work was fmancially supported by National Research Laboratory program (Korean
Ministry of Science and Tschnology).

References

1. Kuma, Shirahama, Izumi, US. Patent 5,753,345 (1998).

2. Kuma, Toshima, Shirahama, Noriaki, Izumi, Horaki, US. Patent RE37,779 (2002)
3. Kuma, Tosimi, Okano, Hiroshi, US.Patent 4,911,775 (1990)
4. Kuma and Hirose, J. Chem. Eng. Japan, Vol. 29, No. 2,376 (1996).
5. Dinnage, Paul A., Tremblay, Gerard, U.S.Patent 5,505,769 (1996)

386
SYNTHESIS OF ZIRCONIA COLLOIDS FROM AQUEOUS SALT SOLUTIONS
AND THEIR APPLICATIONS

KANGTAEK LEE
Department of Chemical Engineering, Yonsei University,Seoul, Korea
E-mail: [email protected]

ALON v. MCCORMICK' AND PETER w.c m 2

Department of Chemical Engineering and Materials Science' and Chemistd,
University of Minnesota, Minneapofis,M N 5.5455, USA
E-mail:;- [email protected]

We monitor the synthesis of submicron zirconia colloids from dissolved ZrQCI2.8HzO using
quasielastic light scattering. We investigate the effects of both the precursor salt concentration and
of the pH on the final colloid size distribution. We find that the pH has the strongest effect on the
final colloid size. These colloids are used in the polymerization-induced colloid aggregation process
to produce micron-range particles, and the performance is compared with commercial colloids.

1 Introduction

Monodispersely-sized submircon zirconia colloids are useful starting material for

ceramics, catalysts, and chromatographic stationary phases. One such process
(polymerization-induced colloid aggregation or PICA process) requires entirely
reproducible aqueous zirconia sols with no surfactants [1,2]. In the PICA process
developed by Iler and McQueston [3], the concentrated (-20 wt. %) 100 nm zirconia
colloids are aggregated by urea-formaldehyde polymerization reaction to produce the
-
porous zirconia particles in the size range of 4 6 pm [ 1,2].
Two classes of precursor have been demonstrated to prepare such zirconia colloids:
salts [4-71 and alkoxides [&lo]. Given the constraints above, the hydrothermal synthesis
from zirconium salts [4,7] is especially attractive. The salt is easier to protect and handle
than alkoxide precursors, and the product colloid does not require the removal of organics
or dispersing agents before firther aqueous processing. Bleier and Cannon [7] showed
that one can easily make 80 nm monodisperse aqueous sols with no dispersing agents.
Unfortunately, the procedures reported to date do not show how one might easily
control independently the average size and the final concentration. For instance, the
process to make 80 nm colloids gives only ca. 2.5 wt.% sol [7], while the PICA process
requires higher concentrations [1,2]. Since dialysis is time consuming and evaporation
would require some care to avoid flocculation, it would be advantageous to produce more
concentrated sols with the hydrothermal salt method while maintaining the average size ca.
100 nm and maintaining monodispersity.
In this paper, we show that the average size is primarily controlled by pH, and we
show that important limitations are imposed by increasing solubility at high initial salt
concentration due to the lower resultant pH. We investigate the effect of salt concentration
and of added acid on the final average size, polydispersity, and yield. We use these
colloids to make PICA particles and make a comparison with the commercially available
colloids.

387
2 Methods

2. I Zirconia Cottoit&@nthesis
Crystalline zirconium oxy-chloride octahydrate (ZrOCI2-8H20)was used as received
from three sources - Aldrich, Acros, and Alfa. ZrOC12-8H20 was dissolved in distilled and
deionized water so that the final concentration was 0.1,0.2, 0.4, and 0.6 M.Dissolution at
room temperature required ca. three hours. Then, the solution was boiled (up to -105 "C
at these concentrations)under reflux to allow reaction.
Samples were taken at different reaction times and immersed in a water bath to
quench the reaction. The following tests were performed size measurement using QELS
(quasi-elastic light scattering, Coulter Model N4 SD)and yield measurement. For QELS,
the samples were diluted with distilled and deionized water and the run time was 400 s.
For the yield measurement, the sample was centrifuged at 12,000 rpm for at least 15 min,
then the mass of the dried solid was measured.
For some samples, a nitrogen sorptometer (Micromeritics) was used to measure the
surface area after drying.

2.2 PICA Reaction of the Synthesized Colloids

When 0.4 M ZrOCI2.8H20 was used, the weight fraction and the pH of the final
colloidal solution were -5 wt.% and -0.5, respectively. To compare the PICA
performance of the synthesized colloids with that of the Nyacol colloids, it was
concentrated by reverse-osmosis. For reverse osmosis, stirred cell from Amicon with an
ultrafiltration membrane (50,000 MW cut-off) was used and pressure up to 50 psi was
applied. After reverse osmosis, water was added to increase the pH. By repeating this as
many times as necessary, 20 wt.% colloids at pH 1.75 were prepared.
For a PICA experiment, urea (U) from Fisher Scientific company was added to a new
30 ml polystyrene reactor. Formaldehyde solution (F) from Mallinckrodt company in a
separate beaker was added to a reactor to start the reaction. ([U]+[F])/[Zr02] and [U]/[F]
ratios were kept at 1.5 and 0.75, respectively. Reaction was quenched with water when the
secondary particles started appearing (-15 min). Particles were washed with water three
times, then with isopropanol twice. The particles were, then, filtered and dried in a
vacuum oven. At this point, the weight of particles was measured. The particles were
burned at 350 OC for two hours to remove polymers on the surface, then sintered at 750 OC
for six hours and 900 "C for three hours. After burning and sintering, the weight of
particles was again measured to get yield and polymer content. The surface area and the
pore size distribution of these particles were measured using Micromeritics nitrogen
sorptometer. S-800 SEM (scanning electron microscopy) from Hitachi was used to
observe the final particles.

3 Results

3. I Zirconia Colloids Synthesis

We find that the synthesis as reported by Bleier and Cannon [7], but changing the salt
concentration, is capable of reproducibly making 100 - 250 nm (average diameter)
zirconia colloids with a narrow distribution (standard deviation always smaller than -50
nm). However, we note that there is a maximum in the final particle size at 0.2 M
ZrOC12.8H20 (for which the pH is 0.9). Figure 1 shows the effect of the initial salt
concentration on the final particle size. Note that there is a correlation between the salt
concentration and the pH of the solution; the pH falls with higher salt concentration
because of zuconia ion hydrolysis. If too much salt is supplied (> 0.4 M),the yield is
limited to L 80 %. The highest concentration we obtain (pH < 0.9) is ca. 5 wt. %.

300

250 - 0

200 - -

150 -
100

50
-
-
i
0' I I I
0.0 0.2 0.4 0.6 0.8
[ZrOCI;8H20] (M)

1.1 0.9 0.5 0.4

PH

Figure 1. Final particle size vs. initial [ZrOC12.8H20] (the axis at the bottom indicates the initial pH of the
solution).
3.2 PICA Reaction of the Synthesized Colloi&
In order to compare the PICA performance of the synthesized colloids with that of
commercial colloids from Nyacol, 5 wt.% colloids were concentrated to 20 wt.% and pH
was adjusted to 1.75 as described in the experimentalsection.
After urea and formaldehyde were added to our colloids, small particles appeared in
-5 minutes, and they grew until the secondary particles appeared in -15 minutes. This is
shilar to the PICA performance of the Nyacol colloids [1,2]. After sintering, the yield
was -7 % and the polymer content was -70 %. The surface area of these particles was -20
m2/g. Figure 2 shows the SEM pictures of final PICA particles made from our colloids;
clearly, they are very porous and show the irregular shapes.

4 Discussion

4. I Zirconia Colloid Synthesis

One spec%= goal was to see whether simply adjusting the salt concentration could
provide starting marerid for PICA process. Of our trials, a solution of 0.4 M
ZrOC12.8H20with HCI is optimal for the procftrction of monodisperse -100 nm
colloids because:
 t 0p.m
Figure 2. SEM micrograph of PICA particles made from the synthesized colloids.

1) It produces a narrow distribution.

2) The fmal concentration of colloids is high, at -5 wt. %.
3) The yield reaches 100% in less than a week.

Since we would ideally like to produce even more concentrated sols (ca. 20 wt. %).
we wish to better understand the effect of the initial salt concentration. It is helpful to
consider the solubility of zirconia in water. The solubility of zirconia is reported by Baes
and Mesmer [ 111. At low pH range the solubility increases sharply with acidity. For the
systems with no added acid, the limited yield - when the initial concentra$on of salt is
higher than 0.4 M (pH 50.4) - may be attributed to the very high solubility of zirconia at
low pH.

4.2 PICA Reaction of the Synthesized Colloids

The differences of the final PICA particles using the synthesized colloids and the
Nyacol colloids are summarized in Table 1. It should be noted that the reactant
concentrations and the burning procedure were optimized for the Nyacol colloids, but not
for our colloids. The different polymerization kinetics in the PICA reactions using our
colloids instead of Nyacol colloids causes the very high polymer content, which also leads
to lower yield, smaller particles, bigger pores, irregular shape, and a broader distribution.
We also ascribe big pores to the unoptimized burning procedure. Thus, in order to
reproduce PICA particles made from Nyacol colloids, it is essential to reoptimize the
reactant concentrations of PICA reactions using our colloids.
Table 1 Comparison of PICA particles using the synthesized and the Nyacol (V-66 batch) colloids

Synthesized colloids Nyacol colloids (V-66)

Yield (%) 7.0 15.0

Polymer content (%) 70.0 40.0

Pore size (angstrom) 1000.0 400.0

Final distribution broad narrow

Shape irregular spherical

Final particle size (prn) -

1.O 4.0 -
3.0 5.0

5 Acknowledgements

This study is supported by Korea Research Foundation Grant (KRF-2001-005-E00030).

References

1. Amen M.J., Kizhappali R., Cam, P.W. and McCormick A.V., J. Muter. Sci.29 (1994)
pp.6123-6130.
2. Sun L., Annen M.J., Lorenzano-Porras F., Can P.W. and McCormick A.V., J.
Colloid Intevace Sci. 163 (1994) pp. 464-473.
3. Iler R.K. and McQueston H.J., US Patent No. 4,010,242 (1977).
4. Blumenthal W.B., The Chemical Behavior of Zirconium (D.Van Nostrand Company,
Inc., New York, 1958) pp. 125-132.
5 . Matsui K. Suzuki H. and Ohgai M., J Am. Ceram. SOC.78 (1995) pp. 146-152.
6. Aiken B. Hsu W.P. and Matijevic E., J. Muter. Sci. 25 (1990) pp. 1886-1894.
7. Bleier A. and Cannon R.M., In Better Ceramics Through Chemisny 11, ed. by C.J.
Brinker, D.E. Clark and D.R. Ulrich (Materials Research Society, Pittsburgh, 1986)
pp. 71-78.
8. Fegley B., White P. and Bowen H.K., Am. Ceram. Bull. 64 (1985) pp. 1115-1120.
9. Bartlett, J.R.,Woolfiey J.L., Percy M., Spiccia L. and West B.O., J. Sol-Gel Sci.
Tech. 2 (1 994) pp. 2 15-220.
10. Lerot L., Legrand F. and De Bruycker P., J Muter. Sci. 26 (1991) pp. 2353-2358.
11. Baes C.F. and Mesmer R.E., The Hydrolysis of Cations (John Wiley & Sons, New
York, 1976) pp. 152-159.

391
COMPARISON OF NANO-SIZED AMPHIPHILIC POLYURETHANE (APU)
PARTICLES WITH SDS, AN ANIONIC SURFACTANT FOR THE SOIL
SORPTION AND THE EXTRACTION OF PHENANTHRENE FROM SOIL

IK-SUNG, AH" AND HEON-SIK, CHOI

Dept. of Chem. Eng., Yonsei University, Seoul, Korea
E-mail: [email protected]

JU-YOUNG, KIM
Dept. of materials engineering, Samchok National University,Samchok, Korea
E-mail: JUYOUNGK@amchok. ac.kr

Understanding of surfactant sorption onto soil is needed to assess surfactant mobility in soil and
surfactant-facilitated transport of hydrophobic pollutants in soiVaqueous systems. Micelle-like amphiphilic
nano-sized polyurethane (MU)particles synthesized from amphiphilic urethane acrylate anionomers could
solubilize a model hydrophobic pollutant, phenanthrene within their hydrophobic interiors. Batch
experiments were conducted with soil slurries to compare APU Sodium Dodecyl Sulfate (SDS), anionic
surfactant for the sorption onto soil. APU particles (K,@.2 mug) were weakly adsorbed onto the sandy
soil compared to SDS ( K d 1 . 3 mug), due to their chemically crosslinked structure. Compared with SDS,
APU particles exhibited the higher extraction efficiency to remove phenanthrene from the contaminated
sandy soil.

1 Introduction

Contamination of soil and groundwater by hydrophobic organic carbons (HOCs) is

caused by leakage fiom storage tanks, spillage, or improper disposal of wastes. Once in
the soil matrix, HOCs can act as a source of dissolved contaminants[1-31. Among HOCs,
PAHs are of special interest because they are strongly sorbed to soil or sediment, as a
consequence, sorbed PAHs may act as a long-term source of groundwater contamination.
So many researchers have been using various surfactants to enhance desorption of sorbed
PAHs from soil through solubilization of sorbed PAHs in surfactant micelles[4-8].
Surfactant-enhanced remediation techniques have shown significant potential in their
application to the removal of PAHs in the soil remediation process. Some of the
disadvantages of these techniques are micelle breakage and loss of surfactant through
sorption to soil. In addition, Surfactant-enhanced desorption and washng process is only
effective when surfactant dose is greater than its critical micelle concentration (CMC),
because most of surfactants molecules below CMC are sorbed onto soi1[6-8, 9-11]. So,
recent research has been directed toward the design of surfactant that minimizes their
losses and the development of surfactant
recovery and recycling technique.
The purpose of this research is to compare micelle-like polymeric particles with a
surfactant with respect to the sorption to soil and the extraction of an organic soil
contaminant. Nano-sized polyurethane (APU) particles synthesized from amphiphilic
urethane acrylate anionomers were used as model micelle-like polymeric particles.
Employing APU particles with various degrees of hydrophobicity, the effects of
hydrophobicity on the soil sorption and the phenandvene extraction from soil of
polymeric particles were studied. Sodium Dodecyl Sulfate (SDS)was used as a model
conventional surfactant. Phenanthrene was used as a model soil contaminant.

392
 2 Materials and Methods

2. I Materials
The soil used in all experiments was obtained from coal-mine region in Samchok,
Korea. Soil sample was air-dried and screened through a US standard No. 10 mesh (2mm)
sieve to remove coarse firagments. The fraction of organic carbon in the soil sample was
determined to be 0.14% from TOC analysis using MULTI N/C-300 Total Carbon
Analyzer (Analytic Jeni. AG., Germany). Phenanthrene. SDS and radio-labeled
phenanthrene (9-14C, 13. IrICV-hnol) were purchased from Sigma Chemical Co.(St. Louis,
MO USA). APU particles were synthesized by polymerization of urethane acrylate
anionomers (UAA) as described by Kim et al. [12]. Table 1 shows the molar ratios of
reagents used in the synthesis of the UAA precursors. The hydrophilicity and the
hydrophobicity of the synthesized polymer can be modified by changing these molar
ratios. APU made from UAA 2:8 has the highest hydrophilicity, while that from UAA 6:4
has the highest hydrophobicity.

Table 1. The molar ratio of reagents in the synthesis of various APU particles
Types of APU Reagents Molar ratio
UAA 218 PTMG/DMPA/TDI/2-HEMA 0.2/0.8/1.5/1.5
UAA5:5 PTMGIDMPAITDI/2-HEMA 0.5/0.5/1.5/1.5
mn-m-p --
UAA 6:4 PTMGIDMPNTDIR-HEMA
----*
0.6/0.4/1.5/1.5
-*,?,**vc-<y

2.2 Methods

2.2.1 Adsorption of SDS or APU particles onto soil

A batch equilibrium method was employed to measure the sorption isotherm of SDS or
APU particles in the soil-water system. Samples were prepared in duplicate by mixing
1.OOg of soil and SDS or APU solutions (IOmL) of various concentrations in glass vials
with an open-port screw cap, which was fitted with a Teflon-lined septum. The vials were
shaken in a mechanical rotary tumbler for 72 h at room temperature. After equilibrium
partitioning was attained, 5 mL of the samples were centrifiged at 15000g and
concentrations of SDS or APU in the supernatants were estimated as total organic carbon
fractions using Pharma Total Organic Carbon Analyzer.

2.2.2 Solubilization of phenanthrene in SDS or APU solutions

Solubility of phenanthrene in SDS or APU solutions was determined using mixture of
radiolabeled phenanthrene and nonlabeled phenanthrene. A concentrated phenanthrene
solution (lOOg/L) was prepared in methylene chloride. 2mL of the concentrated
phenanthrene solution was added into 20mL glass vials equipped with an open-port screw
cap, which was fitted with a Teflon-lined septum. After evaporation of methylene chloride,
SDS or APU aqueous solutions (IOmL) of various concentrations were added into the
vials. The vials were sealed and gently shaken in the rotary tumbler for 5 days. Afier
equilibrium, 5mL of the samples were withdrawn and centrifuged at 15OOOg to separate
solid-phase materials from aqueous-phase materials. 1mL of the supernatant was

393
transferred into scintillation vials containiig lOmL of Ecolume cocktail, and the
concentration of 14C-phenanthrenein the aqueous phase was measured using a Liquid
ScintillationCounter (LSC).

2.2.3 Extraction of phenanthrene sorbed in soil using APU or SDS

ImL of 14C-phenanthreneaqueous solution was added into glass vials containing lg of
soil and shaken in the rotary tumbler for 72 h at room temperature. After sorption
equilibrium was attained, 9mL of SDS or APU solutions of various concentrations was
added into vials, and mixed for 72 h. 5mL of the samples was then withdrawn from each
vial and centrifuged at 15000g. The concentration of 14C-phenanthrenein the supernatant
was measured by the same method in the above section 2.2.2.

3 Results and Discussion

In the sorption experiment of APU particles or SDS onto soil, APU particles exhibited
relatively low degree of sorption onto soil than SDS (Figure1.). It is believed to be due to
the highly cross-linked structure of APU. Sorption isotherm of phenanthrene in the test
soil was shown in Figure 2. As shown in Figure 3, phenanthrene showed the lower
solubility in APU solutions than in SDS solutions. Hence APU particles seem to have the
lower affinity with phenanthrene than SDS with phenanthrene. However, at low
concentrations ofAPU particles or SDS (below 2100mg/L,CMC of SDS), APU particles
showed the higher extraction efficiency of sorbed phenanthrene than SDS did (Figure 4).
At higher concentrations of APU particles or SDS,the extraction efficiency of APU was
almost same as that of SDS,even though APU particles exhibited the lower solubilization
of phenanthrene than SDS micelles in the absence of soil. Extraction of phenanthrene
from soil using interfacial agents seems to depend not on their affiity with phenanthrene
but on the degree of their sorption onto the soil.

2.0 -
-
1.8 ..... ..o...
.. U A A 2:8
t--UAAS:S
1.6 -

3--
1.4 -

1.2 -

4. 1.0 - -.. I
Log (CMCof SDS)
X
0.8 -

0.6 -

0.4 - ...
... .... .......
0.2 !
0 3 4
Log (SDS or A P U Concentration, m g l L )

Figure 1. Distribution of SDS or APU between soil and aqueous phase. (F& =
[SurfJJ([Surfl,*f,) = ([mg of surfactant sorbed in soil/g of soil])/([mg of
surfactant in aqueous solution/mL]*fractional organic carbon content of the
soil)

394
 w5 0.0 0.1 0.2 0.3 0.4 0.5
Equilibrium concentration of phenanthrene in water (mglL)
Figure 2. Sorption Equilibrium of phenanthrene onto soil. (Sfie,, = Equilibrium
concenbation of phenanthrene in soil, C&+, = Equilibrium concentration of
phenanthrene in water)

........ 0 .......

0 3 4 i
Log (SDS or APU particles Concentration, m g L )

Figure 3. Enhanced Solubility of phenanthrene in the aqueous phase containing

SDS or APU particles. (C = Solubility of phenanthrene in the presence of SDS
or APU particles, C,,= Solubility of phenanthrene in the absence of SDS or
APU particles)

395
 0.5 , I

CI

0.4

ui
1
zY 0.3
F
5
(II

B
a 0.2
P,
c

-#
e
.s
0.1
* I
4
Log (CMCof SDS)
I
I
0.0
0 3 4
Log (SDSor APU Concentration, m g h )
Figure 4. Extraction ofphenanthrene sorbed in soil using SDS or APU. (Sphe=
Phenanthrene remaining in soil, Cphe= Phenanthrene remaining in the aqueous
phase containing SDS or APU particles)

Aknowledgements

This study was supported by Yonsei University Research Fund.

References

1. J. H. Harwell, Transport and remediation of subsurface contaminants. D. A. Sabatini,

D. A. and know, R. C. Eds., ACS Symposium Series 491, Am Chemical SOC.,
Washington, D. C. (1992) pp. 124-132.
2. D. M. Mackay, Cherry, J. A, Groundwater contamination: Pump-and
-treat-remediation, Environ. Sci. Technol.,23(6), (1 989) pp. 630-636.
3. J. F. McCarthy, J. M. Zachara, Subsurface transport of contaminants, Environ. Sci.
Technol. 23(5), (1989) pp. 496-502.
4. 1. T. Yeom, M. M. Ghosh, C. D. Cox, Kinetic aspects of surfactants soiubilization of
soil-bound PAHs. Environ. Sci. Technol.30 (1996) pp. 1589-1595.
5. I. T. Yeom, M. M. Ghosh, C. D. Cox and K. G. Robinson, Micellar solubilization of
PAMs in coal tar-contaminated soils, Environ. Sci. Technol. 29 (1995) pp.
3015-3021.
6. D. A. Edwards, Z. Adeel, and R. G. Luthy, Distribution of nonionic surfactants and
phenanthrene in a sedimentlaqueous system, Environ. Sci. Technol. 28 (1994) pp.
1550-1560.
7. C. T. Jafiert, Sediment- and saturated-soil-associatedreactions involving an anionic
surfactant (Dodecylsulfate). 2. Partition of PAH compounds among phases, Environ.

396
 Sci. Technol. 25 (1991) pp. 1039-1045.
8. W. Chu, and W. S. So, Modeling the two stage of surfactant-aided soil washing, Wut.
Res, 35(3), (2001) pp. 76 1-767.
9. K. D. Pennell, L. M. Abriolar, W. J. Weber, Jr, surfactant-enhanced solubilization of
residual dodecane in soil columns. 1 Experimental investigation, Environ. Sci.
Technol. 27 (1993) pp. 2332.
10. C. C. West, J. H. Harwell, Surfactants and subsurface remediation. Environ. Sci.
Technol. 26(2) (1992) pp. 2324-2329.
11. D. A. Edwards, R. G. Luthy, and Z. Liu, Solubilization of polycyclic aromatic
hydrocarbons in micellar nonionic surfactant solutions. Environ. Sci. Technol. 25
(1991) pp. 127-133.
12. J.Y.Kim,C. Cohen. M. L. Shuler, and L. W. Lion, Use of amphiphilic polymer
particles for in-situ extraction of sorbed phenanthrene from a contaminated aquifer
materials. Environ. Sci. Technol.34 (2000) pp. 4 133-4139.

397
 SYNTHESIS OF MESOPOROUS ACTIVATED CARBON WITH IRON ION-
AIDED ACTIVATION

YOSHIMI SEIDA
Institute of Research and Innovation, Takuda 1201, Kashiwa, 277-8501 Japan
E-mail: [email protected]

KAORI WATANABE AND YOSHIO NAKANO

Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technologv,
Nagatsuta 4259, Midori-ku. Yokohama226-8502, Japan, E-mail: [email protected]

Synthesis of mesoporous activated carbon was performed using the ion-exchange resin as a precursor
and iron ion as an activator through heat-treatment in the series of gases. The dependence of the
amount of iron in the resin and the composition of the gas during the heat-treatment on the pore
structure of the produced activated ciirbon was elucidated. The mesoporous activated carbons with
pore size of more than 2nm were obtained in the presence of iron in the resin. The activated carbon
with larger mesopore volume was obtained by the heat-treatment in 100% hydrogen. The effects of
the iron on the synthesis of the mesoporous activated carbon were examined as a function of gas
during the synthesis, condition of the heat-treatment, the amount of introduced iron and the particle
size of the iron compounds produced in the activated carbon to clarify the mechanism of the
mesopore production in the activated carbon with iron.

1 Introduction

Porous carbon materials have been considered as an effective material for many
engineering use such as catalytic support, battery electrode, capacitor and gas storage
material. For such applications, the carbon materials should have high mesopore content
so that ions and large molecules can penetrate and adsorb efficiently in pores. Mesopores
act as the main transport arteries for the adsorbates and contribute significantly to their
adsorption. Reports that study the synthesis of the mesoporous carbons have been
increasing recently [4,6,7,9,10]. Synthesis through activation of precursors with metal
ions is one of the ways to obtain the mesoporous carbons [2,8]. However, there are not
many reports that studied rigorously the effect of the metal ions andor metals on the
mesopore formation. In the present study, synthesis of mesoporous activated carbons was
performed through a heat-treatment of an ion-exchange resin with Fe ion in a series of
gases. The synthesized activated carbons were characterized in terms of specific surface
area, pore size distribution, produced phase and crystallite size of the produced Fe
compounds in the activated carbons. Effects of the condition of heat-treatment
(carbonization) and the iron in the resin on the mesopore formation of the activated
carbons were examined to elucidate a role of the introduced Fe on the mesopore
production.

2 Sample preparation and characterization

An ion-exchange resin was used as a precursor of the mesoporous activated carbon to

prepare a starting material in which some amount of Fe ions were introduced with high
dispersion. Commercially supplied acrylic type cation exchange resin with carboxyl

398
group (IRC-76, Organ0 Co. Ltd., spherical Table 1 Conditions of the synthesis
particle diameter = 400pn) was used in the
present study. Fe ions were introduced into -treatm
the resin through ion exchange from solution {eating
of iron chloride. The amount of introduced Fe No. rate
ions in the resin is shown by their mole Wmin:
hction against the ion-exchange capacity of
the resin (Q/QEc). The dried resins with Fe
were heat-treated at a temperature in an
electric furnace for 3h in a series of gases to
obtain activated carbons. The obtained
activated carbons were then crushed by agate
motor to control their particle size below 5
250pm in diameter. The conditions of the
synthesis are listed in Table 1. Thermal
behavior of the dried resin with Fe ions
(Q/QEC=O.l) was measured by thermo-
gravimetry with simultaneous deferential
thermal analysis (TGIDTA) in hydrogen
atmosphere. Adsorption and desorption
isotherms of the Synthesized activated carbons for nitrogen at 77K were obtained using
Autosorb (Quantachrome Co.). Specific surface area, pore volume and pore size
distribution were calculated based on the isotherms by BET and Barrett-Joyner-Halenda
(BJH) method. Produced phase of Fe compounds in the synthesized activated carbons and
graphitization of the samples were identified using powder X-ray diffractometer.
Crystallinity of both the produced Fe compounds and the graphite were evaluated by full
width at half maximum; FWHM of their diffraction peaks. The smaller FWHM means the
larger crystallite size. Crushed particles of the synthesized activated carbons were
observed by means of scanning electron microscopy with energy dispersive X-ray
microanalysis ( S E W D X , S-3500N7 Hitachi, Co. Ltd.). The resolution of the EDX was
1m.

3 Results

From TG/DTA analysis of the dried resin with Feythe resin was found to show a major
weight loss with an exothermic thermal decomposition at 473-873 K followed by a slight
endothermic weight decrease at above 873K. About 2 0 W ? of the carbonized resin
remained at 1273K.

(a)IW%N, (b)25%H, (c)IW%H,

Fig.1 SEM images of the synthesized activated carbons in a series of gases
 Figure 1 shows the SEM photographs of the activated carbons synthesized in the
series of gases at 973 K (Sample
- - No.3, 7 and 13). The resins lost their initial snherkal
shape after the heat-treatment and showed expanded single carbon b lo ck The
carbonization of the resin occurred through the
liquefaction at 473-87313. Macroporous
structure was observed in the crushed samples
I a-Fe
I

synthesized in 100% H2 (Fig.1c). On the

contrary, non-porous structure with glassy
surface was observed in the crushed samples
synthesized in 100% N2 and 25% H2 (Fig.la,
b). Any iron-based particles were not observed
by the SEMEDX analyses of the surfaces
shown in Fig.1 under their resolution except
for the samples with Q/Q~c=0.41 and 0.53
(No.5,9 and 16). 0 10 20 30 40 50 60 70
Figure 2 shows the XRD diffraction 2 0
fig. 2 XRD pamm of me semples
patterns of the samples No. 3, 7 and 13.
Formation of graphite-like carbon, which was
identified from the diffiction peak at around with Fe
---- lWhH, -
26.2 degree [4], was observed in all the
samples synthesized with Fe. Dominant Fe
compounds produced in the activated carbons
were identified from the diffiction patterns as
FeJC for the samples synthesized in 100% N2
and a-Fe for the ones synthesized in 100% H2,
respectively. The a-Fe produced in the
activated carbons was hardly removed by a
repeating wash using 1N HN03 and N&Cl 1 10 100
Pore diameter [nm]
under ultrasound. The a-Fe was stable against
oxidation for more than one year in the air. Fig3 Pore size dishibution, Dv, of
the synthesized activated carbons
These results indicate that the a-Fe Droduced
in the activated carbons may =be 0.8
immobilized in a carbon matrix.
The pore size distributions, D,, of
the synthesized samples (Sample No.1, o
3 , 7 and 13) are shown in Figure 3. The
samples synthesized with Fe contain
mesopores with 2.0-30 nm in diameter. $ 0.4
The pores with more than 4 nm in 2
diameter increased largely by the g
synthesis in 100% H2. The sample a 0.2
synthesized without Fe did not produce
mesopores. Figure 4 shows the effect of
the introduced amount of Fe and gas Q,Q,

during the heat treatment on the pore 0.10 0.53 0.10 0.41 0.10 0.41
and above nm in Fig.4 Effect of introduced amount of iron on the pore
diameter. The total pore volume volume of the activated carbons synthesized at 973K
increased with the increase of Fe in the in a series of gases
samples synthesized in 100% Nz and
25% Hz. On the contrary, the pore
volume decreased oppositely with the
increase of Fe in the samples
. 0.5
p.
-
,"
synthesized in 100% H2. The increase of . 0.4 d
the pore volume is attributed to the rs
increase of mesopore with more than 4 t .0.3 w
nm as shown in Fig.4. The mesopore can
be developed effectively by the 0 . 0.2 $
synthesis in 100% H2atmosphere with
an appropriate amount of Fe. Figure 5
shows effect of the treatment -
temperature on the specific surface area, 773 873 973 1073
pore volume and the FWHM of the 110 Temperature [K]
reflection of a-Fe for the samples
synthesized with Fe in 100% H2
(Sample No.11, 12, 13 and 14). The a-
Fe was produced by the heat treatment
at above 873K. The specific surface
area and the pore volume increased
largely along with the production of a- 9
-
0.8

Fe. The increase of the pore volume was g 0.6

mostly due to the production of
mesopores. The crystallite size of a-Fe p 0.4
-P
in the samples was calculated to be
15-30 nm from the 110 reflection based a 0.2
on the Scherrer's equation,
The crystallinity of the graphite 0 ,
was low in the samples with high 0 0.2 0.4 0.6 0.8 1
crystallinity of Fe compounds (a-Fe and FWHM of Fe compounds
Fe3C), which indicates that the
graphitization of the activated carbon Fig6 Relationshipbetween the FWHM of the
Fe compounds and the pore volume. FWHM of
may be reduced in the presence of Fe the 110 reflection for mFe and that of the 112
and with the particle growth of the Fe reflection for Fe$
compounds (data are not shown).
Figure 6 shows the relationship between the FWHM of the Fe compounds and the total
pore volume for the series of the synthesized activated carbons. The pore volume showed
a maximum at a particle size of the Fe compounds. The results indicate that there exists
an optimum size of the Fe compounds for the largest mesopore volume.

4 Discussion

The introduced Fe contributes to increase the specific surface area and mesopore of the
activated carbon as shown in Fig.3. The Fe ions migrate in the resin to form nanoparticles
of Fe compounds during the heat-treatment. Along with the formation of the particles of
Fe compounds, the mesopore of the activated carbon may also be produced at the sites
where the Fe compounds migrated. In the case of a-Fe, the a-Fe particles may act as a
catalyst for gasification of the carbons around the particles in Hz atmosphere at above

401
873K [1,3], which would result in the drastic increase of the mesopore in the activated
carbons synthesized in 100% H2. Based on these mechanisms of the mesopore
production, control of the particle size of the Fe compound should be important to
produce mesopores. The size of the particles can be controllable through adjustment of
the introduced amount of Fe in the precursor, dispersion of Fe at the preparation stage of
the precursor and the condition of heat-treatment such as heating rate and treatment
temperature. The conditions should be optimized depending on precursors. We also
confirmed the effectiveness of the present method in selective increase of mesopores of
activated carbons using used coffee beans and tea leaves wastes. The results will be
presented in the next paper.

5 Conclusion

The mesoporous activated carbons with large mesopore volume can be obtained from the
resin exchanged with Fe ion through the heat-treatment in hydrogen atmosphere. The key
factors to obtain the highly mesoporous activated carbon are (1) the introduction of Fe
with high dispersion and (2) the size control of Fe compounds produced during the heat-
treatment through adjustment of the introduced amount of Fe and the condition of heat-
treatment.

References

1. Hermann, G. and Huttinger, K. J., Mechanism of iron-catalyzed water vapor

gasification of carbon, Carbon 4 (1986) pp.429-435.
2. Hu, Z., and Srinivasan, M. P., Mesoporous high-surface-area activated carbon,
Microporous and Mesoporous Materials 43 (200 1) pp.267-275.
3. Mckee, D. W., Effect of metallic impurities on the gasification of graphite in water
vapor and hydrogen, Carbon 12 (1974) pp.453-464.
4. Nakagawa, H., Watanabe, K., Harada, Y. and Miura, K., Control of micropore
formation in the carbonized ion exchange resin by utilizing pillar effect, Carbon 37
(1999) pp. 1455-1461.
5 . Ozaki, J., Mitsui, M. and Nishiyama, Y., Carbonization of ferrocene containing
polymers and their electrochemical properties, Carbon 36(1-2) (1998) pp.131-135.
6. Ryoo, R., Joo, S. H. and Jun, S., Synthesis of highly ordered carbon molecular sieves
via template-mediated structural transformation, J. Phys. Chem. 103(37) (1999)
pp. 7743-7746.
7. Seida, Y. and Nakano, Y., Removal property of activated carbon with mesopore and
dispersed iron nanoparticle for humic substances, Proc. 2nd International Water
Association World Conference,P2 126, 1-6(200 l), CD-ROM version.
8. Tamai, H., Kojima, S., Ikeuchi, M., Mondori, J., Kaneda, T. and Yasuda, H.,
Preparation of mesoporous activated carbon fibers and their adsorption properties,
Carbon (Japanese) 175 (1996) pp.243-248.
9. Yoshizawa, N., Yamada, Y., F w t a , Y., Shiraishi, M., Kojima, S., Tamai, H. and
Yasuda, H.,Coal-based activated carbons prepared with organometallics and their
mesoporous structure, Energy & Fuels 11 (1997) pp.327-330.
10. Zhichang, L., Licheng, L., Wenming Q. and Lang, L., Preparation of pitch-based
spherical activated carbon with developed mesopore by the aid of ferrocene, Carbon
37 (1999) pp.663-667.

402
 SEPARATION OF PEPTIDES FROM HUMAN BLOOD BY RP-HPLC

SUENG K1 LEE, YULIA POLYAKOVA AND KYUNG HO ROW

Centerfor Advanced Bioseparation Technology and Dept. of Chem. Eng., Inha University,
253 Yonghyun-Dong,Nam-Ku, Incheon 402- 751, Korea
E-mail: [email protected]. kr

The biologically active peptides in human blood can adjust the functions of many physiological
systems. The peptides in human blood were separated on the five steps of linear gradient-elution
mode by RP-HPLC with UV detection. The size of commercially available Cischromatographic
column was 4.60X150mm with particle size of 5pa and pore size of IOOA. The mobile phases used
were water in 0.75 % trifluoroacetic acids (TFA) and organic modifier of acetonitrile. The isolation
methods suggested in this work for peptides from the blood were composed of the formation of
immiscible liquid layers and precipitation by centrihge and chemicals of sodium citrate and
trichloroacetic acid (TCA). The some peptides were identified based on the retention times by
previously constructed database. It was experimentally observed that a peptide was degraded and a
new peptide was formed during the storage of Idays.

1 Introduction

The processes of separation of biological-active substances are very important task in

modem chemistry, biology, biochemistry and medicine. Now there is information about
the peptides, which has wide spectrum of biological activity. The proteins and the
peptides prevail quantitatively above large molecules in organism of the man. They have
various functions: catalysis of the ferments, transportation, protection and transfer of the
information, regulation. The information that the human hemoglobin can by a source of
the biological peptides has appeared recently I]. Researches on these peptides become
interesting and urgent. Thus, a study on the separation of the peptides by HPLC will
surely lead to the basis for the various types of research.
Reversed-phase high-performance liquid chromatography (RP-HPLC) is the most
widely used analytical technique for separation and isolation of the peptides [2,3,4].
Peptide separation in RP-HPLC mainly depends on its amino acid composition, but also
on peptide chain length and sequence, as well as on the corresponding stationary phase
and chromatographic condition for the elution. So the mobile phase is a mixture of an
aqueous and organic solvent in which the hydrophobic interaction between the peptides
and the non-polar stationary phase allows to do separation of the peptides. Peptide elution
from the column occurs by increasing the percentage of the organic modifier [ 5 ] . The
mobile phase is essentially fixed since the most separation was conducted with the
shallow gradients using the aqueous buffers modified with the acetonitrile [6,7].
Acetonitrile is generally employed owing to its UV transparency, low viscosity and
volatility. The mobile phase for the peptide chromatography contains an ionic modifier
which acts as counter-ion of any charged amino acids. This ionic modifier increases the
peptide hydrophobicity by forming ion-pair complexes. Trifluoroacetic acid (TFA) has
been commonly used as an ion pair-forming reagent for the peptide isolation. TFA is an
excellent polypeptide solvent and a strong ion-pairing reagent, which is transparent to
ultraviolet light and easily eliminated by lyophilization. The purpose of this work was to
find the separation condition of the low molecular peptides in human blood by RP-HPLC
and their identification. Also the brief degradation mechanism was discussed.

403
2 Experimental

2.1. Reagents
The chemicals used in this study were sodium chloride, magnesium chloride, calcium
chloride, potassium chloride, trichloroacetic acid (TCA), trifluoroacetic acid, a-d glucose,
sodium citrate dihydrate, sodium hydrogen phosphate, hydrochloric acid, and iron (11)
sulfate heptahydrate. These chemicals were purchased from Sigma (St. Louis, MO, USA).
The HPLC grade solvent, acetonitrile was 6om J. T. Baker (PhilipsburgNJ, USA). Water
was distilled and deionized prior to use.

2.2 Apparatus and method

HPLC was performed using Waters 600s solvent delivery system (Waters, Milford,
MA, U.S.A.). 2487 UV dual channel detector of Waters was used and injector (20 &?
sample loop) from Rheodyne. The data acquisition system was Millenium3*(Waters).
Water filtered by Milipore ultra-pure water system (Milipore, Bedford, MA, USA). The
wavelength was fixed at 254 nm and the experiment was performed at room temperature.
The size of the analytical column packed by CI8was 150x4.6mm (5m)(Alltech, USA).
The mobile phase of 0.75% TFA in water and acetonitrile were used in this experiment.
The flow rates of the mobile phase were fixed at 1 ml/min. The constant volume of lop!,
was injected. This experiment was implemented at room temperature. The gradient mode
was employed to isolate peptides. The complete gradient condition was listed in Table 1.
Table 1. Mobile Phase Compositions in Fig. 2 and Fig. 3

Gradient time 0.75% TFA in water

Acetonitrile vol. 'YO
(min) vol. %
0 100 0
4 90 10
7 80 20
10 65 35
20 50 50
30 0 100

2.3. Sample preparation

The venous peripheral blood of the healthy donor by starting material was used. The
sample was prepared in High-Purity Separation Lab., Inha University. The peptides
extracts were obtained in the following way [8,9]. Take a fresh sample of the human
blood, 9 ml, and it was dissolved in 1 ml of the sodium citrate solution. Centrifuge the
blood (10 min, 2280 rmp). Throw away the cell element of the plasma in upper side. Add
0.154 M sodium chloride solution into the part in the lower side with pH 7.4. Repeat 3-5
times with the last procedure. Collect the fiaction of the erythrocytes in bottom side as
shown in (Fig.1 a). Add 15% trichloroacetic acids (TCA) solution to fraction of the
erythrocytes. Centrifuging of the last sample 10 min, 3 120 rmp). Analyze a liquid above a
deposit (Fig. I b).

404
 (a) (b)
Figure 1. Photograph showing the sample preparation process. (a) leukocytes in the upper
side and erythrocytes in the lower side (b) peptides in the upper side and
erythrocytesand proteins in the lower side.

3 Results and Discussion

The peptide extracts by TCA from the fresh human blood and the stored human bloods
during 3, 5 , and 7 days were experimented. The specific character of the peptides
formation is a change in quantity of the peptides in the extract. The accumulation of the
peptides in the extracts probably explained by the fermentative peptides proteolysis,
which is the decomposition process of the proteins and peptides catalyzed by the special
ferments.
The identification of the some peptides was carried out in this work. Some standard
substances of the peptides for the procedure of the identification were used. Other
peptides were identified based on the retention times by previously constructed database.
The qualitative determination of the peptides in the extract was made on the retention
times of the standard peptides substances. The chromatogram of the peptides by fresh
sample is shown in Fig. 2, while that in 7 days is shown in Fig. 3. The five peptides in the
extracts were identified, and for coding structure of the peptides we used one letter for a
code.
 I . , . , . , . , .
0 4 12 16 I
Time (min)

Figure 2. Separation of peptides from human blood by RP-HPLC (fresh sample, injection
volume lo,&)

0.08

- 0.04

0.00

12 18 20

Time (rnin)

Figure 3. Separation of peptides from human blood by RP-HPLC (7days, injection

volume l0,d)

Quantity of chromatographic peaks and their quantitative characteristics during their

preservation remarkably changed. During its storage, the area level of the peak #1 (V-H-
L-T-R-E-E-K-S-A-V) was increased. Initially, the peak #2 was occurred during the
storage after 3 days. At the same time, the area of the peak #3 (V-H-L-T-P-G-E-K-S-A-
V) and the peak #4 (V-A-G-V-A-N-A-L-H-R-R-Y-H) were decreased. But the area of the
peak #5(T-L-S-E-L) was not changed during the storage of the human blood. The number
of the peaks on the chromatogram increased, because the new peptides occurred during
the storage of the human blood. Such phenomenon may be explained by the cells of
erythrocytes which can use the oxygen. The deficiency of oxygen in a preservation of the

406
human blood caused degradation. These conditions had influence on the formation of the
peptides. It is expected that more concrete understanding of peptide-separation will
provide very useful information to the hnction of peptides in human blood.

4 Acknowledgements

The authors gratefully acknowledged for the financial support by Center for Advanced
Bioseparation Technology in Inha University.

5 Reference

1 . A. A. Karelin, M. M. Filippova, 0. N. Yatskin, E. Yu. Blishchenko, I. V. Nazimov,

and V. T. Ivanov, Proteolytic Degradation of Hemoglobin in Erythrocytes Yields
Biologically Active Peptides, Peptide Science, 4 (1 998), 27 1-28 1.
2. V. Casal, P. J. Martin-Alvarez, T. Harraiz, Comparative prediction of the retention
behaviour of small peptides in several Reversed-phase high-performance liquid
chromatography columns by using partial least squares and multiple linear
regression, Analytica Chimica Acta, 326 ( I 996), 77-84.
3. Zukowski J., Pawlowska M., Nagatkina M. Armstrong D. W., High-
performance Liquid Chromatographic Enantioseparation of Glycyl dipeptides and
tripeptides on Native Cyclodextrin Phases, Mechanistic Considerations, Journal of
Chromatography.629 (1993), 169- I 79.
4. T. Harraiz, Sample preparation and Reversed-phase high-performance liquid
chromatography analysis of food-derived peptides, Analytica Chimica Acta, 352
(1997), 119-139.
5. Florance J., Galdes A., Konteatis Z., Kosarych Z., Langer K., Mamtcci C., High
Performance Liquid Chromatographic Separation of Peptide and Amino Acid
Stereoisomers,Journal of Chromatography.41 4 (1 987), 3 13-323.
6. Apffel A., Peptide mapping, Journal of Chromatography. 712 (1995), 177-190.
7. Seiler P., Peptide bank regulatory peptides, Journal of Chromatography A . 852
(1999), 273-283.
8. lvanov V. T., Karelin A. A., Philippova M. M., Nazimov 1. V., Pletnev V. Z.,
Hemoglobin as a source of endogenous bioactive peptides: the concept of a tissue-
specific peptide pool, Peptide Science, 43 (1997), I 7 I - 188.
9. Blishchenko E. Yu., Mernenko 0. A., Yatskin 0. N., Ziganshin R. H., Philippova M.
M., Ivanov V. T., Karelin A. A., Neokyotorphin and neokyotorphin (14):secretion
by erythrocytes and regulation of tumor cell growth, FEBS Lett, 414 (1997), P. 125-
128.

407
 SEPARATION OF ACANTHOSIDE-D IN ACANTHOPANAXSENTICOSUS BY
PREPARATIVE RECYCLE CHROMATOGRAPHY

SEUNG PYO HONG,DEXIAN WANG, KYUNG HO ROW

Centerfor Advanced Biosepration Technology and Dept. of Chem. Eng., lnha University,
253 Yonghyun-Dong, Nam-Ku, Incheon 402-751, Korea
E-mail: [email protected]

To extract and separate Acanthoside-D, a ginseng-like substance contained in Acanthopanar

Senticosus, the optimum operating conditions were experimentally determined in the analytical and
preparative chromatography. Before HPLC applications, acanthoside-1) was firstly extracted from
the powder of the trunk of Acanthopnm Senticosus by ethanol, and then partitioned with n-hexane
to remove impurities. The optimized mobile phase composition and injection volume in analytical
chromatography were water/acetonieildmethanol= 80/14/6vol.% and 20 d,respectively, while
those by preparative recycle HPLC were 70/15/15 vol.% and 2ml, respectively. By recycling the
sample for 4 times, Acanthoside-D with 93% purity was obtained.

1 Introduction

Acanthopanacis Senticosus, a typical plant medicine in Korea, is widely used in many

fields, such as, food, medicine, tea, drinks and cosmetics. Acanthopanax senticosus
(Siberian ginseng) not only shows medical benefit for treating tonic, limbago, neuralgia
and palsy, but also shows acceleration activities of metabolism [l]. The medical part of
the plant is its root, stem and peel of the trunk. The reported components of
Acanthopanax senticosus include sesamin, savinin, lignan glycoside such as Acanthoside
A, B, C, D and Chiisanoside, P-sitosterol, stigmaserol, campesterol, vitamin, mineral and
so on[2-31. Among the above mentioned components, acanthoside-D [(+)-syringaresinol
di-O-P.D-glucoside, Eleutheroside-E, and acankoreoside-D, C34H46018, mp-242OC1, is
regarded as the index component and has been reported to show activities of increasing T-
cell level, reducing cholesterol, activating the prostate, increasing stamina, improving
learning ability, improving the liver function, inhibiting stomach ulcer, improving the
immunity of the organism and inhibiting leukemia[3-4]. So the determination and
preparation of acanthoside-D in Acanthopanacis Senticosus is very important.

Recycle HPLC is a technique that involves the reinjection of unevaporated impure

eluent fractions, as a single injection, back onto the original column to enhance the total
recovery and purity of components from a separation[5]. Recycle HPLC is more powerful
to separate and purify than quantitative and qualitative analysis of the organic compounds
for the synthetic organic chemistry. Also, it can be applied to the chemistry of natural
products, medicine, biochemistry, agricultural chemistry and bioengineering.

The goal of this study is to obtain Acanthoside-D at a high purity and to establish an
optimal separation and purification condition of preparative recycle HPLC by changing
the mobile phase compositions of water, acetonitrile and methanol.
2 Experimental

2.1 Reagents
The Acanthopanm senticosus was purchased from Korea Siberian Ginseng
Association(cily, Korea). Methanol, acetonitrile and twice distilled water were filtered by
using decompressing pump (Division of Millipore, Waters) and filter film(FH-0.5 m).
The extracted sample was concentrated using a rotary evaporator from LABO-THEM
SW 200, Resona Technics Co.(city, Korea)

2.2 Apparatus and Methods

2.2. I Solvent Extraction

5 g of trunk powder of Acanthopanm senticosus was added to 100 !m of ethanol and
put into a reflux condenser. After extracting for 3 hours, the extraction was cooled down
and filtered. Then, it was concentrated and dissolved in 50 !m water. The water solution
was partitioned with 50 !m hexane to remove impurities. After extraction process, the
water layer was collected, dissolved in methanol and filtered by 0.2 m filter film before it
was injected into the HPLC system.

2.2.2 Analysis of Acanthoside-D

The analytical HPLC column used for analysis was packed in 300 x 3.9 mm stainless
column with Lichrospher 1OORP-18 packing(l5 w, Merck Co.). The HPLC used for
analysis consists of a M930 Solvent Delivery Pump, a 486 detector (M 7200 Absorbance
Detector) fiom Young-In Scientific Co, and a Rheodyne injection valve (20 pA? sample
loop). Autochrowin (ver. 1.42, Young-In Scientific Co.) connected to PC was used as
data acquisition system.

2.2.3 Preparative Recycle HPLC

Preparative Recycle HPLC column was a JAIGEL-ODs-AP column(13 m, 2 1.5 x 300
m,JAI Korea). The Preparative Recycle HPLC consists of a PI-SOF Pump(JA1 Korea), a
UV-3 IOB detector(JA1 Korea), a RI-50 detector(JA1 Korea), a Rheodyne injection
valve(3 me sample loop), and a SS-250F2 recorder(JA1 Korea).

3 Results and Discussion

3. I Solvent &traction
Because many kinds of components existed in Acanthopanax senticosus, when the
trunk powder of the plant was extracted by ethanol, a large amount of impurities were
included except for the interest component acanthoside-D. As acanthoside-D is soluble in
water, a partitioning step with waterhexane as the partition solvents was used to remove
the impurities before fiuther refining. In this procedure, non-polar components were
moved into hexane layer and polar components including acanthoside-D were moved into
water layer. The water layer was concentrated, and then it was dissolved in methanol and
refined by the preparative recycle HPLC.
3.2 Analysis of Acanthoside-D
A stainless column packed with Lichrospher IOORP-18 was used to analyze
Acanthoside-D from the solvent extraction. The composition of mobile phase in
analytical HPLC was experimentally determined and it was
water/acetonitrile/methanol=80/14/6 vol.%. From the chromatogram, retention time of
Acanthoside-D was found to be 12 min. Figure 1 shows the analysis of Acanthoside-D
fkom the extraction of the trunk of Acanthopanmr senticosus. n i flow rate of mobile
phase and injection volume were 1 d / m i n and 20& respectively.

300

250 -

-
6
'i
200-

150 -
-
#
100 -
Acanthoride-D

0 5 lo 15 20 25 30 35 1

Figure 1. Analysis of Acanthoside-D from the trunk of Acanthopanmr Senticosus

(water/acetonitrile /methanol = 80/14/6Vol.%,20,d injection volume)

3.3 Preparative Recycle HPLC

To obtain Acanthoside-D in a high-purity and for preparative scale-up, the amount of
injected sample was increased and preparative recycle chromatography was used in this
work. The optimum composition of mobile phase in preparative recycle HPLC was
water/acetonitrile/methanol= 70/15/15vol.%. The flow rate and injection volume were 5
mQmin and 2 d,respectively. Figure 2 shows the chromatogram by preparative HPLC,
where, peak #1 was identified as Acanthoside-D. Peak #1 was obtained by collection of
several experimental runs. After the collection, the dilute samples were concentrated to
3104. The concentrated sample was injected into preparative recycle HPLC to remove
impurities contained. Recycle process was performed to remove tailing each peak, which
enabled to acquire high-purity Acanthoside-D. Figure 3 shows the chromatogram by
preparative recycle HPLC. By recycling the sample 4 times, Acanthoside-D with 93%
purity was finally collected. Figure 4 shows the purified Acanthoside-D analyzed by
analytical HPLC.

410
Figure 2. Separation of Acanthoside-D from the trunk of Acanthopanax Senticosus by
preparative HPLC column
(water/acetonitrile/methanol= 70/15/15 vol.%, 2mI injection volume )

Figure 3. Separation of fraction #1 by recycle preparative HPLC column

(water/acetonitrile/methanol = 70/15/15 vol.%, 3ml injection volume )

4. Conclusion

In this work, by utilizing the optimum mobile phase: water/acetonitrile/

methanol=80/14/6vol.% for analytical HPLC and water/acetonitrile/methanol =
70/15/15vol.% for preparative recycle HPLC, respectively, purified Acanthoside-D with
93% purity has been preparatively obtained.

411
 i ; . ,
10
. I
15
- ,
20
.
I

Time (mia)

Figure. 4. Analysis of the sample containing Acanthoside-D in the 4th recycle

(water/acetonitrile/methanol= 80/14/6 vol.%%,20p.l injection volume)

5. References

1 . The Korea Pharmacopeia, 6th Ed., (1 992), p 1027.

2. Zhao W. M., Qin G. W., Xu R. S., Li X. Y.,Liu J. S., Wang Y.,and M. Freg,
Constituents from the roots of Acanthopanm setchuenensis, Fitoterapia, 70 (1999),
529-53 1.
3. Hahn D. R., Kim C. J., and Kim J. H., A Study on Chemical Constituents of
Acanthopanm koreanum Nakai and Its Pharmaco-biological Activities, Yakhak
Hoeji, 2!3(6) (1989,357-361.
4. Nishiyama N., Kahiko T., Iwai A., Saito H.,Sanada S., Ida Y.,and Shoji J., Effects
of Eleutherococcus senticosus and Its Components on Sex-and Learning-Behaviours
and Tyrosine Hydroxylase Activities of Adrenal Gland and Hypothalamic Regions in
Chronic Stresses Mice, Sho. Zas.,39(3) (1985).
5. John R. C., Kimberly C. J. and Ranmali W., External recycle Chromato- graphy : A
practical method for preparative purifications, Journal of Chromatography A, 462
(1 989), 85-94.

412
 USE OF VARIOUS FORMS O F KRAFT LIGNIN
FOR TOXIC METAL UPTAKE

DELANSON R. CRIST
Department of Chemistv, Georgetown University, Washington DC 2005 7, USA
E-mail: [email protected]

RAY H. CRIST AND J. ROBERT MARTIN

Messiah College, Grantham, PA 17027, USA
E-mail: [email protected]

A promising new material to remove toxic metals from polluted waters is kraft lignin, a by-product of
paper production produced in large quantities. This material has acidic sites (carboxylic acids, phenolic
and enolic hydroxyl groups, and mercapto groups) which can hnction as an ion exchange sites for
metals. One form of kraft pine lignin resulting from an acidification step in paper production is lndulin
AT (hereafter called lignin). One practical problem with use of this powder is difficulty in its separation
from water after uptake. A more eficient form of lignin was obtained by treating a Ca-loaded lignin
power with dimethylformamide(DMF)containing a polysulfone resin. Removal of DMF,which could
be re-used, provided a material that could be rolled into cylinders and then flattened into plates or strips.
Such strips I) were structurally stable after suspension in water for six weeks, and 2) maintained
efficiency for Pb uptake over eight cycles of uptake followed by extraction of Pb by EDTA. Uptake of
Pb and Cd released one mole of Ca for one mole of the metal sorbed. This result clearly shows that the
uptake process is one of ion exchange and not simple adsorption in which no species is released on
uptake. Chips were efficient in removing Cd from a solution passed through a column.

I Introduction

Various materials have been used for biosorbents to remove toxic metals from process or
groundwater. Ones selected by the US Army Corps of Engineers in 1997 [l] for further
study include chitosan, peat moss, seaweed (algae), and lignin. Work in our laboratory has
shown that uptake by peat moss [2] and algae [3] occurs by an ion exchange mechanism in
which an existing proton or metal is displaced in the process:

M2' + 2 (HX) R (MX2) + 2H' K,: = [H']' (MX2) / [M"] (HX)'

M ~ ++ (c~x,)R ( M X ~+) Ca2+ K,, = [Ca"] (MXZ)/[Mz'] (CaXz)

Recently we reported that uptake by a krafi pine lignin also occurs by these ion exchange
reactions [4].
Lignin is a major component of plants where it serves as a binding agent for cellulose
and other materials, and the kraft process of paper production, heating with alkali and
sulfide, produces polyhydroxy phenolic, carboxylic acid, and sulfide functional groups in
a soluble black liquor mixture [5]. Acidification precipitates this modified lignin as a
powder, hereafter referred to as lignin, a 20B30% by product in the manufacture of paper.
Constancy of values of K," and K,, for wide ranges of initial concentrations
demonstrate validity of these displacement processes and prove that uptake in these systems
is not simple adsorption. Furthermore, it was found that the stoichiometry for Sr and Cd
uptake by Ca-loaded lignin is one metal sorbed for one Ca released. This observation for
metals of very different binding strengths is difficult to rationalize with adsorption models,
but is in complete agreement with an ion exchange process.

413
 The finely divided powder (200-300 mp) used in the above study is difficultto separate
from purified water in large scale and is not suitable for column applications. This work
describes a novel formulation of this powder into a solid workable into various forms
effective for heavy metal uptake.

2 Methods

Kraft pine lignin, kindly provided by MeadWestvaco, Charleston, SC as their lndulin AT

product, was converted to a completely acidified form, HLg, by treating a suspension of 6.0
g of lignin in 40 mL of water with dilute nitric acid to pH 1.5 and stirring for 30 min. After
centrifugation,the solid HLg was washed twice with water and to a suspension in 40 mL of
water was added calcium hydroxide powder (50 mg per g HLg, unless stated otherwise).
After stirring for several hours, the resulting CaLg was centrifuged, washed twice with
water, and air-dried overnight.
A plastic material was made by mixing 1.O g of CaLg with 0. I g of clipped, glass wool
and then adding 0.6 mL of DMF which contained 60 mg of a polysulfone resin (Aldrich,
average MW 30,000). Thorough mixing gave a uniform mass of putty-like, plastic material
which could be rolled into a cylinder. Flattening by further rolling gave a strip which could
be cut into various shapes: strips (typically 0.3 x 4 x 10 mm for 0.1 g), chips, or flat pieces.
Hard material can be formed by heating for 20 min at 70 OC and then cooling to RT. This
hard material is referred to as modified lignin MLg.
Concentrationswere determined by atomic absorption (AA) with a Perkin-Elmer Model
2380. For amounts of sorbed metals, the solid was treated with 0.01 M HN03 for 30 min
followed by centrifugation and AA analysis of the aqueous layer.

3 Results show uptake occurs by ion exchange with potential practical applications

3.1 Recovery of DMF

A strip of MLg made with 20 mg of calcium hydroxide was heated to 120 OC and held at that
temperature of 20 min. A stream of nitrogen passing over the sample in a glass tubing was
collected at Dry Ice temperature. The weight of condensed gases was essentially the same
as the amount of DMF used in the sample preparation.

3.2Structuralstabikty in water

A 40-mm long plastic strip was glued to the end of strip MLg. This plastic strip seived as a
pointer to provide more accurate measurements. The MLg strip was extended horizontally
40 mm from the edge of a support block and a small weight (ca. 0.3 g) hung at the junction
to cause a bending in the MLg. The extent of bending was measured as the vertical
deflection of the pointer end and the sample immersed in water. After one week, the sample
was dried at 70 OC for 20 min and re-tested. After six weeks there was no significant change
in the degree of bending.

414
3.3Metal uptake is a rapid ion exchange process

Strips of MLg, ca. 0.2 g each, were immersed in 20 mL of solutions 0.5 to 10 mM in Pb. Pb
uptake occurred with an equivalent amount of Ca released. Similar results were found for
Ca release on Cd uptake. This Cd uptake was rapid, as shown by an experiment in which a
0.1 g MLg strip was immersed in 25 mL of a 0.2 mM Cd solution, and 65 % of the Cd was
sorbed within 10 minutes.

3.4 MLg is re-useable

A 0.1 g strip was suspended in 10 mL of a solution 0.1 mM in Pb. After measurement of Pb

sorbed (as the difference of initial - final Pb in solution), Pb was removed by treatment
Ca-EDTA. The same sample was re-used by adding a fresh solution 0.1 mM in Pb. The
amount of Pb sorbed for each re-use was the same after 8 cycles.

3.5Batch vs. Column Operations

In a batch operation, MLg (0.1 g strip) sorbed 65% of the Cd from 25 mL of 0.2 m M Cd,
but 35 % remained in solution at equilibrium. All the Cd cannot be removed in a one-batch
operation, since released Ca in solution displaces sorbed Cd. In a column operation, various
Cd solutions were passed through a column (10 cm x 1.5 cm) packed with MLg (1.2 g small
chips). N o appreciable Cd could be detected in the eluant after passage of 300 mL of 0.1
mM Cd, then 200 mL of 0.2 mM Cd, and then for most of 200 mL of 0.4 mM Cd. In contrast
to the batch operation, no detectable Cd was found in the eluant after sorption of equivalent
ofalmost 1.5LofO.l mMCd.

4 Discussion

The fact that one mole of Ca is released for one mole of Pb or Cd sorbed onto MLg shows
that the process is one of ion exchange rather than adsorption. This stoichiometry cannot be
explained by the biotic ligand model, also an adsorption model which assumes that metals
act independently, even when metal competition is included [6].
Advantages to this modified lignin include: stable structure; re-usable; recovery of
solvents for re-use; high capacity, determined by the amount of calcium hydroxide used;
rapid; effective in batch or column operations.

References

1 . Bailey S., Olin T. J. and Bricka R. M., Low-cost sorbents: a literature summary,
Technical Report SERDP-97-1(U. S . Army Corps of Engineers, Vicksburg, MS, 1997).
2. Crist R. H., Martin J. R., Chonko, J. and Crist, D. R. Uptake of metals on peat moss: an
ion exchange process, Environ. Sci. Technol. 30 (1996) pp. 2456-2461.
3. Crist R. H., Martin J. R., Carr D., Watson J. R., Clarke J. and Crist D. R. Interaction of
metals and protons with algae. 4. Ion exchange vs. adsorption models and a

415
 reassessment of Scatchard plots; ion exchange rates and equilibria compared with
calcium alginate, Environ. Sci. Technol. 28 (1994) pp. 1859-1866.
4. Crist R. H., Martin J. R and Crist D. R. Heavy metal uptake by lignin: comparison of
biotic ligand models with an ion exchange process, Environ. Sci. Technol. 36 (2002)
pp. 1485- 1490.
5. Marton J. Reactions in alkaline pulping. In Lignins Occurrence, Formation, Structure,
and Reactions, ed. by Sarkanen K. V. and Ludwig C. H. (Wiley, New York, 1971) Ch.
16.
6. Meyer J. S. A mechanistic explanation for the ln(LC5O) vs In(hardness) adjustment
equation for metals, Environ. Sci. Technol. 33 (1 999) pp. 908-912.

416
 REMOVAL OF URANIUM IONS IN SLUDGE WASTE BY
ELECTROSORPTION PROCESS

C. H. JUNG, J. K.MOON,S.H. LEE AND W . Z. OH

Korea Atomic Energy Research Institute, P.o. Box 105, Yuseong Taejon, Korea E-mail:
[email protected]

Y. G. SHUL
Dept.. of Chemical Engineering,Yonsei Universiw, 134 Shinchondong Seoa'aemoongu, Seoul,
Korea
E-mail: [email protected]

A study on the electrosorption of U(VI) onto a porous activated carbon fibers (ACFs) was performed to treat
uranium-containing lagoon sludge. Effective U(VI) removal is accomplished when a negative potential is
applied to the activated carbon fiber(ACF) electrode. For a feed concentration of 10@-350mg/L, the
concentration of U(VI) in the cell effluent is reduced to less than ImglL. The adsorbed uranium could be
desorbed from the ACF by passing a IM NaCl solution through the cell and applying a positive potential on the
electrode.

1 Introduction

For the removal of uranium (VI) fiom waste streams, a variety of physical and
chemical methods such as precipitation, coagulation, ion exchange and adsorption have
been used. But, these techniques have been restricted in application due to their limited
capacity when the concentration of U(VI) in the waste water is relatively high[l]. As an
efficient electrochemical method for the removal of high concentration uranium,
electro-deposition on carbon materials has been extensively investigated and used very
effectively[2]. For uranium having a high reduction potential, electro-deposition is not a
practical method. An alternative to electro-deposition is electro-sorption, that is,
adsorption of the metal cations onto a negatively charged carbon surface.
Electrosorption technique, which may use the electrical potential as the 3d driving
force to the traditional adsorption and ion exchange mechanism, has reversible
characteristics of purifying waste solution by adsorption and concentrating contaminants
by desorption. Carbon materials satis@ the basic requirements for an efficient electrode
material, and have good radiation and chemical-stability.Especially activated carbon fiber
(ACF), which can be easily made into a variety of types (textures or sheet), has a high
specific surface area and electrical conductivity.
In this study, we conducted the experiments on a selective adsorption of uranium
(VI) fiom a high concentration of chemical salt to investigate the technical feasibility of
the electrosorptionprocess using ACF as an electrosorption adsorbent.

2 Experimentals

Flow-through adsorption experiments were carried out using a three-electrode

electrochemical cell where electric current flows parallel to the- solution flow. Working

417
electrode(ACF) was placed on a platinum mesh used as a current collector and supporter.
The counter electrode was platinum wire and the Ag/AgCI electrode was used as the
reference electrode. All the potentials reported in this paper are relative to this reference
electrode. The electro-chemical cell was connected with a potentiostat (EG&G Model
273). The fixed flow rates through the cell were controlled by a peristaltic pump.
Various type of salts such as NaCl, NaN03 and NH4N03 were tested as the supporting
electrolyte.

3 Result and Discussions

Test on the electrosorption with 100 mg/L U( VI) feed onto the ACF felt at various
negative potentials in the range of - 0.3 to - 0.9 V (vs. Ag/AgCl) was carried out. The
uranium concentration in the effluent from this test is shown in Fig. 1.

t- -0.w
0.7 * - -0.w
-
0.6
+-O.w

0.3

Fig. 1. Electrosorption of U(W) with variation of potentials.

Effective U(V1) removal is accomplished at all negative potentials. At a potential of

- 0.3V, the U(V1) concentration in the effluent is reduced to Img/L in 2h and then
increased continuously. At a potential of - 0.5 and - 0.9V, a complete removal of U(V1)
is rapidly reached in Ih and maintained throughout the test. However, in case of
OCP(open4rcuit potential) the effluent concentration of U( VI) increased within 3h and
finally reached the level of the feed, indicating saturation of sorption capacity by the ACF.
From these results, it could be confirmed that the external negative potential exerted on
the ACF electrode has great impact on the adsorption capacity of the ACF.
To determine the electrosorption capacity of this ACF, a long-term test was
conducted with a 350mg/L U(V1) feed at - 0.9V. As shown in Fig. 2, the effluent
concentration is rapidly reached at less than lmg/L and it corresponds to a specific
-
sorption rate of 662M/(g min). The cumulative amount of uranium within 15h is about
552~grniudgACF-

418
 0.1c

0.08

0.a
0

I!
0.04

0.02

0.00

lime, rrin
Fig. 2. Long term test of U(VI) electrosorption.

The adsorbed uranium could be desorbed up to 92% for 1 Oh by passing a 1 .OM NaCl
solution through the cell and applying a potential of +1.2V (Fig. 3).
The electrosorption behavior of U(VI) with a variation of solution pH at -0.9 V is shown
in Fig. 4. As the solution pH was lowered, the adsorbed uranium decreased. At pH 2.1,
the eftluent uranium concentration is 5Omg/L at the beginning but rapidly increases to
200mg/L in about 6h, showing an extremely low uranium sorption capacity. The
electrosorption at pH 3 and 4 shows very similar results except for the slightly higher
uranium removal during the first hour of electrosorption at pH 4. The effect of
electrolyte concentration was studied and the results are shown in Fig. 5 . It can be seen
that uranium adsorption extents and kinetics increased with the increase of electrolyte
concentration. It is due to the increase of ionic mobility and decrease of ohmic voltage
drop resulting from the increase of electrolyte concentration. As ionic mobility

I ' I ' I ' l ' l . 1

100 -
80-

s .
g
.-c.
E '
0
0 '
M -

01 -*-
0 1 0 0 2 0 0 J o o 4 0 0 5 0 0 m

419
 Tim,rnin
Fig. 4. Electrosorption of U(VI) with variation of solution pH.

increases, charge transfer becomes active, which is favorable in electrosorption reaction.

The electrosorption behavior of U(VI) with a variation of eIectrolyte type at -0.9 V is
shown in Fig. 6. In case of using NaNO3 and NH4NO3 as electrolyte, only small amount
of U(VI) ion is removed and adsorbent is easily saturated. This is due to electrochemical
reduction of nitrate present in the solution to nitrite, ammonium or nitrogen gas. After
pretreatment in a suitable condition, U(VI) ion was removed within 30 min completely
and this result was maintained throughout the test.

lime, nin
Fig. 5. Electrosorption of U(VI ) with variation of electrolyte concentration.

420
 Fig. 6. Efect of pretreatment on the efficiency of Uraniumelectrosorption.

4 Acknowledgements

This project has been carried out under the Nuclear R & D Program finded by the
Ministry of Science and Technology

References
1. Lenhart, J.J., Figueroa, L.A., Honeyman, B.D., Kaneko, D., Colloids Surf. 120
(1997) pp. 243-254.
2. Farmer, J.C., Bahowick, S.M.,H a m , J.E., Fix, D.V., Martinelli, R.E., Au, A.K.,
Carroll, K.L., Energ. Fuels 11 (1997) pp. 337-347.

421
 ION EXCHANGE CHARACTERISTICS OF PALLADIUM FROM A
SIMULATED RADIOACTIVE LIQUID WASTE

S. H.LEE, C. H.JUNG, J. K. MOON, J. H.KIM AND H.CHUNG

Atomic Energy Research Institute, P.O. Box, 105, Yusong,Taejon 305-600,Korea

E-mail: [email protected]

Radioactive high-level liquid wastes contain significant quantities of platinum group metals (PGM)
such as Pd(Il), Rh(lll) and Ru(1ll). The PGM are produced as fission products in nuclear reactor. In
this study, batch and column experiments were carried out to investigate the ion exchange
characteristics of Pd(I1) including the effects of the ionic group of ion exchangers and the
concentration of nitric acid by various anion exchangers such as IRN 78 and Dowex 1x8 and also the
elution characteristics of Pd(1l) by various eluents. Anion exchangers such as Dowex 1x8 with the
ionic group of quaternary methyl ammonium had a higher capacity than anion exchanger such as
IRN 78 with amine group for the adsorption of Pd(l1). Especially, new type anion exchanger, AR-OI
with quaternary and tertiary benzimidazole groups showed very strong. Pd(I1) could be easily eluted
by 0.5 M thiourea and 0. I M nitric acid mixed solution.

1 Introduction

The irradiation of nuclear fuels in power reactors leads to the production of atoms of a
wide range of fission products, ranging in atomic mass from 70 to 160. These fission
products include three of PGM, i.e., palladium (Pd), rhodium(Rh) and ruthenium(Ru).
The PGM are valuable and important as noble metals, but their natural resources are
rather limited. Due to the increase in utilization of PGM in the automotive and dental
industries, in electronic and electrical devices, in the production of ultrapure hydrogen,
and as an industrial catalyst. The demand for PGM continues to grow in a steady manner.
The noble metals generated by fission can be an important alternative resource to meet
the increasing demand [4]. In order to separate the PGM from high-level radioactive
liquid waste(HLLW), various recovery methods have been studied, for example, lead
extraction, solvent extraction, precipitation, adsorption and ion exchange methods
[ 1,3,4,5,7]. Recently, Kondo, Y et a1.[4] investigated precipitation behavior of PGMs
from simulated high level liquid waste. And Wei, Y.Z.,et a1.[7] reported that the anion
exchanger AR-0 1 with a benzimidazole group especially shows significantly strong
adsorption of Pd(I1) from nitric acid solution.
In this work, batch and column experiments were carried out to investigate the ion
exchange characteristics including the effects of the ionic group of ion exchangers, the
concentration of nitric acid, the feed rate and also the elution characteristics of Pd(1I) by
various eluents for obtaining the optimal separation conditions.

2 Experimental method

Commercial strongly basic anion exchangers such as IRN-78 and Dowex 1x8 were used
as anion exchangers for these experiments. The resins were pretreated with 10 % sodium
hydroxide solution, washed with distilled water, converted into nitrate form with nitric
acid, and dried overnight at 50'c in a convection oven. Reagent grade nitric acid was
diluted with distilled water to obtain a desired concentration varying from 0.1 to 7 M. The

422
nitric acid solution of Pd(I1) was prepared by dissolving Pd(N4)2 into the nitric acid
solution.
In batch experiments, 1 g of an anion exchanger and 20 ml of a platinum group metal
containing solution were put in a glass flask and the flask was set in the mechanically
shaking water bath maintained at 20'c or 609: for 24 hours. It was then taken out and
centrifuged for 5 minutes at 3000 rpm, and the metal concentration in the solution was
measured by inductively coupled plasma atomic emission spectroscopy (ICP Model:
Jobin Yvon JY38 plus, polychromator, JYOCY).
The column experiments were carried out in a glass column of 1.5cm internal
diameter and 30cm length filled with 5g of ion exchange resin. Pd solution was percolated
through the packed column at a flow rate of 1.O ml/min controlled by a peristaltic pump
(EYELA SMP-21, Japan). The effluent samples were collected at regular intervals by the
fraction collector (Model: Adventec SF-2 100) and analyzed for Pd concentration by ICP.
The distribution coefficient (&) of metal is defined as the ratio of the metal
concentration in the resin of the ion-exchanger to that in the solution [2,7]. Thus, Kd
value was calculated by the following equation,

where C, is the content of metal in 1 g of resin and C2 is the content of metal of 1 ml in

the solution.

3 Results and Discussion

The sorption kinetics of Pd(l1) by anion exchangers such as IRN-78 and Dowex 1x8
are shown in Fig. 1. The results show that the rate of ion exchange of Pd(l1) by an anion
exchanger was very rapid and reached an equilibrium state within 1 hr. The ionic groups
of ion exchangers acted as metal ion exchange sites. Pd(I1) was strongly adsorbed from
dilute nitric acid solutions.

f 50
0 40
30 -0- :Dower 1x8

0 5 10 15 20 25 O m 5
Time (hr) Time (hr)
Fig. I . Adsorption raw of Pd(ll) on various anwn cschangcrs Fi.2. Ellkc1ofwnic gmup on the adsorpton of Pd(O) on
INiVi 8cidconccnvaion.:O.l M.Tcmpcnlure: 20"CI various anion exchangers N i t k acid concenmatan:
6 hi kmperdture: 60 'C 1

The ion exchange reaction of anionic complexes in nitrate media is generally expressed as
following equation [7],

*
VA(R*.NO<)+ pd(No3)~- (v~R+'Pd(Noj)3-)+ NO<

423
where ,'R Pd(N03)i, and vA denote the fixed ionic group, the counter ion(anionic
complex of Pd), and the charge number of the counter ion, respectively.
Fig. 2 shows the effect of ionic groups on the adsorption of Pd(II) from nitric acid
solution onto the several anion exchangers. These results were compared with those from
the anion exchanger named AR-01 with quaternary and tertiary benzimidazole groups and
the several conventional anion exchangers such as IRA 900 and IRA-93ZU cited in the
literature [7]. As shown in Fig. 2, anion exchangers such as Dowex 1x8 and Dowex 2x8
with an ionic group of quaternary methyl ammonium have higher ion exchange capacity
than anion exchangers such as IRN 78 and IRA-93ZU with a conventional amine group.
The anion exchanger AR-01 with a enzimidazole group especially shows significantly
strong adsorption of Pd(II) from nitric acid solution, the distribution coefficient (Kd)
value for the adsorption of Pd(II) was about 2000-3000 at 0.1 M of nitric acid solution
171.
Fig3 shows the effect of the nitric acid concentration on the adsorption of Pd(I1) in
the range of 0.1 to 7 M by Dowex 1x8 and IRN 78. the optimal nitric acid concentration
was shown to be in 2 3M. -
The ion exchange characteristics of Pd(II) in packed column filled with 5g of anion
exchange resin at 2.0 M nitric acid solution are shown in Fig. 4. The ion exchange
capacity of Pd(II) by Dowex 1x8 was higher than that by IRN 78 at 2.0 M nitric acid
solution. Breakthrough by IFW 78 occurred at 40 ml volume treatment when Pd
concentration was 172 ppm, while breakthrough volume by Dowex 1x8 was 100 ml. This
result shows a similar trend to the results in batch experiments, Dowex 1x8 with ionic
group of quaternary methyl ammonium has higher ion exchange capacity than IRN 78
with conventional amine group.

0.4

0 1 2 3 4 5 6 7 8
0.2

0.0
0
'
I

50
1
100
-c-
4-

150
: IRN 78
:Dower 1x8

200 250 300 350

Contentratiin of HNO, (N) Treated Volume, ml
Fig. 3 . Elfeciofniirk acid concentralion on the idrowlion Fig. 4. Breakthrough curve for Pd(ll) sdrorplion by anion
alPd(ll) O n a n a n c x s h 8 n g c n I T em p en lu re: 20 "c exchsngen [ Pd con.: 172 pprn. nitric acid con.:
2.0 M. feed rate: 1 mumin 1

Fig. 5 shows the effect of feed rate on the adsorption of Pd(1I) in the packed columns
by Dowex 1x8. As shown, the ion exchange capacity of Pd(I1) decreased as feed rate
increased. When feed rate was 1 ml/min, breakthrough point occurred at 100 ml treatment
and the bed was saturated at 3 10 ml treatment. When the feed rate increased 3.5 mumin,
however, the breakthrough volume for Pd(l1) adsorption decreased to 48 ml and the bed
was saturated at 150 ml treatment.
The elution characteristics of Pd(I1) in packed bed filled with 5 g Dowex 1x8
saturated with Pd(I1) by various eluents are shown in Fig. 6.
As shown in Fig. 6, the capacity for the elution of Pd(I1) by 0.5 M thiourea and 0.1 M
nitric acid mixed solution was high. Pd was easily and

424
 0.8
"O

0.4 ,;!Lr , ,

I?
02

0.0
0 50 loo 1so 2w 250 3 0 0 m
Treated Volume, mi Treated Volume. ml
Fig. 5. Effect of faad rate for Pd(ll) adsofption by ion exchangers Fig. 6. Elution curve8 for Pd(ll) saturated column by various
[ Pd conuntntion: 172 ppn. nitric acid urn.: 2 M] eluents [feed rate: Imlhnin.. resin: 5 g Dowex 1x81

perfectly eluted by 0.5 M thiourea and 0.1 M nitric acid mixed solution. However, the
elution capacity of Pd(I1) by low concentration of nitric acid was very low.

4 Conclusions

Ion exchange and elution characteristics of Pd(I1) by several anion exchangers in the
batch and column were investigated. Based on experimental results, the ionic groups of
anion exchangers affected the ion exchange capacity for the adsorption of Pd from nitric
acid solutions significantly. Anion exchangers such as Dowex 1x8 with the ionic group of
quaternary methyl ammonium had a higher capacity than anion exchanger such as IRN 78
and IRA-93ZU with amhe group for the adsorption of Pd(I1) from nitric acid solution in
the batch and column experiments. Especially, new type anion exchanger, AR-01 with
quaternary and tertiary benzimidazole groups showed very strong adsorption of Pd(I1)
compared with other conventional anion exchangers. The optimal nitric acid
concentrationwas shown in the 2 3 M. -
The elution characteristics of Pd(I1) in the batch and packed column showed that Pd
could be easily eluted by 0.5 M thiourea and 0.1 M nitric acid mixed solution. The
capacity for the elution of Pd(II) by 0.5 M thiourea and 0.1 M nitric acid mixed solution
was high compared with the other eluents such as nitric and hydrochloric acid solutions.
Based on these experiments, to separate Pd(l1) effectively by an ion exchange method
from radioactive liquid waste, the development of anion exchanger with the high ion
exchange capacity for Pd(I1) may be important.

References

1. Beamish, F.E., A critical review of methods of isolating and separation the six
platinum metals, Tuluntu, 5, (1960), pp 1-35.
2. Helfferich, F.,Ion Exchange, McGraw-Hill, New York (1962).
3. Jensen, G.A. et al. .J., Recovery of noble metals from fission products, Nucl. Technol.,
65, (1984), ~~305-324.

425
4. Kondo, Y and Kuboto, M., Precipitation Behavior of Platinum Group Metals from
Simulated High Level Liquid Waste in Sequential Denitration Process ,J. Nucl. Sci.
Technol.,29(2), (1992), pp140-148.
5. Navratil, J.D., Ion exchange technology in spent fuel reprocessing,
J. Nucl. Sci. Technol, 26(8), (1989), pp735-743.
6. Rizvi, G.H. et al., Recovery of fission product palladium from acidic high level waste
solution, Sep. Sci. Technol.,3 1(13), (1 996), pp 1805-18 16.
7. Wei, Y.Z.,et al.,"Adsorption and Elution Behavior of Platinum Group Metals In Nitric
Acid Medium", Proceeding of IEX'96, 174, Royal Society of Chemistry (July, 1996).

426
 APPLICATION OF CHARACTERIZATION PROCEDURE FOR COMPLEX
MIXTURE ADSORPTION IN WATER AND WASTEWATER TREATMENT

SEOUNG-HYUN KIM', TAE-WONKIM', DAE-LIM CHO',

DAE-HAENG LEE' AND HEE MOON'

'Faculty of Applied Chemistry,Chonnam National University, Gwangiu 500- 757, Korea

E-mail: [email protected]
2
Water quality Research Institute, Gwangju Waterworks 503-350.Korea
E-mail: [email protected]

The background organics in water and wastewater are necessary to fractionate into groups or
components according to the difference in adsorbability. In this study, the background organics were
!hctionatedin terms of the adsorptive strength described by the Freundlich isotherm constants k and
n with the assumption that the fractionated components differ in the value of k but have the same
value of fl based on IAST (Ideal Adsorption Solution Theory) using binomial concentration
distribution.A simple CharaCterLationprocedure for water and wastewater to chantcterizeDoc in t e r n
of adsorbabilitywas proposed and applied to three types of organic mixture contained in different water
sources. The composition of each organic mixture was successfully evaluated to describe the IAE
(Integral adsorption experiments)data of the total organic carbon using this characterization procedure.

1. Introduction

In order to estimate the adsorptive capacities of DOC (Dissolved Organic Carbon) on

activated carbon, the determination of single substance isothenns is not sufficient.
Competition phenomena give rise to a different adsorption behavior of DOC in complex
multi-component mixtures like natural waters or effluents compared to single solute
solutions. This necessitates an evaluation of the adsorption characteristics under more
realistic conditions to predict for adsorption amount and kinetics of DOC depending on
initial influent concentrations in wastewater treatment [I]. Membrane MF
(Microfiltration) is a drinking-water treatment and water reuse process, which is
particularly suitable for the removal of suspended solids, especially bacteria, algae, and
protozoa [2]. However, MF is less successful for the removal of dissolved contaminants
such as DOC. Adsorption behavior on PAC of the DOC permeating through CFMF
(Crossflow-Microfiltration) was investigated. A numerical analysis of the adsorption
isotherm data according to the IAST (Ideal Adsorbed Solution Theory) confirmed these
results. Competition phenomena give rise to a different adsorption behavior of DOC in
complex multi-component mixtures like natural waters or effluents compared to single
solute solutions [3,4]. The common approach used to overcome this difficulty is to group
all the contaminants present together and to characterize their presence by a single
surrogate quantity such as BOD (biological oxygen demand), COD (chemical oxygen .
demand), and TOC (total organic carbon). The presence of a variety of substances with
different adsorption affinity in an aqueous solution to be treated by adsorption implies that
one must consider the competitive interactions among the various substances in predicting
the kinetics of the adsorption process.
In the present study, the competitive effects of DOC on the adsorption were
investigated at equilibrium and shorter contact times by PAC in the membrane-PAC
hybrid system. The three carbons studied were commercially available, and have been

427
considered for use in wastewater and water treatment plants within Korea. The conditions
undex which they were studied can be related directly to practical applications in terms of
carbon dose and contact time. It can be shown that the availability of advanced
characterization techniques offers the potential for predicting the adsorbability of DOC
without precise knowledge of the chemical structures present in the water and wastewater.

2. Experiments

Adsorption experiments with PAC (James Cumming & Sons LTD, Australia) were carried
out by varying the amount of activated carbon dosed to 1 L of 0.45 pm filtered sample.
The PAC-HA, PAC-WB and PAC-CB based on coconutshell, wood and coal were used in
this study.
The materials chosen for the investigation in wastewater and surface-water were
DOC. DOC concentration was measured using W-Persulfate TOC analyzer
(DOHRMANN Phoenix 8000). Batch adsorption experiments were performed at different
RPM (50, 100, 150 and 200) to obtain the data required for the design and operation of
membrane-PAC hybrid system for the treatment of real-wastewater and surface-water. A
schematic diagram of the membrane-PAC hybrid system used in this study is shown in Fig.
1. The suspension of PAC particles and wastewater was delivered from a stock tank to a
CFMF cell by peristaltic pump. The dimensions of the filtration channel in the CFMF cell
are 6 cm, 0.6 cm and 0.036 cm of length, width and thickness respectively. The CFMF
cell has 9 filtration channels and the total membrane area is 3.24 X lo4' m2.A
spiral-mixing device was made by winding PVC tube with diameter of 6 mm around a
column 110 rnm in diameter. The PAC and solution were sent together into a mixing
device and went through to the CFMF cell.

Fig. 1. Experimental Set-up for hybrid system with membrane-PAC

3. Theoretical modeling

The main modification in the characterization procedure was to assume a simple discrete

428
distribution function to represent a number of pseudospecies with the same Freundlich
exponent. The initial concentration of each pseudospecies can be assigned by a binomial
function in term of Freundlich constant. Furthermore the competitive adsorption between
species was estimated by a conventional equilibrium theory such as the IAST (ideal
adsorbed solution theory) [3,4].

N
CT =cci
i=l

zi = cioI [ ( n I n K J + ( M I V ) ( r I I n ) ] (3) c(n

N

i=l / nK
CiO
Y + (M I V /n)
-1=o (4)

Ci is the equilibrium concentration of the i th component, and n is the reciprocal of

the Freundlich exponent (for all the components). M and V are the mass of the
adsorbent and volume of solution. Cio is the initial concentration of the i th
component. Zi is the mole hction of the i th component in the adsorbed phase. ll
is the dimensionless spreading pressure defined as (nA I R T ) , where n is the spreading
pressure. A is the adsorption surface area per unit mass of the adsorbent, R is the law
constant., and T is the absolute temperature. II can be found from equation (4) with
n , K i , Cio, M and V known. Once II isknown, Zi can be foundequation(3).
One can readily determine Ci (i =1, 2, ..., N ) with Zi and ll . Once the exponent
value is properly assigned for a given system, the characterization can be carried out
straightforwardly based on a binomial distribution. Then the DOC fraction of the j-th
species in the original solution may be represented as equation (5). Here each species, j ,
is specified by a Freundlich coefficient., K ,which is assigned by the above equation (6).
One may assume that all the pseudospecies obey the Freundlich expression like Eq. (7) to
simplify the subsequent computation work required in both the characterization and the
adsorption calculations. The adsorption rate of adsorbate by a PAC grain is linearly
proportional to a driving force using the LDFA (Linear Driving Force Approximation)
model, defined as the difference between the surface concentration and the average
adsorbed-phase concentration. The IAE (Integral Adsorption Experiment) was used batch
adsorption data as the basis for characterizing solutions of unknown compositions.

4. Results and Discussion

A characterization results was showed for a wastewater and surface-water in removing

DOC using three PACs as shown in Figs. 2, 3, 4 and 5. Characterization results
parameters of wastewater and surface-water on PAC showed in Table 1 and 2. A
characterization procedure in this study is quite suitable for the wastewater to get
information depending on the initial distribution of DOC hction. From the results
obtained in this work, kinetic experimental data were predicted on the assumption that the
diffusion coefficients were unchanged during the experiments ( k, =1.0-2.0X loas).

429
Predictions based on this characterization procedure in batch-adsorber were in good
agreement with experimental data as shown in Fig. 6. The application of the
characterization procedure will be very beneficial in designing and simulating of
membrane-PAC hybrid system systematically to remove DOC from natural water and
wastewater as shown in Fig. 7.
Table 1. Characterization results of real-wastewater on PAC

100 ND
F(%) = - [Gev- Gcal)/ ,G
], ( N D :Number of data)
ND m=l
Table 2. Characterization results of surface-water on PAC

1000

0 HA-Wastewater
0 WB-Wastewater

I00

s?
t

10

't
0
1
0.1 I 10
0 0.02 0.04 0.06 0.08 0.1
CT,mgRrDOC
M N , kgIm3

Fig. 2. Calculated and experimental IAE Fig. 3. Adsorption isotherm simulated data
data for real-wastewater on PACs for real-wastewater on PACs

430
 4

o WBSurfacewater
0 HASurfacewater
3
100

L
6 2
B
10
1

I I I l l 1 I I I I I I U
1 I
I I I I
0
0 0.02 0.04 0.06 0.08 0.1 0.1 1 10
W ,k@ Cr. +DOC

Fig. 4. Calculated and experimental IAE Fig. 5. Adsorption isotherm simulated data
data for surface-water on PACs for surface-water on PACs
4

3.5 0 RPM= 50
0 RPM=100 2
0 RPM=150
3 - WFA+ChPracterizatioA

2.5 1.5
& P
u
u 2
B 1.5
8 1

1
0.5

Oe50 0
0 10 20 30 40 SO 60 0 20 40 60 80
Tim, nin Time on stream min.

Fig. 6. Effect of DOC concentration on Fig. 7. Membrane-PAC hybrid experimental

mass transfer rate @AC-WB= 150 mgL) data (Contact time=4 min, MPS=0.65pm)

5. Acknowledgements
This study was supported by a research fund provided by the ARC (Australia) - KOSEF
(Korea) joint research.

References
1. Y. Matsui, A. Yuasa, F. Li, and F. Li, J. Environ. Eng., 124, 11, (1998) p. 1099.
2. Maccormick A. B., The application of microfiltration in water and wastewater
treatment, In Modern Tech. in Water and Wastewater Treatment, CSIRO Publishing,
Australia, 1995, p. 45.
3. H. Moon, H. C. Park, and C. Tien, Chem. Eng. Sci., 46, 1. (1991) p. 23.
4. Chi Tien, Adsorption Calculations and Modeling, Butterworth-Heinemann, USA,
1994, p. 167.

431
 SURFACE CHARACTERISTICSOF MCM-41 ON Cr(1ll) AND Cr(VI)
ADSORPTION BEHAVIORS '

S. J. PARK, B. R.JUN, AND M.HAN

tAdvanced Materials Division, Korea Research Insfihrte of Chemical Technology,P. 0.Box 107,
Yusong, Taejon 305-600, Korea
te-maii :psjin Wict.re.kr

In this work, the MCM-41 mesoporous materials were prepared from hydrothermal synthesis using gel
mixture of sodium silica solution as silica source and cetyltrimethylanunoniumchloride (CTAMCI) as
template. The surface and the structure properties of MCM-41 were determined by pH, acid/base
value, and XRD measurements. NZadsorption isotherm characteristics, including the specific surface
area ( S B ~ ) ,total pore volume (VT),and average pore diameter (Rp), were determined by BET and
Boer's I-plot methods. Also, the adsorption amounts of Cr(lll) and Cr(V1) ion on MCM-41 were
measured using ICP-AES.As a result, the surface of MCM-41 was acidic in naturc. It was found that
the presence Si-0-Si and Si-OH groups that easily were dehydrogenated by hydrogen bonding
between hydroxyl groups on MCM-41 surface in aqueous solution. Also, the adsorption amount of
Cr( 111) on MCM-41 had higher than that of Cr(VI). This result was probably explained that surface
hydmxyl group of MCM-41 enhanced adsorption of Cr( 111 ) cations, as it suppressed the adsorption of
Cr(Vi) anions.

1 Introduction

Problems associated with the removal of heavy metals in effluent waters from several
different industries have become controversial issues in the world. Especially, chromium is
one of the undesirable heavy metals because it affects human physiology, accumulates in
the food chain, and causes several ailments [ 1,2].
The methods employed for the removal of Cr ions from wastewater include
precipitation [3], ion-exchange [4], membrane [ 5 ] , and solvent extraction [6]. However,
these treatment methods have not widely practiced due to their high cost and for low
feasibility small-scale industries.
Two classes of materials that are used extensively as heterogeneous catalysts and
adsorption media are microporous (pore diameters 5 -20 A ) and mesoporous (-20 - 500
A ) inorganic solids [7]. In 1992, the discovery of a new family of mesoporous molecular
sieves designated M41S had been reported by researchers at the Mobil Coporation [S]. The
MCM-41 material possesses a uniform hexagonal array of linear channels constructed with
a silica matrix like a honeycomb. These materials have very high surface area, ordered pore
structure, and extremely narrow pore sue distribution. Pore size can be controlled by
intercalation of layered silicates with a surfactant species [9]. Because of their large areas
and well-defined pore size and pore shape, these materials have great potential in
environmental and industrial processes.
Brown et ul. [ 101 reported that selective Hg2+adsorbent could be synthesized by using
silica, also Dai et ul. [l 11 synthesized selective Cu2' adsorbents the adsorption of bents
using MCM-type mesoporous silica. However, little research has been reported for noble
metal ions using mesoporous silica.
The main objective of this work was to investigate the surface and adsorption
properties of the MCM-41 by using pH, acidhase value, and XRD and BET isotherms.
Also, it was assessed how these properties affect the adsorption of Cr( II1) and Cr(V1).

432
2 Experimental

2. I Synthesis of MCM-41
A MCM-41 has been synthesized following a procedure of Ryoo and coworkers [ 121.
The fust procedure used colloidal silica (Ludox HS40,39.5wt???SO*, 0.4 wt?? Na20, and
60.1 wt?? H20,du Pont) as the silica source and 25% cetyltrimethylammonium chloride
(CTMACI, Aldrich) solution as surfactant. A solution of sodium silicate was prepared by
combining 1.00 M NaOH solution with colloidal silica. The sodium silicate solution was
dropwise added to a polypropylene bottle containing a mixture of 28 wt?? aqueous NH3
solution and CTMACI solution with vigorous magnetical stirring at room temperature. This
mixture was heated for 4 d at 373 K. The pH of the reaction mixture was adjusted to 10.2
by dropwise addition of 30 wt?? acetic acid with vigorous stirring. Subsequently, the
precipitated product, MCM-4 1 with CTMA template, was filtered, washed with distilled
water, and dried in an oven at 370 K. The product was calcined in air under static
conditions using a muffle furnace at 770 K for 24 h.

2.2 Characterization
The pH of MCM-41 was measured with ASTM D 3838. About 1.Og of dry MCM-41
was added to 20 ml of distilled water, and the suspension was shaken overnight to reach
equilibrium. Then the sample was filtered, and pH of the solution was measured. The
surface fbnctional groups of the samples were determined by Boehm's titration method
[ 131. In the case of acid value, about 1.Og of the sample was added to 100 ml of 0. I N
NaOH solution and the mixture was shaken for 24 h. Then the solution was filtered through
a membrane filter and titrated with 0. IN HCl. Likewise, the base value was determined by
converse titration.
XRD patterns were obtained at room temperature, using a Rigaku Model D M X Ill
B (CuK, X-ray source). All samples were scanned under the same conditions (28 = 1.2 -
10"). N2 isotherm was measured by using an ASAP 2400 (Micromeritics) at 77 K. Prior to
each analysis, the samples were outgassed at 298 K for 6 h to obtain a residual pressure of
less than I O 3 torr.

2.3 Adsorption of Chromium

The adsorption efficiency of Cr ( 111) and Cr (VI) on the MCM-41 sample was
obtained by using stoppered flasks containing 0.lg of MCM-41 in 100 ml of aqueous
-
solution of chromium chloride (CrC13 6H20) and sodium chromate (Na2Cr04* 4H20),
-
respectively. Also, the initial pH of the solutions (CrC13 6H20 and Na2Cr04 4H@) -
were adjusted to about 3 and 5 by using 0.1 N HCI and 0.1 N NaOH. The amount of
chromium adsorbed was determined by ICP-AES.

3 Results and Discussion

It is well known that surface properties should play an important role in the process of
gasAiquid adsorption on MCM-4 1.

433
 Table I summarizes the experimental surface properties of the MCM-41 determined
by pH and Boehmstitration. As a result, it could be seen that the surface of MCM-4 1 was
acidic in nature. It was found that the presence Si-0-Si and Si-OH groups that easily were
dehydrogenated by hydrogen bonding between hydroxyl groups on surface MCM-41 in
aqueous solution.
Table 1. The pH and Acid-Base Values of MCM-41

XRD pattern of the MCM-41 sample synthesized in the present work was presented in
Figure 1. The XRD pattern was comprised of one very intense line, three weak lines, which
could be indexed to (loo), (1 lo), (200), and (210) diffraction lines characteristic of a
hexagonal structure of MCM-4 1, respectively. This feature was generally attributed to the
formation of Si-0-Si bonds by condensation of Si-OH groups present in the as-synthesized
material. The hexagonal unit cell (a,,) could be calculated by Eq. (1)

a,, = 2d,,J3

As a result, hexagonal unit cell (a,,) value, containing pore walls, WBS about 48.3 A .

I 110 24.2
2w 20.9

I , . , . . . , . . (
2 4 6 8 10
28
Figure 1. Typical X-ray diffraction patterns of synthesizedMCM-41 sample.

The pore structure parameters were included in Table 2. It was found that the BETS
surface area and the pore volume were 803 m2/g and 0.83 cm3/g, respectively. Also, the
BJH pore size distribution shows MCM-41 material with a quite narrow pore diameter
distribution centered around 25.8A. Thus, the framework walls are 22.5A thick.
Adsorption behavior of Cr(lll) and Cr(VI) ions from aqueous solutions of chromium
chloride in the concentration of 50 mg/L on the samples are shown in Figure 2. All
adsorption isotherms were also carried out by adjusting the initial pH of solution by using
0.1N HCI and 0.1N NaOH and without adding any buffer to control the pH constantly
[ 14,151.

434
 Table 2. NJ77K Adsorption Data of the MCM-41

MCM-41 803 1414 0.8264 25.8 22.5

a : Specific surface area from BET-method by Nd77K adsorptionb: BET constant

' : Total pore volumed: BJH Average pore diameter' : Pore wall thickness ( A )

The initial adsorption rate increased rapidly. Also, it was seen that adsorption amount
of Cr( 111) on MCM-41 had higher than that of Cr(VI). This was due to the presence Si-0-Si
and Si-OH groups that were easily dehydrogenatedby hydrogen bonding between hydroxyl
groups on surface MCM-41 in aqueous solution. Thus, This result was probably explained
that surface hydroxyl group of MCM-41 enhanced adsorption of Cr(lll) cations, as it
suppressed the adsorption of Cr(VI) anions.

0
0 D
Time (mill)
Figure 2. Adsorption of Cr(II1) and Cr(VI) on MCM-41 as a knction of contact time.

4 Conclusions

In this work, the effects of the surface and adsorption properties of the MCM-41
sample on chromium adsorption were investigated. As a result, It was found that the
presence Si-0-Si and Si-OH groups that were easily dehydrogenated by hydrogen bonding
between hydroxyl groups on surface MCM-41 in aqueous solution. Also, adsorption
amount of Cr( 111) on MCM-41 had higher than that of Cr( VI). This result was probably
explained that surface hydroxyl group of MCM-41 enhanced adsorption of Cr( 111) cations,
as it suppressed the adsorption of Cr(Vl) anions.

435
 References

1. R. C. Bansal, J. B. Donnet, and F. Stoeckli, Active carbon, Marcel Dekker, New York
(1988).
2. S. J. Park and Y. S. Jang, Pore structure and surface properties of chemically modified
activated carbons for adsorption mechanism and rate of Cr(VI), J. Colloid Interjiuce
Sci. 249 (2002) 458-463.
3. A. T. de Matos, M. P. F. Fontes, L. M. da Costa, and M.Z. Martinez, Mobility of
heavy metals as related to soil chemical and mineralogical characteristics of Brazilian
soils, Environl. Pollur. 111 (2001) pp. 429-435.
4. J. T. Matheickal and Q. Yu, Biosorption of lead (11) and copper (11) from aqueous
solutions by pre-treated biomass of Australian marine algae, Bioresour. Technol. 69
(1999) pp. 223-229.
5. G. C. Steenkamp, H. W. J. P. Neomagus, H. M. Krieg, and K. Keizer, Centrifugal
casting of ceramic membrane tubes and the coating with chitosan, Sep. PuriJ:Technol.
25 (2001) pp. 407-413.
6. D. Dong, X. Hua, Y. Li, and Z. Li, Lead adsorption to metal oxides and organic
material of hshwater surface coating determined using a novel selective extraction
method, Environl. Pollur. 119 (2002) pp. 3 17-32 1.
7. K. S. W. Sing, D. H. Everett, R. A. W. Haul, L. Moscou, R. A. Pierotti, J. Rouquol,
and T. Siemieniewska, Reporting physisoption data for gadsolid systems with special
reference to the determination of surface area and porosity, Pure Appl. Chem. 57
(1985) pp. 603-619.
8. C. T. Kresge, M. E. Leonowics, W.J. Roth, J. C. Vartuli, and J. S. Beck, Ordered
mesoporous molecular sieves synthesized by a liquid crystal template mechanism,
Nufure 359 (1992) pp. 7 10-712.
9. R. Ryoo, C. H. KO, and I. S. Park, Synthesis of highly ordered MCM-41 by
micelle-packing control with mixed surfactants, Chem. Commun. (1 999) pp. 14 13
-1414.
10. J. Brown, L. Mercier, and T. J. Thomas, Selective adsorption of Hg2+ by
thiol-fhctionalized nanoporous silica, Chem. Commun. (1 999) pp. 69-70.
11. S. Dai, M. C. Burleigh, Y. H. Ju, H. J. Gao, J. S. Lin, S. J. Lin, S. J. Pennycook, C. E.
Barnes, and Z. L. Xue, Hierarchically imprinted sorbents for the separation of metal
ions, J. Am. Chem. SOC.112 (2000) pp. 992-993.
12. J. M. Kim, J. H. Kwak, S. N. Jun, and R. Ryoo, Ion exchange and thermal stability of
MCM-4 1, J. Phys. Chem. 199 (1 995) pp. 16742- 16747.
13. H. P. Boehm, Chemical identification of surface group, Adv. Curul. 16 (1966) pp.
179-225
14. S. J. Park and W. Y. Jung, Adsorption behavior of chromium (111) and (VI) on
electroless Cu-plated activated carbon fibers, J. Colloid Inrerjiuce Sci. 243 (2001) pp.
3 16-320.
15. S. J. Park and J. S. Kim,Anodic surface treatment on activated carbons for removal of
chromium (VI), J. Colloid Interjiuce Sci. 239 (2001) pp. 380-384.

436
INFLUENCE OF ANODIC OXIDATION OF ACTIVATED CARBON FIBERS ON
THE REMOVAL OF HEAVY METAL IN AQUEOUS SOLUTION

S. J. PARK, Y.M. KIM, AND J. R. LEE

Advanced Materials Division, Korea Research Institute of Chemical Technology,
P. 0.Box 107, Yusong, Taejon 305-600.Korea
E-mail :psjin @krict.re.kr

In this work, the effect of anodic oxidation treatments on activated carbon fibers (ACFs) was studied
in the context of Cr(VI). Cu( It), and Ni( II ) ion adsorption behaviors. Ten wt?! phosphoric acid and
sodium hydroxide were used for acidic and basic electrolytes, respectively. Surface properties of the
ACFs were determined by XPS. The specific surface area and the pore structure were evaluated from
nitrogen adsorption data at 77 K. The heavy metal adsorption rates of ACFs were measured by using
a UV spectrometer and ICP. As a result, the anodic treatments led to an increase in the amount of
total acidity by an increase of acidic functional groups such as carboxyl, lactone, and phenol, in spite
of a decrease in specific surface area, due to the pore blocking by increased acidic functional groups.
The adsorption amount of anodic oxidation treated ACFs was increased and the adsorption capacity
improved in order of Cr(VI) > Cu( II ) > Ni( It ). It was probably accounted that the surface functional
properties on ACFs had a main effect as compared to the sbucture properties.

1 Introduction

In recent years, activated carbons fibers (ACFs) because of their high surface area,
microporous character, and the chemical nature of their surface have been considered
potential adsorbents for the removal of heavy metals from industrial wastewater [I-31.
The properties of ACFs are determined by their microstructure, it is therefore important to
investigate the microstructure of ACFs in terms of specific surface area, micropore
volume, pore size distributions, surface chemistry and so on. Also, the adsorption
properties of carbonaceous adsorbents are dependent on not only the porous structure but
also the surface chemistry [3,4].
It is known that surface oxides on carbon surfaces are formed by means of thermal
treatment at high temperature, ozone treatment, and liquid treatment of the chemical
aqueous solution [5]. Also, anodic oxidation of fibers in electrolytes can produce a variety
of chemical and physical changes in the fiber surfaces [6].The main aim of oxidation of
carbon surfaces is obtaining a more hydrophilic surface structure with a relatively large
number of oxygen-containing surface groups [7]. The existence of surface functional
groups on carbons such as carboxylic, phenolic, lactonic and acid anhydrides has been
postulated as constituting the source of surface acidity [8,9]. Also, these surface groups
make the carbon surface hydrophilic and enhance its ion exchange capacity so that these
carbons are potential adsorbents for the removal of metal ions from industrial and
domestic wastewater.
In this work, ACFs are modified by anodic oxidation treatments with acid (H3P04)
and base (NI&OH) to obtain oxygen-containing groups. Surface hctional groups on the
carbon surfaces are investigated by XPS and acid-base values. Nz adsorption isotherm
characteristics, including the specific surface area, micropore volume, and total pore
volume, are determined by BET. And the adsorption amounts of Cr(VI), Cu( II ), and
Ni( II) ions on ACFs are using UV/vis spectrophotometer (UV) and inductively coupled
plasma atomic emission spectrometer(ICP-AES).

437
2 Experimental

2. I Materials and Anodic oxidation

The pitch-based carbon fibers were prepared by Kureha (Japan) and activated by
steam diluted with nitrogen at 900 c . These ACFs were washed with deionized water and
dried overnight at 80c (pure-ACFs). The ACFs were subjected to electrolytic reaction in
an aqueous solution of 10 wf??H3P04 and NH40H, whereby negative ions were attracted
to the surface of the ACFs acting as an anode, thereby modifying the ACF surfaces. The
ACFs prepared from these procedures were named as A-ACFs and B-ACFs, respectively.
A cathode graphite plate was also submerged in the electrolyte solution. And the
conditions of the surface treatment were processed in an electro-bath at 7 A for 10 min.

2.2 Surface Functional Group

The surface properties of ACFs were measured by XPS (X-ray photoelectron
spectroscopy, ESCA LAB MK II ). The XPS was collected using a MgK X-ray source
(1253.6 eV). The pressure inside the chamber was held below 1 X 1O 9 torr.
The surface functional groups containing oxygen were determined according to
Boehm titration [8]. One gram of ACFs was placed in 100 ml of the following solutions:
sodium hydroxide, sodium carbonate, and sodium bicarbonate. The solutions.were sealed
and shaken for 24 h and then filtered. Twenty milliliters of the filtrate were pipetted and
titrated with 0.1 N HCI. The number of acidic sites was determined under the assumptions
that NaOH neutralizes carboxylic, lactonic, and phenolic groups; that Na2C03neutralizes
carboxylic and lactonic groups; and that NaHC03neutralizes only carboxylic groups.

2.3 Pore Structure of ACFs

Specific surface area was calculated fiom the Brunauer-Emmett-Teller (BET)
equation for N2 adsorption at 77 K (Micromeritics, ASAP 2010) [lo]. The t-method of de
Boer was used to determine the micropore volume [l I]. The pore size distribution curves
of micropores were obtained by the Horvath-Kawazoe(H-K) method [121.

2.4 Ahorption Properties of Cr( v,Cu(lT), and Ni(lT)

0.05 g of each of the activated carbon fibers were placed in contact with 150 ml
-
solutions of 20 ppm concentrations of sodium chromate (Na2Cr04 4H20), cupric
chloride (CuC12 * 2H20), and nickel chloride (NiC12 6H20) for the adsorption of Cr(VI),
Cu( 11), and Ni( II) ions. The initial pH of the Cr(VI), Cu( II), and Ni( II) solutions was
adjusted with 0.1 N HCl and NaOH. The bottles were sealed with parafin film and then
shaken on a time sequence (10,20,30,40,50,90, and 180 min) at 25 c at a frequency of
100 strokedmin using a shaking bath. At the end of the reaction period, each reaction
mixture was filtered to separate the Supernatant and activated carbon fibers. The adsorbed
amount of Cr(VI) was measured by using the pink color complex developed between
diphenyl carbohydrize and chromium ions in an acidified solution with a UV/vis
spectrophotometer(UV S-2100, Scinco) at an absorbance of 540 nm [3].

430
 1200 900 600 uw)

Binding Energy (eV)

Figure 1. XPS survey o f activated carbon fibers.

The residual copper and nickel ions were analyzed by ICP-AES (Jovin-Yvon Ultima-
C). The amount of adsorbed heavy metal ions was obtained by calculating the difference
of each concentration of heavy metal ions before and after adsorptions.

3 Results and Discussion

3. I Su$ace Characteristics and ACFs Pore Structure

Figure 1 shows the XPS survey scan spectra of the nontreated and the anodic
oxidation treated ACFs. The CISand Ols peaks of ACFs are appeared in 284.6 and 532.8
eV, respectively [13]. As a result, the relative intensity of Ols peak is increased by the
anodic oxidation treatment. This is clearly attributed to the increasing of oxygen groups
on the ACFs surfaces.
Boehm titration results are presented in Table 1. As seen in Table 1, oxygen
functional groups of surface-treated ACFs are increased. These results can be explained
that the anodic oxidation treatment of the ACFs surfaces produces various oxygen
functional groups, i.e., carboxylic, lactonic, and phenolic groups.
Table 2 shows microstructural properties, including specific surface area, micropore
volume, and total pore volume of the ACFs. The specific surface area of the pure-ACFs,
A-ACFs, and B-ACFs are 1944, 1430, and 1937 m2g-', respectively. That is, the BET'S
specific surface area is decreased by 26% for A-ACFs compared to the pure-ACFs, but B-
ACFs have no significant change. Also, total pore volume and micropore volume of
surface-treated ACFs are decreased. This is due to the increase of oxide functional
groups, which are attributed to the block of the micropores.

Table 1. Results of Boehm Titration.

Carboxylic (meq/g) Lactonic (meq/g) Phenolic (meq/g)
Pure-ACFs 3 40 70
A-ACFs 40 230 140
B-ACFs 4 80 70

439
 Table 2. Textural Characteristicsof Anodic Oxidation Treated ACFs.

Pure-ACFs A-ACFs B-ACFS

BET surface area [m2-g-'1 1944 1430 1937
Micropore volume [cm3. g-'1 0.787 0.243 0.343
Total pore volume [cm3.g-'1 1.027 0.560 0.701

3.2 The Adsorption of Heavy Metal Ion

The results of percentage removal of metal ions with contact time for the initial
concentration of metal ions are shown in Figure 2. It is observed that the initial adsorption
rate increases rapidly. The percentage adsorption of metal ions increases with increasing
the agitation contact time and the percentage adsorption of metal ions of surface-treated
ACFs is always superior to that of pure-ACFs. This is because at higher dose of adsorbent
due to the increased surface functional groups, more adsorption sites are available,
causing higher removal of metal ions. Also, A-ACFs show lower percentage adsorption
of metal ions than B-ACFs. It seems to be due to the pore blocking by surface oxide
groups, resulting in decreasing the specific surface area.

0 40 80 1 2 0 1 6 0 : 40 80 IP la 0
Tl(min)

Figure 2. Adsorption of metal ions onto

activated carbon fibers as a
function of contact time.
4 Conclusions

In this work, the anodized ACFs were studied in the adsorption characteristics in
terms of the microstructures and surface functional groups. In the results of XPS, acid-
base values, and BET, the specific slirface area of surface-treated ACFs was decreased,
whereas, oxygen-containing functional groups of surface-treated ACFs were increased.
As expected, the increased surface fhctional groups led to an increase of the adsorption
of heavy metal ions. In case of A-ACFs, the adsorption of heavy metal ions was increased
by an increasing of oxygen-containingfunctional groups, in spite of a decrease of specific
surface area. In conclusion, the surface functional properties of ACFs made a major
important role in increasing the heavy metal adsorption in aqueous solution.

References

1. M. Suzuki, Activated carbon fiber: Fundamentals and applications, Carbon 32 (1994)

pp. 577-586.
2. J. B. DOMet, R. Y.Qin, S. J. Park, Ryu S. K., and Rhee B. S., Scanning tunnelling
microscopy study of activated carbon fibers, J. Muter. Sci. 28 (1993) pp. 2950-2954.
3. S. J. Park,B. J. Park, and S. K. Ryu, Electrochemical treatment on activated carbon
fibers for increasing the amount and rate of Cr(VI) adsorption, Carbon 37 (1999) pp.
1223- 1226.
4. Z. Ryu, H. Rong, J. Zheng, M. Wang, and B. Zhang, Microstructure and chemical
analysis of PAN-based activated carbon fibers prepared by different activation
methods, Carbon 40 (2002) pp. 1 144-1147.
5. J. B. DOMet and R. C. Bansal, Active Carbon, 2nd ed., (Marcel Dekker, New York,
1990) pp. 145-204.
6. S.J. Park and J. S. Kim, Anodic surface treatment on activated carbons for removal
of chromium(VI), J. Colloid Interface Sci. 239 (2001) pp. 380-384.
7. J. W. Shim, S. J. Park and S. K. Ryu, Effect of modification with HN03 and NaOH
on metal adsorption by pitch-based activated carbon fibers, Carbon 39 (2001) pp.
1635-1642.
8. H. P. Boehm, Chemical identification of surface groups. In Advances in Catalysis,
ed. by Weisz P. B. 16 (1966) pp. 179-287.
9. S. J. Park and Y. S. Jang, Pore structure and surface properties oftchemically
modified activated carbons for adsorption mechanism and rate of Cr(VI), J. Colloid
Interface Sci. 249 (2002) pp. 458-463.
10. S. Brunauer, P. H. Emmett, and E. Teller, Adsorption of gases in multimolecular
layers, J. Am. Chem. Soc., 60 (1 938) pp. 309-319.
11. J. H. de Boer, B. G. Linsen, T. Plas, and G. J. Van Zonder, Studies on pore systems
in catalysts, VII. Description of the pore dimensions of carbon blacks by the t-method,
A Card. 4 (1 965) pp. 649-653.
12. G. Horvath and K. Kawazoe, Method for the calculation of effective pore size
distribution is molecular sieve carbon, J. Chem. Eng. Jpn. 16 (1983) pp. 470-475.
13. S. J. Park and W. Y. J u g , Effect of KOH activation on the formation of oxygen
structure in activated carbons synthesized from polymeric precursor, J. Colloid
Interjiace Sci. 250 (2002) pp. 93-98.

441
 KINETICS AND DIFFUSION PROCESSES FOR REACTIVE DYE
ADSORPTION BY DOLOMITE

STEPHEN J ALLEN, G A V M M WALKER, LMDA HANSEN AND JULIE-ANNE HANNA

School of Chemical Engineering, Queens Universiw Belfast,Be&st BT9 5AG. Northern Ireland.
E-mail: [email protected]

A wastewater treatment technique has been investigated, for reactive dye removal, in batch kinetic
systems. Results indicate that charred dolomite has the potential to act as an adsorbent for the
removal of reactive dye from aqueous solution. The effect of initial dye concentration, adsorbent
mass: liquid volume ratio, and agitation speed on removal have been determined with the
experimental data mathematically described using empirical external mass transfer and intra-particle
diffusion models. The experimental data show conformity with an adsorption process, with the
removal rate heavily dependent on both external mass transfer and intra-particle diffusion.

1 Introduction

Dolomite is a sedimentary rock-forming material composed of minerals with a trigonal

bar 3 symmetry. The general formula of this group is AB (C03) t , where A can be
calcium, barium and/or strontium and B can be iron, magnesium, zinc and/or manganese.
The ordered layering of different or nonequivalent ions causes a loss of the two-fold
rotational axes and mirror planes that are present in the calcite group structure [3]. Here,
charred dolomite was used for the removal of reactive red dye fiom aqueous solution.
Dolomite treatment has been reported as a removal method for metal ions [9]. In the
thermal processing (calcining) the MgC03component decomposes at around 800OC. The
product of partial decomposition contains CaC03 (calcite) and MgO and shows an
increase in specific surface area and pore volume [1 I].

2 Methods

The dolomite used has a typical chemical composition of 44%MgC03and 53% CaC03.
The raw dolomite was charred in a furnace at 850C for a period of 6- 18 hours producing
a CaC03/Mg0 porous structure. Specific surface area was measured using BET nitrogen
adsorption employing a Sorptomatic 1900.
A reactive dye, Levafix Brilliant Red E4BA was used. The research investigates the
effect of: mass-volume ratio, initial dye concentration and agitation speed. For variation
in mass-volume ratio, the mass of the dolomite ranged from 1.06g to 4.25g i 0.002g in
1.7 dm3 of solution (dye concentration of 400 mg dmj, impeller speed 400rpm). For
variation in concentration, the initial concentration of dye solution ranged from 400 mg
dm-3to 1000 mg dmm3(mass volume ratio of 17g per 1.7dm3of solution, impeller speed
400rpm). For variation in yitation rate, the speed ranged from 100-5OOrpm (mass
volume ratio of 17g per 1.7 dm of solution, dye concentration of 400 mg dm).

442
3 Results and discussion

An empirical model was used to calculate the external mass transfer coefficient, kF [4,12].
The change in fluid phase concentration with time Ct is related to kF ,concentration at the
surface C, and specific surface S, by:
dCt = -k, .S, .(C,
- - C,)
dt
A solution to the model follows: since at t = 0, CS+ 0, and Ct + CO

Hence at t = 0 a plot of CdC0 will yield a slope of -kFSs. The plot of CJC0 against time
can be curve fitted and the slope as t + 0 can be calculated fiom the curve fit relationship
and not just the first two data points.
The intra particle diffusion model (single resistance) assumes external mass transfer
is significant only at initial stages. The internal diffusion parameter, k, is calculated from
the plot of dye adsorbed qt versus square root of time [S].
qr = k(to.) (3)
The BET analysis on dolomite and charred dolomite is shown in Table I. There is a
rise in surface area on charring of the dolomite due to the calcining process which
effectively creates a porous structure.

Table 1. BET nitrogen adsorption analysis. Dolomite charring temperature = 800C.

Dolomite Sample SunFace Area (m2 g-)

raw dolomite 0.7
6 hr charred dolomite 19.5
12 hr charred dolomite 23.0
18 hr charred dolomite 36.0

The effect of initial dye concentration on rate of adsorption is shown in Figure 1.

Removal rate is rapid as t + 0 with between 50% and 90% occurring within the first
minute after which there is a decrease in the rate. The rate of removal increased with
increase in mass to volume ratio and with agitation rate. The effect of agitation rate on the
rate of adsorption suggests that the removal rate is rapid as t 3 0 with between 70% and
90% removal occurring within the first minute. kf values were determined using the
empirical model outlined above for t + 0 (using equation 2) and are illustrated in table 2
in which M,is the mass of dolomite per unit volume of solution.
Results suggest that k, is independent of mass to volume ratio, indicating that the use
of the empirical adsorption model is valid. For variation in initial dye concentration,
there appears to be an increase in kf with decrease in C,. The effect of increasing the
initial dye concentration serves to decrease the initial rate of adsorption onto the sorbent
surface. Here, it would seem that interactions between solute molecules in solution and

443
the relative increase in competition for available adsorption sites override the increase in
driving force for the adsorption urocess at elevated dye concentrations.
0.0

0.8

0.7

0,s
N
O
-0.5

0.4

0.3
0.2

0.1
A

0
.. .
0

*
0

.
0
: It
0

Figure 1. Concentration decay curves for variation in initial solution concentration

Table 2. Effect of process variable on external mass transfer coefficient.

Concentration kf (cm s) Agitation kf (cm s-l) M,(g dm) kf (cm s-l)

(mgdm) (rpm)
1000 0.0260 100 0.045 1 0.63 0.1425
800 0.0352 200 0.0507 1.25 0.1535
600 0.0555 300 0.0679 2.50 0.1249
400 0.0834 400 0.0834

kfin this work is higher that that calculated elsewhere using other adsorbents, and
would indicate that mass transfer in the reactive dye - charred dolomite system is very
high [I]. The magnitude of kffor variation in agitation rate indicates that kfincreases with
increase in impeller speed. Since kf = D/6, where 6 is the effective thickness of the
Nernst layer which is smaller the greater degree of agitation, then a linear plot of In kf
versus In (rpm) would be expected [6]. The correlation in Figure 2 is described by the
equation ks = 3.58 x 10-2(rpm)0.14 .Rate of internal mass transfer is in most cases the
rate-determining step, more notably in the case of large molecular weight dyes [lo]. The
single resistance model used assumes diffusion occurs in the pore structure of the
adsorbent and that this diffusion is described by Ficks second law. The intra particle
diffusion parameter, k, is defined as the gradient of the solid phase concentration qs,
versus the square root of time. This line can often be linearised into regions representing
different stages of mass transfer of adsorbates into adsorbents [7]. The initial stage
represents external mass transfer, with the following stages representing intraparticle
diffusion in the macro, meso and micropore structure of the adsorbent. In figure 3, after
the fmt minutes where external mass transfer is the dominant process, there is a single
linear relationship between qs and to.. The single linear stage (table 3) would indicate that
intra particle diffusion occurs in pores of a similar magnitude. The results of the BET
analysis indicate a low surface area and pore volume, hence much of the internal structure
of the charred dolomite consists of macropores, with little or no branched pores.
Table 3. Effect of process variable on intra-particle diffusion parameter.

Concentration (mgdm) k (mg rni~~-.~) M,(g dm) k (mg minaO)

1000 5.4 1 0.63 0.98
800 3.77 1.25 4.76
600 1.93 2.50 8.25
400 0.98

Results indicate the rate of intra particle diffusion is dependent upon initial dye
solution concentration with an increase in C, giving a corresponding increase in
magnitude of k. In the case of intra particle difision being the only rate determining step
it was found that k varied with the square root of the initial concentrations used [2, 121.
The effect of an increase in sorbent mass is an increase in the surface area for adsorption.
The trend in k was possibly due to the effect of the increase in surface area increasing the
value of k which is a rate parameter and not a diffusion coefficient. Results from other
investigations into the removal of metal ions indicated that the removal from solution was
related to the volumes of precipitation of their hydroxides formed in the presence of
dolomite [8,9]. In this investigation there was no evidence of a precipitate.

4.0 5.0 h(rpm) 6.0 7.0

-2 ,

3.5 J I

Figure 2. Correlation between external mass transfer coefficient and agitation rate

45

rmfs
0630

25

Figure 3. Solid concentration versus square root of time for different initial concentration

Conclusions

Batch kinetic data on the removal of reactive dye from solution using thermally charred
dolomite have been well described by empirical external mass transfer and intra-particle
diffusion models. It was found that external mass transfer and intra-particle diffusionhad
rate limiting effects on the removal process which were attributed to the relatively simple

445
macropore structure of the charred dolomite particles. These analyses, plus a visual
inspection would appear to indicate that adsorption is the likely removal mechanism in
the process.

4 References

1. Allen S.J., McKay G and Khader K.Y.H., Intraparticlediffision of a basic dye during
adsorption onto sphagnum peat. Em Poll 56 (1989) pp. 39-50.
2. Allen S.J.,Brown P., McKay G. and Flynn O., An evaluation of single resistance
transfer models in the sorption of metal ions by peat. J Chem Technol BiotechnolS4
(1 992) pp. 27 1-276.
3. Boynton R.S., Chemistry and Technology of Lime and Limestone. (Intersci, New
York, 1967).
4. Furusawa T. and Smith J.M., Fluid-particle and intraparticle mass transport rates in
slunies. Ind Eng Chem Fundum 12 (1973) pp. 197-203
5. Furusawa T. and Smith J.M., Intraparticle mass transport in slurries by dynamic
adsorption studies. AIChEJ 20 (1 974) pp. 88-93
6. McKay G. and McConvey I.F., The external mass transfer of basic and acidic dyes
by wood. J Chem Technol Biotechnol31 (1 98 1) pp. 401-408
7. McKay G., Blair H.S. and Gardner J., The adsorption of dyes in chitin 11.
intraparticlediffusion processes. J Appl Polymer Sci 28 (1983) pp. 1767- 1778
8. Staszczuk P., Stefaniak E. and Dobrowolski R., Characterisation of thermally treated
dolomite. Powder Technol 92 (1997) 257
9. Stefaniak E., Dobrowolski R. and Staszczuk P., On the adsorption of chromium V1
ion dolomite and dolomitic adsorbents. Adrorption Sci and Technol 18 (2000) pp.
107-115.
10. Walker G.M.and Weatherley L.R.,Adsorption of acid dyes from aqueous solution-
the effect of adsorbent pore size distribution and dye aggregation. Chem Eng J
(2001) (accepted).
11. Walker G.M., Hanna J-A and Allen S.J., Investigations into the adsorption of reactive
dye, phosphate and nitrate by dolomite. (Fundamentals of Adsorption 7, 2001,
Nagasaki Japan.
12. Weber W.J. and Moms J.C., Kinetics of adsorption on carbon fiom solution. JSun
Eng Div A X E (1 963) April pp.3 1-58.

446
 PERMEATE FLUX BEHAVIOR DURING MICROFILTRATION OF
PROTEIN-ADSORBED MICROSPHERES IN STIRRED CELL

YONGSU CHANG, SUNG-WOOK CHOI, TAI-GYULEE, SEUNGJOO HAAM

AND WOO-SIK KIM
Depart of Comical Engineering, College of Engineering,
Yonsei Universiw, Seoul 120-749,Korea
E-mail:[email protected]

A study on the variation of permeate flux was performed in a stirred cell charged with
carboxylated microspheres (CM-I, 2,3,4)to investigate the effect of the surface charge density (Nc:
0.45, 5.94, 9.14, and 10.25), the stirrer speed (300,400 and 600 rpm), and protein concentration
(0.1, 0.25, and 0.5 g/L) under constant pressure. It was found that the permeate flux of the bared
microspheres was dependent on the surface charge density and the stirrer speed. High permeate flux
was obtained in the condition of high surface charge density and high stirrer speed due to the
increase of the electrostatic repulsive force between microspheres.
The permeate flux variation and continuous separation of mixed solution of BSA and BHb were
also investigated using a stirred cell charged with low carboxylated microspheres (CM-I) at different
pH of 6.8 (near the PI of BHb) and 4.5 (near the PI of BSA). In the single component study, the
permeate flux was increased after the step input of BSA solution while it was not increased after the
step input of BHb solution. In the binary study, the permeate flux was increased at pH 4.5 not at pH
6.8 because BSA-adsorbed microspheres occurred to the agglutination. As the low carboxylated
microspheres were being covered with BSA at pH 4.5, microspheres losing their statbility were
agglutinated each other and therefixe the effective particle size was enlarged.

1 Introduction

Polymeric materials are widely utilized in biomedical and biochemical applications

such as solid-phase immunoassays, blood cell separations, and column packing reagents
and, thus protein adsorption on polymeric surfaces has become a center of attention.
Particularly, microspheres, defmed as fine polymer particles having diameters in the
range of 0.1 to several microns, can be used as functional tools by themselves or by
coupling with viocompounds. To apply these microspheres to the liquid chromatographic
separation, the particles must be large enough to achieve high glow rates and also be
porous to achieve high load capacities. To achieve this, we contrived a new separation
system using a stirred cell with a membrane filter. The stirred cell plays the role of a
chromatographic column as well as a continuous stirred tank reactor (CSTR). A
microfiltration membrane filter lets the protein pass through whilst returning microspheres.
Direct use of uncoupled microspheres in the form of dispersion although they are
submicron in size. For high selectivity without ligands, the interaction relationships
between protein and microspheres should be completely elucidated. The aim of this work
was to investigate the effect of surface charge density of microspheres on the permeate
flux and the flux behavior of protein adsorbed microspheres and the optimal conditions of
separation.

447
2 Experimental

2.1 Materials
2.1.1 Proteins
Bovine serum albumin (BSA) and bovine hemoglobin (BHb) were purchased from
Sigma Chemical Co., USA, catalog number A-7906 and H-2500, and no finther
purification was performed.
2.1.2 Monomers
Styrene (s) and methacrylic acid (MAA) were obtained from Junsei Chemical Co.,
Japan, are stabilized by monoethyl ether hydroquinone (MEHQ), therefore, an inhibitor
remover column (Aldrich Chem. Co., USA, catalog number 306312) was used for
removing MEHQ.

2.2. Preparation of microspheres

All microspheres were prepared using soap-free emulsion polymerization.
Carboxylated microspheres were prepared by batch copolymerization. Batch
copolymerization was performed with main monomer, S, lwt?! of co-monomer, MAA,
and 0.4WtOh of initiator, KPS. All weight fiactions were based on the amount of main
monomer, S. These microspheres were named CM-1 for low carboxylated PSRMAA,
respectively.

2.3. Permeate flux Procedure

A stirred cell equipped with a 0.22Cm membrane filter was charged with 30 mL of latex,
the dispersion of microsphere. The specific surface area was adjusted to 0.19 m2 per 1mL
and the ionic strength was calibrated to 0.0 1. At the constant stirrer speed, buffer solution
was introduced into the stirred cell until steady state flux was attained. Protein solutions
were introduced with step of pulse injection. The permeate flux was measured
continuously with an electronic balance (Precision plus, Ohaus Co., USA) by a data
acquisition system. The electronic balance was connected to a PC through a RS 232C
interface. The surface charge density of microspheres was varied as 0.45, 5.94, 9.14 and
10.25, and the stirrer speed was varied as 300,400 and 600rpm.

3 Results and discussion

3.1. Steady-state permeate flux

Fig. 1. shows the steady-state permeate flux with respect to various stirrer speeds and
surface charge densities. As the stirrer speed was faster, the higher permeate flux was
observed because of the higher shear stress at the surface of a cake layer. The permeate
flux was proportional to the surface charge density of microspheres, in the case of the
higher surface charge density, the repulsive force became larger and the cake resistance
decreased.

448
3.2. Porosity of a cake layer
The cake porosity, obtained from the steady-state flux, is shown with the surface charge
density in Fig. 2. It was observed that the porosity of a cake layer tended to increase as the
surface charge density increased. The stirring effect was investigated with monodisperse
microspheres of different surface charge density. It was observed that the stirrer speed was
proportional to the porosity of a cake layer.

3.3. Flux variation of protein-adsorbed microspheres

3.3,l. Single component
After step injection of BSA solution, the permeate flux was increased with time until the
breakthrough point (at 200 mL), but it turned to decrease after that point. This flux
behavior was not observed in the conventional filtration system. But the permeate flux
was not decreased after step injection of BHb solution.

3.3.2. Binary Component

3.3.2.1 Flux variation at pH 6.8
The permeate flux of BSA and BHb mixtures was performed at pH 6.8 (Fig. 4). At this pH,
BSA molecules with negative net charge were not adsorbed onto the suface of
carbowylated microspheres by electrostatic repulsion, while BHb molecules were
adsorbed by hydrophobic force low-modified microspheres having the different charge as
BHb at a pH 4.5 exhibited low selectivity. This is due to the increase of hydrophobic
interactions with BSA and electrostatic interactions with BHb.

4 Conclusion

The experimental investigation on the permeate flux variation of carboxylated

mirospheres with or without proteins was performed at different surface charge densities,
stirrer speeds, and protein concentration. The permeate flux was significantly depended on
the surface charge density of microspheres. To obtain large permeate flux during
microfiltration using stirred cell, the high surface charge density was required. The
increase of the surface charge density caused the effective radius of the microsphere to
increase, consequently, the porosity of a cake layer to become large. The permeate flux of
protein-adsorbed microspheres was influenced by the colloidal stability. BHb-adsorbed
microspheres still have their stability while BSA-adsorbed microspheres lost stability and
were agglutinated into other larger primary particles.

Acknowledgement

This work was supported by Korea Research Foundation Grant (KRF-2001-005-E00030).

449
References

1. Y. Inomate, T. Wada,H.Handa, K, Fujimoto and H. Kawaguchi, JBiomater, Sci.

Polym. Edn. 5,293 (1994).
2. Y. Inomate, Y. Kasuya, K. Fujimoto, H. Handa and H. Kawaguchi, Colloids Su@ B:
Biointerfaces 4,23 1 (1 995).
3. J.Y. Yoon, Ph.D. Dissertation, Yonsei Univ., Seoul (1998).
4. J.-Y. Yoon, W.S. Kim, J. ColfoidZnterfaceSci.177,613 (1996).
5. C.A. Haynes and W.Norde, J Colloid Interfacesci., 169,313 (1995).
6. S.W.Choi, J.Y.Yoon, S. Haam, J.K. Jung, J.H. Kim and W.S.Kim, Journal of
colloid and interface science, 228,270-278 (2000).

450
 -
20 0.32
-- I I d 300 rpm
400 rpm
0.30

a
1.2
-- 0.28

t 1.0
0 0.2a
'8
e
g 2
8
0.8
0.24
0.6

g 0.4
3
0 0.2
0.22

0.0 0.20
200 300 400 500 600
0 2 4 e

Stirrer speed (rpm) Surface c h a r g e density (Nc, lnm'2)

Fig. 1. Steady-state permeate flux Fig. 2. Calculated porosity of a

with different stirrer speeds and cake layer with surface charge surface
charge densities densities

1.2
,
1.o 1

,
I
1
0 50 1W 150 200 250 300 0 50 100 150 200 250 300

Time (min) Time (min)

Fig.3. Single Component Fig. 4. Permeate flux Behavior

permeate flux curve with time at the different
concentration

45 1
 Surface fractional dimensions of the adsorbents from industrial sludge

J. H,You, H. M. Wu and Z. X. Fang

Department of Chemical Engineering and Material Engineering,ChangGung

University,Tao-Yuan,Taiwan,R.O.C.

Fractal geometry has been used to describe the structure of porous solid. The surface fractal
dimension D was calculated h m their nitrogen isotherms using both the fractal isotherm
equations derived from the FHH theory based on thermodynamics. Organic sludge is rich in
carbonaceous organic material; therefore it has the potential to be converted into activated
carbon if pyrolysed under controlled conditions or with some chemical treatment. The
objective of this research work was to use the surface fractal theory to analyze the adsorbent
from industrial sludge at various conditions. Effects of activated temperature, activated time
and activated agent on the specific surface area and pore volume surface of adsorbents were
investigated the surface fractal theory. The surface roughness of adsorbent was also
investigated based on the hctal analysis of their nitrogen adsorption isotherms. The
activated temperatures were controlled at 500 - 8OO'c for 0.5 to 2 hours. The various
concentrationsof activated agents (KOH and H 3 W ) were controlled at 0 to 7M.
From the experimental results, the BET surface areas and pore diameters of the sludge
adsorbents are 0.2 - 499 m2/g and 19 - 188 A. The D values of the sludge adsorbents from
H 3 W activation are 1.1 - 2.2, indicating the presence of a regular and smooth surface. The
D values of the sludge adsorbents with steam activation are 2.2 and 2.6, indicating a partially
irregular surface and semi-microporous structure. The D values of the sludge adsorbents from
KOH agent or/and steam activation are 2.7-2.9 (Van Der Waals forces are dominant) and
1.9-2.8 (surface tensions are controlling factors). The sludge adsorbents are suggested that
have strong micropore structures and irregular surfaces, when the sludge chars were activated
by a steam and 7M KOH agent for 700C. Furthermore, the optimum activation condition for
the sludge adsorbent with 7MKOH and steam occurs at 700'cfor lhr.

Key Words: fractal dimension, KOH, sludge, adsorbent, activated carbon.

1.Introduction
Activated carbon is a porous carbon material. It has been produced from many
different carbonaceous materials, including coal, wood, coconut shells, and peat. The
major manufacturing processes of the activated carbon are carbonization and activation.
Organic sludge is rich in carbonaceous organic material; therefore it has the potential to
be converted into activated carbon if pyrolysed under controlled conditions or with some
chemical treatment. The few studies 11-3 1 have been made on the pyrolysis behavior of
the organic sludge and on the physicochemical properties of the reclaimed adsorbent. The
effects of preparation condition parameters such as the concentration of activating agent,
heating temperature, dwell time, heating rate were investigated.
Several authors 14-51 have studied the effects of KOH on carbonization of
carbonaceous materials. Marsh et al. indicated that the oxygen of alkali could remove
cross-linking and stabilizing carbon atoms in crystallites at the activation temperature
-
range of 550 900C. After activation reaction, a new structure of the microporosity of the
activated carbon was created when potassium salts and carbon atoms from the internal
volume of the carbon were removed. Later, Otowa et al. also pointed out that high

452
temperature and high KOWcarbon ratio produced large pores in the carbon structure, due
to the presence of K20,derived from KOH, expanded the carbon atomic layers. When the
temperature exceeded 700 c ,a considerable amount of K was formed by the reduction of
K20 with carbon. Since the inner carbon atoms were consumed, pores were formed in the
structure. Carlos et al. indicated that chemical activation of olive-mill wastewater with
KOH produced activated carbons with much lower ash content, higher nitrogen surface
area and much better developed porosity than in the case of either its chemical activation
with H3P04 or its physical activation with C02 at 840C.
Fractal geometry has been used to describe the structure of porous solid and
adsorption on heterogeneous solid surface 16-81 . The surface fractal dimension D was
calculated from their nitrogen isotherms using both the fractal isotherm equations derived
from the FHH theory. The Frenkel-Halsey-Hill (FHH) adsorption isotherm applies the
Polanyi adsorption potential theory and is expressed as:
In N=constant + S lnA
where N is the amount adsorbed, and Anis the adsorption potential defined as:
A=-AG=RTh(I/X).
For smooth surface, the parameter S is assumed to be equal to -1/3, while for a
fiactal surface, S is a function of the surface fractal dimension, D. If the van der waals
attractive forces are dominant between adsorbent and adsorbate, then S is equal to
(Ds-3)/3 (91 . For higher surface coverage where the adsorbent-adsorbate interface is
controlled by the gadliquid surface tension, S would be equal to (D-3) 10,ll I .
According to the FHH model, on a In N vs. In A plot, the slop of the straight-line part
should be equal to S.
The objective of this research work was to use the surface fractal theory to analyze
the adsorbent from industrial sludge at various conditions. Effects of activated
temperature, activated time and activated agent on the specific surface area and pore
volume surface of adsorbents were investigated.

2.Experimental
A dewatered sludge was collected from the wastewater treatment of PET
manufacture process. The sludge was put in oven at 103 Cfor 24 hours. The water
content of the industrial sludge was about 75%. The dry sludge was put into a quartz tube
located in the furnace for carrying out the pyrolysis process. The pyrolysis was conducted
in a fixed bed, using nitrogen to carry away expelled gases. The temperature was
controlled at 600 C for 2 hrs.
The product, the sludge char, after the carbonization process was activated. The
chemical activation with activated agents (KOH and H3P04) orland physical activation
with C02 were used in the activation process. The activated temperatures were controlled
at 500 - 8OOC for 0.5 to 2 hours. The various concentrations of KOH and H3P04were
controlled at 0 to 7M.The various concentrations of H3P04 were controlled at 0 to30 %.
Nitrogen adsorption and desorption isotherms for the sludge adsorbent were
measured using the standard N2-BET test (Micromeritics Instrument Corporation
ASAP2000). The properties of the sludge adsorbents were also characterized by the BET
surface area method. The surface fractal dimension D was calculated from their nitrogen
isotherms using both the fractal isotherm equations derived from the FHH theory.

453
 3. Results and Discussion
The BET surface areas, pore diameters, and fractal dimensions of the sludge
adsorbents with H3P04 at various activation conditions are shown in Tablel. The BET
surface areas and the pore diameters of the sludge adsorbents are 0.3- 9.4 m2/g and 58 to
188 A, respectively. Comparing the properties of the sludge adsorbents from H3P04
activation with those of the commercial activated carbons show that the BET surface areas
of the sludge adsorbents are very less and the pore diameters of the sludge adsorbents are
very large, (2) the adsorption potentials of the sludge adsorbents from H3P04 activation
are very poor. From the data of sample 1 1 and sample12 in Table 1, when the
carbonization sludge was activated with steam, significant changes are observed for the
BET surface areas and the pore diameters of the sludge adsorbents. The BET surface
areas and the pore diameters of the sludge adsorbents are 131- 217 m2/g and 27 to 59 A,
respectively. The specific surface areas of the sludge adsorbents from steam activation are
10- 1000 times higher than those of the sludge adsorbents from H3P04 activation.

Table 1. The specific surface area, pore diameter, and fractal dimension of the sludge
adsorbent with H3P04 at various activation conditions.

Note:
1. The assumption that Van Der Waals foeces are dominant between the adsorbate
and the adsorbent
2. S: the slope of the line of In(VNm) against In(1n PO/P) for sludge adsorbent
3. D: the surface fractal dimension

In generally, the value of the fractal dimension, D, from the fractal FHH model is
from 2 to 3. But in this study, when the sludge char was activated by only H3P04 at
various activation temperatures and times, most slopes, S, of the lines of ln(VNm) against
ln(1n POP) for the.N2 adsorption isotherm data of the sludge adsorbents are smaller than
-1/3. The D values of the sludge adsorbents are smaller than 2. These are derived based
on the assumption that Van Der Waals forces are dominant between the adsorbate and the
adsorbent. The D values of the sludge adsorbents from H3P04 activation are 1.1 2.2, -
indicating the presence of a regular and smooth surface. The D values of the sludge
adsorbents with steam activation are 2.2 and 2.6, indicating a partially irregular surface
and semi-microporous structure.
The specific surface areas, pore diameters, and surface fractal dimensions of the
sludge adsorbents with KOH at various activation conditions are shown in Table2. As in
shown, when the sludge chars were activated by KOH agent, the specific surface areas and
the pore diameters of the sludge adsorbents are 19 - 36 m2/g and 58 to 98 A, respectively.
When a steam and KOH agent simultaneously activated the sludge chars, the specific
surface areas and the pore diameters of the sludge adsorbents are 133 - 499 m2/g and 19
to 32 A, respectively. It is also observed that the optimum activation condition of the
sludge char, based on the highest specific surface area, was found to be with 100% steam
and 7MKOH agent at 700'c for 1 hr. The pore diameters of the sludge adsorbents
decrease as the surface area of the sludge adsorbent increased. The phenomena could be
ascribed to the fact that the sludge chars would not decompose completely to provide
enough specific surface area at a shorter activation time, a lower activation temperature,
or a lower concentration of KOH, however that a sintering effect resulted in the reduction
of the surface area at longer activation time or a higher activation temperature, or a higher
concentration of KOH. Moreover, the result of analysis for the surface areas and pore
diameters of the sludge adsorbents in Table 2 shows that the parameter of a steam than
KOH agent for activation process is more important.

Table 2. The specific surface area, pore diameter, and fiactal dimension of the sludge
adsorbent with
KOH at various activation conditions.

S: the slope of the line of ln(VNm) against h(ln POP) for sludge adsorbent
D1: the assumption that Van Der Waals foeces are dominant between the adsorbate
and the adsorbent.
D2: the assumption that a higher surface coverage with gaslliquid surface tension as
the controlling factor at the adsorption interface.

When the sludge chars were activated by KOH agent orland a steam at various
activation temperatures and times, most slopes, S, of the lines of In(V/Vm) against h(ln
POm) for the N2 adsorption isotherm data of the sludge adsorbenst are larger than -1/3.
The D values of the sludge adsorbents are almost larger than 2.5, whether these are
derived based on the assumption that Van Der Waals forces are dominant or surface

455
tensions are controlling factors. The D values of the sludge adsorbents fiom KOH agent
orland steam activation are 2.7-2.9 (Van Der Waals forces are dominant) and 1.9-2.8
(surface tensions are controlling factors). From the Figurel, the D values of the sludge
adsorbents increase as pore diameter decreased. The sludge adsorbents are suggested that
have strong micropore structures and irregular surfaces, when the sludge chars were
activated by a steam and 7M KOH agent for 700'c. According to N2 adsorption isotherm
data of sludge adsorbent with KOH at various activation temperature, Fig. 2 shows that
plots of ln(V/Vm) against ln(hPO&) exist a strong linear correlation. The slopes, s, of the
lines of ln(V/Vm) against h(ln POR) for sludge adsorbent gradually increase as the
activation temperature increased. Furthermore, the optimum activation condition for the
sludge adsorbent with 7MKOH and steam occurs at 700 'c for lhr.

3 -
29 - a0
0 0
428 -+ 0
8 27 + 0
'1 26
s = + +
24
2.3 +
I22
& 21 +VanDerWallsfO~dominant
2
1.9 , Q ~ l m s k t l ~

Figure 1. The correlation of pore diameter and fractal dimension of the sludge
adsorbent from KOH activation process

A 7MKOH0.5hr600C X 7MKOHOShr700'C

I .6 -...A
1.5 -- ' - - .A

-
1.4
1.3
. = -0.1 1 3 6 r + 0 . 4 1 9 5
- -.A
P = - 0 . 2 9 1 ~+ 0 . 4 2 8 8
e 1.1 R' = 0 . 9 4 23
2 o.;
;..---.......,.
=x' --.....x- ..-....x . ,
--. A
. .- P
R' = 0 . 9 8 3 7
0.R
0.7
-------l__
-
.
---
0.6
o,5 . y I - 0 . 0 7 7 6 ~+ 0 . 4 3 6 6
0.4 .- R'= 0 . 9 5 4 8
0.3
0.2 -
0.1 : A
0 '
-5 4.5 -4 4.5 -3 -1.5 -I -1.5 .I 4.5 0 0.5 I
In( In P O P )

Figure 2. FHH plots for sludge adsorbent at various activation temperatures

4.Conclusions
In this study, the BET surface areas and pore diameters of the sludge adsorbents are
0.2 - 499 m2/gand 19 - 188 A. From the point of view of the specific surface area and
pore diameter, it was concluded that the optimum activation condition for the sludge
adsorbent with 7MKOH and steam occurs at 700'cfor Ihr.From the analysis of hctal
dimensions(D) using the FHH theory, the D values could be used to describe the structure

456
of porous solid. The D values of the sludge adsorbents from H3P04 activation are 1.1 -
2.2, indicating the presence of a regular and smooth surface. The D values of the sludge
adsorbents with steam activation are 2.2 and 2.6, indicating a partially irregular surface
and semi-microporous structure. The D values of the sludge adsorbents from KOH agent
orland steam activation are 2.7-2.9 (Van Der Waals forces are dominant) and 1.9-2.8
(surface tensions are controlling factors). The sludge adsorbents are suggested that have
strong micropore structures and irregular surfaces. However, the D values of the sludge
adsorbents increase as pore diameters decreased.

Reference

1. .J. H. Tay, X.G. Chen, S. Jeyaseelan, N. Graham, A comparative study of

anaerobically digested and undigested sewage sludges in preparation of activated
carbons, Chemosphere 44 (2001) 53-57
2. J. H. Tay, X.G. Chen, S. Jeyaseelan, N. Graham, Optimising the preparation of
activated carbon from digested sewage sludge and coconut husk,, Chemosphere
44(200 1)45-5 1
3. Chiang, P.C., You, JH., Use of sewage sludge for manufacturing adsorbents, Can.
J. Chem. Eng. 65( 1987) 922-927
4. Carlos Moreno-Castilla et al.,Chemical and physical activation of olive-mill waste
water to produce activated carbons, Carbon 39(2001) 14 15-1420
5. Jia Guo and Aik Chong-Lua,Effm of surface chemistry on gas -phase adsorption by
activated carbon prepared from oil-palm stone with pre-impregnation, Separation
and Purfication Technology 18 (2000) 47-55.
6. Nasrin R. Hhlili et al.,determination of fractal dimensions of solid carbons from gas
and liquid phase adsorption isotherms,Carbon 38 (2000) 573-588
7. W. Rudzinski, Lee Shyi-Long, Yan Ching-Cher, T. Panczyk, A fractal approach to
adsorption on heterogeneous solid surfaces. I. The relationship between geometric
and energetic surface heterogeneities. Journal of Physical Chemistry B, 105
10847-10856 (2001).
8. W. Rudzinski, Lee Shyi-Long, T. Panczyk, Yan Ching-Cher, A fractal approach to
adsorption on heterogeneous solids surfaces. 11. Thermodynamic analysis of
experimental adsorption data. Journal of Physical Chemistry B, 105 10857-10866
(2001).
9. Avnir D, Farin D, pfeifer P., Molecular fiactal surfaces,. Nature 1984; 308:
26 1-263
10. Avir D, jaroniec M.,An isotherm equation for adsorption on fractal surfaces of
heterogeneous porous materials, Langmuir, 1989;5 (6)143 1-1433
11. Yin Y,Adsorption isotherm on hctally porous materials, Langmuir 1991;7
(2):216-2 17

457
 ADSORPTION OF ACIDIC PEPTIDE
ON
CROSSLINKED CHlTOSAN FIBER:EQUILIBRIA

NOBORU KISHIMOTO*
Department of Material Science, WakayamaNational College of Technology
77 Noshima, Nada-cho. Gobo-shi, WakayMla644-0023. JAPAN
E-mail: noboruC2 wakayaha-nct.ac.jp

HIROYUKI YOSHIDA
Department of Chemical Engineering, Osaka Prefecture University,
1-1, Gakuen-cho,Sakai-shi, Osaka 599-8531. JAPAN
E-mail: yoshida @chemeng.osakajk-u.ac.jp

Adsorption of glutathione, which is an acidic peptide. on the crosslinked chitosan fiber (ChF)
appeared technically feasible. The experimental equilibrium isotherm (q-C curve) for adsorption of
glutathione on ChF was independent of the initial concentration of glutathione. But the adsorbed
amount of glutathione on ChF was effected by the pH value of the solution on the q-pH curve,
significantly. It appeared that the adsorption of glutathione was correlated by the Langmuir equation
well.

1 Introduction

In the food, medical, and pharmaceutical industries, amino acids, peptides, and
proteins play very important roles. Their purities must be high because of using for
human body. For separation and purification of amino acids, peptides, and proteins,
adsorption and ion exchange processes are generally used in these industries.
In our previous works [l-31, we reported experimental and theoretical equilibrium
isotherms for adsorption of L-glutamic acid in the single component system on
polyaminated highly porous chitosan (hereafter called PEI-CH), weakly basic ion
exchanger, and crosslinked chitosan fiber. We found that the adsorption of L-glutamic
acid, which is a kind of acidic amino acid, was controlled by the acid/base neutralization
reaction between neutral L-glutamic acid (zwitterion, A 3 and those adsorbents.
In this work, we investigate the possibility of using ChF for adsorption of acidic
peptide. The experimental equilibrium isotherm and mechanism for adsorption of acidic
peptide is presented and discussed.

2 Experiments

We used the crosslinked chitosan fiber (hereafter called ChF) in this experimental
study. ChF was fabricated by Fuji Spinning Co., Japan. Fig.1 shows the unit molecular
structure of chitosan which was transformed from chitin by deacetylation. Chitin is a
natural biopolymer which is contained in the shell of arthropods. Chitosan was
crosslinked to make an adsorbent with acid, alkaline, and chemical proofs. The
fabrication method of ChF was presented elsewhere.[5,6].
Glutathione (Peptide Institute Inc., Japan) was used as the acidic peptide in this
experimental work. It has important role in biochemical oxidationheduction reaction.
The structure is shown in Fig.1.
~

*Author to whom correspondencethis paper should be addressed.

458
 CH*OH CH20H

0 ~ \ CH2CH 0 ~

H ?IH
H NH2
\0/
EGDE-Residue Chitosan EGDE
0
II
0
II
yoo-
HOOC-H2C-HN- C- CH-NH- C- (CH2)2- FH
I NH~+
CH2
I
SH Glutathione (;type )

Fig.1 Unit molecular structures of Chitosan, EGDE, and Glutathione.

The equilibrium isotherms for adsorption of peptide were measured

by the batch method. Equilibrium was fully reached in 4 days. The
solution for the peptide was analyzed with a Shimadzu Liquid
Chromatograph Model LClOATvp and a Shimadzu Fluorescence HPLC
Monitor Model FLD-1. The pH of the equilibrium solution was analyzed
with a Horiba pH meter Model F-23.
The adsorbed-phase concentration of peptide was calculated according to eq. (1).

where Co and C are the initial &d equilibrium concentrations of the peptide in the
solution (kmol/m3), respectively. q denotes the equilibrium concentration of the peptide
in adsorbed phase. (kmoVkg dry fiber). V and Ware the volume of the solution (m3) and
the weight of adsorbent (kg dry fiber), respectively. All experiments were carried out at
298 K.

3 Results and Discussion

Fig. 2 shows equilibrium isotherm for adsorption of glutathione on ChF. The

experimental equilibrium isotherm for adsorption of glutathione is independent of the
initial concentration of glutathione. Glutathione may be adsorbed by chemisorption and
the equilibrium isotherm may be expressed by the Langmuir equation:

where Q is saturation capacity of glutathione (molkg) and K shows the equilibrium

constant ( m 3 h o l ) . The solid line in the figure shows the Langmuir isotherm. The
data is correlated well by eq.(2) using Q and K in Table 1. The coefficient of correlation
is 0.99 1.
 '1 r
298K
I

q=QKC/(1+KC)
\
I

1 p 3 . 91 molkg -
,
~ = 7 . 2x -
3 ~02m3flun~i
OO 0.01 0.02 0.03

Fig.2 Equilibrium isotherm for adsorption of glutathione on ChF.

Table 1 Langmuir coefficients for adsorption of Glutathione on ChF

pH K[m3/moll &[moVkgdry fiber] Fig.No.
No adjustment 7.23 X 102 3.91 2
3.2k0.2 3.22 X 102 4.63 4

5 - I , - . .

Co=O.bl [kmoVm3],298K -
4- f V=l x 10-5[m3] -
*it W=l X 10"[kg]
t.-
?
I
I
t
t,
5 -
I I

9
c -
1 , I
II

1 - 7 ,' ,'.' :* -
- 74. ',
d
,,a

Theoretial concentration distributions of glutathione

Fig. 3 shows the theoretical concentration distributions of glutathione in the liquid
phase and the effect of pH on the equilibrium adsorbed amount of glutathione on ChF.

460
The pH of the solution was adjusted by using HCI or NaOH aq.. When 2<pH<4.5,
pH4.8, and pH>10.5, glutathione is adsorbed on ChF. In other pH regions, it is not
adsorbed on ChF.
Glutathione dissociates in the aqueous solution as follows:

where A', A*,A-, A2; and A3- are R,,(-COOH)(-COOH)(-SH)(-NH3+),

R&-COO)(-COOH)(-SH)(-NH3+), R,,(-COO)(-COO-)(-SH)(-NH3'), Rp
(-COO-)(-COO-)(-S-)(-NH3+),and %(-COO-)(-COO-)(-S-)(-NH2), respectively.
For example, the structure of A* is shown in Fig.1.

In the region of 2<pH<4.5, the A*type of glutathione is dominant in the solution. So

that type of glutathione may be adsorbed on ChF. Yoshida et al. [I J and Yoshida and
Kishimoto [2] reported that the adsorption of L-glutamic acid on PEI-CH, weakly basic
ion exchange resin, and ChF is controlled by the acidbase neutralization reaction between
the carboxylic group of the neutral L-glutamic acid (A? and the fixed amino group of the
adsorbents. This may suggest that glutathione is also adsorbed according to the
following, an acidbase neutralization reaction:
R-NHz + &(-COO-)(-COOH)(-SH)(-NH33
-
- K
R-NHC * (-OOC-)R,(-COO-) (-SH)(-NH3+) (7)
where R-NH2 denotes the fixed amino group of ChE

A 0.03
~ = 3 . 2x
2 1o2m3km~i 0.04
rn 0.05
0
' 0.01 0.02 0.03 0.04 0.05
C[I~OI/~~]
Fig. 4 Equilibrium isotherm for adsorption of glutathione on ChF at the constant
pH (=3.2+0.2).

461
 Fig. 4 shows the equilibrium isotherms for adsorption of glutathione on ChF at the
constant pH (=3.2M.2). The experimental adsorption isotherms of glutathione were
correlated well by the Langmuir equation (2) by using Q and K are listed in Table 1.
The coefficient of correlation is 0.966. Since the equilibrium isotherm is independent of
the initial concentration, the mechanism for adsorption of glutathione may be
chemisorption. Since in this pH region, type A* of glutathione is dominant in the
solution, it is selectively adsorbed on ChF. We investigated the equilibrium isotherms
for adsorption of glutathione at the pH=l .O, 2.0, 6.4, and 1 1.8*0.2 in this work. Those
isotherms were unfavorable rectangular, and glutathione was adsorbed little on ChF at
those pHs.

4 Conclusion

Adsorption of glutathione, which is an acidic peptide, on the crosslinked chitosan fiber

(ChF) appeared technically feasible.
(1) The experimental equilibrium isotherm (q-C curve) for adsorption of glutathione on
ChF was independent of the initial concentration of glutathione.
(2) The adsorbed amount of glutathione on ChF was effected by the pH value of the
solution on the q-pH curve, significantly.
(3) It appeared that the adsorption of glutathione correlated by the Langmuir equation
well.

References

[I] H. Yoshida, N. Kishimoto, and T. Kataoka, fnd. & Eng. Chem. Rex, 34, 347-355
(1995).
[2] H. Yoshida, and N. Kishimoto, Chem. Eng. Sci.,50,2203-2210 (1995).
[3] N. Kishimoto and H. Yoshida, Sep. Sci. & Tech.,30,3 143-3163 (1995).
[4] H. Yoshida, N. Kishimoto, and T. Kataoka, fnd. Eng. Chem. Rex, 33,854-859 (1994)
[5] H. Yoshida, A. Okamoto, H. Yamasaki, and T. Kataoka, Fundamentals of Adsorption,
Kodansha, Tokyo, 767-774(1993)
[6] H. Yoshida, A. Okamoto, and T. Kataoka, Chem. Eng. Sci., 12,2267-2272(1993).

462
 REMOVAL OF SALT AND ORGANIC ACIDS FROM SOLUTION USED TO
SEASON SALTED JAPANESE APRICOTS (UME) BY COMBINING
ELECTRODIALYSIS AND ADSORPTION

W. TAKATSUJI
Industrial Technology Center of Wakayama Prefecture, 60 Ogura, Wakayama 649-6261,Japan
E-mail: [email protected]

H. YOSHIDA
Department of Chemical Engineering, Osaka Prefecture University,I - I Gakuen-cho Sakai
599-8531, Japan
E-mail: [email protected]

With the aim of repeatedly reusing the solution used to season salted ume (Japanese apricot), we
investigated its desalting and deacidification by electrodialysisand adsorption. Although NaCl and
acids could be easily removed from used seasoning solution by electrodialysis, useful substances
such as amino acids were removed at the same time. Chitopearl CS selectively adsorbed citric acid
in the used seasoning solution without also adsorbing useful substances. The equilibrium data of a
citric acid, which was a main organic acid, were correlated by the Langmuir equation. The
saturation capacity decreased with increasing concentrations of NaCl in the solution but the
equilibrium constant did not change. The breakthough curves of citric, malic and acetic acids in
used seasoning solution on Chitopearl CS were investigated. Chitopearl CS can be regenerated
using 1 N NaOH solution.

1 Introduction

Pickled ume (Japanese apricot), a traditional delicacy in Japan, are prepared in the
following manner. Ume h i t are first preserved in salt for about 2 months (ume
fruitsalt = 4 kg:l kg). The salted ume are then washed with water and immersed in a
seasoning. NaCl and organic acids are extracted from the salted ume into the solution
during the seasoning process, which means that about 30 % of the solution has to be
replaced before it can be reused. As new seasoning solution is expensive and used
solution contains considerable amounts of amino acids and sugars, an economical means
of treating used seasoning solution for reuse is greatly desired. The removal of NaCl and
organic acids from used solution by combining electrodialysis and adsorption, without at
the same time removing useful components, were investigated.

2 Experimental Section

A laboratory electrodialysis system (CS-0 type; Asahi Glass Co., Tokyo) was used. The
system consisted of cation-exchange membranes (CMV; 14 sheets) and anion-exchange
membranes (AMV; 10 sheets). The ion exchange resin used in this study was Chitopearl
CS, which was the effective for removing organic acids [l, 21. The physical properties of
Chitotearl CS were shown in Tablel. Adsorption isotherms were measured by the batch
method. The resin particles were contacted with each solution at 298 K for 7 d. The
breakthrough curves were measured using the column (1 cm i.d.) in which Chitopearl CS
beads packed at 298 K.

463
 Table 2 Compositions of new
Table 1. Experimental physical

3 Result and Discussion

As shown in Table 2, NaCl and total acids concentrations in used salted ume seasoning
solution were 1.5 and 5 times new seasoning solution, respectively. However, total
nitrogen concentration in the used salted ume seasoning solution was largely remained.
Figure 1 shows the removal of NaC1, total acid (TA), and total nitrogen (TN) at 10 V.
These results showed that while electrodialysis could remove NaCl fiom used seasoning
solution, the removal of acids was slow and was accompanied by decreasing TN. It is
thought the acid removal rate might be slow because the molecular weight of citric acid
(192 Da), which exists abundantly in the solution, is larger than those of NaCl(58 Da) and
glutamic acid (147 Da).

0
0 1 2 3 4
Time (h)
Figure 1. Changes in each component by electrodialysis.
Wi.0, initial weight of i component.
wN&lO = 184g; wT&O = 52.5g; wTN,O = 2.65g

2.5 7 1 500

0.01 0.1 1
CN.CL(kmol m-9
Figure 2. Effect of NaCl on saturation capacity (q,) and
equilibrium constant (K).

464
Figure 2 shows the effect of NaCl on the adsorption isotherm for citric acid on Chitopearl
CS. K and qowere calculated by the Langmuir isotherm equation.
40KC
'= 1 +KC
where q, C, K, and qodenote the adsorbed phase concentration of organic acid (kmol m-3
wet resin), concentration of organic acid in the liquid phase (kmol m-3), equilibrium
constant (m3 kmol-I), and saturation capacity (kmol m-3 wet resin), respectively. qo
decreased with increasing CN&l,but K was constant. When CN&l= 0.1 km01/m3 (5.8 kg
m-3),40 was reduced to about half the value at CNacl= 0 lanol m-3and the resin could not
adsorb citric acid when > 1 kmol m-3. The above results showed that the lower the
concentration of NaCl was, the greater the adsorption of organic acid became. We
therefore used electrodialysis to first reduce the NaCl in the used seasoning solution.
NaCl was reduced almost linearly by electrodialysis, for 3 h, at which time its
concentration was only 0.2 kg m-3. The removal of organic acid by CS was then tested
using this solution.

0 10 20 30 40 50 60
V/w ((masolution)(g wet resinP)
Figure 3. Degree of removal of each component from used seasoningsolution desalted
by electrodialysis (CMV + AMV, 3h, 1OV)prior to adsorptionon CS.
0 0.2 kg mS;
Wi.0, initial concentration of i component. W ~ t . =
WTAO 15 kg ni3; Wm.0 = 0.89 kg m-3W~m.0= 19 kg mJ

Figure 3 shows the degree of removal of each component from the solution. TA
decreased but TN and direct reduce sugar (DRS), which were useful substances, did not.

1.2

- 0.6
3
0.4
g0-
0.2
0
0 0.02 . 0.04 0.06 0.08 0.1
c (lunol m3)
Figure 4. Adsorption isotherms of citric, malic and acetic acids in used salted
ume seasoning solution on CS.
0 ; citric acid, A; malic acid, 0 ; acetic acid

465
Figure 4 shows the adsorption isotherm of citric, malic and acetid acids in the solution on
CS. A citric acid was selectively adsorbed compared with other organic acids. This
result is consistent with the adsorption on a weakly basic exchanger described previously
[31.

1.4 I I I I

1.2

0.8
0 citricacid virgin
0.6 0 citricacid Ztimcs
0 ciaicaeid Jtimcs

0.4

0.2

0
0 1000 2000 3000 4000 5000
Time (sec)
Figure 5. The breakthrough curve for adsorption of citric, rnalic and acetic acids on CS.
Uf,superficial velocity; H, bed height; C, fluid phase concentraion

Figure 5 shows the breakthrough curves for adsorption of organic acids in the solution on
CS packed column. They are little affected by the repeat time. Figure 6 shows the
eluent curves using 1 kmol m-3 of NaOH. Acetic acid adsorbed on CS was desorbed
with increasing the adsorption of citric acid and malic acid. Subsequently malic acid
adsorbed on CS was desorbed with increasing the adsorption of citric acid. Finally only
citric acid in used seasoning solution was adsorbed.

R e p l d NnS
0 cilncacid virgin
0.3 0 citricacid 2 rim
0 citricacid 3 limv
A mvlic acid virgin
V mdicacld 2 times
A nidicacid 3 rimer
0.2 0 aucticnd virgin

-E 0 irtlicacid
prrlicacid
2 times
3 times
*8
- Us = 0.05 rm
L, 0.1

0
I200
Time (sec)
Figure 6. The eluent curves of citric, malic and acetic acids.

466
References

1 . Takatsuji, W. and Yoshida, H., Removal of organic acids from wine by adsorption on
weakly basic ion exchangers: equilibria for single and binary systems, Sep. Sci.
Technol. 29 (I 994) pp. I473 - 1490.
2. Takatsuji, W. and Yoshida, H., Adsorption of organic acids on polyaminated highly
porous chitosan: equilibria., Ind. Eng. Chem. Res. 37 (1998) pp.1300 - 1309.
3. Takatsuji, W. and Yoshida, H., Adsorption of organic acids on weakly basic ion
exchanger: equilibria for binary systems, A. I. Ch. E. J. 44, (1998) pp. 12 I6 - 122 1.

467
 STUDIES ON THE ONE-COLUMN ANALOGUE OF A FOUR-ZONE SMB

Y.S. KIM, C.H. LEE, Y . M. KO0

Department of Biological Engineering, ERC for the Advanced Bioseparation Technology,
Inha University,Incheon 402-751, Korea
E-mail: ymkoo@jnha. ac.kr

P.C . WANKAT
School of Chemical Engineering, Purdue University, West h f q e t t e , IN 47907. USA

An one-column process was developed to simulate a four-zone SMB using just one column and four
storage tanks. The basic principle of this process is identical to a four-zone SMB. A systematic
design for one-column process was carried out for the separation of two amino acids, phenylalanine
and tryptophan. The process was compared with a four-zone SMB by computer simulation using
Aspen Chromatographym. Operating parameters were obtained h m the triangle theory. The
simulation yielded the product purity of the one-column process to be 92.0% and 85.6% for
phenylalanine and tryptophan, respectively, while 92.8% and 87.8% for the fowzone SMB. The
recovery yields of the one-column process were 93.0% and 84.0% for phenylalanine and tryptophan,
respectively, while 94.0% and 85.5% for the four-zone SMB. The low purity and yield of the
one-column process were due to the fact that the concentration profiles are destroyed in the fluid
phase after mixing in the collection tanks, but maintained in the solid phase. The one-column
process would be usefbl to reconstruct an existing conventional chromatographyto a SMB process.

1 Introduction

The simulated moving bed (SMB) technology was developed in the early 1960s by UOP
for large-scale hydrocarbon separations in plants [l, 21. Recently, SMB is extensively
used for the separation of enantiomers [3,4],amino acids [5,6], and protein desalting [7].
The SMB system simulated counter-current movement; this is done with a series of
packed bed by periodically advancing the input (feed, desorbent) and output port
(raffinate, extract) along the solvent flow direction.
The solute of lower affinity with adsorbent moves along the fluid direction, while
solute of higher affinity with adsorbent moves along the counter direction [8]. The main
advantage of SMB system is the fact that it is a continuous process with high purity and
recovery, and low eluent consumption. For successful separation, appropriate zone flow
rate and switching period should be selected. The design methods of SMB have been
proposed in the literatures [9-111. In this study, the one-column process was simulated
using single column and four tanks to compare with the conventional SMB system. The
operating conditions were obtained by equilibrium theory, assuming a negligible mass
transfer resistance.

2 Theory

2.1 Design of SUB process

The operating parameters are chosen to achieve successful operation, such as maximum
productivity, maximum yield, minimum desorbent consumption. There are two major
methods to obtain these operating parameters: one is the periodic moving port analysis

468
19-1 11 and the other is the continuous moving bed analysis [12, 131. The latter is
represented by the triangle theory. The optimum operating conditions for an SMB in this
study was obtained from the triangle theory [lo, 14, 151, where the separation
performance is controlled by the flow-rate ratios between each zones of the SMB unit:
Q,t* - v&*
mj =- v I-&
where mj is the flow rate ratios in the four zones of the SMB, Q, is the volumetric flow rate
in zone j , V is the volume of the column, t* is the switching time, and E* is the total
column porosity.
The triangle theory is acknowledged as one of the most effective design tool for SMB.
Although this theory neglects mass transfer resistance and axial dispersion, it provides a
useful approximation of the real SMB behavior.

2.2 Mass Transfer Parameter Estimation

When solute concentrations are sufficiently low and isotherms are in the linear range, the
following correlation can be used to estimate the lumped mass transfer coefficient (Kf)
[l 11:
1
-=
R2
+- R
K, 15&,D, 3k,
where R is the sorbent particle radius, Dp is the effective intraparticle diffisivity, and kf is
the film mass transfer coefficient. kf is estimated from the correlation of Wilson-
Geankoplis [16]. The Wilke-Chang correlation was used to estimate the Brownian
diffisivity, D. The effective pore diffisivity was obtained from the Mackie-Meares
correlation [ 171:

where E, is intraparticle porosity. The axial dispersion coefficients for each component
were estimated from the following correlation [ 181.

siEz si si
Where E, is the axial dispersion coefficient, v, is the fluid velocity, &, is the
interparticleporosity, and Re is Reynolds number.

3 One-column process

A conventional SMB can be divided into four zone; zone I is between desorbent and
extract, zone II is between extract and feed, zone III is between feed and raffinate, zone
IV is between raffinate and desorbent as shown in Figure 1. Each column plays the role of
different zone during each period between switching time. For example, column 3 plays
the role of zone III during the fKst switching period, zone I1 during the second

469
Figure lschematic diagram of a four-zone SMB process during four switching periods. The weakly adsorbed B
(rafinate) moves up in zone 111 while strongly adsorbed solute A (extract) move down in this zone. For each
switching period in four column four zone SMB,columns play the role of different zone.
period, zone I during the third period, and zone IV during the fourth period. One column
can perform the role of four zone during one cycle of four switching time. This is the key
concept of the one-column process. Only one column and two tanks are needed in a period.
To provide liquid with appropriate concentration to a column during a cycle, four tanks
are needed. A simplified flowsheet of the one-column process is given in Figure 2.

t+4 tew + +t- I b,,t e l

Figure 2. A simplified flowsheet of one-column process using a single column and four tanks.

470
4 Discussion

The separation of two amino acids [5, 61, phenylalanine and tryptophan, was tried using
one-column process. The adsorption isotherms of two amino acids were obtained by
multiple frontal chromatography [ 5 ] . The equilibrium adsorption data were fitted by
Langmuir isotherms as follows;
1 .74
4 Phr = 1 + 0.01995 C,, + 0.1285 C ,
c, qTrp =
11 .95 c,
1 + 0.01995 C,,he+ 0.1285 C,,
where Phe and Trp are abbreviationsof phenylalanine and tryptophan, respectively.
The basic principle of one-column process is identical to four-zone SMB. The
performance of the process for the amino acids separation was compared with four-zone
SMB by computer simulation using Aspen ChromatographyTM. The system and operating
parameters are listed in Table 1. It was set that T2, T3 and T4 are initially filled with
desorbent and T1 is empty in the simulation. Liquid in each tank is ideally mixed. Liquid
of the average solute concentration in a tank is introduced into the column. The simulated
concentration profile of two amino acids in the one-column process is presented in Figure
3.

Table 1. System and Operating Parameters for one-column system and four column four-zone SMB.

System parameter r- ~ ~ G r a t i n g parameter

column internal diameter (cm) 2. 5 tryptophan 1

liauid viscosity (cP) 0.89 flow rate (mi/rnin) zone I 8.28

mass density of eluent (Kg/m3) 1000 zone II 2.29

Interparticle porosity (E iJ 0.35 zone Ill 7.29

intraparticle porosity (c p) 0.55 zone IV 2.36

adsorbent particle radius (v rn) 21 1 feed 5

mass transter coefficients raffinate 4.93

phenylalanine 1.83
(l/min)
extract 5.99

desorbent 5.92
tryptODhan 1.83
switching period (mid 53.82

The simulation results of one-column process and four columns four zone SMB were
presented in Table 2. The purity and yield are defined as follows;

where E and R are abbreviations for extract flow rate and rafinate flow rate, respectively.

471
 0 500 loo0 1500 2wo m . 3wo
Time (min)

Figure 3. Simulated concentration profile of one-column process. Solid line is for tryptophan and dashed line
is for phenylalanine.

Table 2. Simulation results of one-column process and four columns four-zone SMB.

Purity(%) Yield(%)

Phe Trp Phe Tip

4 columns SMB 92.8 87.8 94.0 85.5
one column SMB 92.0 85.6 93.0 84.0

The results using the Aspen Chromatography simulator showed the product purity of
one-column process of 92.0% and 85.6% for phenylalanine and tryptophan, respectively,
while 92.8% and 87.8% for four-zone SMB. The recovery yields of one-column process
were 93.0% and 84.0% for phenylalanine and tryptophan, respectively, while 94.0% and
85.5% for four-zone SMB. The low purity and yield of the one-column process comparing
with those of four-zone SMB was due to the fact that the concentration profiles are
destroyed in the fluid phase after mixing in the collection tanks,but maintained in the
solid phase. The one-column process can be applied to reconstruct an existing
conventional chromatography to a SMB process. As the operating conditions of this work
have not been optimized based on purity and yield, the optimization and experimental
work with the two amino acids using one-column process are under way to compare with
the simulation results.

5 Acknowledgements

This study has been supported by the ERC for the Advanced Bioseparation Technology,
KOSEF.

472
References

1. Broughton D. B., Molex: Case History of a Process, Chern. Eng. Prog., 64 (1968)
pp.60.
2. Broughton D. B., Neuzil R. W., Pharis J. M., and Brearley C. S., The Parex Process
for Recovering Paraxylene, Chem. Eng. Prog., 66 (1970) pp. 70.
3. Ching C. B., Chu K H., Hidajat K., and Ruthven D. M., Experimental Study of a
Simulated Counter-current Adsorption System: VII. Effects of Non-linear and
Interacting Isotherms, Chem. Eng. Sci.,48 (1993) pp.1343.
4. Rodrigues A. E., Loureiro J. M., and Pais L. S., Separation of I7l-bi-2-naphthol
enantiomers by continuous chromatography in simulated moving bed, Chem. Eng.
Sci.,52 (1997) pp. 245-257.
5. Wu D., Xie Y.,Ma Z. and Wang N. -H. L., Design of SMB Chromatography for
Amino Acid Separations,Ind. Eng. Chem. Res. 37 (1998) pp. 4023-4035.
6. Xie Y.,Wu D., Ma Z. and Wang N. -H. L., Extended Standng Wave Design Method
for Simulated Moving Bed Chromatography: Linear Systems, Ind. Eng. Chem. Res.
39 (2000) pp. 1993-2005.
7. Hashimoto K., Adachi S. and Shirai Y., Continuous Desalting of Proteins with a
Simulated Moving-bed Adsorber, Agric. Biol. Chem. 52 (1988)pp. 2 161.
8. Wankat P. C., Rate-Controlled Separations, Glasgow/London (1994).
9. Ruthven D. M. and Ching C. B., Counter-Current and Simulated Counter-Current
-
Adsorption Separation Processes, Chem. Eng. Sci.44 (1 989) pp. 10 1 I 1038.
10. Storti G., Mazzotti M., Morbidelli M. and Carra S., Robust design of binary
counter-current adsorption separation processes, A.I.Ch.E.J. 39 (1993) pp. 47 I.
11. Ma Z. and Wang N. -H. L., Standing Wave Analysis of SMB Chromatography:
Linear Systems, A.LCh.E.J. 43 (1997) pp. 2488-2508.
12. Ching C., Ruthven D. M. and Hidajat K., Experimental Study of a Simulated
Counter-Current Adsorption System: 111. Sorbex Operation, Chem. Eng. Sci. 40
(1985) pp. 1411.
13. Hashimoto K., Adachi S., Noujima H. and Maruyama A., Models for separation of
glucose-fructose mixture using a simulated moving bed adsorber, J. Chem. Engng.
Japan. 16 (1 983) pp. 400-406.
14. Gentilini A., Migliorini C., Mazzotti M. and Morbidelli M., Optimal operation of
simulated moving-bed units for non-linear chromatographic separations 11.
Bi-Langmuir isotherm, J. Chromatogr. A. 805 ( I 998) pp. 37-44.
15. Mazzotti M., Storti G. and Morbidelli M., Operation of simulated moving bed units
for nonlinear chromatographic separations, Margalef R., J. Chromatogr. A . 769
(1 997) pp. 3-24.
16. Wilson, E.J. and Geankoplis, C. J., Liquid Mass Transfer at Very LOWReynolds
Numbers in Packed Beds, Ina! Eng. Chem. Fundam. 5 (1 966) pp. 9.
17. Mackie, J.S. and Meares, P., The Diffusion of Electrolytes in a Cation Exchange
Resin Membrane, Proc. Roy. SOC.London, Ser. A 267 (1955) pp. 498-506.
18. Slater, M.J., The Principles of Ion Exchange Technology, Butterworth, Heinemann,
Boston. (1991).

473
 ADVANCED FLUE GAS TREATMENT BY NOVEL DE-SOX
TECHNOLOGY OVER ACTIVE CARBON FIBERS

MASA-AKI YOSHIKAWA
R & D Department, Osaka Gas Co., Ltd. 6-19-9, Torishima, Konohana-ku,Osaka
554, Japan
E-mail: yosltikmQosakapas. co.iu

AKlNORl YASUTAKE
Nagasaki Research & Development Center, Mitsubishi Heavy Industries, Ltd.5-717-1,
Fukuhori-machi, Nagasaki 851-03, Japan

ISAO MOCHIDA
Institute of Advanced Material Study,Kyushu Universiv 6-1, Kasugakoen, Kasuga-shi, Fukuoka
816, Japan

It becomes obvious that SOX in flue gas can be removed continuously over active carbon fibers (ACF)
subjected to surface treatment such as heat treatment. Pitch based ACF shows highest activity for continuous
removing of SOX among several types of ACFs. It is considered that Pitch based ACF can make surface
hydrophobic property higher than other types. The flue gas treatment technology using ACF is a semidry
oxidative de-SOX process which is effective around temperature of 273-353 K. In addition, this process can
produce by-products such as sulfuric acid, and sulfates, and is applicable in the field of flue gas treatment to
which the conventional de-SOX technologies could not be applied for economical reasons. Therefore, this new
technology is expected to be useful in reducing environmental load not only in Japan but also globally.

1. Introduction

There is worldwide and increasingly the discharge of the sulfur dioxide with the use of
fossil fuel. Especially, the discharge in the Asia region increases recently, but the
environment countermeasure has been retarded for the problem of economical efficiency.
Though in the advanced nation, the limestone gypsum method is mainly used as stack gas
desulfurization facility, it is the method in which both cost of equipment and operational
cost are expensive.
On the other hands, it becomes obyious that SOXand NOx in flue gas can be removed at
room temperature by using active carbon fibers (ACF) subjected to surface treatment such
as heat treatment [l, 21. The flue gas treatment technology using ACF is a semidry
oxidation type de-SOX method which is effective even around room temperature. In
addition, this technology enables by-products such as sulfuric acid, sulfates, nitric acid,
and various nitrates to be recovered, and is applicable in the field of flue gas treatment to
which the conventional de-SOX method, such as the limestone gypsum method, could not
be applied for economical reasons.

2. Experimental
2.1 Characteristicsof ACF catalyst
To evaluate the basic characteristicsof de-SOX over ACF, comparison was made between
4 kinds of commercialized ACFs under the untreated condition. Table 1 shows the
comparison of the properties of ACFs. Because the structures such as an external
surface area, specific surface area, and pore size are unified in the manufacturing process,
there is slight difference in structure of raw materials. But the compositions of raw
materials significantly vary with types; phenol-type (Kuraray Chemicals co.,ltd.) or

474
cellulose-type (Toyobo co.,ltd.) ACFs have high oxygen content, PAN-type (Toho tenax
co.,ltd.)ACF has high nitrogen content, and pitch-type (Adall co.,ltd.) ACF has the largest
carbon content, which indicates that the compositions of raw materials may exert great
influence on the chemical properties of ACFs

Table 1
Comparison of the properties of ACFs made of different raw materials
Pitch type ACF PAN type ACF Phenol type ACF Cellulose type ACF
Composition formula of raw materials [CluHs$lO]n [C3NH3ln [&Hs~Oliln [CsH~oOsln
Carbon content in raw material (%) 93. I 67.9 76.6 44.4
Diameter of fiber (pm) 10 to 18 7 to 15 9toll 15 to 19
External surface area (m2/g) 0.2 to 0.6 0.9 to 2.0 1.0 to 1.2 0.2 to 0.7
Specific surface area (m*/g) 700 to 2500 500 to 1500 900 to 2500 950 to I500
Pore volume ( d g ) 0.3 to 1.6 0.22 to 1.2
Pore diameter (nm) to 2 2 to4 1.8 to 4.4 1 to 1.8
Tensile strength (kg/mmz) 10 to 20 20 to 37 30 to 40 6 to 10
Elastic modulus (kg/mm) 200t01200 7000t08000 1OOOto1500 1000 to 2000
Elongation (76) 1.0 to 2.8 to 2.0 2.7 to 2.8
PH 7 7
Ignition temperature (C) 460 to 480 470 470
Benzene adsorbing amount (46) 22 to 68 17 to 50 22 to 90 30 to 58
Iodine adsorbing amount (mg/g) 900 to 2200 950 to 2400
Methylene blue decolorizing ability to 400 to 300 to 380

. In order to improve the performances of desulfirization, high temperature heating

treatment in the inert gas, such as nitrogen, were carried out at several conditions,
described later.

2.2 Desulfbization reaction tests

Desulfirization reaction was examined in a fixed bed flow reactor at 25-50 C ,with a
flow rate 400 mumin of lOOOppm SO2, 5% 02,and 10-15% H 2 0 in N2 over 2 g of ACF
catalyst to adjust W/F (Catalyst weight by Flow rate) to 5 * 10 g midml. SO2
concentration were analyzed continuously at the inlet and outlet of the reactor using
Infrared SO2 Gas Analyzer (SOA-7000, Shimadzu Co.,Ltd.). Steady-state SOz conversion
was observed when the SO2concentration at the outlet of the reactor became stable.

3. Results and Discussion

Fig. 1 shows the change of de-SOX performances of commercialized 4 kinds of
untreated ACFs with elapse of time. All the ACFs used have the same specific surface
areas of about 1,000 m2/g. The reactions were performed under the same condition that
the inlet SO2 concentration was 1000 ppm, the reaction temperature was 25 C ,and water
content 1O%, which means super-saturated condition.

475
 100

80

8 60 -A.Phenol-ACF
0

8 40

20

0
0 5 10 15
Reaction time(h)

Fig. 1 Basic characteristicsof de-SOX reactions over various ACFs

The Y axis in the figure indicates the SO2 breakthrough ratio (outlet SO2 concentration /
inlet SO2 concentration). The figure shows the de-SOX activities of these 4 kinds of
ACFs in decreasing order: pitch-type, PAN-type, phenol type, and cellulose-type ACFs.
The figure also revealed that the de-SOX ratio was kept stable even after 15 hours.
Like the above, the results of investigation on basic characteristics of de-SOX reaction
over ACFs revealed that the pitch type ACF is most suitable for the performance. Because
the 4 kinds of ACFs are scarcely different in structures including the specific surface area
and pore size, chemical properties of them seemed to affect the performance of de-SOX
reactions. Considering that the pitch type ACF contains the largest carbon content and
the least content of both oxygen and nitrogen in the raw material, and ACF with less
oxygen groups may be suitable for de-SOX reactions, we investigated to achieve high
perfomance of ACFs.
Therefore we devised to reduce oxygen groups of ACF as much as possible in order to
accelerate this reaction. So we tried to remove the oxygen groups on the surface by
carrying out high temperature heating treatment of pitch-type ACFs in the nitrogen. Fig.
2 shows the result of de-SOX when ACF with 1,060 m2/gof specific surface area was heat
treated in the temperature range from 600 to 9OOOC in the nitrogen. The reaction condition
was nearly the same as the basic characteristics mentioned above, but the W/F (contact
time) was reduced by half to 2.5 * g midml.

476
 I00

80

8 60
8
3 40

20

0
0 5 10 15
Reaction time(h)

Fig. 2 Change of the de-SOX performance by heat treatment temperature

The figure clarified that the de-SOX performance was improved as the heat treatment
temperature rose, and in 900 "C, it reached the performance over 80%. The de-SOX
performance was stable after the reaction initiation over 50 hours, and as the product
material, sulfuric acid was continuously recovered. The material balance between
removed SO2 and recovered sulfuric acid agreed in the propotion over 98% .

In the de-SOX over ACF, the reaction mechanism is considered like the following.
First, SO2 is adsorbed onto the active sites on the surface of ACF, and oxidized with
oxygen in flue gas into SO3 . Sulfuric acid is formed by hydrating the SO3 with water in
flue gas immediately,then finally the sulfuric acid is absorbed in water and desorbed from
the s u h c e of ACF. In the avobe-mentionedreaction proces, the desorption of sulfric acid
from the ACF surface is considered to be a rate controlling step, and therefore it is most
important to increase the desorption rate. The fact that condensation of water in flue gas
onto ACF doesn't inhibit adsorption of SO2 but accelerates the reaction is the
confmation-finished by the different experiment.
On the other hand, it is also known that pore diameter and surface area of ACF do not
decrease by the heat treatment at 1000C [3]. The heat treatment increases markedly the
electrical conductivity of ACF [4]. The image potential of an SOz molecule with the
graphitic pore wall should stabilize the dipole-oriented structure in micropores of ACF,
then it intensifies the catalytic activity for the SO2oxidation.
The oxygen function groups of ACF surface inhibits the desorption of generated
sulfh-ic acid by its hydrogen bonding ability, and the de-SOX reaction is inhibited.
Therefore, the heat treatment of pitch-type ACFs in the nitrogen removes surface oxygen
functin gruops to make the hydrophobic surface, and it becomes the high activity catalyst
for the desulfurizationreaction.

477
 References
1. Mochida et al., Continuous Removal of SO2 in the Model Flue Gas over PAN-ACF
with Recovering Aqueous HzSO~, Chemistry Letters, pp. 1899 - I902 (1 993)
2. Mochida et al., Oxidation of NO into NOz over Active Carbon Fibers, Energy & Fuel,
Vol. 8, No. 6,1341 (1994)
3. T. Ohlubo et al., High-temperature treatment effect of microporous carbon on orderd
structure of confined S02, Chem. Phys. Letters, 329, pp.71-75 (2000)
4. M. Ruike et al., J. Colloid Interface Sci. 207, pp.355 (1998)

478
ADSORPTION OF NATURAL GAS COMPONENTS ON ACTIVATED CARBON
FOR GAS STORAGE APPLICATIONS

I.A.A.C. ESTEVES, M.S.S.LOPES,P.M.C. NUNES, M.F.J. EUSEBIO, A. PAIVA, J.P.B. MOTA*

Departamento de Quimica, Centro de Quimica Fina e Biotecnologia, Faculdade de CiPncias e
Tecnologia. Universidade Nova de Lisboa. 2829-51 6 Caparica,Portugal
Single-component adsorption equilibria on activated carbon of the n-alkanes (4x4 and of the
odorant tert-butyl mercaptan were measured at the operating conditions expected in a large-scale
facility for adsorbed natural gas (ANG) storage. The experimental data were correlated success-
hlly with the Adsorption Potential theory and collapsed into a single temperature-independent
characteristic curve. The obtained isotherm model should prove to be very useful for predicting
the adsorption capacity of an ANG storage tank and to size and optimize the operation of a
carbon-based filter for ANG applications.

1 Introduction

It is well-known that the nature of NG composition has a negative impact on ANG tech-
nology due to the deterioration of the adsorbent capacity on extended cyclic operation. '
Although this problem has been addressed by some authors, either experimentally or the-
oretically, the single- and multi-componentadsorption data reported to date, as they apply
to natural gas, are still rather scarce.
The main purpose of the present work is twofold (i) to report an extensiveset of single-
component adsorption isotherm data of the more common natural gas components on ac-
tivated carbon, and (ii) to present a means of extrapolatingthe measured data to higher
alkanes in order to be able to span the whole composition of a typical natural gas. There
is experimental evidence that for the n-alkanes series this can be done using the Adsorp-
tion Potential theory, as demonstratedrecently by Holland et al. on Westvaco BAX- 1 100
carbon, and assumed previously by US.^.^

2 Theory

The Adsorption potential theory has been used extensively for correlatingadsorption equi-
libria on porous adsorbents. It states that for a given gas-solid system the volume of the
adsorbed phase, W, is a function of the adsorption potential,
w =qvm = W@),
- (1)
where q is the amount adsorbed at equilibrium and V,( 7') is the adsorbed-phase molar
volume, which is temperature dependent. The adsorption potential is defined as
7 = R T ln(P,/P), (2)
where P is the equilibrium pressure at temperature T and P , ( T ) is the saturated vapor
pressure of the adsorbate. At high pressure, however, P and P, should be replaced by
the corresponding fugacities, f and fs,to correct nonideal gas behavior. The functional
relationship between & and W is characteristic of the particular gas-solid system and is
usually referred to as the characteristiccurve.
*Author to whom correspondence should be addressed (e-mail: [email protected]).

479
 The Potential theory is especially usell for predicting single-component adsorption
equilibria from a limited set of experimentalmeasurements. Firstly, the characteristiccurve
is temperatureindependent accordingto the theory. Therefore,only adsorptionequilibrium
measurements at one temperature are necessary to obtain the characteristic curve, and this
is sufficient to describe the adsorption at all temperatures for the same gas-solid system.
Secondly, in many cases the theory can be generalized if an affinity coefficient, p, is used
as a shifting factor to bring the characteristiccurves of all gases on the same adsorbent into
a single curve. Under this assumption, Eq. (1) can be replaced by

which is characteristic of a given adsorbent and can be equally applicable to all adsor-
bates. The practical applicabilityof this feature is limited by our ability to predict affinity
coefficients,but this has been greatly improved by Wood' who has recently published an
extensive compilation of experimental p values for gases and vapors and showed that p is
highly correlated to the molecular parachor accordingto
p = 8.27 x 10-3(Parachor)o.90. (4)
Under conditions below the critical temperature, T,,of the adsorbate, Vm is assumed
equal to the molar volume of the saturated liquid at system temperature. Above T , the
adsorbedphase is ill-defined, and this has led to different approximationsbeen proposed for
V,. Likewise, in the supercritical region the concept of vapor pressure does not exist and
f s in Eq. (2) must be replaced by a pseudo-vaporpressure.In the present work we followed
the suggestionsof Agarwal and and estimated f , and Vm at temperaturesabove
Tc as follows:

where f c is the critical pressure of the adsorbate, Tb and vb are the temperature and molar
volume of the liquid adsorbate at the normal boiling point, and 52 is an estimate of the
thermal expansion coefficient of the adsorbate in a superheated liquid state, which for light
gases is SZ 2 . 5 ~ K-'.

3 Experimental section

The carbon selected for study is a coal based, high activity (109 %CTC), extruded carbon
(2 mm diameter pellets) from Sutcliffe Speakman Carbons Ltd (SSC). The carbon was
characterized using N2 adsorption at 77 K and mercury porosimetry. An analysis of the
chemical nature of the carbon surface using the PZC method and Bohem titration revealed
that its surface is mostly clean of acidic groups.
The adsorption equilibria of the n-alkanes C 1 x4and tert-butyl mercaptan (TBM) on
SSC carbon were measured gravimetrically using a high-pressure magnetic suspension
balance from Rubotherm, driven with in-house-developedLabview software. TBM is cur-
rently one of the more widely used odorants for natural gas. All gases employed are CP-
grade and were obtained from Praxair Portugal Gases, SA.
The adsorption data of the n-alkanes were measured accordingto standard procedures:
the adsorption chamber was pressurized with the pure gas at approximately the desired
pressure and then sealed. Then, the pressure and the change in the mass (weight) of the
carbon sample were recorded until equilibrium was reached. On the other hand, obtaining
reliable adsorption equilibriumdata for TBM at the very low concentrationsusually found
in natural gas (0-25 mg/Nm3) was a more challenging task. In this work they were deter-
mined by recording the weight change of the carbon sample upon saturation with various
feed concentrationsof TBM diluted in an inert carrier. The measurements consisted of ex-
posing the adsorbent sample to a single calibratedmixture of TBM and helium (6.74ppmv)
circulating at different pressures. This was realizable in practice because in our adsorption
apparatusthe pressure can be spanned and accuratelymeasured over a wide range. Prelimi-
nary adsorption tests with pure helium showed that its adsorption was negligible compared
with that of TBM,even at the low partial pressures considered in this work, and that helium
could be safely used as the carrier gas without the need for correction of the results.
The flow of the feed stream and the pressure in the adsorption chamber were both fixed
using a mass flow controller, located at the inlet port, and a back-pressure regulator placed
at the outlet. Each measurement was started by circulating the feed gas through the sample
chamber and continued until the recorded weight change attained a steady value. These
experimentstook considerablymore time than those of the alkanes using the conventional
gravimetric procedure. During each experiment the feed stream to the adsorption cham-
ber was routed through a coil immersed in a second temperature-controlled thermostatic
bath in order to maintain consistent isothermal conditions at the entrance to the adsorption
chamber. When the sample weight attained a steady value, the inlet port was slowly closed
while maintaining a constant pressure in the chamber by means of the back-pressureregu-
lator. Then, a final reading of the weight change of the sample was taken. Since this final
reading was taken with the inlet port closed, the measured value of the sample weight is
not erroneouslymasked by the drag exerted by the circulatinggas during the equilibration
phase.

4 Results and discussion

The gravimetric measurements, as well as all other conventional adsorption methods, rather
than giving the total amount adsorbed, q, give just the amount in excess, qex,with respect
to that of bulk gas occupying the same volume as the adsorbed phase. However, to apply
the adsorption potential theory the required variable is q and not qex.In this work, q was
estimated from the measured qex value by assuming that the adsorbed-phase volume is
, ~ = 0.850 cm3/g, determined from N Z
equal to the total pore volume of the ~ a r b o n Vs
adsorption at 77 K:

4 = qex + VSP& (6)

where pg is the density of the bulk gas phase at the equilibrium pressure and temperature.
The success of Wood's correlation, Eq. (4), led us to use it for estimating the /? values
for the adsorbates C3+, which are clearly vapors at ambient temperature;on the other hand,
the /? values for the adsorbates C2-, which have critical temperatures close to or below
ambient, had to be obtained experimentally by superposing the corresponding adsorption
data onto the characteristic curve of the higher alkanes. We believe that the reason for
this is the difficultyin calculatingdensities of condensed,adsorbed phases at temperatures
approachingand exceeding critical temperatures.

481
 0

-1
h
M
\
-2

1-3
-$ 4

-5

-6
0 1 2 3 4
I 0-4 x Q (J/mol)

Figure I . Experimental characteristic curve of adsorption. W(#),

for normal alkanes q-C4 and TBM on ac-
tivated carbon: 0 , CI (81= 0.47); x, C2 (82 = 0.65); A, C3 (83 = 0.76); 0 , C4 (84 = 0.93); 0,TBM
(&BM = 1.05).

Figure 1 displays the experimental characteristic curve obtained and the B values em-
ployed to generate it. The existence of very little scatter in the data demonstrates that the
isotherms of the various adsorbates were successfully correlated as a single temperature-
independent characteristic curve. This fact corroborates the applicability of the Potential
theory to the carbon under study.
To obtain a workable isotherm model for future process simulation, the experimental

0.20
o . 1 6 v i

0.12
0.20
0.08

0.10
0.04

p 0.00 0.00
2 0 2 0 4 0 6 0 8 0 1 0 0 0 10 20 30 40
P 0.50

0
0.20
0.30 . 4 O F I

0.10

0.00 0.00
0 5 10 15 0 1 2 3 4

P (bar)

Figurc 2. Singlc-component adsorption equilibria of the normal alkanc scrics q - C 4 on SSC carbon. Symbols
denote experimental data and lines represent predictions from the D-A isotherm model with parameters given by
Eq. (8) and pl = 0.47, 82 = 0.65.83 = 0.76,84 = 0.93. The average error over 364 data points is 6.8%. The
temperatures (K) are (Cl) 284,298,323; (C2) 287,299,314,324; (C3) 284,298,323; (C4)273,288,298,324.
 0.09
I I
0.06

b
0.03
I
0
0.0 0.8 1.6 2.4
104xP (bar)

Figure 3. Low-pressure adsorption equilibrium for TBM on SSC carbon. Symbols represent experimental data,
and lines represent predictions from the D-A isotherm model with parameters given by Eq. (8) and &BM = 1.05.
The avetagc error over 2 1 points is 4.1%.

characteristic curve was successfully fitted using the Dubinin-Astakhov (D-A) equation.
For a pure component,this isotherm model can be written as
W = Wsexp(-vb"), (7)
where Wsis the specific pore volume and y and n are parameters. The regressed values are
Ws= 0.856cm3/g, y =4.402~ (J/mol)-'*, n = 1.09. (8)
The isotherm model is compared directly to the experimental adsorption data in Figures 2
and 3. The plots show that there is good agreement between model predictions and the
experimental data for all of the adsorbates. Furthermore, the regressed value of W s is in
excellent agreement with the total pore volume determined from N 2 adsorption at 77 K. It
is also somewhat surprising to observe that most of the TBM measurements were in the
nonlinear range of the isotherms even at the low concentrationsconsidered in this study.

Acknowledgements

The authors thank SutcliffeSpeakmanCarbons for providing the carbon samples and char-
acterizing them using N2 adsorption and mercury porosimetry. This work was supported
by the European Community through contract No. ENK6-CT-2000-00053. I. E. acknowl-
edges the financial support of FCT/MCT (grant PRAXIS XXL/BD/19832/99).

References

1. T.L. Cook et al, in Carbon Materialsfor Advanced Technologies,ed. T.D. Burchell

(Pergamon Press, 1999).
2. C.E. Holland, S.A. Al-Muhtaseb and J.A. Ritter, Ind. Eng. Chem. Res. 40,338 (2001).
3. J.P.B. Mota, AIChE J. 35,986 (1999).
4. J.P.B. Mota and A.J.S. Rodrigo, Ind. Eng. Chem.Res. 39,2459 (2000).
5 . G.O. Wood, Carbon 39,343 (2001).
6. R.K. Agarwal and J.A. Schwarz, Carbon 26,873 (1988).
7 . S. Sircar, Ind. Eng. Chem. Res. 38,3670 (1 999).

483
 PREDICTION OF BREAKTHROUGH CURVES FOR TOLUENE AND
TRICHLOROETHYLENE ONTO ACTIVATED CARBON FIBER

JEE-WON PARK,SANG-SOON LEE

Dept. of Applied Chemishy, Dongduk Womens University

YOUNG-WHAN LEE, DAE-KI CHOI*

Environment & Process Technology Div., Korea institute of Science and Technology*

E-mail :dkchoi@st,re.kr

Adsorption dynamics of toluene and trichloroethylenewere investigated using a fixed bed of ACF at
the isothermal condition. Breakthrough curves of experimental and theoretical results reflected the
effects of the experimental variables such as the partial pressures for adsorbate and the interstitial
bulk fluid velocities. It is obvious that high feed concentration also makes steeper breakthrough than
low feed concentration. Also when the feed concentration is lower, a longer time is required for
reaching the column saturation. A dynamic model based on the linear driving force mass transfer
model, which was suggested to represented column dynamics, showed that the prediction agreed
remarkably with the empirical data of breakthrough curves.

1. Introduction

To date, activated carbon is the most universal adsorbent for VOCs control. However,
some disadvantages for the application of activated carbon include its flammability,
difficulty in regenerating high-boiling point solvents and required humidity control. On
the positive side, activated carbon fiber has uniform size and dimension, higher adsorption
capacity, faster adsorption and desorption rates than activated carbon, and ease of
handling [1,2]. These features obtain adsorptive system size reduction and added adsorbed
vapor selectivity. In these respects, activated carbon fiber, as alternative to activated
carbon inefficiencies, is an excellent micro-pore adsorbent.
In this study, we focused ow attention on investigating the adsorption dynamics in
column packed with activated carbon fiber. By optimizing the breakthrough curve data
with a mathematical model, effective overall mass transfer coefficient was obtained. And
it can be given reasonable predictions compared with the experimental data of
breakthrough curve.

2. Methods

2. I Materials
Pitch-based activated carbon fiber, Nano-10 (Nanotechnics Co., Korea) [3], was used
as the adsorbent. Using the adsorption of nitrogen at 77 K, the BET specific area for
Nano- 10 was measured by an automatic sorption analyzer (Quantachrom Autosorb- 1).
Nano-10 has a BET surface area of 1453 m2-g-',total pore volume of 0.4096 cm3*g-'and

484
mean pore diameter of 22 A. Prior to the experiments, the adsorbent was kept in a drying
vacuum oven at 423 K for more than 12 h to remove impurities.

2.2 Fixed bed operation

A fixed bed adsorption unit was examined to study adsorption dynamics in column
packed with activated carbon fiber. The packing characteristics of adsorption column and
properties of packed adsorbent are given in Table 1. Details of the equipment and
operating procedures were described in the previous publication of Yun and Choi [4].
Table 1. Packing characteristics and properties of column used.

lo-10

Particle density, kgm 1.6

Bulk density, kgm- 0.136
Particle porosity 0.21
Bed void fraction 0.35
Column diameter, cm 1.7
Column length, cm 3.1

3. Data Handling

3. I Mechanisms and overall mass transfer coeficient

Generally, there are four steps are included in the mass transfer mechanisms of the
adsorption process. These steps are fluid-film transfer, pore diffusion, surface adhesion,
and surface diffusion. The rate of surface adhesion for physical adsorption on the surface
of porous adsorbents is very rapid, enough to be assumed instantaneous relative to the
other transfer rates [ 5 ] .
The complex heterogeneous structure of fiber is replaced with an effective
homogeneous structure having an equivalent rate of mass transfer. Coefficient of mass
diffusion through the effective homogeneous activated carbon fiber is defined as the
effective diffisivity (D&. Deffis then used to explain the mass transfer of adsorbed vapors
through the original activated carbon fiber complex structure [6].

Film transfir coeficient:

-
(Nsc)0.33
0.357(NRe)0.64 DM
, 3<N~,<2000 (1)
kf =E 2%

The Knuhen difision coefficient:

485
 The surface d i e i o n coeflcient: D,= D,,,
exp(-E,, 1R T ) (3)

The effectivepore dfisivity: (4)

The resulting equation of the overall mass transfer coefficient could be expressed as
follows:

3.2 Mathematical modelfor dynamics

To represent the adsorption dynamics in column, the linear driving force (LDF)
approximation model for overall mass transfer coefficient was applied. The LDF model
for gas adsorption dynamics is frequently and successfully used for analysis of column
dynamics because it is simple, analytic, and physically consistent [7]. We assumed that the
velocity of the gas in column is constant, and radial temperature, concentration and
velocity gradients within the bed are negligible in this model. With the ideal gas-law
assumption, the set of equation for this work is as follow:

Mass balancefor component i:

a2yr i3yl i3yl RT I-& an,

- 01.-
az2 +-+-+-+-
az at P E
PpX=O

Mass transfer rates:

dn.
)= k , ( q0 - n i ) (7)
at

Adsorption equilibrium :
mP
Toth N=
(b + P' )"I

where m, b, and tare isotherm parameters.

Boundary condition at z=O and z=L andfor P O :

486
 Although the model equation included the axial dispersion coefficient (DL), plug flow
was approximated by assigning a very large value to the Peclet number (S/DL). This is
because the effect of axial dispersion is quite negligible in a small column and the model
with the second derivatives can give more stable numerical results.

4. Results and Discussion

Activated carbon fiber has much steeper breakthrough curve than that of other
adsorbent in that adsorbate vapors in an activated carbon fiber with only micropores reach
the adsorption sites through micropores without the additional diffusion resistance of
meso- or macropores, and it was required for very short time to reach C / CO= 1 [8].
Breakthrough curves of toluene and trichloroethylene reflected the effect of fed
concentration and flow rate of solvent-laden gas as shown in Figure 1. It is obvious that
high feed concentration also makes steeper breakthrough than low concentration. This is
due to the fact that the higher bulk concentration can make the greater concentration
driving force between bulk and solid phase. And, when the feed concentration is lower, a
longer time is required for reaching the column saturation.
Adapted a nonlinear regression technique, the results of calculation were given the
following relationships:

kToluene = 0.091x [P]o.920[~]o.5@

kTrichloroeihyletw = 0.0712 x [P]o.769[~]o.671

In Eqn. (1 1) and (12), k is overall mass transfer rate coefficient, P is partial pressure (in
kPa), and u is interstitial bulk fluid velocity (in d s ) in column. Because of the lack of
experimental data, the correlation can only be available in the experimental range, the
partial pressure of 0.1 1 to 0.35 kPa for toluene and trichloroethylene, and the interstitial
bulk fluid velocity of 0.22 to 0.46 d s .

References

1. Chung D. D. L., Carbon Fiber Composites, Buttenvorth-Heinemann,Boston, 1994.

2. Blocki S. W., Hydrophobic Zeolite Adsorbents: A Proven Advancement in Solvent Separation
Technology. Environ. Prog. (1993), pp. 226-238.
3. Kim Y. M., Korea Patent No. 10-1999-0047363, (1999)
4. Yun J. H. and Choi D. K., Equilibria and Dynamics for Mixed Vapors of BTX in an Activated
Carbon Bed, AIChE Journal, 45 (1999), pp.751-760.
5. Ruthven D. M., Principle of Adsorption and Adsorption Processes, John Wiley & Sons, 1984
6. Mehrad L., Mark J. R. and Massoud R. A., Modeling Effective Diffusivity of Volatile Organic
Compounds in Activated Carbon Fiber, Environ. Sci. Technol.,35(2001), pp.613-619

487
7. Sircar S. and Hufton J. R., W h y Does the Linear Drving Force Model for Adsorptio Kinetics
Work?, Adrorption, 6 (2000), pp. 137-147.
8. Motoyuki S., Activated CarbonFiber: Fundamentals and Applications, Carbon, 32 (1994), pp.
577-586.

1.o

0.8

0.2 $= 0.6
U

0.4

02

0.0 0.0
0 0 10 20 30 40 50
Time Elapsed, m h Time Ela& mm

1.o

0.8

9
U
0.6 ?-
U

0.4 0.4
0 0.15mls
0 0.25 mls
0.2 0.2 0.35 mls

0.0 0.0
0 20 40 80 80 40 80 80
Time Elapsed, mi. Time Elaplcd. m h

Figure 1. Experimental and predicted breakthrough curves for toluene and trichloroethylene

488
 CATALYTIC REDUCTION MECHANISM OF NITRIC OXIDE OVER
ACFdCOPPER CATALYST

S.J. PARK,B. J. KIM, AND Y. S. JANG

Advanced Materials Division, Korea Research Institute of Chemical Technology,
P. 0.Box 107, Yusong, Taejon 305-600. Korea
E-mail: [email protected]

A widely used method for activated carbon fibers (ACFs)/Cu catalyst is nitric oxide reduction.
which includes the process involved whenever nitric oxide (adsorbate) is brought into contact with a
solid (adsorbent), and then reduced into nitrogen and oxygen. In this work, the catalytic reduction
mechanisms of NO over ACFdCu prepared by electrolytic copper plating has been studied. It is
found that copper content on carbon surfaces increases with increasing the plating time. As an
experimental result, nitric oxide is converted into the nitrogen and oxygen on ACFs and ACFdCu
catalyst surfaces at 5OOc. Especially, the surfaces of ACFdCu catalyst were found to scavenge the
oxygen released by catalytic reduction of NO, which could be explained by the presence of another
nitric oxide reduction mechanism between ACFs and ACFdCu catalysts.

1 Introduction

There has been considerable effort to solve and diminish ecological and
environmental problems such as air pollution, acid rain, soil pollution, and etc [I]. It has
been well known that the acidic gases, such as, SO2 and NO, are major components of
air pollution. Carbons have been used as not only a reducing agent itself but also
catalyst support for the purpose of removing pollution sources such as NO,, SO,, and etc
[2-41.
The catalytic conversion of NO into O2and N2 by using carbon with the metal (Ni,
Fe, Cu, K, Ca, Pd, Sn, and Na) prepared by impregnation and precipitation with an
aqueous solution of metals has been investigated for a long time. It has been found that
the carbon supported with copper is the one of the best catalytic effect to transform NO
into O2and N2 either in the presence of O2 or in the absence of 0 2 [5-71. In this system
above, oxygen over the ACF/Cu is used as not only a modifier to play a role in changing
metallic copper into copper oxides known as active species for NO removal, but also an
accelerator to produce CO to be reducing agent, resulting in promoting NO conversion
~~91.
The electro-copper plating from an aqueous solution of a copper ion has an
electrochemical mechanism involving the electron transfer with an external source of
electric current. Copper is autocatalytically deposited on the carbon surface as a
function of reaction time [101.
In the present study, a novel surface treatment of microporous carbon, so-called
electro copper plating which has been used in the application fields is provided to
introduce copper onto activated carbons. And, the catalytic reduction mechanisms of
NO over ACFdCu prepared by electrolytic copper plating have been studied.

489
2 Experimental

2. I Materials and Sample Preparation

Virgin ACFs, AW2001, supplied by Taiwan Carbon Co. was used in this work. To
sensitize the surface of ACFs were immersed in a solution composed of acidified tin
chloride solution at concentration of 5 gel-' for 5 min at room temperature, and followed
by a rinse with distilled water. The ACFs were successively subjected to a copper
electroplating in copper plating solutions with varying treatment time.

2.2 Measurements
The identification and amount of copper on the activated carbons were evaluated
by energy dispersive X-ray analysis (EDX, Hitachi, S-246ON) and inductively coupled
plasma (ICP, Atomscan 25, Thermo Jarrell Ash Co.), respectively. For the test of NO
conversion of ACFs, fixed-bed column made of quartz was used after N2purge at 300C
for 30 min before reaction. The outlet gas was checked by a NOx meter (Ecom@A plus,
ECOM America Ltd.) with time on stream for the catalytic reaction NO-Cu in the
presence of oxygen. The conversion of NO was estimated from the concentration of NO
at the input and output of the column [1 1,121.

3 Results and Discussion

Fig. 1 shows the copper contents on the ACFs/Cu catalysts. It is found that virgin
ACFs show no copper peak. Cu5 and CulO, however, show copper peaks, and it is
increased with plating time. This result shows that copper contents are depended on
plating time if other conditions are same.
Fig. 2 shows the NO conversion on the ACFs/Cu catalysts by copper electroplating.
Nitric oxide is converted into the nitrogen and oxygen on ACFs and ACFs/Cu catalyst
surfaces at 500C. As an experimental result, it shows that all of the samples prepared
by electrolytical copper plating represent higher NO conversion than that of unplated
one. In the presence of 02,the higher the content of copper on the ACFs is, the higher
the conversion is, until the reaction time on stream of about 20 h.
Figs. 3 and 4 show the X-ray diffraction profiles of the ACFs/Cu catalysts before
and after catalytic reactions, respectively. The copper metal (Cu), which the diffraction
patterns revealed around 26 43 and 50 on ACFs/Cu, is oxidized to CuzO (28 36 and 42)
during NO catalytic reduction process. The surfaces of ACFs/Cu catalyst are found to
scavenge the oxygen released by catalytic reduction of NO, which can be explained by
the presence of another nitric oxide reduction mechanism between ACFs and ACFs/Cu
catalysts.

490
 ACFdCulO

2 111 1.10 1.1 .#a .

.U ,.om 7.1 ..
1 *.P
-
110 %.la
-
%la .
.
la *.la
-*.lo-" .I *.la la 1
. u.1. Y.U

Figure 1. EDX (energy dispersive X-ray) data on ACFs/Cu catalysts by Cu electroplating.

Y
s
0.
1000
-
- ACFs
ACFICu5
ACF/CulO
750
C
.-
0
Y

E
8
L
500

8
2 250

0
0 50 100 150 0

Reaction time (min)

Figure 2. Time dependence of the NO conversion on the ACFs/Cu Catalysts by copper electroplating (reaction
temp.: 500C).

It is clearly seen that the copper species of Cu5 and CulO before NO test is only
metallic copper. And the CuO peak intensities increase with increasing the reaction time
on stream, which is implied by the fact that the sintering of CuO is taking place. From
the results above, the state of metallic copper is changed into copper oxides by oxygen
in this system, resulting in high NO conversion.

491
 I
0 20 40 0
2 Theta (degree)

Figure 3. X-ray diffraction profiles of the ACFslCu catalysts by electrolytic copper plating: (a) ACFs, (b)
ACFslCuS, (c) ACFslCulO.

I I
0 20 40 60
2 Theta (degree)

Figure 4. X-ray diffraction profiles after the NO-ACFslCu catalysts reaction at 500C: (a) ACFs, (b) ACFsICuS,
(c) ACFsKulO.

4 Conclusion

In this work, the copper content and the net heat of adsorption due to the amount of
copper content on ACFs are increased with increasing the plating time, resulting in
improving the NO conversion as the reaction time on stream increases. As the
experimental results above, it is seen that the copper metal deposited on the ACFs
appears to be increasing the ability of NO removal for the catalytic reactions of NO-Cu
and NO-C in the presence of oxygen. From which, it is clearly found that CulO is a very
prominent catalyst for removing NO. Consequently, it is said that electrolytical copper
plating on ACFs is a useful method in the preparation of the AC/Cu catalysts for NO
conversion.

492
References

1. S. Calvert and H.M. Englund, Handbook of Air Pollution Technology, John Wiley &
Sons, New York (1 984).
2. S. J. Park, In: Interfacial forces and fields: theory and applications, Ed J. P. Hsu,
Marcel Dekker, New York (1999).
3. S. Lowell and J. E. Shields, Power surface area and porosity, Chapman & Hall,
London (1993).
4. B. J. Park, S. J. Park, and S. K. Ryu, Removal of NO over copper supported on
activated carbon prepared by electroless plating, J. Colloid Znterface Sci. 217 (1999)
pp. 142-145.
5. G. Horvath and K. Kawazoe, Method for the calculation of effective pore size
distribution in molecular sieve, J. Chem. Eng. Jpn 16 (1 983)
6. P. Denton, A.Giroir-Fendler, Y. Schuurman, H. Praliand, C. Mirodatos, and M.
Primet, A redox pathway for selective NO, reduction: stationary and transient
experiments performed on a supported pt catalyst, Appl. Catal. A: General 220
(2001) pp. 141-152.
7. T. W. Chung and C. C. Chung, lncrease in the amount of adwqtion on modified activated
carbon by using neutron flux irradiation,Chem. Eng. Sci. 54 (1999) pp. 1803-1809.
8. E. Ruckenstein and Y. Hu,Catalytic reduction of NO over Cu/AC, tnd. Eng. Chem.
Res. 36 (1997) pp. 2533-2539.
9. S. J. Park and J. B. Donnet, Evaluation of the distribution function of adsorption site
energies based on the Fermi-Diracs law in a monolayer,J. Colloid Interface Sci. 200
(1 998) pp. 46-5 I .
10. S. 3. Park and K. D. Kim, Adsorption behaviors of C02 and NH3 on chemically
surface-treatedactivated carbons, J. Colloid tntet$ace Sci. 212 (1999) pp. 186-189.
11. Z. Zhu, Z. Liu, S. Liu, and H. Niu, Catalytic NO reduction with ammonia at low
temperatures on V205/ACcatalysts: effect of metal oxide addition and S02, Appl.
Catal. B: Environmental 29 (2001) pp. 267-276.
12. K. E. Noll, V. Gounaris, and W. S. HOU,Adsorption technology for air water
pollution control, Lewis, Michigan (1992).

493
 NO REMOVAL OF ACTIVATED CARBON FIBERS
TREATED BY CU ELECTROPLATING

S. J. PARK, J. S. SHIN, AND J. R. LEE

Advanced Materials Division, Korea Research Institute of Chemical Technology,
P. 0. Box 107, Yusong, Taejon 305-600, Korea
E-mail: [email protected]

In this study, activated carbon fibers (ACFs) deposited by copper metal were prepared by
electroplating technique to remove nitric oxide (NO). The surface properties of ACFs were
determined by FT-IR and XPS analyses. Nd77K adsorption isotherm characteristics, including the
specific surface area, micropore volume were investigated by BET and t-plot methods respectively.
And, NO removal efficiency was confirmed by gas chromatographic technique. From the
experimental results, the copper metal supported on ACFs appeared to be an increase of the NO
removal and a decrease of the NO adsorption efficiency reduction rate, in spite of decreasing the
BETS specific surface area, micropore volume, and micro-porosity of the ACFs. Consequently, the
Cu content in ACFs played an important role in improving the NO removal, which was probably due
to the catalytic reactions of C - N W u .

1 Introduction

Atmospheric pollution caused by SO, compounds from the combustion of fossil

fuels, 03,COZY volumetric CH compounds, and NO, has become as matter of growing
world wide concern in recent years. Nitrogen oxides contribute to acid rain and
photochemical smog, and they can cause respiratory problem. Therefore, nitrogen
oxides should be controlled and removed from the source in order to keep the earth
clean and as-received for our health [l-51.
Usually improvement combustion processing or after combustion treatment are
used nowadays for NO reduction. However, they are some problems as like a complex,
expensive setup, harmness gas emission, and corrosion metal. In recent years, to
overcome these problems, some researchers have reported that NO is reduced more
effectively use of the adsorption characteristics of activated carbons (ACs) and
activated carbon fibers (ACFs) [6-81. Also, some researchers are studying for NO
reduction using metal supported ACs and ACFs by impregnation, metal plating,
deposition, and so on [9-131. However, metal supporting methods on ACs and ACFs in
a second and their NO removal efficiency are not studied yet systematically.
The objective of the present work is to investigate the surface and adsorption
properties of the electrolytic Cu-plated ACFs by using pH, FT-IR, and BET isotherms
and to discuss the influence of NO removal on plating time and copper content of the
ACFs studied.

2 Experimental

2.1 Materials and Electrolytic Cu-plating

For the present investigation,the ACFs were supplied by Taiwan Carbon Co. AW2001

494
(weight: 45 g/m2, thickness: 0.3 mm, and specific surface area: 2121 m2/g). Cu-deposited
ACFs were prepared by Cu electroplating at constant velocity and same electric current.
Electrolytic copper plate and graphite club were used as the anode and cathode, respectively.
The samples were denoted in as-received, Cu-2, Cu-5, Cu-10, and Cu-20 with different
plating times, such as, 0,2,5, 10, and 20 seconds, respectively.

2.2 XPS Analysis and ACFs Pore Structure

The surfaces of ACFs were analyzed using a VG Scientific LAB MK-I1 X-ray
photoelectron spectrometer (XPS). The spectra were collected using a M g b X-ray source
(1253.6 eV). The pressure inside the chamber was held below 5x10-' torr during analysis.
Both survey XPS spectra are recorded at a 45 O take-off angle.
Nitrogen isotherms were measured by using an ASAP (Micromeritics) at 77K.
Prior to each analysis, the samples were outgassed at 573K for 10 - 12 h to obtain a
residual pressure of less than torr. The amount on nitrogen adsorbed was used to
calculate specific surface area, and the micro pore volumes determined from the BET
equation [ 141 and t-plot method [151, respectively. Also, the Horvath-Kawazoe model
[161was applied to the experimental nitrogen isotherms for pore size distribution.

2.3 NO Conversion Test

For the present experiment, gas chromatograph (DS6200 model, Donam Korea) and
thermal collect detector were used to measure NO conversion. Reactor temperature was
sustain constantly at 500C using P.I.D. temperature controller (UP-350,Yokokawa) and
gas flow rate was maintained 10 mVmin by mass flow controller; GMC 1000, MKS). All
samples were heated under Helium purge at 150C for 1 hour to remove residual H 2 0
before NO conversion test. The NO conversion was determined from the concentration of
NO at the outlet reactor. Prior to each analysis, NO standard curve was gained using 300,
600, and 1000 ppm NO gas.

3 Results and Discussion

3. I Surface Characteristics and ACFs Pore Structure

XPS spectra of as-received and copper-plated ACFs are shown in Figure 1. As

expected, the as-received ACFs show a Cjs peak and a substantial 0 , s peak at 284.6 and
532.8 eV, respectively [17]. Ols peak of as-received is probably due to the intrinsic
surface carbonyl or carboxyl groups. Otherwise, Ojs peak of copper-plated ACFs are
probably due to the deposition of more active forms, such as, CuO, Cu(OH), and copper
metal on the inactive carbon. From the XPS results, the carbon content of copper-plated
ACFs is decreased when the ACFs were plated with copper, whereas the oxygen and
copper contents of ACFs were higher than those of as-received. The active groups on the
ACFs surfaces after copper plating can contribute to change the polarity and fhctionality
of the ACFs surfaces.

495
 I
1200 900 Mx) 300 0
Binding energy (ev)

Figure 1. NO conversion of the electrolytic Cu-plated activated carbon fibers.

An understanding of porosity of an adsorbent can be achieved by the construction of

an adsorption isotherm. The adsorption isotherms of the adsorbents were measured lo6 -
10 tom range of relative pressure at 77K.
Although the results are not presented here, the samples are approximately the Type 1
isotherms having well developed micropores according to the BETS classification [ 14,171.
It is evident that most of the pore volumes of the samples are filled low relative pressure,
indicating these ACFs are highly microporous. After a sharp increase of relative pressure,
the isotherms show a very small increment in the further adsorption as indicated in the
fraction of microporosity of Table 1.

Table 1. Textural properties of the electrolytic Cu-plated activated carbon fibers.

as-received cu-2 cu-5 cu- I0 cu-20

BET surface area
1709 1620 1534 1060
(mag-)
Micropore volume
1.145 0.906 0.83 1 0.809 0.526
(cm.g-)
Total pore volume
1.216 0.997 0.9 12 0.889 0.572
(cm3.g-)
Fraction of 92
94 91 91 91
Micropore (%)
Average pore diameter
12.7 12.9 12.7 12.9 12.3
(A)

Table 1 shows the structural properties of the ACFs studied. As shown in Table 1, all
samples have welldeveloped micropore and microporosity of Cu plated ACFs are decreased
as compared that of as-received. Also, the volume Gslction of micropom is decreased with
increaSing of Cu content on ACFs. It can be seen that the copper plating shows pore-blocking
phenomena owing to deposition of copper metal on ACFs [13]. Specific surfice areas are
decreased result h m micropore volume h m t-plot method was decreased.

496
3.2 NO Conversion
NO conversion of the electrolytic Cu-plated ACFs at 500c is shown Figure 2. NO
conversions of all the specimens are outstanding to the as-received ones. In addition, it
can be seen that NO conversion on the as-received is sharply decreased within a few hours,
and it reaches at about 14% after 180 min [3]. The reasons for this result are attributed to
the NO-C reactions with the time on stream at 500 c . Whereas, NO conversions of Cu-2
and Cu-5 are slowly decreased within a few hours, and then, reaches at a constant
conversion [181. Also, in the case of Cu-20, NO conversion is almost constant at more
than 99%, which can be attributed to the existence of optimal reduction of NO into O2and
N2 between NO-Cu and NO-C reactions. It is then considered that the metallic catalytic
system of the adsorbents is very predominant in which copper metal is activated for
removing NO at 500 c.

4 cu-20

0 50 100 150 200 250 300

Reaction time (min)

Figure 2. NO conversion of the electrolytic Cu-plated activated carbon fibers.

4 Conclusions

In this work, the effects of the surface properties and pore structure of the copper
plated ACFs on NO removal were investigated. As experimental results, tHe copper metals
appeared to be an increasing of the characteristics of NO removal in spite of the
decreasing of adsorption properties, including specific surface area, micropore volume,
and microporosity of the ACFs. In addition, NO conversion was linearly increased with
increasing the copper content in the ACFs. It could be concluded that the copper content
in ACFs played an important role in improving the NO removal, due to the catalytic
reactions of C-NO-Cu. It is then noted that copper electro-plating treatment on ACFs is a
useful technique for the removal of NO with a viewpoint of high functional metallic
catalytic system.

497
References

1. K. E. Noll, V. Gounaris, and W. S. Hou, Adsorption technology for air water

pollution control, Lewis, Michigan (1992).
2. S. Calvert and H. M. Englund, Handbook of Air Pollution Technology, John Wiley &
Sons, New York (1984).
3. B. J. Park, S. J. Park, and S. K. Ryu, Removal of NO over copper supported on
activated carbon prepared by electroless plating, J. Colloid Interface Sci. 217 (1999)
pp. 142-145.
4. Z. Zhu, Z. Liu, S. Liu, and H. Niu, Catalytic NO reduction with ammonia at low
temperatures on V20dAC catalysts: effect of metal oxide addition and SO2, Appl.
Catal. B: Environmental 29 (2001) pp. 267-276.
5. Y. Hu and E. Ruckenstein, The catalytic reation of NO over Cu supported on mesocabon
mimbeadsofullrahighsurfkcearea,J.Catal. 172(1997)pp. 110-117.
6. E. Ruckenstein and Y. Hu,Catalytic reduction of NO over CdAC, Ind Eng. Chem.
Res. 36 (1997) pp. 2533-2539.
7. P. Denton, A.Giroir-Fendler, Y. Schuurman, H. Praliand, C. Mirodatos, and M.
Primet, A redox pathway for selective NO, reduction: stationary and transient
experiments performed on a supported pt catalyst, Appl. Catal. A: General 220
(2001) pp. 141-152.
8. S. K. Ryu, S. Y. Kim, N. Gallego, and D. D. Edie, Physical properties of silver-containiing
pitch-based activatedcarbon fibers, Carbon37 (1999) pp. 1619-1625.
9. T. W. Chung and C. C. Chung, Increase in the amount of adsorption on modified activated
carbon by usingneutron flux irradiation,Chem. Eng. Sci. 54 (1999) pp. 1803-1809.
10. S. J. Park and K. D. Kim, Adsorption behaviors of C02 and NH3 on chemically
surfaceaeatedactivatedcarbons,J. Colloid InterJace Sci. 212 (1999) pp. 186-189.
11. S. Lowell and J. E. Shields, Power surface area and porosity, Chapman & Hall,
London (1993).
12. S. J. Park and J. B. Donnet, Evaluation of the distribution function of adsorption site
energies based on the Fermi-Diracs law in a monolayer, J. Colloid Inteflace Sci. 200
(1998) pp. 46-5 1.
13. S. J. Park and Y. S. Jang, Interfacial characteristics and hcture toughness of
electrolytically Ni-plated carbon fiber-reinforced phenolic resin Matrix composites, J.
Colloid Interface Sci. 237 (200 1) pp. 9 1-97.
14. S. Brunauer, P. H. Emmett, and E. Teller, Adsorption of gases in multimolecular
layers, J. Am. Chem. SOC.60 (1938) pp. 309-319.
15. B. C. Lippens and J. H. de Boer, Studies on pore systems in acknowledgements. V.
The t-method, J. Catal, 4 (1965) pp. 3 19-323.
16. G. Horvath and K. Kawazoe, Method for the calculation of effective pore size
distribution in molecular sieve, J. Chem. Eng. Jpn 16 (1983) pp. 470-477.
17. S. J. Park and Y. S. Jang, Effect of micropore filling by silver and anti-bacterial
activity of activated carbon fiber surfaces treated with AgN03, J. Korean Ind Eng.
Chem. 13(2002) pp. 1-7.
18. V. I. Parvulescu, P. Oelker, P. Grange, and B. Dehon, NO decomposition over
biocomponent Cu-Sm-ZSM-5 zeolites, Appl. Catal. B: Environmental 16 (1 998) pp.
1-17.

498
THE APPLIANCE STUDY OF 02-PSA IN THE OXYGEN ACTIVATED SLUDGE
PROCESS

S. H.LEE, P. S. YONG, H. M. MOON AND D. S. PARK

Daesung Ctyogenic Research Institute, Daesung Sans0 Co., Ltd.
781-I Wonsi-dongAnsan Gyeonggi-do 425-090 Korea
E-mail: shlee@astopia. co.kr

The oxygen activated sludge process is well known as a treatment method for effectively
high-concentrationwastewater. But it is true that many wastewater treatment facilities have avoided
the introduction of the 02-PSA equipment, because it is expensive and its operating cost is high.
With a view to design and operate the oxygen activated sludge system for more competitive cost, we
carried out the experiment and process analysis on the wastewater treatment process including
02-PSA. As a result, we found it is more cost-effective to use 60% concentration oxygen than high
concentrationoxygen for the wastewatertreatment.

1 Introduction

The oxygen activated sludge process using oxygen for wastewater treatment is being
generally used in U.S., Japan, etc., and it is well known that its efficiency is also good. But
the reason why the oxygen activated sludge process is not used on a large scale is that the
cost of oxygen feed is higher than the cost of air feed.
Generally, the methods to provide oxygen into wastewater are divided into two. One of
them is to inject liquid oxygen through a pipe line or tank-lorry, and the other is to inject
the gaseous oxygen by 02- PSA(Pressure Swing Adsorption)[l]. In this study, in the
case of injecting oxygen produced by the 02-PSA, the oxygen recovery of PSA for
oxygen concentration, the amount of oxygen transferred into wastewater, and wastewater
treatment efficiency for oxygen concentration were examined.

2 Theory

2. I 0 2 - PSA Process
The maximum purity of 02-PSA is 95% due to the separation of argon in produced gas,
and usually it is 90-93%. The purer the 02-PSA is, the lesser its oxygen recovery
becomes, so the flux of product gas decreased. That means the drop in economical
efficiency of 02-PSA[2]. The concentration of 02-PSA can be generally changed by
tuning P f f ratio(Purge/Feed ratio), and the more PIF ratio increases, the smaller the flux
of product gas becomes. So, when producing highly concentrated oxygen, the power cost
to produce I kg oxygen increases[3].

2.2 Oxygen Activated Sludge Process

The oxygen activated sludge process is a method applying highly concentrated oxygen to
activated sludge process, and it can be used more widely and has higher treatment
efficiency comparing with the activated sludge process. This oxygen process can maintain

499
high sludge concentration, and its retention time is shorter, and it occupies smaller place
comparing with the air activated sludge process. This enables one to reduce the initial
construction cost. Generally, the oxygen process is 2-3 times of MLSS(Mixed Liquor
Suspended Solid) comparing with the air process, and even operating it for 1/3 of the
general retention time, its overall treatment ratio is same as that of air process or better.
Therefore, the oxygen process makes an efficient operation possible in short retention
time, and it provides more oxygen per unit power and makes an operation possible in the
high concentration of dissolved oxygen as well. Furthermore, because the power with
small shearing stress can operate in high concentration of dissolved oxygen, the sludge
which is well deposited and concentrated can be produced, and the amount of excess
sludge decrease. Generally the activated sludge process has a slow treatment speed and
produces much excess sludge. Because this excess sludge treatment costs a lot, it should
be considered by terms of operation expense. The minimal DO(Disso1ved Oxygen) of
activated sludge is about OSmgA, and it is well-known that 1-2mgA or more would be
proper. The OUR(0xygen Uptake Rate) increases at around the entrance of drain, and
decreases low near the exit of drain[4].

Rehm Activated Sludge

I"
w&ta s w e
Figure 1. The schematicdiagram of oxygen activated sludge process using 02-PSA

The Closed type in the oxygen activated sludge process is a method increasing the
efficiency of oxygen by shutting the upper part of aeration tank tight and separating the
aeration tank into 3-4 parts. As a diffuser, the surface aerator is used, and the 02-PSA is
used as an oxygen producer. The open type is a method not covering the top of aeration
tank,and its aeration tank is divided into several parts like the closed type.The open type
used a diffuser specially manufactured to increase the efficiency of oxygen, etc., and
comparing to the closed type, its initial installation cost is low and its power expense
needed to dissolve oxygen is also low, and it is easy to maintain. But the efficiency of
oxygen seems to be worse than the closed type.

500
3 Methods

3. I Oxygen Activated Sludge Pilot Plant Test

A pilot plant with 1Om3/&y wastewater treatment capability was manufactured. It consists
of raw collection tank,aeration tank,settling chamber and effluent tank, and the aeration
tank is divided into 3 parts. The top of aeration tank can be opened and closed, and the
flow of bubble and the shape of flock in the aeration tank can be observed through a
window. And by attaching a return pipe, materials can be returned from the settling
chamber to the aeration tank,and its drainage becomes easy by attaching a drain valve on
each part of aeration tank.Surface aerator was attached on each part of reaction tank, and
each aerator has an inverter to control the speed of revolution, and the height of aerator
can be changed. Analytical devices are 3 DO meters, MLSS meter, pH meter, and ORP
meter, etc. These are connected to a computer to automatically save their data, and for
sample analysis, CODcr, CODw, TN, TP, etc. were measured with Hach(USA) reagent.
Wastewater used for the experiment was from the joint wastewater disposal plant in the
complex of H-pharmaceutical company, and in the wastewater CODM,, was
1000-3000mg/l. All test devices and analysis equipment were installed indoors, and
oxygen aeration was used in the daytime when wastewater is discharged, and in the
nighttime, the air aeration was used to increase the activity of microbes.

3.2 02-PSA Pilot Plant Test

Based on the results of adsorption test and breakthrough test, etc. oxygen with 93% purity
was produced by using ZMS(Zeo1ite Molecular Sieve) 5A adsorbent, and its flow was
1.5Nm3/h.The air input flow was 22Nm3/h,and the recovery rate was aimed to be 30%.
The 02-PSA pilot plant consists of a compressor for air supply, freezer, after-cooler for
cooling, air receiver tank, 02-product tank,air actuated valve, control box, etc. On the
adsorption tower, 4 RTDs are installed from its upper part to its lower part to change
temperature when adsorbing or removing it. To control feed, purge and product, MFM
was installed for each function.

4 Conclusion

Fig. 2 shows the efficiency of wastewater treatment and the recovery of OZPSA
according to the oxygen concentration. As Fig. 2 shows, the higher the oxygen
concentration becomes, the lower the recovery of 02- PSA becomes. But, the efficiency
of wastewater treatment was about 96% in 60% oxygen concentration, and it was 88% in
90% oxygen concentration. This shows that the highly concentrated oxygen is not
essential for the oxygen activated sludge process. Namely, when using a product gas with
60% oxygen concentration, the efficiency of wastewater treatment is better. Fig. 3 shows
the amount of oxygen transferred into wastewater and total dissolved oxygen according to
the oxygen concentration. The oxygen transferred amount into wastewater gradually
increases when its solubility increases because of high oxygen concentration. But it is
observed that the total dissolved oxygen does not increase continually according to the
oxygen concentration, and it decreases when the oxygen concentration rate reaches about
60%. This is because when the oxygen concentration increases, the recovery rate of

501
02-PSA decreases, so the amount of product gas is reduced, and the total oxygen amount
decreases. That means even though the oxygen concentration is high, the recovery of
02-PSA can decrease, then the amount of product gas produced in the 02-PSA becomes
reduced. Finally the total oxygen dissolved in wastewater is reduced.
In the case of pharmaceutical wastewater, it shows when its oxygen concentration was
60%, the amount of total dissolved oxygen is the most. According to the condition and
concentration of wastewater and the recovery of 02-PSA used in other wastewater
treatment plants, the optimal, and most economical oxygen concentration and operating
conditions will be decided. This study shows that oxygen with 90% concentration or
higher doesn' t have the best treatment efficiency and economical value.

Figure 2. Oxygen recovery ratio and Wastewatertreatment efficiency for oxygen concentration

F ' -A- TOTAL DISSOLMD OXYGEN J -

K.18 -

0,.
4- OXYGEN TRANSFERRED AMOUNT "
A*AA4Tkk .
fo
F,,
z - A' I A 'A - 1s
3
0
. A
. 9
w
w
3 12-
A
9
- 1122

;
A
A
0 ' . X3
ut
A'
$00.9
$ 9- -1 0 .99 a
w
K
-06 5:
I- 0 3 -03 A
z
0
9 00 00 I-
a, 40 60 80 1w
0
0, CONCENTRATION[%]

Figure 3. Oxygen transferredamount and total dissolved oxygen for oxygen concentration

502
References

1. R.E. Speece, and J.F.Malina, Applications of commercial oxygen to water and

wastewater system, Center for research in Water Resources, 1973,~.
16
2. G.Reib,Gas Separation&Purification,8(2),(1994),95-99
3. A.P Hinderink, and H.J.Van der Kooi, Chem. Eng. Sci, 51(20), (1996) 4693
4. G.Tchobanoglous, Wastewater Engineering treatment, disposal and Reuse, 3rd.Ed.,
MetcalfkEddy, Inc, 1993, p.100

503
 APPLICATION OF SOLID ADSORBENT FOR VOC MONITORING SENSOR

OK JIN JOUNG AND YOUNG H A N KIM

Dept. of Chemical Engineering, Dong-A University, Pusan 604-714, Korea
e-mail: [email protected]

Coating the molecular sieve on one electrde of a quartz crystal oFcillator is conducted to fabricate a
gas sensor in this study. The detection performance of the sensors with two different pore sizes of
the molecular sieve is examined from their implement in the determination of methanol, ethanol and
n-hexane concentration in air. From the performance test, the ruggedness, sensitivity and selectivity
of the sensor are investigated. From the variation of oscillation frequency of the sensor, it is found
that the variation is proportional to the concentration of the organic vapors. The sensors show good
recovery to the initial state when fresh air is provided. While the activated carbon coated sensor
dcesnt show selectivity among different organic materials, the molecular sieve coated sensor gives
selective detection from methanol and ethanol mixed vapor.

1 Introduction

One of modification techniques of the quartz crystal sensor for the determination of
organic vapor is organic film coating on the surface of the electrode. Yet the organic film
coating has problems with its implementation and maintenance in spite of its good
selectivity. Because the adjustment of film thickness is difficult, the reproduction of the
film-coated sensors is also dificult and the sensitivity of the sensors depends largely upon
the film condition. In addition, they are not utilized in the environment of large
temperature variation.
Carbon coated sensors have high stability and endure harsh environment in application,
though they are not selective to organic substances. Kim et al. [l] investigated the
stability of a carbon-coated sensor to find that it is very stable and easy to regenerate by
simple thermal treatment. The sensor is fabricated from the process under very high
pressure, and therefore it is difficult to fabricate and expensive. Chao and Shih [2]
introduced a simply prepared carbon sensor. The coating of fullerene (C,) is readily
made from dropping a drop of dissolved carbon in solvent and drying without any binding.
However, its maintenance is not simple and hard to be utilized in harsh condition. In
another study [3], hllerene derivative has been applied to detect various organic
molecules.
Coating the molecular sieve on one electrode of a quartz crystal oscillator is conducted
to fabricate a gas sensor in this study. The detection performance of the sensors with two
different pore sizes of the molecular sieve is examined from their implement in the
determination of methanol, ethanol and n-hexane concentration in air. From the
performance test, the ruggedness, sensitivity and selectivity of the sensor are investigated.

2 Experimental

2. I Sensor Preparation
An AT-cut quartz crystal having base frequency of 8 MHz (Sunny Electronics Co.,
Korea) is utilized to prepare the sensor. The electrode of the crystal is silver finished. Two

504
pore sizes of 4 and 5 A molecular sieve (Yakuri Pure Chemical Co., Japan) having pellet
size of 1/16 inch is used, and for binder the phenol resin (Novolac, Dong Kwang Chem.
Co., Korea) having average molecular weight of 400, purity of 99% or above and particle
size of 200 meshes with 13 % hardener of hexamethylenetetramine is added.
The coated crystal is cured in a drying oven of 160 "C for an hour to make the particles
bound on the surface. An SEM photograph of the crystal is exhibited in Fig. 1. In the
figure,dark particles are binder resin and grayish particles are molecular sieve. At lower
left comer there is a large particle of molecular sieve and resin bound.

Figure 1. An SEM photograph of molecular sieve coated sensor.

2.2 Experimental Procedure

The experimental setup is demonstrated in Fig. 2. Air is provided from a blower and a
portion of the air is vent through a bleeding valve. An objective organic vapor is obtained
by passing some of the air to be fed to the measuring cell through an Erlenmeyer flask of
0.2 L capacity containing liquid of the organic substance.
The adjustment of organic vapor concentration is carried out from manipulating the
valves installed in the line of gas flow. The flow rate of mixed gas flow during frequency
measurement is maintained at 0.4 Llmin. all the time by controlling the valve installed at
the bottom of the flow meter. Experiment is conducted at room temperature.

Figure 2. Experimental setup.

505
3 Results and Discussion

The detection performance of the molecular sieve coated sensor is examined from the
measurement of frequency variation while different concentrations of organic vapor
contained air are contacted to the sensor surface. While organic vapor contained air flows
continuously with constant flow rate of 0.4 L/min., the variation of frequency is
monitored and the outcome is converted to the organic concentration. In order to examine
the process of adsorption and desorption of the organic vapor on the molecular sieve
coated on the sensor surface, fresh air and organic substance contained air are alternately
provided.
Figure 3 shows the experimental result of the determination of methanol with 5A
molecular sieve coated sensor. In the beginning, fresh air is supplied, and then methanol-
contained air is provided. The concentration of methanol is measured by taking the
sample of 1 mL with a sampling syringe from the gas flow line and by analyzing it with a
gas chromatograph. The decrease of frequency indicates that the coated molecular sieve
adsorbs the methanol and there is a mass change on the sensor surface. After the decrease
is settled, fresh air is fed to desorb the methanol adsorbed on the molecular sieve. The
frequency increases again, and it retums to a little more than the initial value. Because the
experiment is conducted in a flow system, some drift in the initial frequency is
accompanied. After the frequency with the air is kept at a steady value, sample flow of a
higher concentration is introduced to determine frequency variation with different
concentration of methanol. These measurements are conducted for four different
concentrations of methanol- contained air in flow. The incompleteness of frequency
recovery with fresh air supply is linearly increased for the last three runs, but the
frequency shift with methanol is distinctive and proportional to the concentration of
methanol.

-800 awrc.inppm
-m
0
75

2 M x ) 4 w o 6 o o o 8 o o o
t i (set.)
1
Figure 3. Variation of 6equency with various concentrationof methanol contained air.

In case of ethanol, the sensitivity of the sensor to ethanol-contained air is about 1/5 of
that of methanol. Because the 5A molecular sieve is designed for the separation of the
molecular size of methanol, ethane and propane or smaller, the 5A molecular sieve coated
sensor gives much less sensitivity with ethanol contained air. Unlike other solid particle
coated sensor, such as activated carbon sensor, the molecular sieve coated sensor has
selectivity to the molecular size of detecting material. The outcome presented in Fig. 3
indicates that the sensor coated with 5A molecular sieve satisfactorily discern methanol
vapor from bigger molecules, but it does not separate from small molecules. When 4 A
molecular sieve coated sensor is implemented to detect methanol, the same result of
measuring ethanol with 5 A molecular sieve sensor is yielded as shown in Fig. 4. In other

506
words, provided that two sensors of the 4A and 5%
molecular sieve coating are implement
at the same time, the determination of methanol from the mixture of various species
having different sizes of molecule is available.
When it is considered that the experiment of this study is conducted in a flow system,
the ruggedness and sensitivity of the sensor are satisfactory for the application in a harsh
environment. Because the sensor is produced fiom thermal treatment, it can be
implemented in dry environment unlike an organic film sensor and the regeneration of the
sensor by heat treatment is possible.

-250
oncinm

-mo taa 20m 3ooo urn0

tirn ( u c l

Figure 4. Frequency variation of 4A molecular Figure 5. Methanol concentration and frequency

sieve coated sensor. shift.

Figure 5 shows the relation between methanol concentration and frequency shiA with 5
A molecular sieve coated sensor. This indicates the measured frequency shift gives the
concentration of organic vapor.

4 Conclusion

In order to give selectivity to a solid particle coated quartz crystal sensor having
ruggedness and stability with easy maintenance, the coating of two difference pore sizes
of molecular sieve is exercised here. The applicability of the sensor as a gas sensor is
examined by monitoring the adsorption and desorption of organic vapors on the
molecular sieve. The measurement is conducted in a system of flowing air to prove the
ruggedness of the sensor in practical implementation.
From the frequency measurement for different concentrations of organic substance
contained in air, it is shown that the sensor can detect the concentration of different sizes
of organic materials. A good recovery of the sensor to the initial condition is shown, and
the ruggedness is proved from implementing the sensor in a flow measurement system.
The sensor is produced from readily available materials with a simple process, and it is
easily regenerated fiom heat treatment.

5 Acknowledgements

Financial supports h m the Korea Science and Engineering Foundation and the Korea
Research Foundation are gratefully acknowledged.

507
References

1. J.-M. Kim, S.M. Chang, Y. Suda and H. Muramatsu, Stability study of carbon
graphite covered quartz crystal, Sensors and Actuators, A 72 (1999) pp. 140-147.
2. Y.-C. Chao and J.-S. Shih, Adsorption study of organic molecules on fullerene with
piezoelectric crystal detection system, Anal. Chim.Acta, 374 (1998) pp. 39-46.
3. J.-S. Shih, Y.-C. Chao, M.-F.Sung, G.-J.Gau, C.-S. Chiou, Piezoelectric crystal
membrane chemical sensors based on fullerene C60, Sensors and Actuators B 76
(2001) pp. 347-353.

508
 PSA FOR SOLVENT RECOVERY WITH USY-TYPE ZEOLITE;
AN EXPERIMENTAL AND A SIMULATION STUDY

KAZUYUKI CHIHARA, TAKASHI KANEKO,TADAHIRO AIKOU, SHIREN ODA

Department of industrial chemistry, Meiji university
1-1-1 Higashimita, Tama-kx, Kawasaki,214-71
e-mail [email protected]

Recovery of solvent vapor (dichloromethane, CH2C12)from air by pressure swing adsorption (PSA)
was studied, using two columns packed with high silica zeolite and resin as adsorbent. Gravimetric
measurements were made for dichloromethane on the adsorbents. Computer calculations were
carried out to simulate the experimental results using the Stop40 method to show the calculated
results coincide well with experimental results. The method is usefil to predict the performance of a
solvent recovery system operated by PSA.

1 Introduction

The possibility of the PSA process to separate eficiently and economically the
components of some mixed-gas streams have made the process to be used many other
applications, as suggested in many references.
As the PSA process is operated at lower temperature, without heating in regeneration,
it was thought not to be advantageous as for solvent recovery. As a result, it was not fully
studied yet, compared with the recovery by thermal regeneration.
In this study, the solvent recovery by PSA with resin adsorbent and high silica zeolite
was tested. As a first step, CH2C12was chosen as solvent and the following items were
examined.
1) Isotherms and dynamics of CH2C12 adsorption for commercial resin and high silica
zeolite adsorbents were measured by a flow type gravimetric method.
2) CH2C12recovery from air was tried to obtain basic data, using a PSA system, consisted
of two columns.
3) A numerical calculation of PSA operation were performed with the Stop 62 Go method
and results were compared with the experimental results.

2 Experimental

2. I Adsorption equilibrium
Adsorption equilibrium was measured by a flow type gravimetric method
A high silica zeolite sample PQ-USY (SiO2/AI2O3=7O)and resin sample EZOP,
respectively, were used as adsorbents. These samples were pretreated to remove moisture,
in nitrogen gas stream @I2) at 473 K for 12 hours for PQ-USY, at 353K for 12 hours for
EZOP, respectively.
The weight of this sample loaded in a hunging quartz basket was recorded by a chart
recorder through a differential transducer.
The nitrogen gas was introduced through a bubbler to load CHzC12solvent and the gas
was diluted again with nitrogen stream. Thus obtained solvent-laden gas of constant flow

509
rate (1 00 cm3/min.) and various constant concentrations was fed to the measuring device
as feed gas through two mass flow controllers (for diluting line and for concentrating
line). Concentrations of feed gas were determined by a gas chromatograph (SHIMAZU
GC-7A). Finally, the weight increase due to adsorption for each sample (EX20P and PQ-
USY) at equilibrium was measured.
Adsorption isotherms of CH2C12on each sample at 303.15 K are shown in Fig. 1.
Solid lines in Fig. 1 are those correlated with Langmuir equation. Equilibrium parameters
for each sample are shown in Table. 1. From the results, the sample PQ-USY was found
to be better to use for PSA operation.

,-52.0
Table 1. Equilibrium parameters
of adsorbents
kEX20P J Saturated amount Adsorption equilibrium
adsorbed : q%/d constant : K [i/kPa]
PQ-USY 0.260 0.1250
EX2OP 0.220 0.0718

fig. 1 Adsorption Isothcnns of PQ-USY and EX2OP 9" KP

Langmuir equation : q =,li~p
for CHZCIZ(303.15K)

2.2 Experimental procedure

PSA unit was consisted of two columns. First, a column pressure was raised by stream
of gaseous feed to 200 kPa (pressurization step). Then product gas was taken out through
pressure control bulb (adsorption step). The other column was evacuated and regenerated
at lower pressure (40 kPa) by feeding purge or backwashing flow, taken from the product
gas. The roles of two beds were reversed periodically. Pressurization, adsorption,
evacuation and backwashing purge were set to 20, 700,20 and 700 seconds, respectively.
Solenoid valves were controlled by a key board programmer to achieve the sequence. Air
from compressor passed through a silica gel column to make the air dry. The dried air was
introduced to a bubbler to load CH2C12vapor and was mixed with the dried air stream for
dilution. The solvent-laden air of constant flow rate (1000 cm3/min.) and constant
concentration (0.8 ~01%)was fed to one of the columns as feed gas. The flow rate of the
gas was measured by a mass flow meter.
The adsorption columns were of 27 mm in i.d. and 300 mm in length. A certain
amount of regenerated adsorbent was packed into each column, as shown in Table. 2.
Temperature was set to 303.15 K by use of a constant temperature bath. A thermocouple
was inserted into the column at 15 cm from the bottom. The concentration changes in
product gas were measured until a cyclic steady state was obtained. Concentration of
feed, product and exhaust gases were determined by a gas chromatograph (SHIMAZU
GC-7A) equipped with a thermal conductivity detector.

510
 Table 2. PSA operating conditior
Organic solvent CH?CI?
(concentration) (8000ppm)
Dimension of the column 27.0 mm O X 300 mm
Amount of adsorbent 42.0 g (EXZOP)
in the column 51.2 g (PQ-USY)
Feed flow rate 1000 ml/min
Pressure swing rate 2.0 atm , 0.4 atm
Ambient temperature 303.15 K

3 Result and Discussion

3.1 Break through curves and simulation

As for ylksav (Bed density/Overall mass transfer coefficient), one of the key
parameters of PSA simulation, the simulated break through curves for various ylksav by
Stop-Go method were compared with the experimental curve as shown in Fig.4 to obtain
a reasonable value. There, break through operation and simulations were equal to
pressurization step and adsorption step of the PSA operation. Langmuir constant and
saturated amount adsorbed were equal to the result of adsorption equilibrium experiment.
The other parameters and operation conditions were set to be equal to those of the PSA
operation as Table. 2. From this comparison, in the case of PQ-USY, suitable value of
y h a v were 150 sec, in the case of EX20P, this is 20 sec.
PQ-USY EX20P
1 . 2 ~ ~ - 1.2 1

0 100 200 300 400 500

Time [min] Time [minl

Fig. 2 Comparison between experimental and simulated break through curves

3.2 PSA operation

3.2.1 Comparison between the results of PQ-USY and H 2 0 P

Fig. 5 shows the relation between yield against purity of product gas for PQ-USY and
for EX20P, respectively at a cyclic steady state. Yield is defined as follows:
Yield [-3 = Total amount of product gas per cycle / Total amount of feed gas per cycle
and the purity is defined as follows:
Purity[-]=l-C/Co

511
where C is the product concentrationof CHzC12, and Co is feed concentration of CH2CI2.
For PQ-USY and EX2OP, the larger the purge gas flow rate, the higher the purity was
obtained and the lower the yield was observed.
As there existed more data of PQ-USY around the most suitable point than EX20P, it
is clear that PQ-USY is better than EX20P in view point of solvent recovery PSA
operation.

Parameter in the diagram is

Purge feed ratio (R)
90 92 94 96 98 100
Purity [%I
Fig. 3 Yield againstpurity

3.3 PSA simulation

Experimental data were compared with the results of simulation by Stop-Go method
as shown in Fig. 4, 5. Fig. 4, 5 were obtained for dynamic steady operation, regarding
four conditions of the purge gas ratio(R), i.e., R=0.50,0.97,1.46 and 2.72, respectively.
Fig. 4 shows changes of concentrations of CH2C12 in product gas with time in cases
with various purge ratios. Fig. 5 shows the relation between yield and product purity.
From these comparisons, it can be said that the experimental behavior can be simulated
rather well by this simulation method.

-
0.25

-
A
0.20
0.15 Simulation
1.46
0.05 2.72
Purge feed ratio (R)
0.00
0 10 20 30 40 50 60 0-
70 80 90 100
Cycle [-I purity of product gas [%I
Fig. 4 Changes in the concentrations of product Fig. 5 The relation between yield and product
gases for various purge ratio (PQ-USY) Purity (PQ-USY)

3.3.I Prediction of Axial Distribution

Since the experimental results are well simulated, this Stop-Go method can be
possible to predict the behavior in the column, such as the distribution of the amount
adsorbed in the axial direction. Fig. 6 tells the distribution of amount adsorbed at various
positions with purge. This result might be useful in considering axial distribution of the
amount adsorbed in PSA operation and to decide an effective length of column.

512
 Bottom I
50

P
Z3O
O

ZO

10

n
0 10 20 30 40 50

Cycle [-I
Fig. 6 Transition of amount adsoM at various positions with purge (Rd.97)

4 Conclusion

The product gas and the exhaust gas were measured for dichloromethane-laden air
feed by PSA system using a resin adsorbent and a high silica zeolite as adsorbent. The
PSA method for solvent recovery with each adsorbent seems to be technically feasible.
The high silica zeolite gave better performance than the resin adsorbent, as expected from
their isotherms.
Also, It was confirmed that the experimental data could be well simulated by the Stop-
Go method. Further, it might be possible, to simulate the condition within the column,
such as the distribution of the amount adsorbed, and to estimate the effect of operational
factors.

513
 AZEOTROPIC ADSORPTION OF ORGANIC SOLVENT VAPOR
MIXTURE ON HIGH SILICA ZEOLITE, EXPERIMENTAL & SIMULATION

KAZUUKI CHIHARA*, KAZUNORI HIJIKATA ,HIDEAKI YAMAGUCHI

HIROWKI SUZUKI AND YASUSHI TAKEUCHI

Department of Industrial Chemistry, Meiji Universig

Higashimita, Tam-hu, Kawasaki 214-71. JAPAN

*EL&F&M4-934-719 /E-mail: [email protected]

Fixed-bed adsorption experiments of laboratory-scale were carried out to m o v e organic solvent vapors by
several types of adsorbents. Binary adsorption equilibria of azeotmpic mixture-HSZsystems showed two
azeotropic points. Those experiment data were compared with molecular simulation by the Grand Canonical
Monte Calm (GCMC) method. Experimental results for single and binary component system, including
azeotropic mixture systems, could not be in agreement with the simulation result correlated satisfactorily. But it
turns out that the tendency of the simulation result is the same as the experiment result

1 Introduction

Discharge of organic substances into air has been strongly prohibited since some
decades ago, to preserve comfortable natural environment. Though hydrocarbons,
alcohols and chlorinated hydrocarbons had been used as the degreasing agent in
industries, these solvents may also affect to our environment, and it is necessary to
remove them from air as much as possible. This study was performed aiming at
presenting useful data for the design of adsorption processes, especially the removal of
the solvent vapors as above from air by adsorption. The soIvents used were
trichloroethylene (TCE) and iso-propanol (PA).
Molecular simulation has now become powerfil means for the study of adsorbed
molecules in high silica zeolites, and GCMC method is especially u s e l l for predicting
adsorption equilibria. However, idormation on forcefield parameters and charges are
often inadequate, even in systems where the structure is well known.

514
2 Experimental

2.1 Fixed-Bed Adsorption Experiment

Solvents used were toluene (TOL), benzene (BEN), iso-propanol (PA), n-propanol
(NPA), ethanol (EtOH), tetrachloroethylene (PCE), and trichloroethylene (TCE),
respectively, and their binary mixtures were also prepared.
The adsorbents used in this study were high silica zeolite (HSZ) (Y-type from
TOSOH Corp., Ltd Japan.).
Experiments were carried out using a flow method, for BEN-EtOH*, NPA-TOL*,
NPA-PCE*, PA-TCE*, BEN-PA, and BEN-NPA systems, respectively (*: means
azeotropic mixture).
Adsorbents were packed in a glass column of 0.1Om length and 0.0156m i.d.
Experimental conditions were as follows: linear flow rate; 0.2 d s , influent concentration;
in the range of 0.004-0.2 moVm3, and temperature; 298 K.

2.2 Molecular Simulation

Cerius2 (MSI Inc.) was used throughout the simulations. Adsorption equilibria was
carried out by GCMC method for same systems of experiments. Adsorbent model was
pure silicious Y type that was same type as experimental adsorbent. Simulation forcefield
parameters were new forcefield parameter obtained by Mellot et al"]. Solvent charges
were determined with Charge-Equilibration method, respectively.

3 Result and Discussion

3.1 Breakthrough Curves

Fig.1 shows several types of breakthrough curves obtained for PA-TCE -Y-type
zeolite system. For this system, reversal of the order of breakthrough (turn over) occurred
twice at concentrations of 0.25 and 0.75 mole fractions of P A , respectively. When the
mole fractions were 0.25 and 0.75, the mixture of two components behaved as if it was a
single component system as shown in Fig2 (B) and (D). For other azeotropic mixture
systems, the turnover occurred only once. The breakthrough curves for other systems
always showed so-called constant pattern behavior for the whole concentration range.

515
 Fig.1. Several types of breakthroughcuryes observed for IPA-TCE-Y-type system

3.2 Adsorption Equilibrium

For binary adsorption equilibrium of PA-TCE -Y-type system, the component that
showed higher amount adsorbed between the two components reversed twice as expected
from the breakthrough curves. For this system, it is reported that the azeotropic point of
vapor-liquid equilibrium is only one, but two azeotropic points were observed in the
adsorption equilibrium. When the saturation vapor pressures of each component were
almost the same, appearance of two azeotropic points were reported for vapor-liquid
equilibrium. It is thought, therefore, that the phenomenon occurred in adsorption
equilibrium for these systems can be ascribed to the fact that the saturation vapor pressure
and the boiling points of each component were almost the same. It seemed the
phenomenon was influenced by experimental temperature and concentration, because the
different result was obtained when different experimental concentration (1000-1 5OOppm,
2000-40OOppm, 5000-S5OOppm) were used. Binary adsorption equilibria for these cases
are shown in Fig.2 in the form of X-Ydiagram. Relation between mole fraction of IPA
and amount of absorbed is shown in Fig.3. (Experiment conditions are the same as the
figure of the center of Fig.2)
Even in these cases, binary adsorption equilibrium could be expressed partly by the
Markham-Benton equation.

516
 1.o

3 0.8
0
0.6
3.i
000
0;.
0.2

"."
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

X (=CO/CO-Total) X (=CO/CO-Total) X (=CO/CO-Total)

Fig. 2 Comparisonof adsorption equilibrium for experimental concentration

1000-1500,2000-4000,5000-5500ppm on PA-TCE-Y-type zeolite system

0 0.2 0.4 0.6 0.8 1.0

mole fraction of P A

Fig.3 Relation between mole fraction of P A and amount of adsorbed

4 Molecular Simulation

Fig.4 and Fig.5 show adsorption isothenns for single component systems, obtained
from fixed-bed experiments and molecular simulation, respectively. Adosorption
equilibria were simulated well. Except EtOH system, quantitative order of amount
adsorbed was good agreement with experimental data. As for BEN and TOL systems, the
amount of adsorbed for simulations were lower than experimental data. So it is necessary
to examined van der Waals parameter for benzene ring. Fig.6 and Fig.7 show adsorption
equilibria For binary component systems, obtained from fixed-bed experiments and
molecular simulation, respectively. These are examples, which show azeotropic
adsorption"]. Especially, IPA-TCE, BEN-EtOH systems show two azeotropic points.
Result of simulation shows only one azeotropic point. More investigative is necessary.

517
 -
- ECOH
PA
WA
TOL
4 F a
TCE

01
A BEN
-
BEN

om1 0 01 01 I
o.mt omt 0.01 0.1 I 10 tm
C[mol/m3] C[mol/m3]
Fig.4 Adsorption isotherms by Fig.5 adsorption isotherms by
Fixed-bed experimental Molecular simulation
1.0 I wl 1.o
.i

0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
X[=C/CO-totall X[=C/CD-totall
Fig.6 Adsorption equilibrium for Fig.7 Adsorption equilibrium for
IPA-TCE Y-type zeolite system IPA-TCE acid site model system
(experimental) (molecular simulation)

5 Conclusion

Binary adsorption equilibrium except azeotropic mixture-HSZ systems could be

correlated by Markham-Benton equation for the whole concentration range, and the break
times could be estimated well by using the Extended-MTZ-Method. For azeotropic
mixture-HSZ systems, the equilibria and the break times could be correlated and
estimated only for a part of the all concentration range. Then, two azeotropic points
appeared in the adsorption equilibrium for PA-TCE -Y-type system.
In this study, equilibria were measured using fixed-bed adsorption experiment.
Molecular simulations by GCMC method did not go much well. Furthermore,
examination is required.

6 References

1. 1. K. Chihara et al., AIChE, AMU~ Meeting, (1993)

2. 2. K. Chihara et al., AIChE,Annual Meeting, (1995)
3. 3. K. Chihara et a]., Proceeding of the 13'hInternational Zeolite Conf, (2001)

518
 VOC ENRICHMENT BY A VSA PROCESS WITH CARBON BEDS

JAEYOUNG YANG, MINWOO PARK AND JAY-WOO CHANG

R&D Center, SK Eng. & Construction Ltd, Chongro-ku.Seoul 110-300, Korea
E-mail:[email protected]

CHANG-HA LEE
Dept. of Chem. Eng., Yonsei University, Seodaemun-ku, Seoul 120-749, Korea
E-mail: [email protected]

A VOC enrichment system incorporating a vacuum desorption of carbon beds was evaluated for an
acetone recovery. The adsorbent was regenerated by vacuuming and purging the beds. The pilot
plant was composed of two carbon bed filled with Pellet-BG or BPL from Calgon Corp., a
liquid-ring vacuum pump, two desiccant beds, and a condensation system. Acetone laden air(12000
ppm) was treated with two carbon beds which emits clean air containing less than 500 ppm of
acetone. With this VSA process, acetone enriched to 8Y0,which means around 6 times as rich as the
feed concentration. The enrichment depends highly on adsorption isotherms, especially Henry
constant. As many VOCs which show stiff Langmuir-type isotherms on carbons, desorbing acetone
from adsorbents by vacuuming andor purging the beds were not likely to be as effective as heating
up the adsorbents in a sense of working capacity. Therefore, vacuum swing adsorption system turns
out to be competent in recovering high-vapor-pressure VOCs with a moderate Henry constant on a
used adsorbent. A numerical simulation results using Aspen ADSIM were also provided

1 Introduction

VOCs emitted from chemical processes need to be treated properly using several kinds of
environmental processes including oxidation, adsorption, and condensation technologies,
etc.[2]. Among these technologies, adsorption process could be the fmt candidate for the
recovery of VOCs since it can enrich VOCs in an effective way. Therefore, adsorption
beds are, in general, incorporated in the VOC recovery systems which also include
distillation towers andor condensation systems andor absorption towers to form hybrid
systems[ I].
In this study, the steam regeneration method widely used for the on-site regeneration
of carbon is not employed to avoid.some negative effects such as fire, side reactions,
fouling and so on. Carbon regeneration by vacuuming or purging has not been
comprehensively used until recently except gasoline recovery systems. Its mainly because
VOCs are strongly adsorbed components onto activated carbons and, so, hardly desorbed
only by lowering total pressure in the beds. Therefore, in order to evaluate the feasibility
of vacuum desorption for the enrichment of VOCs, four-step VSA process, which is
similar to the four-step Skarstrom cycle, had been studied with acetone as a VOC and
Pellet-BG as an adsorbent.

2 VSA process models for the simulation

VSA processes were simulated using the Aspen ADSIM, a commercial adsorption process
simulator from AspemTech Co.,with the following governing equations:

519
 a'Tg C aTg C, aTg aV,
- k g E -+ 3V,p, + E -p, -+ P - + HTC * ap(Tg - T,)
az2 y y at az

where, Ji is an adsorption rate. Along with these equations, The Darcy's law was applied
to describe a momentum balance in a bed. In addition, valve equation was employed to for
the pressure driven model. The dead volume and a knock-out drum were implemented by
void tanks.

3 Experiments
IP, * 1 ~ ~ e * ~4"3 ~ ~ a
wi =
1+ I P ~* ep6lTn * pi"
In Figure 1, the VSA Pilot is composed of the following four major sections: (1) two

aw. -
A - MTcsi(w,: -Wi)
at
vertical carbon beds. (diameter: 0.4 m, height : 1.O m); (2) liquid-ring vacuum pump with
a knock-out drum,(3) two Silica-gel dryers; (4) a condensation system incorporating a
chiller. The feed to this pilot plant is the acetone-laden air transmitted from an acetate
production process by a series of two ring blowers with cooling systems. The feed air
contains around 1.2% acetone with a variation of W.596 irregularly.
A VSA cycle was made up of four steps: pressurization, adsorption, vacuum, and
vacuum purge. The step times for the four steps are IOOs, 700-SOOs, ~ O S ,550-65Os,
respectively. As can be seen from step times, the vacuum step time is not long compared
with the vacuum purge step since acetonelcarbon system is not favorable to desorption
only by vacuuming the beds. The pressure during a vacuum purge step was around 150
todabs.), which significantly depends on purge rates.
Water was selected for the sealing liquid for which EG(Ethy1ene Glycol) could be a
candidate. EG shows a very low vapor pressure but too high solubility for acetone. As
acetone is less soluble in water than in EG, in spite of higher vapor pressure, water was
used for sealing liquid and the moisture included in the emitted stream from the pump was
supposed to be captured in dryers.

520
 Figure 1. Schematic diagram of a pilot VSA system.

4 Results and discussion

4. I Adsorption and desorption characteristics of a carbon bed.

In general, many breakthrough experiments have been done to get some breakthrough
curves under several different conditions for the sake of bulk separation or purification.
However, in the case of VOC abatement or recovery, feed rate might be the only major
operating parameters, having direct relationship with adsorption step time. Feed rate was
set to 80 NCM/hr. At this feed rate, the breakthrough happens at 15 min. after the
beginning of feed injection and the curves have a long tail until saturation, as usually
observed in VOC breakthrough curves[3].
As vacuum swing operation depends largely on the successfbl regeneration, the
performances under some different regeneration conditions were assessed to determine a
regeneration condition. Figure 2 shows that vacuum-only regeneration is not sufficient or
nearly impossible to regenerate the carbon beds, showing lower performance than even
the atmospheric purge. Therefore, it is sure that purging is essential for the regeneration
and too much purging rate deteriorate the regeneration performance because it elevates the
vacuum purge pressure. As an example, when the purge rate of 300 Vmin was employed,
the bed pressure surged to 480 torr. From these basic experiments, the regeneration
performance could be estimated through a regeneration index that shows a magnitude of
driving force for regeneration.

Table 1. Operating conditions and regeneration index of regeneration runs

B 300 480 0.888

C 200 239 1.188
D 200 760 0.374
E(base) 100 142 1 .ooo

521
 Figure 2. Regeneration curves of a carbon (Pellet BG) bed saturated with feed stream beforehand.

4.2 VSAprocess and its perjormance

The experimental and simulation results to evaluate the effects of elongated

adsorption step time together with purging step time show that more than 1100 s step time
results in polluted clean air, although the enrichment performance increases
asymptotically with step times (not shown in this paper). From the basic regeneration
experiments, the purge rate was crucial operating parameter. The corresponding VSA
results are shown in Figure 3. Deep vacuum with low purge rate or shallow vacuum with
high purge rate is not satisfactory in respect of enrichment, which means that there might
be an optimum condition within a considered boundary of purge rate. This suggests that
regeneration depends especially on partial pressure of acetone rather than total vacuum
pressure as regeneration index implies. In the long run, comparing to the TSA process for
the same purpose, a VSA process is not satisfactory in view of adsorbent productivity.
This is mainly because of characteristics of adsorption isotherm that show a strongly
favorable Langmuir type (high Henry constant). In this case, it is not easy to regenerate
adsorbents through vacuum and purge. One of the successhl commercial VSA process
includes gasoline recovery system, in which gasoline shows moderate Henry constant and
make it possible to regenerate easily by vacuum or purge. Therefore, Henry constants of
adsorption isotherms, as well as safety problems, determine the productivity of VSA
process.

522
 Figure 3. The effects of purge rate on the acetone concentration in the enriched stream and the clean air

5 Acknowledgements

The financial support of Korea Institute of Industrial Technology is gratefully

acknowledged.

References

K.,Pressure Swing Adsorption Cycles for Improved

1. Liu Y.,Ritter J. A., and Kaul B.
Solvent Vapor Enrichment.AIChE J. 46(2000) pp 540-55 1.
2. Ruddy E. N. and Carroll L. A., Select the Best VOC Control Strategy. Chem. Eng.
hog. July(1993) pp. 28-35.
3. Yun J.-H., Choi D.-K., and Moon H., Benzene adsorption and hot purge regeneration
in activated carbons. Chem. Eng. Sci, 55(2000) pp. 5857-5872.

523
 ISOBUTANE PURIFICATION BY PRESSURE SWING ADSORPTION

SANG-SUP HAN, JONG-HO PARK,JONG-NAM KIM AND SOON-HAENG CHO

Separation Process Research Center, Korea Institute of Energy Research
71-2, Jang-dong, Yusung-ku, Daejeon 305-343, KOREA
E-mail: [email protected]

This study is on the development of high-purity isobutane production from isobutane-enriched

stream by gaseous adsorption technology. Isobutane purification from C4 mixture, in which not only
isobutane, but also n-butane and several kinds of CJ olefins in small or in trace are involved, is very
difficult by a traditional distillation method because of their close relative volatilities between
constituting components. The continuous layered 3-bed process in which was comprised of six steps
as follows; pressurization-1 by the cocurrent effluent gas from the other bed, pressurization-2 by
isobutane product, adsorption, wcurrent depressurization, countercurrent blowdown, and low
pressure purge by isobutane product, was applied. From the experiment, isobutane product with over
99.9% purity and with the trace levels of olefin components could be obtained at ambient
temperature. Silver impregnated clay prefers to CMS for the removal of C4 olefins

1 Introduction

Isobutane is used as a solvent in an ethylene polymerization process, as an aerosol

foaming agent, as a polystyrene foaming agent, and also in the production of alkylate
used to enhance octane number. A large amount of isobutane is contained in LPG, and
also naphtha cracked gas contains this component. Normally, C1,Cz, C3,and C4 parafins
and olefins are present in mixtures containing isobutane. The mixtures are mainly
composed of close boiling C4 constituents and the relative volatilities of constituting
components are almost the same. Therefore high-purity isobutane purification from the
mixture using general distillation is almost impossible. One of the methods used to
produce high-purity isobutane is that olefins in the mixture are hydrogenated with
catalytic reaction, by which olefins are converted to the paraffins, and then an
isobutaneh-butane stream is separated using distillation. The concentrated isobutane is
passed through a filter to remove impurities and the product is again through a dryer. In
this method, very high capital and operation costs are incurred. Adsorption process to
produce enriched isobutane shown in previous patent [I], are limited to mixture
containing only 2 or 3 components and the separation mechanism is based on
linearhranched characteristics of the constituting components. There is no adsorption
process to separate and purifjl an isobutane, from a mixture containing more than 5
components of paraffins and olefins.
In this paper, we present the adsorptive isobutane purification method with two kinds
of adsorbents. One adsorbent is used to remove C4 olefins and the other to remove light
paraffms. We also discuss the applications of two kinds of adsorbents for the removal of
C4olefins in a viewpoint of process performance.

2 Methods

2. I Adsorbents
Two kinds of functional adsorbents are used to which one is for selective adsorption
removal of paraffins and the other for C4 olefins. The former is zeolite A and the latter

524
carbon molecular sieve (CMS) or modified clay. Adsorbents of zeolite A and CMS are
already commercialized and modified clay for olefin selective removal was prepared in
our laboratory. Physical properties of these are summarized in Table 1. Preparation
method of modified clay can be found in other literatures [2,3,4,5].
Table 1. Physical properties of three adsorbents used
Particle size & Average pore BET surface Remarks
Adsorbent
7 e area mZ/ *
Zeolite A
-10 mesh, 5.9 463 Commercial
spherical
CMS
- mes
&t%dA 6.0 339 Commercial
Modified Clay 8 ~ f e ~ ~40~ ~ 120 Prepared
*Analyzed by nitrogen with ASAP 20 10 Analyser (Micromerjtics Co.)

2.2 Measurements of adsorption equilibria and kinetics

Adsorption equilibria and uptakes were gravimetrically measured with two balances, i.e.,
Cahn 1100 microbalance and Magnetic Suspension Balance (Rubotherm Co.). Adsorbents
were loaded and regenerated in the balance under the flow of ultra high purity Helium gas
-
at 150 35OoC, then adjusted to the measuring temperature and evacuated to lo4 mmHg
by using a turbo molecular pump. Adsorbates used were as follows; propane (min.99.5),
isobutane (min.99.5%), n-butane (min.99.5%), I-butene (min.99.5%), isobutene
(min.99.5%) and t-2-butene (min.99.8%).

2.3 Purijcation unit and step configuration

Two bench-scale versions of isobutane purification units are designed and fabricated to
operate automatically. One unit has the adsorption column dimension of 5Omm(ID),
15OOmm(L) in which adsorbents of zeolite A and CMS are loaded. The other has the
adsorption column dimension of 25mm(ID), 1000mm(L)in which zeolite A and modified
clay adsorbents are loaded.
The units include three adsorption beds, a vacuum pump to regenerate each
adsorption bed, a feed buffer tank,a C4 fuel buffer tank and a product surge tank (Figure
1). This surge tank plays a role in accepting the purified isobutane fiom adsorption bed
and sending the purified gas to an adsorption bed as pressurization gas and purge gas. The
process is comprised of several steps as follows; an adsorption step where feed gas is
supplied to the bottom of an adsorption bed and the purified isobutane stream fiom the
top is sent to the surge tank, a cocurrent depressurization step by which a (partially)
purified gas in a upper part of a bed is sent to the bottom of the other bed which has
undergone a low pressure purge step, a countercurrent blowdown step where regeneration
with a vacuum pump takes place in the bed which has undergone the adsorption and
cocurrent depressurization steps, a low pressure purge step where part of high-purity
isobutane product is countercurrently flowed to the bed while the vacuum pump removes
any desorbed impurities in the bulk void of the bed, a pressurization step which is
accomplished by a off-gas stream fiom the other bed which is undergoing the cocurrent
depressurization step, finally a supplementary pressurization step in which part of
isobutane product stored temporarily in the surge tank is sent to the bed to meet the
pressure of the adsorption step. Figure 2 depicts step configuration for the continuous
process.

525
 A:feedbuffertank B:productsurgetank C:Gfueltank
Figure 1. Schematic diagram of isobutene purification unit

lsoborpnc

Figure 2. Step configuration

3 Results

3.1 Adsorption isothenns

Adsorption isotherms of propane and C, components on three adsorbents are plotted in
Figure 2. Isotherm patterns of propane, n-butane, l-butene and t-Zbutene on zeolite A are
very sharp in a low pressure region of less than lOOmmHg at 30C (Figure 3-(a)).
However isobutane isotherm shows almost the linearity in the region of experimental
pressure. In case of CMS adsorbent, isobutane cant easily penetrate into the pore and
resultantly shows a molecular exclusion effect at 30C (Figure 3-(b)). The other
components are considerably adsorbed on CMS. Uptakes of all the above C4 components
on C M S are much slower than the corresponding components on zeolite A. -
Adsorbed amount of 1-butene at 80C is four times larger than n-butane on modified
clay by silver impregnation (Figure 3 4 ~ ) ) .Such a high selectivity of l-butene over n-
butane is mainly due to the presence of Ag+ions for Ir-complexation.Adsorption amounts
of other C4 pa&ins and olefins are inferred to be almost the same with that of n-butane
and 1-butene, respectively. Uptakes of n-butane and l-butene on modified clay are surely
faster than the corresponding components on CMS. But isobutane component can be
considerably adsorbed on modified clay.

526
 2.2 , 1

i
L

-.-A
1.4 _..-.. +,.__.--.a.
__.._I

... .
o zoo um goo 8 o o i a o o 0 100 200 a00 400 5M) M)o 100 800
Pressure (mmHg)

Figure 3. adsorption isotherms of

C, components on (a) zeolite A at
30C. (b) CMS A at 30C and (c)
Ag+-impregnatedclay at 60C.

3.2 Process performances

In the first, zeolite A and CMS adsorbents were loaded in each bed as explained in the
section of purification unit. For a continuous operation at ambient temperature, adsorption
and desorption pressure were allowed to 970mmHg and 25mmHg, respectively. The
duration of a cycle was 600sec. In this case, feed composition was as a follow; methane
0-400ppm, ethane 0-400ppm, propane 0.4-4.3%, 1-butene 200-800ppmv, isobutene
0.07-O.11%, t-2-butene 200ppm, n-butane 2.6-6.7%, and isobutane 92.8-95.6%. By
these conditions, isobutane product purity and recovery were 99.9% and 84%,
respectively, and an isobutane productivity based on total adsorbent amount was
3.1mol/kg/hr. C4olefins in the product could be controlled to less than 100ppm.
In the second, zeolite A and modified clay adsorbents were loaded in each bed as
explained in the section of purification unit. For a continuous operation at ambient
temperature, adsorption and desorption pressure were allowed to 870mmHg and 25mmHg,
respectively. The duration of a cycle was 270sec. In this case, feed composition was as a
follow; ethane 0-OS%, propane 2.5-7.5%, 1-butene 0.05-0.15%, isobutene 0.1-0.25%,
t-2-butene 2OOppm, n-butane 3.5-KO%, and isobutane 86.0-89.0%. By these conditions,
isobutane product purity of over 99.99% (not detectable by gas chromatography) was
obtained with the recovery of 66%. A productivity based on total adsorbent amount was
5.2molkg/hr. The performance of this process is graphically shown in Figure 4. Modified
clay was better than CMS in point of isobutane productivity and removal efficiency of C4
olefms.

527
 99.9 -

99.8 -

99.7 -

99.6 - O

88.5 I css
SBM) 3800 4otM 4200 4400 4600
Feed flow rate (sccrn)
Figure 4. Performance of isobutane purification process with zeolite A and modified
clay.

4 Discussion

Gaseous isobutane purification process from C4 mixture by adsorption has been

developed. For the purification, two adsorbents are applied. Zeolite A is used for the
removal of paraffin components, and CMS or silver impregnated clay for the removal of
C4 olefins. The continuous 3-bed process operated at ambient temperature was comprised
of six steps, i.e., pressurization-1 by the cocurrent effluent gas from the other bed,
pressurization-2 by isobutane product, adsorption, cocurrent depressurization,
countercurrent blowdown, and low pressure purge by isobutane product.
From the process experiment, isobutane product with over 99.9% purity can be
continuously obtained with the trace levels of olefin components. Silver impregnated clay
prefers to CMS in viewpoint of product purity and productivity.

5 Acknowledgements

This work was partially supported by the National Research Laboratory (NU)
program,
the Ministry of Science and Technology (MOST).

References

1. Volles W.K. and Cusher N. A., Normal butane hsobutane separation, U.S.Patent No.
460806 1 (1986).
2. Han S.S., Kim J.N., Cho S.H., Choudary N.V., Kumar P. and Bhat S.GT., Adsorbents
for light alkane/alkene separation, The 8th APCChE Congress, Aug. 16-19, Seoul,
Korea (1999) pp. 1777-1780
3. Cho S.H., Han S.S., Kim J.N., Choudary N.V., Kumar P. and Bhat S.GT., Adsorbents,
methods for the preparation and method for the separation of unsaturated
hydrocarbons for mixed gases, US. Patent No. 6315816 B1 (2001).
4. Cho S.H., Han S.S., Kim J.N., Park J.H., Yang J.I. and Beum H.T.,Adsorbent
preparations and applications for C4 olefin separation from mixtures, Korean Patent
Application No. 2001-88953 (2001)
5. Cho S.H, Han S.S., Kim J.N., Park J.H. and Rhee H.K., Adsorptive ethylene recovery
fiom LDPE off-gas, Korean J. Chem. Eng., 19 (2002) pp. 821-826

528
 CHARACTERISTICS OF GAS SEPARATION BY USING ORGANIC
TEMPLATING SILICA MEMBRANE

JONGHO MOON AND CHANG-HA LEE+

Department of Chemical Eng., Yonsei UniversityShinchon-dong, Seodaemun-gu,Seoul, 120-749,
Korea
E-mail: leech@onsei. ac.kr

SANGHOON HY UN
Department of Ceramic Eng., Yonsei UniversityShinchon-dong, Seodaemun-gu,Seoul, 120-749,
Korea

Separation characteristics of inorganic membrane depend on pore property of top layer material of
composite membrane. Unsupported top layer materials made by TPABr (Tetrapropylammonium
bromide) and MTES(Methy1triethoxysilane) have comparatively larger adsorption capacity than
a-AlzO3 support material. It means that surface diffusion plays an important role in inorganic
membrane permeate mechanism. This tendency could be confirmed by membrane separation
experiments in mixed gas systems, composed of Nz.COZ,CH4 and Hz. In general case of single
molecule permeation experiment in inorganic membrane, light molecule, like Hz, has higher
selectivity than heavier molecules(C02 and C h . ) . But, in this study, heavier and strongly
adsorbable molecules, like CO2 and CH4,have higher separation factor(COdH2=5-7, CH&z k
4-8) due to the surface diffusion.

1 Introduction

Silica membranes have received extensive attention in recent years because of their
excellent chemical and thermal stability, especially in the application of gas separation
and catalytic membrane reactor processes. And the separation of high purity HZfrom the
mixed gas, is very important to convert the chemical energy to the electric energy, such as
fuel cells. The final objective of this study is to understand the adsorption and separation
mechanism in the MTES templating composite silica membrane, which can get high purity
Hzfrom C 0 2 and CH4 mixture.

2 Experiment

Preparation of membranes
Organic templating membranes were prepared by a dip coating on tubular a-Alz03
support or y-A1203 /a-AIZ03 composite support (outside diameter, 10mm; thickness,
1.Omm, length 1OOmm; and mean pore diameter 0.1pm) by TPABr and MTES sols. After
several times of dip coating, membranes without pinhole products were prepared.

Adsorption Measurement
Due to the adsorption property of top layer on membrane (templating sols; TPABr or
MTES), adsorption experiments should be executed before membrane permeation
experiment. And adsorption experiments were performed at the 293K-3 13K temperature
range and Oatm-O.8atm pressure range for various gases by gravitational method using
electric Cahn balance(Cahn 2000, Cahn instrument, INC). From this experiment,

529
diffusionaltime constants and equilibrium parameters were obtained.

Membrane Permeation Measurement

In this study, we used the modified Wicke-Kallenbach cell which is tubular membrane
cell type. Permeation measurements were performed in the 293K-373K, Oatm-5atm range
for H2,N2, C02 and C b . Feed gas and retentate gas were controlled by MFC(Mass Flow
Controller, Tylan Co.) and BPR(Back Pressure Regulator). Permeate gas flux was
measured by soap bubble flow meter, MFM (Mass Flow Meter, Teledyne Co.) and wet
gas meter. Especially, MFM was used to measure kinetics of membrane permeation.
Separated and retentate gas composition was analyzed by on-line GC(HP 5890 11, TCD
type). Helium was used as carrier gas and sweeping gas. Temperature was detected by
RTD(Hanyoung. Co.) at inlet, inner cell and skin of cell. Pressures were detected by
pressure transducers(Dec0 Co.) at inlet and permeate part.

3. Theory and Mathematical Model

Ceramic membrane is the nanoporous membrane which has the comparatively higher
permeability and lower separation factor. And in the case of mixed gases, separation
mechanism is mainly concerned with the permeate velocity. The velocity properties of gas
flow in nanoporous membranes depend on the ratio of the number of molecule-molecule
collisions to that of the molecule-wall collision. The Knudsen number Kn=h/d, is
characteristic parameter defining different permeate mechanisms. The value of the mean
fiee path depends on the length of the gas molecule and the characteristic pore diameter.
The diffusion of inert and adsorbable gases through porous membrane is concerned with
the contributionsof gas phase diffusion and surface diffusion.
Adopting the dusty gas model(DGM) for the description of gas phase mass transfer and
a Generalized Stefan-Maxwell(GSM) theory to quantiw surface diffusion, a combined
transport model has been applied. The tubular geometry membrane mass balance is given
in equation (1).

Equilibrium flux and kinetics of membrane can be calculatedby equations (1) and (2).

N,o, = --RT
I [Di + $ P + ( l - ~ ) q ,
-1-
bD" ap
I+bP ar

All parameters used in equations (1) and (2) were calculated from single gas adsorption
and membrane permeation experiment. The numerical method for solving above equations
were MOL(Method Of Line). These calculations were executed using LSODE
solver(F0RTRAN code).

530
3 Result and Discussion

Figure 2(a) contained the comparison of amount of C02 adsorption between a-A1203
support and CMS(Carbon Molecular Sieve, Takeda Co.) . Owing to larger pore(about
lOOnm) and lower chemical affinity, a-A1203 support has trivial adsorption capacity,
compared with CMS.This result means that surface diffusion caused by adsorption could
be neglected in a-A1203support. Therefore, only Knudsen and viscous difisions were
considered as permeate mechanism of a-AI2O3support. In this paper, single component
adsorption experiments were executed in order to investigate surface diffusion effect of
silica membrane. From the above description, surface diffusion in permeate mechanism
could be neglected due to the non-adsorbable property of a-A1203support. Figure 2(b)
shows equilibrium isotherm curve of MTES templating silica unsupport at 293K.
Langmuir and L-F parameters were obtained from this equilibrium data. Figure 2(c) shows
C02 uptake curve of TPABr at 293K. Diffusional time constants in low pressure range
were calculated from adsorption rate equation using modified Ficks second raw and its
numerical method was presented by Ruthven et a1[6].
From the permeate experiment of membrane, the value of the mean pore radius of
support and other properties can be calculated. The real value of pore radius of a-A1203
support(measured by mercury porosimetry) was near loom, and the calculated value of
mean pore radius from experiment was 101nm(N2)and 1 16nm(C02).These were in a
good agreement with porosimetry measurement and permeate method calculation.[4]
However, in case of light gas, H2, showed some deviation from the theoretical results.
From these data, adsorption data and membrane permeate data, we could analyze
permeate and separation mechanism of organic silica membrane. Figure 3(b) shows
permeate rate in N2/MTES templating silica membrane at 293K. This curve was well
fitted to equation (2). Permeate flux arrived with 40 sec. And Figure 3(c) shows predicted
tans-membrane pressure profile which was calculated by DGM and GSM model with
Langmuir parameters and other membrane parameters.[3,6]

4 Conclusion

After dip coating on support by organic templating sol, for example TPABr and
MTES, pore size was controlled under lOA . Therefore, it was expected sieving effect or
surface diffusion effect on usupport(t0p layer of membrane) because of its narrow and
small pore. In the mixture of adsorbarblehon-adsorbable molecules, silica membrane was
very selective for adsorbable one, due to its surface diffusion and blocking effects.
Adsorbable molecules, for example C 0 2 and CH4, adhered on surface and surface
diffusion takes place by the hoping of molecules between adsorption sites. And they
blocked the other non-adsorbable molecules, for example COz molecules block H2
molecules. Then other vacant sites were occupied with adsorbable molecules and then
surface concentration of adsorbable molecules became higher and higher. Finally
adsorbable molecules permeated through the membrane and non-adsorbarble molecules
were retentated.

531
5 Acknowledgements

The financial support of KOSEF(Korea Science and Engineering Foundation, RO1-

1999-00198)is gratefully acknowledged.

6 Reference

1. Yang, R.T.,Gas separation by adsorption processes. Butterworths, Boston, 1987

2. E.A.Mason, A.P.Malinausks, Gas transport in porous media : The dusty-gas model,
Elsevier, 1983
3. Axel Tuchlenski et al, An experimental study of combined gas phase and surface
diffusion in porous glass, Journal of Membranes Science 140(166-184),Elsevier,
1997
4. Petr Uchytil, Gas permeation in ceramic membranes Partl. Theory and testing of
ceramic membranes., Journal of Membrane Science 97(139-144),Elsevier, 1994
5. A.J.Burggraaf, L.Cot., Fundamentals of inorganic membrane science and technology,
Elsevier, 1996
6. Karger, J. & Ruthven, D.M, Diffusion in Zeolite and other microporous solids, John
Wiley & Sons, 1992

Fig. 1. Schematic apparatus for experiment (a) adsorption(ektric cahn balancc)measurcmcntsystem and (b)
membrane transport system

-w

Fig. 2. (a) comparison of C@ gas adsorption equilibrium isotherm curve of a-AIp3 support and CMS at
293& (b) equilibrium isotherm curve of MTES unssuport at 313K (b) CO2 stepwise uptake curve of top layer
of membrane(TPABr sol) at 293K.

532
Fig. 3. (a) permeate flux and separation factor in H d C a mixtute/MTES silica composite membmne system at
293K. (b) peameate rate of Nz in MTES silica composite membrane at 293K (c) predicted pressure profile of
C a on TPABr silica membrane at 4abn

533
 ADSORBER DYNAMICS OF BINARY AND TERNARY HYDROGEN
MIXTURE IN ACTIXATED CARBON AND ZEOLJTE SA BEDS

MIN-BAE KIM.JOONG-SUCK KIM,AND CHANG-HA LEE

Department of Chemical Engineerlng, Yonsei University
134 Shinchon-dong, Sectdaemun-gu,Seoul 120-749,KOREA
E-nuzik [email protected]

CHAN-HWI CHO
Depamnent of Chemical Engineering, Youngdong University

To recover I-& from the mixture gas, the adsorption dynamics of binary and ternary
hydrogen mixture in activated carbon and zeolite 5A beds were investigated
experimentally and theoretically through breakthrough and desorption
experiments. It can be helpful to determine the operating variables of H, PSA
process. The breakthrough experiments were performed in the range of 5 - 9 atm
adsorption pressure and 4.5 - 9.1 LSTP/min feed flow rate. And desorption
experiments were performed through 2 to 4 LSTP/mln at constant pressure of 1.5
atm. To understand the adsorption characteristics and the thermal effects by the
heat of adsorption. a non-isothermal dynamic model incorporating mass, energy.
momentum balances was used. A good agreement between model prediction and
experimental result was obtained.

1 Introduction

Fixed-bed adsorption process has been an important unit operation for purification and
bulk separation of gas mixture, especially by pressure swing adsorption (PSA). In recent
years, PSA process has been increasingly commercialized for air drying, hydrogen
purification, air separation, and various other separations [l]. One of the successful
applications of a PSA process is hydrogen recovery because a high-purity H2 product with
high productivity can be obtained from a well-designed PSA.
To design and develop PSA process, detailed analysis of the fixed-bed dynamics
must be preceded, because the key step to developing an optimum PSA process lies in the
design and operation of the adsorption step. Also, on the condition that the theoretical
models of the adsorption breakthrough can predict its experimental results well, it can be
used to investigate the breakthrough dynamics in a bed and predict PSA performance
without any specific experiment.
In this study, breakthrough and desorption experiments were performed in
activated carbon and zeolite 5A beds to study the adsorption characteristics of binary and
ternary hydrogen mixture gas (H2/CH, : 43/57 and H2/CO/C02 : 39.3/35.4/25.3 ~01.96).
The ternary mixture is the pseudo composition of effluent gas from incinerator for
pyrolysis gasification while the binary mixture is the converted mixture gas from LNG by
the cold plasma reaction. The ultimate purpose of this study is to find out the optimum
condition for obtaining high purity hydrogen from H2 mixtures by using PSA process.
The experimental results and dynamics in the bed were analyzed by using a non-
isothermal dynamic model incorporating mass, energy, and momentum balances. Although
the temperature increase inside a bed is undesirable, in case of bulk separation, the
temperature variation during the adsorption process is inevitable. Therefore, the

534
adsorption dynamics in the bed were investigated according to the relationship between
the concentration and temperature profiles.

2 Experimental

The binary (H2/C& : 43/57 vol.%) and ternary (H2/CO/COz : 39.3/35.4/25.3 vol.%)
hydrogen mixtures as pseudo composition was used as feed gas. And two kinds of
adsorbents, activated carbon (Calgon Co., PCB,6-16 mesh) and zeolite 5A (W. R. Grace
Co., 4-8 mesh) were used. Prior to each experiment, the adsorbents were regenerated for
more than 12 hours at 423 K and 588 K, respectively.
The schematic diagram of an experimental apparatus is shown in Figure 1. The
adsorption bed was made of stainless steel of 100 cm length and 3.71 cm ID. Effluent
stream was sampled between the back-pressure regulator and the product tank and was
analyzed by using a mass spectrometer (Balzers, QME 200). All the data including
concentration, temperature, pressure and flow rate were saved on the computer.
The breakthrough experiments were performed through the adsorption pressure of
5 to 9 atm, and feed flow rate of 4.5 to 9.1 LSTP/min. And the desorption experiments
were performed through 2 to 4 LSTP/min at constant pressure of 1.5 atm.
The equilibrium parameters used in this study have been taken from independent
single-component experiments using volumetric method at several temperatures.
Adsorption isotherms on each adsorbent are shown in Figures 2 and 3.

3 Mathematical model and simulation

Calculations of multicomponent adsorption responses are too highly complicated to solve

analytically because the equations that determine the behavior of this system are partial
differential-algebraic equations (PDAEs) with a large number of variables. To understand
the adsorption dynamics, a non-isothermal dynamic model is applied to the system with
the following assumptions: (i) the flow pattern in the bed can be described with the axial
dispersion plug flow model; (ii) the solid and gas phases reach a thermal equilibrium
instantaneously; (iii) radial concentration and temperature gradients inside the adsorption
bed are negligible; (iv) axial conduction in the wall can be neglected; (v) the ideal gas law
can be applied. The equilibrium of mixture was described by extended Langmuir-
Freundlich model and the sorption rate into an adsorbent pellet by linear driving force
(LDF) model [1,2]. To calculate the pressure drop along the bed, Ergun's equation was
used [3,4]. For boundary conditions Danckwerts equations were used. These equations are
as follows;
(1) Component and overall mass balance;

d2C duc 1-& d q.

i=l
(2) Overall energy balance;

535
 - pB 2ei-
i
dqi 2h.
+-+T
d t RE,
-T,) =o
(3) Energy balance at the wall;
dTW
p w ( c p ),, 4,-= 2 d B i h i (T-T w ) - 2MBoho (Tw -Tm
dt
4, = n ( ~ &- R ; ~ )
The gPROMS-modeling tool was used to obtain the numerical solution of dynamic
simulation. And the results of simulation were stable for the range of conditions used in
this work.

4 Resultsanddiscussion

As can be seen in Figures 4 and 6, the breakthrough curves had long tail due to
temperature variance in the bed. Therefore, energy balance was considered in simulation
of adsorption dynamics. From the adsorption equilibria of H2, C&, CO, and C02,
temperature dependent isotherm parameters of Langmuir-Freundlich type were obtained.
The experimental values under various operating conditions like adsorption pressure and
feed flow rate were compared with predicted ones using linear driving force model.
In the adsorption experiment of binary mixture (Figure 4), breakthrough time of the
activated carbon bed was longer than the zeolite 5A bed, moreover the zeolite 5A bed had
more tailing effect than the activated carbon bed.
In the desorption experiment (Figure 5), the effect of the feed flow rate had similar
effect with the breakthrough experiment.
Whereas CO was the first breakthrough component followed by C02 at intervals of
hundreds of seconds in the breakthrough experiments of ternary mixture (Figure 6).
Because of the relatively high heat of adsorption of C02 compared with that of CO,
temperature increased at mass transfer zone 0 of C02 noticeably. Especially, CO
showed the rolling-up phenomenon, which is due to the competitive adsorption of more
strongly adsorbed component, C02, and this phenomenon is usually Observed in a multi-
component and an equilibrium control system. As concentration wave fronts of CO and
C02 propagate along the bed, sinusoidal temperature excursion was observed. The first
temperature peak was caused by the adsorption of CO and the latter by that of C02.
Excellent agreement between model predictions and experimental results was
obtained.

5 Conclusions

The adsorption dynamics of binary and ternary hydrogen mixture in activated carbon and
zeolite 5A bed was studied by experimentally and theoretically through breakthrough and
desorption experiments. Energy balance is an essential element for accurate adsorption
process modeling in case of bulk separation. Especially in ternary system, sinusoidal

536
temperature excursions were caused by the large difference of wave velocities of each
component. The roll-up of CO and separation of mass transfer zones behveen CO and COz
led to stepwise breakthrough curve of H2in the activated carbon bed.

6 Acknowledgements

The financial support of Korea Institute of Environmental Science and Technology (2002-
11105-0004) is gratefully acknowledged.

References

1. R.T. Yang, Gas separation by Adsorption Process, Butterworths, Boston, 1987.

2. D.M. Ruthven, S. Farooq, and K.S.Knaebel, Pressure Swing Adsorption, VCH
publishers, New York, 1994.
3. E.S. Kikkinides and R.T. Yang, Chem. Eng. Sci., 48, (1993) 1545-1555.
4. J. Yang, M.-W. Park,J.-W. Chang, S.-M. KO,and C.-H. Lee, Korean J. Chem. Eng.,
15, (1998) 211-216.

537
 . .
0 5 1 0 1 5 2 0 2 5 3 0 3 5
Pmwun (am)

Figure 1. Schematic diagram of apparatus Figure 2. Adsorption isotherms on

for breakthrough experiments. activated carbon at 293.15 K.

Figure 3. Adsorption isotherms on zeolite Figure 4. Effect of feed flow rate on

5A at 293.15 K. activated carbon bed (binary mixture).

:h
o r m X m % m n a m
TmM b H
Figure 5. Effect of flow rate on zeolite 5A Figure 6. Breakthrough curves under 7 atm
bed under 1.5 atm desorption pressure adsorption pressure and 6.8 LSTPImin
(binary mixture). feed flow rate (ternary mixture).

538
 Temperature Programmed Adsorption (TPA) of Various Hydrocarbons on
Adsorbers of Honeycomb Type

Dae Jung Kim, Jae Eui Y ie, Ji Man Kim

School of Chemical Engineering and Biotechnology, Ajou University,Suwon 442- 749, Korea
Email : [email protected],[email protected],[email protected]
Young Sam Oh
Research Center,Korea Gar Corp., Incheon 406-130, Korea
Email :[email protected]

TPA characteristics of two adsorbers of honeycomb type for various hydrocarbons were evaluated. In
this study, methyl alcohol, acetone, acetaldehyde, 224 trimethylpentane, n-octane and toluene were
chosen as the hydrocarbons of cold start. The effect of the hydrocarbon components and oxygen
concentration on TPA behavior was studied. According to the precious metal loading and the
presence of 0 2 , the adsorption and desorption amount were decreased, while the conversion
efficiency of hydrocarbons was increased. In case of hydrocarbons with oxygen, the thermal
decomposition appeared to be in the order of methanol, acetaldehyde and acetone.

1 Introduction

Technologies have been developed to reduce the hydrocarbon emissions during cold
start.One of the technologies, the hydrocarbon adsorber [ 1-31 has been most attractive to
automotive makers because it has the advantage in view of cost and performance.
Hydrocarbons of the cold start depend on the condition of a vehicle, fuel and driver.
About 100 hydrocarbon species are present in the exhaust of the cold start. They consist
of about 10% methane, about 30% alkenes such as ethylene or propene, about 30%
alkanes such as pentane or hexane, about 20% aromatics such as toluene or xylene, and
about 10% other species. Therefore, a hydrocarbon adsorber has to show a good
selectivity for the hydrocarbons. In order to remove the adsorbed hydrocarbons effectively,
the hydrocarbon adsorber needs to have an additional function, hydrocarbon conversion.
Until now, in order to select a hydrocarbon adsorber with higher hydrocarbon
trapping and conversion efficiency, experimental tests using vehicle and engine
dynamometer has been employed. A model, Temperature Programmed Adsorption (TPA),
was proposed by Kim, et al. [4-51 to save cost and time. The model has an advantage to
analysis adsorption, desorption and conversion of hydrocarbons simultaneously.
In this study, TPA characteristics of two adsorbers of honeycomb type for various
hydrocarbons were evaluated. Methyl alcohol, acetone, acetaldehyde, 224
trimethylpentane, n-octane and toluene were chosen as the hydrocarbons of cold start. The
effect of the hydrocarbon components and oxygen concentration on TPA behavior was
studied.

2 Experimental

2. I Adsorbers and hydrocarbons

Two adsorber samples used in this study were obtained by coating washcoat onto a

539
cordierite honeycomb ceramic substrates (cell density of 62 cells/cmz (400 cells/in2), wall
thickness of 0.165 mm, 19mm(D)x30mm(L)).The washcoat was consisted of H-ZSM-5,
--A1203 and base metals (Ba, Ce, Zr)used in a conventional three-way catalysts. The
SVAI ratio of H-ZSM-5 was 150/1. In order to study the effect of the presence of precious
metals on the adsorber, Pd and Rh were used. The first adsorber had the 150/1 WA1 ratio
adsorbent with impregnated Pd/ Rh in a 10/1 ratio, and it was named as HA #l. The
second one had the same SVAl ratio adsorbent without precious metals, and it was named
as HA #2. For adsorbers, washcoat loading amount was same. When impregnating
precious metals, PdClz and RhC13. 3H20 were prepared as the precursors of Pd and Rh.
All samples were dried at 423.15 K for 5 hrs and calcined at 823.15 K for 4 hrs.
The components of hydrocarbons generated from engine during cold start are
dependant on fuels, vehicles and driving conditions. In this study, methyl alcohol
(Aldrich, 99%), acetone (Aldrich, 99%), acetaldehyde (Aldrich, 99%),
224trimethylpentane (Aldrich, 99.8%), n-octane (Aldrich, 99%) and toluene (Aldrich,
99.95%) were chosen as the hydrocarbons of the cold start, and were used without any
treatment.

2.2 TPA
The apparatus for TPA was shown in detail elsewhere. For all experiments the
dimension of a sample (20 mm(D)x30mm(L)) was identical. Before doing TPA
experiment the sample was oxide at 673.15 K for 1 hr in zero air (1 Umin), and then it
was purged with Nz(1 Umin) at 673.15 K for lhr and cooled down to 303.15 K in Nz(1
Vmin). After that, the sample was heated in lo00 ppm hydrocarbon /N2 mixture (1 Vmin)
at 303.15 K to 573.15 K with the ramping rate of 1 Wmin. In addition, to evaluate a
conversion of hydrocarbon, Ozwas supplied with the concentration of 1.47%.
The concentrations of hydrocarbon at the inlet and outlet of the reactor were
measured using Fl'IR (Nicolet) with a mercury-calcium-telluride(MCT) detector, which
was cooled by liquid nitrogen and gas cell (Infrared Analysis), and with 16 scans and a
resolution of 2 cm-'. The concentrations were also checked using GC (Hp589Oplus) with
FID (flame ionization detector) and HP plot column. The vapor pressures of the
hydrocarbons were calculated by using Reid equation [6].

3 Results and Discussion

Figure 1 shows TPA curves of 224 trimethylpentane, toluene and n-octane on HA #4.
For all cases, the inlet concentration and flow rate was kept to be constant with lo00 ppm
and 1 Vmin under the absence of 0 2 in supply gas. This figure is represented according
to temperature and relative concentration. The relative concentration means the ratio of
outlet concentration (Cout) to inlet concentration (Cin). The value of 1 stands for the full
saturation of a hydrocarbon on the adsorber. The value over 1 represents the sum of
supplied hydrocarbon and desorbed hydrocarbon from the adsorber. The order of
breakthrough appeared to be n-octane, 224 trimethylpentane and toluene. This result may
be attributed to the difference of interaction intensity of the hydrocarons and the adsorber.
Over 330 K all hydrocarbons started to desorb from the adsorber. Although the gas flow
did not contain 02, the value under 1 appeared for all hydrocarbons. The value under 1
appeared around 390 K for toluene, around 500 K for 224 trimethylpentane and nactane.
This result may be attributed to the structure effect of the adsorber and the oxidation
reaction of hydrocarbons and adsorbed oxygen during the pretreatment.

540
 1.6 540

h 500
c

3
Y 1*2
0
460

g 0.8 420
0
-e 380
=
U

K
(A
0.4
340

0 300
0 40 80 120 160 200 240
Time(min)
Figure 1 TPA curves of 224 trimethylpentane, toluene and n-octane on HA #4 under the absence of 02 in
supply g=

1.6 540
4Acetone
4Methanol
500
+-Acetaldehyde
n
Y
460 w
t3
U
420

-s! 380
E"
=a
Q)
U
(A
c
0.4
340

0 300
0 40 80 120 160 200 240
Time(min)
Hgure 2 TPA curves of acetone, methanol and acetaldehydeon HA #4 under the absence of 0 2 in supply gas.

Figure 2 shows TPA curves of acetone, methanol and acetaldehyde on HA #4. The
conditions of the inlet concentration and the flow rate were same as Figure 1. Although
oxygen was not supplied, the relative values of three hydrocarbons moved to "0" with the
increase of the temperature.

541
 500
? 1.6
B
B 460
L5

Y
g. 1.2
0
g!
3
C
0 420

.-
0.8
E"
-2a 0.4
U

+-224TMP
380 f

-+Toluene 340
+Acetaldehyde .
0 300
0 40 80 120 160 200 240
Time(min)

Figure 3 TPA curves of 224 trimethylpentane, toluene and acetaldehyde on HA #5 under the absence of 02 in
supply gas.

1.6 I 1 540

1 HA #4
224TMP
-A-0% 02
-E-0.8%02
500
n

460
p!
a
.I-
5 0.8 420 f
0
a9
.->
U 380
E"
$
2 0.4
a 340

0 300
0 40 80 120 160 200 240
Time(min)
Figure 4 ' P A curves of 224 trimethylpentane on HA #4 with respect to 0 2 concentration

It may be attributed to the thermal decomposition of the hydrocarbons on precious

metals and the oxidation reaction of the hydrocarbons and adsorbed oxygen. The order of
thermal decomposition appeared to be methanol, acetaldehyde and acetone.
Figure 3 shows TPA curves of 224 trimethylpentane, toluene and acetaldehyde on
HA #5. The experimental conditions were same as Figure 1. In comparison of Figure 1
and Figure 3, TPA curves of 224 trimetylpentane and toluene were similar, but TPA
curves of acetaldhyde were different. This result notes that the thermal decomposition of
the hydrocarbons, which have oxygen in the structure of molecule, occurs mainly on the

542
precious metals.
TPA curves of 224 trimethylpentane on HA #4 according to different 02
concentrations are shown in Figure 4. According to the increase of 02 concentration, the
adsorption amount of 224 trimethylpentane was decreased little, but the hydrocarbon
conversion stated at the lower temperature and was advanced over 460 K.

References

1. Williams, J.L., Patil, M.D. and Hertl, W.,By-Pass Hydrocarbon Adsorber System for
ULEV, SAE, paper No. 960343 (1 996).
2. Engler, B. H.,Lindner, D.,Lox, E. S., Ostgathe, K., Schafer-Sindlinger, A. and
Muller, W., Reduction of Exhaust Gas Emissions by Using Hydrocarbon Adsorber
Systems, SAE, paper No. 930738 (1993).
3. Yamamoto, S., Masushita, K., Etoh, S. and Takaya, M., In-Line Hydrocarbon (HC)
Adsorber System for Reducing Cold-Start Emissions, SAE, paper No. 2000-01-0892
(2000).
4. Kim, D.J., Son, G.S, KO and S.H., Yun, S.W, Characteristic Study of Aged HC
Adsorber for Automobile, Thories and Applications of Chem. Eng., 4 (1999) pp.
2669-2772.
5 . Son, G.S., Yun,S.W.,Kim, D.J., Lee, K.Y. and Choi, B.C., A study on the Reaction of Cold
Start Hydrocarbon from Gasoline Engines Using Hydrocarbon Adsorbers, KSME International
Journal, 14 (2000)pp. 699 -703.
6. Reid, RC., Prausnitz, J.M. and Poling, B.E., The properties of Gases & Liquids,
(McGraw-Hill, New York, 1988).

543
 Non-Isothermal Dynamic Adsorption and Reaction in Hydrocarbon Adsorber
System

DAE JUNG KIM, JAE EUI YIE

School of Chemical Engineering and Biotechnologx Ajou University. Suwon 442-749, Korea

WANG GEUN SHIM, HEE MOON

Faculty of Applied Chetnhtty, Chonnam National University, Gwangiu 500- 757. Korea

A new mathematical model was developed to predict TPA behaviors of hydrocarbons in an adsorber
system of honeycomb shape. It was incorporated with additional adsorption model of extended
Langmuir-Freundlich equation (ELF). LDFA approximation and external mass transfer Coefficient
proposed by Ullah, et. al. were used. In addition, rate expression of power law model was employed.
The parameters used in the power model were obtained directly From the conversion data of
hydrocarbons in adsorber systems. To get numerical solutions for the proposed model, orthogonal
collocation method and DVODE package were employed.

1 Introduction

A hydrocarbon adsorber system [l], which has been used for reduction of cold-start
hydrocarbons, adsorbs the hydrocarbons in the initial period of cold-start. Generally, the
adsorber system is consisted of adsorber followed by light off catalyst. As the surface
temperature of the adsorber is heated up, the adsorbed hydrocarbons are released slowly
from the surface, and then converted to inert gases by oxidation reactions on adsorber or
light-off catalyst.
To predict effectively the adsorber system, it is important to understand the behaviors
of hydrocarbons on adsorber and light-off catalyst during cold-start. Kim, et al. [2]
proposed the experimental model, namely TPA (Temperature Programmed Adsorption),
to simulate the behaviors of hydrocarbons on adsorber or light-off catalyst during
cold-start.
In this study, a new mathematical model was developed to predict the experimental
TPA behaviors with reaction, and it was incorporated with additional adsorption model of
extended Langmuir-Freundlich equation (ELF). LDFA approximation and external mass
transfer coefficient proposed by Ullah, ,et. al. were used [3]. Also, rate expression of
power law model was employed [4]. The parameters used in the power model were
obtained directly from the conversion data of hydrocarbons on adsorber or light off
catalyst [5]. In this study, to get numerical solutions for the proposed model, orthogonal
collocation method and DVODE package were employed [6].

2 Experimental

In this study, a hydrocarbon adsorber system was composed of an adsorber of HA #4

and an light off catalyst of LOC # I . In addition, Toluene (Aldrich, 99.95%) and
224trimethylpentane (Aldrich, 99.8%) were used as hydrocarbons of cold start. The
experimental procedure of TPA is described elsewhere [5].

544
3 Theoretical

Equations for describing non-isothermal and adiabatic adsorption with reactions

consideringassumptions [5] are as follows.

Mass balance for component i in gas phase is derived by

Mass balance for componenti in monolith is expressed by

Energy balance in monolith is represented by

Tg =T, =To + a - t
Initial conditions for O<z<L and t=O are as follows
y, =o; 4, =o
Boundary condition for F O and 0 0 is as follows
Y , =y/o
Adsorption isotherm of Langmuir-Freundlich for multi-component is defined by
qm,b,(PY,,,)"I
qI = "
1 + Cb,(FyJ,JJ

where, b, = b,, exp(-)-

RTS
Subscript i refers to hydrocarbons under consideration (224 trimethylpentane,Toluene).

Oxidation scheme of hydrocarbons adapted in this study can be represented as

follows.
C,H, + 9 0 2 + K O 2 + 4 H 2 0
CgHl, + 12.502 8c02 + 9H2O
For rate expression power model was used as follows.

545
4 Results and Discussion

Figure 1 shows experimental and predicted TPA results for single component on LOC
#l. Hydrocarbon used was 224 trimethylpentane and toluene. Fluid velocity (7.57E-2 d s )
and concentration of supplied O2 concentration (0.18 Wa) for each hydrocarbon were
same. In the result of adsorption equilibrium, the amount of toluene adsorbed seemed to
be higher than 224 trimethylpentane. As can be expected, toluene was emitted more lately.
However, Toluene was converted more rapidly than toluene over 380 K.

Loc#1
-
O 224TMP
To1
/1 500
? 1.6
e --- Pre-224 TMP
Pre-To1
3 460
g 1.2
0
g!
3
E
0 420 5
Q)
0.8 P
.-W 380
-2m 0.4
U
c
340

0 300
0 40 80 120 160 200 240
Time(min)

Figure 1 Experimental and Predicted TPA results of 224 trimethylpentane and toluene on LOC # 1 in presence
of 1.8% 02 ( L = 0.03 m, U = 7.57E-2 ds).

It agreed well with isothermal dynamic adsorption results and conversion results at
steady state. The predicted results by the proposed model described well the experimental
results.
Experimental and Predicted TPA results for binary components, 224 trimethylpentane
and toluene, on HA #4 and LOC #1 were shown in Figure 2 and Figure 3, respectively.
The experimental condition of Figure 2 was identical to that of Figure 3.

546
 1.6 540
I
-
0' 1.2
/

500
CI
h 460 Y
0"
T e!
a

0
0
0.8 420 1
a,
.-c> 380 a,
;
!-
$ 0.4
a 340

0 300
0 40 80 120 160 200 240
Time(min)
Figure 2 Experimental and Predicted "PA results for binary components (224 trimethylpentane / toluene) on
HA #4 in presence of 1.8% 0 2 (-L= 0.03 rn, U = 7.57E-2 ds).

1.6 I 3 540

-
6 1.2
o
To1
TMP
500
cs
h 460 sE
Oeci a
0.8 420
0
.-a,>
Y
380
E
a,
c
m 0.4
3
U 340

0 300
0 40 80 120 160 200 240
Time(min)
Figure 3. Experimental and Predicted TPA results for binary components (224 trimethylpentane / toluene) on
LOC#1 in presence of 1.8% 0 2 ( L = 0.03 m, U = 7.578-2 d s ) .
Total concentration of two hydrocarbons in gas phase was 2000 ppm (0.2 kPa). The
mole fiaction of 224 trimethylpentane and toluene was assumed of 0.5 and 0.5,
respectively. The concentration of supplied O2 concentration was 0.18 kPa, and the
velocity was 7.57E-2 d s . In all cases, 224 trimethylpentane was emitted first, but the
reaction was started lately. It matches to the results of adsorption equilibrium and
conversion experiments. The proposed model described the experimental data properly.

547
References

1. Yamamoto, S., Masushita, K., Etoh, S. and Takaya, M., In-Line Hydrocarbon (HC)
Adsorber System for Reducing Cold-Start Emissions. SAE. 2000, paper No.
2000-01-0892.
2. Kim,D.J., J. W.Kim, J. E. Yie and H. Moon, Temperature-ProgrammedAdsorption
and Characteristics of Honeycomb Hydrocarbon Adsorbers. Ind Eng. Chem. Res.,
2002,41,6589 -6592.
3. Tronconi, E. and Beretta, A., The Role of Inter- and Intra-Phase Mass Transfer in the
SCR-DeNOx Reaction over Catalysts of Different Shapes, Catal. Todqy, 1999, 52,
249-258.
4. Lopez, E., Errazu, A.F., Bono, D.O. and Bucala, V., Altanative Designs for a
Catalytic Converter Operating under Autothermal Conditions, Chem. Eng. Sci, 2000,
55,2143-2150.
5. Kim, D.J. Non-Isothermal Dynamic Ahorption and Reaction in Hydrocarbon
Adrorber Systems. PhD. Thesis, Department of Chemical Engineering, Ajou
University, 2002.
6. Finlyson, B. A. The Method of Weighted Residuals and Variational Principles,
Academic Press, New York, 1972.

548
 SORPTION OF U(VI) ONTO GRANITE: KINETICS AND REVERSIBILITY

M.H.BAlK AND P.S. HAHN

Korea Atomic Energy Research Institute, I50 Dukjin-dong Yuseong-gu,Daejeon 305-353, Korea
E-mail: [email protected]

In this study, surface sorption experiments of U(V1) onto the surfaces of a Korean granite rock are
carried out in order to investigate the kinetics and reversibility of U(V1) sorption as a function of pH
and surface types such a fiesh intact surfaces and natural fracture surfaces. It has been found that
the effect of pH is significant in the sorption of U(V1) onto the both types of granite surfaces. It is
noticed that the surface sorption coefficients of U(V1) for the natural fracture surfaces are greater
than those of the fresh intact rock surfaces due to the higher content of secondary minerals which
acted as stronger sorbents. The kinetic sorption results are fitted with a two-step first-order kinetic
sorption model and the results show a good agreement between the experimental results and the
model. The desorption results show that the sorption process of U(V1) is a little irreversible for the
two types of granite surfaces depending on pH and surface type. It is also shown tiom an X-ray
mapping study that mica and chlorite mainly contribute to the sorption of U(VI) onto the natural
fracture surfaces and fresh intact surfaces of granite, respectively.

1 Introduction

Sorption data for radionuclides in contact with minerals typical of those found in
water-bearing fractures in rock are needed in the safety assessment of deep geological
repositories. Furthermore, an understanding of the factors that influence the sorption
characteristics of the radionuclides is essential when we apply an empirically derived
sorption value outside the range of the experimentalparameters.
The surfaces in old hctures, which have been in contact with moving groundwater,
may have a quite different mineral composition from the surrounding host rock. This
fracture coating material could be the results of weathering and alteration of the rock or
precipitates and crystallization products from the groundwater [l]. It was also reported
that the sorption on natural hcture surfaces is more effective than on fresh core samples,
evidently because of the higher cation exchange capacity of the altered minerals on
fracture surfaces [2].
Most of the sorption experiments for rock materials have been performed using
crushed rocks [3], and only a limited number of sorption studies were performed for rock
coupons or rock cores [4]. The aims of this study are therefore to investigate the kinetics
and reversibility of U(V1) sorption upon the natural fracture surfaces and fresh intact
surfaces of a Korean granite rock, respectively, and to investigate which constituent
mineral of the granite rock contributesto the sorption of U(V1).

2 Experiment

The granite rock used in the study was sampled from a granite quarry site located at
Dukjeong-myun, Kyongki-do, Korea. Rock samples used in the experiments were sections
of drilling cores containing natural fracture surfaces and machined fresh intact surfaces on
one side. The drill cores were cylindrical, about 10 cm in height and about 10 cm in
diameter. In this study, U(V1) in the nitrate form (U02(N03)2-6H20) was used as a sorbing
nuclide. The concentration of uranium was measured by ICP-MS (Varian, Ultramass 700).

549
The surface area of the natural fracture surface was measured using a 3dimensional laser
scanner (Inter Tech Co., Ltd). The surface area of the fracture surface was measured as
79.02 cm'. An experimental setup for the surface sorption experiments used in this study
is shown in Fig. 1.

Figure 1. Experimentalsetup for surface sorption ofnatural rock fracture.

Before granite cores were mounted to the sample holder they were kept in contact
with 0.01 mom NaC10, solution, which is adjusted to a desired pH, for about one month.
Before injecting the uranium tracer solutions, the solutions were adjusted to desired pHs
(i.e., pH 5.5, 7.0 and 8.5) using NaOH or HClO, solutions. The tracer solution was then
pumped to a narrow space between the granite core and the acryl cap at a flow rate of
about 1.0 d m i n . All sorption experiments were performed at about 23&2"C under an
ambient condition. The solution samples for the measurement of uranium concentration
were taken in a kinetic process up to 2 weeks.

3 Results and Discussion

3.1 Kinetics
The surface sorption capacity is usually expressed as a surface sorption coefficient. In
general, the surface distribution coefficient of U(V1)between the solution and the granite
surface, K. (cm), is defined as:

where [VJ (moVcm') is the concentration of sorbed uranium per unit surface area of a
granite surface, and [v,] (moVcm3)is the concentration of uranium in the solution phase,

550
I U , ~( movcm3) is the initial concentration of uranium, v (cm3)is the solution volume in
contact with granite surface, and A (cm2) is the surface area of the granite surface
contacting with solution.
The kinetic results of the surface sorption of U(V1) both onto the fresh intact surfaces
and onto the natural fracture surfaces at different initial pH values of 5.5, 7.0, and 8.5 is
shown in Fig. 2. The results show that it takes about 7 days to reach a constant percent of
U(V1) sorbed (i.e., a steady state), regardless of the starting pH. It is also shown that the
equilibrium percent of U(V1) sorbed strongly depends on the pH of the tracer solution.

50
n n n
U U

$40 +MS-2 (pHz7.0)

2
2 t-MS-3 (pH=8.5)
2 30
h
4 F S - 1 (pH=5.5)
$I
5 -8-FS-2 (pH=7.0)
2 20
c.
8
2 -
k 10

0
0 5 10 15 20
Time, days
-
Figure 2. Kinetic results for U(V1) sorption onto the fresh intact surfaces and the natural fracture surfaces
(MS: fresh intact surfaces, FS: natural fractured surfaces).

The kinetic sorption data can be interpreted based on the assumption that the uranium
sorbs on the surface of granite according to a first-order reaction. The time dependence
can be given as:
pt = P, ( 1-eekt)
and this equation is linearized as:
In( 1 - Pt/Ps) = -k.t
where Pt and P,are the percents of uranium sorbed on the fracture surface at time t and at
steady state, respectively, k (l/day) is a kinetic sorption rate constant, and t is time in days.
-
The plot of ln(1 PJPs) against t should be linear for a first-order kinetic reaction.
The rate constants are estimated by the slopes of lines fitted by a least squares method
and listed in Table 1. It is concluded that a two-step first-order kinetic behavior dominates
in the surface sorption of U(V1). The relatively slower second-step may be due to such
effects as a diffusion-controlled sorption onto the fracture surface of micropores and a
mineral dissolution of the granite surface [ 191. It is noticed from Table 1 that the reaction
rates do not greatly depend upon pH although the amount of U(V1) sorbed onto the granite
surface is greatly dependent on pH.

551
Table 1. The results of first-order kinetic fitting for the U(U) sorption onto the fresh intact surfaces and the
natural fracture surfaces.

Test pH kl (l/d) k2 (1 /d)

MS- 1 5.5 1.015 0.3625
MS-2 7.0 0.9763 0.3662
MS-3 8.5 0.6897 0.3364
FS- 1 5.5 0.8888 0.2702
FS-2 7.0 1.0627 0.2878
FS-3 8.5 0.863 1 0.4464

3.2 ReversibiIiQ
The results of sorption and desorption of uranium are shown in Table 2. As shown in
Table 2, the desorbed percentages of uranium from the granite surfaces are somewhat
greater than sorbed percentages of uranium on granite surfaces. Thus, the results show that
the sorption process of U(V1) is a little irreversible for the two types of granite surfaces
depending on pH and surface type. This may be due to the fact that small amount of a
mineral such as chlorite mainly contribute to the sorption of uranium and the uranium
sorbed on this mineral is hard to be desorbed from the mineral surface.

Table 2. Steady state results of sorption and desorption.

Percent of U(V1)
Test pH Sorptiontype K*(cm)
sorbed or desorbed (%)
MS- 1 5.5 Sorption 22.857 1.303
Desorption 14.328 4.190
MS-2 Sorption 26.555 1.930
7*0 Desorption 20.420 8.304
MS-3 8.5 Sorption 17.694 0.775
Desorption 12.190 5.525
FS-1 Sorption 32.668 2.242
53 Desorption 25.693 8.95 1
Sorption 46.586 16.233
FS-2
70 Desorption 41.880 21.849
Sorption 24.370 1.131
FS-3 8.5
Desorption 16.050 4.688

3.3 MineralogicaI eflect

The mineralogical composition of the fracture surface was determined by point counter
methods using electronic microscope and x-ray diffraction. It was observed that the
fracture surface was mainly composed of chlorite, biotite, quartz, and calcite. The fresh
intact granite was mainly composed of quartz, plagioclase, K-feldspar, biotite, hornblende,
and a small amount of sphene and opaque phases. We investigated the distribution of
U(V1) on the surface of granite by x-ray mapping after sorption experiments. For example,
the result of x-ray mapping for the natural hcture surfaces is shown in Fig. 3. It is noticed
lkom the result that chlorite which is present as a minor mineral mainly contributes to the

552
sorption of U(V1) onto the natural hcture surfaces. On the other hand, it is noticed fiom
the x-ray mapping study that U(V1) is preferentially associated with mica on the machined
fiesh surfaces of granite.

Figure 3. The result of X-ray mapping which shows the U(V1) distribution on granite surface. (Ksp:
K-feldspar, PI: Plagioclase, Qz: Quartz, CI: Chlorite)

4 Acknowledgements

This work was supported by the Ministry of Science and Technology, Korea.

References

1. Skagius K., Neretnieks I., Porosities and diffisivities of some nonsorbing species in
crystallinerocks, WaterResour. Res. 22 (1 986) pp. 389-398.
2. Muuronen S., Khilrtfinen E. -L., Jaakkola T., Pinnioja S. and Lindberg A., Sorption
and diffusion of radionuclides in rock matrix and natural fi-acture surfaces studied by
autoradiography, Mat. Res. SOC.Symp. Proc. 50 (1985) pp. 747-753.
3. Stenhouse M. J. and Pettinger J., Comparison of sorption databases used in recent
performance assessments involving crystalline host rock, Radiochim. Acta 66/67
(1994) pp. 267-275.
4. Vandergraaf T. T. and Abry D. R. M., Radionuclide sorption on drill core material
fiom the Canadian Shield, Nucl. Technol. 57 (1 982) pp. 399-4 12.

553
 ADSORPTION OF URANIUM(VI) ON KAOLINITE:
SPECIATION AND MECHANISM

M. J. KANG+, B. E. HAN, AND P. S. HAHN

Radioactive Waste Disposal Team, Korea Atomic Energy Research Institute,
P. 0.Box 105, Yuseong, Daejeon, 305-600, Korea
E-mail:[email protected]

The uranium speciation in underground aqueous conditions was investigated by equilibrium model
calculations as well as precipitation experiments. Complexation reactions involving the aqueous and
solid-phases of uranyl hydroxide, uranyl hydroxyl carbonate, uranyl carbonate, and uranyl oxide
were considered in the model calculation using MINTEQA2 geochemical code. It is found From the
results of equilibrium calculation that the dominant species is UOz2' at pHs 5 or below. Uranium is
precipitated as s p i e s of B-Ua(OH)2 at a neutral pH. This agrees with the results of precipitation
experiments. The aqueous phases having negative charge such as U02(0H)3-, (UOZ)ZC~(OH)~-, and
U02(C03)34 are dominant s p i e s at a high pH. The batch-type adsorption experiments were carried
out at pHs 4-10 under different Co2 conditions. Adsorption behavior of U on kaolinite can be
explained by the complexation reaction on mineral surface and the precipitation of uranium. The
adsorbed amounts and precipitated amounts have the same tendency as a knction of pH. However,
the absolute amount of adsorbed U is higher than that of precipitated U. This diffetence is mainly
caused by the surface complexation of UOz2' or other uranium species. The rapid decrease of the
adsorbed amount in the high pH range can also be caused by the anionic uranium species mentioned
above.

1 Introduction

The hazardous elements such as heavy metal and radionuclide can migrate and enter the
terrestrial and aquatic environments when the wastes are disposed of at an underground
repository. The important phenomenon to limit the migration of the elements in the
geosphere is their adsorption on surrounding solids [l]. The adsorption behavior greatly
depends on the chemical species of the elements as well as the surface property of the
solids [2]. Uranium is a widespread environmental contaminant resulting from mining and
manufacturing activities related to nuclear power. Chemical properties of uranium are
very complex and not clearly known at present. Chemical behaviors of uranium including
precipitation and adsorption are greatly influenced by the conditions of natural aquatic
systems [3]. Underground matrices are mainly composed of rock-forming materials
together with a smaller amount of oxide and clay minerals. Oxide and clay minerals have
larger adsorption potential than rock-forming materials [4]. For this reason, the studies on
the adsorption of heavy metal and radionuclide with oxide and clay minerals have been
identified as one of the important research works associated with waste disposal [5,6].
This study investigated the adsorption behavior of uranium on kaolinite under various
aqueous conditions. The uranium adsorption was studied with particular emphasis on its
chemical speciation and adsorption mechanism. Equilibrium distribution of aqueous and
solid-phase uranium species was predicted by a model calculation using geochemical code.
Experimental investigation of the partitioning of aqueous and solid-phase uranium was
also performed using a precipitation method under different pH and COz partial pressures.
The adsorbed amount of uranium on kaolinite as a function of pH was investigated and the
adsorption mechanism was proposed.

554
2 Methods

2.I Modeling of uranium speciation

Equilibrium distributions of aqueous and solid-phase uranium at various pH and COZ
conditions were calculated by the computer code MINTEQA2 [7]. The version 3.1 1 of
MINTEQA2 contains 63 kinds of complexation reactions of U with ligands and the
stability constants of each reaction. The U species, considered in these complexation
reactions are U O F and U4+.The ligands such as hydroxide, chloride, carbonate, fluoride,
sulfate, phosphate, and silicate are included. In this study, 20 kinds of complexation
reactions of U were added to the code to increase the reliability of the model calculations.
The stability constants in the code were also updated. New complexation reactions and
stability constants were referred to by the studies of Grenthe and Bond [8,9]. In this model
calculation, complexation reactions of U O F with hydroxide and carbonate ions were
considered. The species of v4' and ligands such as chloride, sulfate and phosphate were
not included considering our experimental conditions.

2.2 Experiments for precipitation or adsorption of uranium

Batch-wise experiments were performed at a fixed temperature of 25 "C under three
different gaseous conditions. The gaseous conditions tested were nearly zeroo/, 0.03%,
10% C02 partial pressures. Experiments under nearly zero% CO2 partial pressure, which
were performed to simulate C02-fiee condition, were carried out in the inert-gas glove
box. The glove box was filled with N2 gas with 99.999% purity. 0.03% and 10% C 0 2
partial pressures were to simulate atmospheric air and excess C 0 2conditions, respectively.
10% C02 condition was maintained by continuously bubbling 10% C02 gas through the
UO? solution. The 1 X 10" M U O F solution was prepared using U02(N03)2.6H20.The
ionic strength was constant at 0.01 M NaC104. The pHs of solution were varied from 4 to
10. Kaolinite was added to the prepared UO? solution and the reaction solutions were
then shaken for 3 days. The mineral to solution ratio was 1 g 5 . The liquid and the
precipitate were separated by using 0.45m membrane filter of cellulose acetate. The
conditions of precipitation experiments are controlled by the same methods as those of
adsorption experiments except for mineral. The concentration of U in the liquid was
determined by using ICP-MS.

3 Results and Discussion

3.1 Modeling of uranium speciation

Equilibrium distributions of uranium species by the model calculation are shown in
Figures 1 and 2. Figure 1 is the calculated result of U speciation under 0% C 0 2 condition.
Uranium almost exists as U O F at pHs 5.5 or below. Uranium also exists as a form of
U020H' and (U02)40H),' between pH 5 and 6.5. Uranium is precipitated as a form of
8 -U02(0H)2(s) between pH 6 and 9. The distribution percentage of the solid-phase is
over 8OYo. Uranium exists as UOz(OH)i at pHs 9 or above. The dominant solid-phase is
uranyl hydroxide because of the C02 free condition. The distribution of U species under
air condition is shown in Figure 2. Uranium exists as UO? at pHs 5.5 or below and as
U02OH+and (U02)40H),' between pH 5 and 6.5. Uranium is precipitated as species of
I3 -U0~(0H)~(s) between 6 and 7.5. The maximum percentage of the solid-phase is 54%.

555
This solid-phase disappears at pH 7.5 and U species including carbonate is formed above
pH 7. The dominant species in the pH range of 7-9 and above 9 are (UOz)&03(OH)i and
UOZ(CO&~,respectively. The aqueous phases of uranyl ion, uranyl hydroxyl carbonate
and uranyl carbonate are formed as the pHs of solution increase. The solid-phase is uranyl
hydroxide around pH 7. It is found from the equilibrium model calculations that the
dominant species at pHs 5.5 or below is uranyl ion although the COz conditions were
varied. Uranium is precipitated as a hydroxide form of I3 -UOz(OH)z(s)at a neutral pH.
The aqueous phase of uranium hydroxide, hydroxyl carbonate and carbonate are dominant
species at a high pH. These species have anionic charge.

3 4 5 6 7 8 9 10

PH
Figure 1. Calculated distribution of uranium species in the aqueous and solid-phase of 1x10" M solution
equilibrated under 0%C a condition.

100

80

60

40

20

0
3 4 5 6 7 8 9 10
DH

Figure 2. Calculated distribution of uranium species in the aqueous and solid-phase of 1x10" M solution
equilibrated under air condition.
3.2 Precipitation of uranium
The results of precipitation experiments are shown in Figure 3. The amounts of

556
solid-phase uranium in the UO, solution under 0% C02conditions was less than 30% of
initial amount of U at pHs 5 or below. The amounts of solid U increased rapidly in the
range of pH 5-9. The precipitated U above 70% was maintained in the pH ranges of 7-10.
In the case of air condition, the precipitated U was less than 40% in all pH ranges of 4-1 1
except for 54% of precipitated U at pH 6.8. The amount of solid is relatively small in the
low pH range. The amount of solid-phase increased with the pH of solution and then
decreased rapidly in the high pH range. Comparing these experimental results with the
calculated distribution of U species in Figures 1 and 2, the amounts of precipitated U
agree with the line of solid-phase of D -U02(OH)2(s). The precipitated U increases
rapidly and then decreases with the increment of pH. The precipitated amounts are
relatively small in the low and high pH range. There are maximum percentages in the pH
range of 7-9 and 6.5-7 under 0% C02and air conditions, respectively.

3.3 Adsorption of uranium on kaolinite

The adsorbed amounts of U as a function of pH under different COz conditions are also
shown in Figure 3. Under nearly 0% C02 condition, less than 40% of initial U was
adsorbed below pH 5. The adsorbed amount of U increased rapidly as the pH increased.
The adsorbed percentage reached more than 95% of initial U above pH 7 and then the
amount was constant in the pH range fiom 7 to 9.5. But the adsorbed amount began to
decrease above pH 9.5. When the C02 condition is air atmosphere, 28% of initial U was
adsorbed at pH 4.4. The adsorbed amount increased sharply in the pH range of 4-7. The
adsorbed percentage decreased slowly at pH 7 or above.
Adsorption characteristics of metal on mineral can be explained by two parameters
that influence metal behavior. One parameter is the chemical species of that metal and
another is the surface property of mineral. The chemical species of metal varies as the pH
of solution changes [2]. It is known that the progress of adsorption at permanent charge
site is different from that at the edge site of AlOH and SOH structure. Ion-exchange
reaction is a main process at the site of mineral having the permanent charge. This process
occurs at relatively low pH ranges. The adsorbed amount by ion-exchange is not varied
with pH and it is affected by other cations [lo]. Surface complexation occurs mainly at the
edge site and greatly depends on the solution because the surface charge of functional
groups varies with the pH [1 11.
The adsorbed amounts in Figure 3 do not show these phenomena related to
ion-exchange reaction. As shown in the results of Figure 3, the significant increase of
adsorbed amount is found in the pH range of 4-7. This results from the surface
complexation between chemical species of U and hydroxyl groups of mineral surface. As
the pH of the solution increases, functional groups at the edge of kaolinite have negative
charges and surface complexes are then formed by the reaction of anionic surface and
uranium species. The dominant species of U in the pH range of 4-6 is UO, as shown in
Figures 1 and 2. The uranyl cation has a good affinity with anionic surface. In the pH
range of 6-9, the dominant species of U is solid-phase of P-U02(0H)2. Therefore, the
adsorbed amount is mainly contributed with the precipitated uranium. As shown in Figure
3, the adsorbed fraction of U is higher than the precipitated fraction. This excess adsorbed
amount corresponds to the adsorbed amount by surface complexation of uranyl ion or
other uranium species. At higher pH ranges, the adsorbed amount decreased rapidly. This
is explained by anionic species such as U02(0H)3-, (U02)2C03(OH)3-and U02(C03)3, as
shown in Figures 1 and 2, which have no adsorption capacity with mineral surfaces. To
con fm the chemical species of uranium, the aqueous and solid-phase of uranium are to

557
be characterized by LIF(TRLFS) spectroscopy additionally. This method is also very
usehl to explain the phenomena in mineral surfaces.

-
8
v
100

E
.$ a0
E
3
V
60
e
0)

2
u
$ 40

:g- 20
0)
.
d

!s
a
0
3 4 5 6 7 8 9 1 0 1 1
PH
Figure The amount of precipitated or adsorbed uranium as a function of pH in the l x D" M solution
equilibrated under OO? and air condition.

References

1. Lieser K. H., Radionuclides in the geosphere: sources, mobility, reactions in natural

waters and interactions with solids, Radiochim. Acta, 70/71 (1995) pp. 355-375.
2. Stumn W., Chemistry of the solid-water interface: processes at the mineral-water and
particle-water interface in natural systems (John wiley & sons, 1992).
3. Kim J. I., Chemical behaviour of transuranic elements in natural aquatic systems,
Handbook on the physics and chemistry of the actinides, Freeman A. J. and Keller C.,
eds. (Elsevier science publishers B. V., 1986).
4. Sposito G., The chemistry of soils (Oxford univ. press, 1989).
5. Payne T. E., Lumpkin, G. R., and Waite T. D., Uranium"' adsorption on model mineral,
Akorption of metals by geomedia (Academic press, 1998).
6. Hyun S. P., Cho Y. H., Hahn P. S., and Kim S. J., Sorption mechnanism of U(1V) on a
reference montmorillonite: binding to the internal and external surfaces, J. Radioanal.
Nucl. Chem. 250(1) (2001) pp. 55-62.
7. Allison J. D., Brown D. S., and Novo-Gradac K. J., MINTEQA2/PRODEFA2, A
geochemical assessment model for environmental systems: version 3.0 user's manual,
EPA/600/3-91/021 (U. S. EPA, 1991).
8. Grenthe I., Chemical thermodynamics of uranium: NEA-TDB (OECD, North-Holland,
1992).
9. Bond K. A., HATCHES: A reference thermodynamic database for chemical
equilibrium studies, Nuex Report NSS/R379 (Nirex, 1997).
10. Dzombak D. A., and Morel F. M. M., Surface complexation modeling: hydrous ferric
oxide (John wiley & sons, 1990).
1 1. Stumn W., Part 1 The solid-solution interface, Aquatic surface chemistry (John wiley
& sons, 1987)

558
HIGH-TEMPERATURE ADSORPTION OF HAZARDOUS METAL CHLORIDES
USING ACTIVATED KAOLINITE

H. C. YANG, J. S. YUN, Y. J. CHO AND J. H. KIM

Nuclear Fuel Cycle R&D Group, Korea Atomic Energy Research Institute, Dukjinding 150,
Daejeon, 305-353, Republic of Korea, E-mail: [email protected]

This study investigated the potential of activated kaolinite as a high-temperature sorbent for the
capture of gaseous toxic metals from hot flue gas. Activated kaolinite granules in the size range of
300-400ptn were obtained from the calcination of natural clay kaolin. A kaolinite granule of this
study is comprised of 2-3 ptn grains separated by large pores, through which the metal vapors easily
diffuse. Sorption tests by passing cadmium, lead and cesium chlorides carrying flue gas through the
packed bed of kaolinite granules were performed in the temperature range of 700-9OO0C. An
increase in the temperature resulted in an increase in the capturing rate, but there was no effect of
temperature on the equilibrium uptake. Of the tested three metal species, cesium was most
preferentially adsorbed onto activated kaolinite. However, a half of captured cesium appeared to be
physically-sorbed cesium compounds, CsCI. Although captured quantities of cadmium and lead
were relatively small, most captured species were chemically-sorbed metal mineral complexes such
as PbAlzSiZ08, CdAlzSi208, CdzAlzSiz09. The results of this study suggest that activated kaolinite
granules having small grains separated by large pores is suited as a sorbent for the capture of
cadmium and lead chloride vapors from hot flue gas.

1 Introduction

High-temperature thermal treatment of hazardous waste offers a reduction in volume as

well as a conversion of toxic organic constituents to harmless or less harmful forms [I].
However, hazardous metals can neither be generated nor destroyed in the waste thermal
process, but they can be transformed both chemically and physically [2]. There is
therefore a potential for hazardous metals to emit if they vaporize at high temperatures [3].
Many matals and their salts will form vapors at temperatures reached by flame and
post-flame zones of a combustion chamber. When the vapors cool, they condense to form
submicron particles, which tend to be relatively difficult to capture in air polution control
equipments. These emissions of submicron metallic particles have been identified as one
of the greatest health risks associated with waste incineration [4].
One of the promising technologies to reduce the emissions of volatile metals is the
in-situ capture using an inorganic sorbent at high temperatures before they condense out
into fine particles [5,6]. It has recently been suggested that many volatile hazardous
metals including cadmium and lead can be reactively scavenged by inorganic sorbents
171.
The present study is to develop the better understanding of the intricate reaction
between activated kaolinite granules and vapors of cadmium, lead and cesium chlorides.
The capturing mechanisms were observed from the analysis of pre- and post-sorption
sorbent samples. The effects of temperature and metal vapor diffusion on the capturing
rate were observed fiom the analysis of long time experimental sorption data.

2 Methods

The natural kaolin was calcined at 9OOOC and activated kaolinite granules in the 300-400

559
pm size range were obtained for the sorbent materials. CdC12,PbC12and CsCl used in this
study were powdered high-purity chemical reagents (Aldrich Co., >99.9%).The principle
of this experiment is to pass a simulated flue gas, which includes lead or cadmium vapor,
through the high-temperature bed of kaolinite granules.
The experimental system, which is shown in Figure 1, mainly consists of a flue gas
supplying system, a sorption reactor assembly and a metal vapor scrubbing train. All
experiments were performed with a metal vapor-carrier gas of 20 LPM (liter per minute).
The composition of the flue gas entering the sorption bed, which includes metal vapor,
was 8% water, and 16% oxygen by volume, with the remainder consisting of nitrogen.
The metal vapor concentrations were controlled from the knowledge of the averaged
vaporization rates determined by the metal vaporizing thermo-gravimetric furnace
(MAC-QOO,LECO Ltd.) and the flue gas flow rate.
The amount of adsorbed metal was determined by the weight difference of the pre-
and post-sorption sorbent. The structural and morphological change of kaolinite granules
during sorption tests was investigated by the powdered X-ray diffraction (XRD) analysis
(Philips, Xpert MPD) and scanning electron microscopy (SEM) (JEOL, JXA 8600).

I
1 4 - 1

Figure 1. Experimenta' Set-up of high-temperature metal vapor sorption system: 1. Flow meter 2. Needle
valve 3. N2 4. 0 2 5. Steam generator 6. Gas mixer 7. Valve 8. Thermo gravimetric furnace 9. High
temperature sorption bed 10. Thermocouple 11. Filter 12. Impingers 13. Furnace controller 14.
Silicagel bed 15. Vacuum pump 16. Dry gas-meter

3 Results and Discussion

3. I Sorbent Structures
SEM microphotographs of the activated kaolinite granules are shown in Figures 2a. The
activated kaolinite granules include fine metakaolinite grains in the size range of about 2-3
pm, relatively coarse feldspar and muscovite grains. Figure 2a illustrates that the activated
kaolinite granule is an agglomerate of numerous fine metakaolinite grains and thus has a
porous structure. The results of a BET analysis of calcined kaolin agreed with the
microscopic observation. The measured BET surface area of activated kaolinite was 15.5
m2/g. No micro pores were found by BET analysis and the averaged pore diameter is
200.5 A. These results show that the macro pores of calcined kaolin were well developed
by agglomeration of fine metakaolinite grains but metakaolinite grains themselves have no
porous structures. The scanning electron microphotographs of metal-sorbed sorbents are
shown in Figure 2b-d. Adsorbed lead, cadmium and cesium were evenly distributed with

560
the surface and inside the metakaolinite grains in the activated kaolinite granules. The
distribution of cadmium, lead and cesium was similar to the initial distribution of
metakaolinite grains in the activated kaolinite granules.

Figure 2. SEM microphotographs of pre- and post-sorption kaolinite: a) activated b) lead-sorbed, c)

cadmium-sorbed, and d) kaolinite grains

3.2 Sorption reaction

The changes in the mineral compositions of the sorbent samples by metal sorption are
illustrated in Figure 3a-d. The calcinations of this raw kaolin cause the dehydration
reaction converting kaolinite minerals (AI2O3'2SiO2'xH20) into metakaolinite
(AI2O3'2SiO2). Figure 3c-d shows that metakaolinite included in the calcined kaolin acts as
a chemical sorbent for lead and cadmium capture. A kind of lead aluminum silicate
(Pb0AI2O3'2SiO2),two kinds of cadmium aluminum silicates (CdO A1203'2Si02and
2Cd0A1203.2Si02),and a kind of cesium aluminum silicate (Cs20A1203'2Si02) were
found in the powdered XRD patterns of each metal-sorbed sorbent grains. Therefore, the
following reaction schemes can be suggested for the high-temperature metal- capturing
mechanisms.
AI2O3'2Si02+ CdC12(g)+ H20(g) CdO A1203.2Si02+ 2HCl(g) (1)
Cd0A1203'2Si02+ CdC12(g) + H20(g) --* 2Cd0AI2O3'2Si02+ 2HCl(g) (2)
AI203'2Si02+ PbC12(g)+ H20(g) Pb0AI2O3,2SiO2
-+ + 2HCl(g) (3)
AI2O3'2SiO2+ 2CsCl(g) + H20(g) Cs20 A1203'2Si02+ 2HCl(g)
-+ (4)

3.3 Metal uptake

Figure 4 shows metal uptakes as a function of sorption time at 1173 K. No additional
metal uptakes were found after about 60 hr. Of the three metal vapors, cesium exhibited
the largest equilibrium uptake. In addition to capturing the metal, ideal sorbents would
retain metal species in the sorbent matrices when they are disposed of. The results of the

561
 A
a) raw kaolin

A Anorthite
[7 Halloysite
b) activated A 0 Kaolinite
kaolinite 0 Montmorilonite
V Muscovite
D Illite
a arts
c) lead-sorbed
OD 0
kaolinite I I?

d) cadmium-sorbed
kaolinite
A CdO A1203.2Si02
+ 2Cd0-AIzO3.2SiO2

e) cesium-sorbed
kaolinite

6 10 15 20 25 30 35 40 45 90

Figure 3. XRD pattterns of pre- and post-sorption activated kaolinite

extraction of metal-sorbed sorbent by the toxicity characteristic leachability procedure
(TCLP) are shown in Table 1 [8]. Fractional leachability in table 1 represents the ratio of
water-soluble metal to total captured metal. A half of captured cesium was wat&-soluble.
The results of powdered X-ray diffraction (XRD) showed the presence of water-soluble
CsCl, which is physically-sorbed cesium species in the presence of chlorine. Although
relatively small quantities of lead and cadmium were captured, most captured cadmium
and lead was water insoluble. This suggested that most captured cadmium and lead
species were in the form of water insoluble metal-mineral complexes, such as
CdOA120j2SiO2 and PbOA12032Si02,which are the products of reaction between metal
vapors and activated kaolinite (metakaolinite).

Table 1. Result of TCLP extraction test of hlly saturated sorbent

Captured metal species Cd Pb cs

Fractional leachability 0.54% <0.1% 48.6%

562
 -
5 2 . 5 ~ 1 0 L cs I
--d 2 2.0~10
-0E
E
v
1.5~10
e
$
L.
1.0~10
n
a
5.0~10~
c
e
I 0.0 I I I I
0 20 40 60 80
Time (hr)
Figure 4. Metal uptakes as a function of time at 1173 K (PwU=175,Ppbci2=175,,Ppbc1p175)

4 Conclusions

Activated kaolinite granules are comprised of 2-3 pm grains separated by large pores,
through which the metal vapors easily diffise. Diffused cadmium and lead chlorides react
with sorbent to form water insoluble metal-mineral complexes. (PbA12Si208,CdA12Si20s
and Cd2AI2Si209).Activated kaolinite grains of 300-400 pm having large pore volumes
and small grains are suited as sorbent for the capture of cadmium and lead vapors from
hot flue gas. Most captured cadmium and lead was present as water-insoluble and
environmentally non-hostile products. Although cesium was more preferentially adsorbed
onto activated kaolinite compared to other two metal vapors, a half of captured cesium
appeared to be water-sorbed cesium compounds, CsCI.

References

1. Agnihotri, R., Chauk, S., Mahuli, S., and Fan, L.S., 1998. Selenium removal using
ca-based sorbents : reaction kinetics. Environ. Sci. Technol., 32, 1841-1846.
2. Barton, R. G., Clark, W. D., and Seeker, W. R., 1990. Fate of metals in waste
combustion systems. Combust. Sci. and Tech., 74,327-342.
3. Ho, T. C.,Tan, L., Chen, C. and Hopper, J. R., 1991. Characteristics of metal capture
during fluidized bed incineration. AZChE Symposium Series, 87, 1 18-126.
4. Linak, W.P. and Wendt, J. 0. L., 1993. Toxic metal emissions from incineration:
mechanisms and control. Progr. Energy Combust. Sci. 19,145-1 85.
5. Mahuli, S., Agnihotri, R., Chauk, S., Ghosh-Dastidar, A. and Fan, L. S., 1997.
Mechanisms of arsenic sorption by hydrated lime. Environ. Sci. Technol., 31,
3226-3231.
6. Sherwood, T. K., Pigford, R. L. and Wilke, C. R., 1975. Mass Transfer, McGraw-Hill
Chemical Engineering Series, 3 19-3 1 1.
7. Uberoi, M. and Shadman, F., 1991a. Sorbents for removal of lead compounds fiom
hot flue gases. AIChE Journal, 36,3226-323 1.
8. U. S. EPA., 1993. Operational Parameters for Hazardous Waste Combustion Devices.
EPA/625R-93/008,62-63.

563
 PROBING THE CUT-OFF FOR INTRACRYSTALLINE ADSORPTION ON
ZEOLITES: PORE MOUTH ADSORPTION

REFIK OCAKOGLU', JOERI F.M. DENAYER', JOHAN A. MARTENS', GUY B. MARM3

AND GIN0 V. BARON'
I
Department of Chemical Engineering, Vrije Universiteit Brussel,
Pleinlaan 2, B-1050 Bmssel, Belgium
Joeri.denaver(iivub.ac.be

* Center for Surface Chemistry and Catalysis,

Katholieke Universiteit Leuven, Kasteelpark Arenberg 23, B-3001 Leuven, Belgium
'Laboratorium voor PetrochemischeTechniek
Universiteit Gent, Krijgslaan 281, B-9000 Gent, Belgium

1 Abstract

The pulse chromatographic technique was used to study low coverage adsorption of linear
and monobranched alkanes in the C5-CS range on two different 10-membered ring
zeolites, ZSM-22 and ZSM-23. Henry adsorption constants, enthalpies of adsorption, pre-
exponential factors of the van't Hoff equation and separation factors were determined. A
detailed interpretation of the experimental data confirmed the previously proposed pore-
mouth adsorption mode for branched alkanes on zeolite ZSM-22, where these molecules
do not have access into the depth of the pores, but rather point their linear part herein.
Zeolite ZSM-23 shows adsorption properties between zeolite ZSM-22 and zeolite ZSM-5,
which is a shape selective zeolite where normal and monobranched molecules have both
access into its pore system.

Key words: adrorption,pore-mouth, ZSM-22, ZSM-23,ZSM-5

2 Introduction

Since their discovery, microporous materials such as zeolites found major application
fields in processes like separation, ion exchange and catalysis. Their uniform pore size
and pore architecture are at the basis of separation processes whereas the chemical
composition of these materials makes them unbeatable candidates to be used as a catalyst
or an ion exchanger. Regardless of which process is used, the molecules engaged are
adsorbed on the surface according to their molecular structure and properties. The
bulkiness of the molecule compared to the pore size of the microporous material decides
if or not the molecule can be trapped in the depth of the porous framework, thus there
exists cases where molecules with larger diameters than the pore size are not able to enter
the pores. This makes the microporous materials acting as a sieve in molecular level and
they are hence referred to as molecular sieves.

564
 Zeolites are crystalline aluminosilicateswhose general formula is:

where A represents mono- (M), di- (D), or tri- (T) valent exchangeable cations, B the
tridimensional framework with an intracrystalline void space consisting of channels and
cages that may sometime be interconnected, and C molecules sorbed in their
intracrystallinevolume which are either water, organic templates used during synthesis,
or sorbed species which can also be reactants, intermediates, and products for catalytic
reaction'. This uniformly defined structure gives zeolites a large versatility in the use of
many reactions. Hydrocracking and hydroconversion are amongst the most important
applications where zeolites with their unique properties are used as host materials. It is
not surprising that zeolites with typical diameters close to those of hydrocarbons exert
shape selective effects. Zeolite Y is perhaps the most extensively zeolite used in
hydrocracking reactions, but it does not exert high selectivity towards a specific
branching due to its large pore system. Ten-ring zeolites with smaller pore structures like
ZSM-5 and ZSM-22 are more suitable for isomerization reactions in the way that they can
exhibit shape selectivity. They possess the optimal pore shape and size to suppress
undesired cracking reactions. Among the 10-membered ring zeolites, ZSM-22 is found to
be the most selective catalyst for the conversion of n-alkanes into 2-methylbranched
isoalkane~~*~. Different approaches were established to find an unambiguous explanation
for this selectivity including transition state selectivity*,product shape selectivity3and a
pore-mouth, key-lock adsorption-reaction models*6,in which catalytic transformations of
n-alkanes occur specifically in the pore openings, with the molecule only partially inside
the confinement of the pore. The branch is created in the first lobe of the pore. The deeper
the n-alkane penetrates into the pore, the stronger the interaction, explaining the
preference for branching at carbon atoms near the end of the chain. A schematic
representation of pore and pore-mouth adsorption modes is given in Figure 1.

Figure 1. Pore and pore-mouth adsorption.

Until now this new adsorption-reaction mechanism is observed only on zeolite ZSM-
22. In order to reinforce the information on such a specific pore-mouth adsorption mode it
may be very helpful if it could be extended to other structures. In this line, zeolite ZSM-
23 is perhaps the best candidate with a pore diameter and pore structure very close to that
of zeolite ZSM-22. In this paper we investigated the low coverage adsorption
properties of normal and branched alkanes in the C5-C8 range on zeolite ZSM-23 with
the pulse-chromatographic technique in order to verify if pore-mouth adsorption also
occurs on this 10-membered ring zeolite.

565
3 Experimental Procedures and Materials

The 10-membered ring zeolites (ZSM-22 and ZSM-23) were kindly provided by
Prof. Martens (COK, KULeuven). Both of the zeolites have unidimensional pore
structures without any intersection. The crystals are needle-like shaped for both materials.
Zeolite ZSM-22 (belonging to the TON family) has free pore dimensions of 0.44 X 0.55
nm and zeolite ZSM-23 (MTTfamily) has free pore diameters of 0.45 X 0.42 nm. The
framework structures are sketched in Figure 2. The low coverage adsorption properties
were determined with the pulse chromatographic technique. The details of the
experimental method are discussed elsewhere. The Henry constant was determined fiom
the first moment of the response curve on the TCD detector after injection of an alkane
trace. Adsorption enthalpy and entropy were obtained by fitting the temperature
dependence of the Henry constant to the van? Hoff equation.

ZSM-22 viewed along [OOl] ZSM-23 viewed along [ 1001

Figure 2. Framework structures of ZSM-22 and ZSM-23.

4 Results and discussion

Related to their similar pore diameter and pore structure, unsurprisingly the Henry
adsorption constants for linear alkanes are very close to each other on zeolite ZSM-22 and
ZSM-23 (Table 1). Somewhat higher constants are obtained for 2- and 3-methylbranched
alkanes on ZSM-23 compared to zeolite ZSM-22. The adsorption constants of linear
alkanes are obviously higher than branched alkanes on the two cases. The separation
power of a zeolite between a linear and a branched hydrocarbon may be given by the
separation factor (a),which is the ratio of Henry constants of linear and branched
molecules at a certain temperature. a values at 523 K are given for both zeolites in Table
1. For comparison, values for ZSM-5 are also included, which is one of the most popular
shape selective catalyst used in isomerization reactions. From this table it can be seen that
both ZSM-22 and ZSM-23 have higher separation constants compared to ZSM-5. The
zeolites can be listed in the following order with respect to their separation capacity
between linear and 2- and 3-methylbranched alkanes: ZSM-22 > ZSM-23 > ZSM-5. In
narrow pore structures such as zeolites ZSM-22 and ZSM-23 it is very probable that
linear alkanes with smaller kinetic diameters have more access to the available adsorption
sites compared to the more bulky branched molecules. This may be regarded as the first

566
reasoning that explains the large separation factors between those molecules. That the a
values are much higher on ZSM-22 and ZSM-23 than those on zeolite ZSM-5, which
already exerts shape selectivity between linear and branched molecules, may indicate that
different adsorption mechanisms, such as pore mouth adsorption for branched molecules,
may be active on the former two materials.

ption properties and separation factors on different 10-membered ring zeolites.

ZSM-22 ZSM-23 I ZSM-22 ZSM-23 ZSM-5. I ZSM-22 ZSM-23 ZSM-5' ZSM-22 ZSM-23
5.6XlO6 5.W106 I 63.3 61.6 51.1 2.lXl 0-l2 313x10"'
1.4XlOJ 17.1 13.0 66.8 z.ix10"~1 . 1 ~ 1 0 ' ~ ~
3.W10J 89.4 03.5 19.6 3.?xio-" 1.1~10'~~
8.4X10J 100.6 92.5 90.1 7.5X10-'5 4.8X10-'4
1 . w r o ~ 1.lX1o6 4.4 2.9 1.4 49.3 51.9 56.1 1.5X10'" 2.9X10"'
2.5X106 5.4 3.6 1.7 60.0 61.3 88.8 2.6X10'" 1.9X10-'3
6.1X106 5.3 3.8 1.9 18.1 17.0 18.4 1.5xiO"~ 2.1x10-~~
1.2X10J 6.1 3.3 1.9 81.3 88.8 88.6 2.4X10-" 3.4X10"
2.1x104 6.6 4.2 24 59.4 69.2 66.0 2.4X10-" 4.4X10'"
4.6X106 7.0 4.3 2.0 72.0 79.7 78.0 3.0X10-" 9.8X10-'4
9.9Xl04 2 . 0 ~ 1 0 ~ 0.5 4.1 2.0 84.5 88.8 88.5 3.6X10"' 43x10'"
a. Separalion fadm dculated fmm rehmnce 8 usin0 the van't HM relalion,enthelpiesfw ZSM-5 are fmm reference 8

The adsorption enthalpies were extracted form the slope of the individual van't
Hoff plots. The values obtained are summarized in Table 1. For pentane, almost the same
adsorption enthalpy is measured on ZSM-22 and ZSM-23. For higher n-alkanes, the
difference in adsorption enthalpy seems to become larger. For both zeolites a linear
relationship exists between the number of carbon atoms and the adsorption enthalpy for
both linear and branched hydrocarbons. On ZSM-22 the adsorption enthalpy increases
with about 13 kJ/mol per carbon added regardless the presence and the position of a
branch and this value is about 10 W/mol for zeolite ZSM-23. A further remark is that the
enthalpies of linear alkanes are clearly higher than those of branched molecules on the
two zeolites, where the differences are larger on zeolite ZSM-22 (1 3- 17 kJ/mol for ZSM-
22 and 3-7 kJ/mol for ZSM-23 - Table 1). On zeolites with well-defined pore diameters
and pore structures the adsorption enthalpy depends on the force field created by the
framework and applied on the guest molecule. These forces are mainly dispersion-
repulsion type in the case of apolar hydrocarbons'. Thus, the narrower the diameter of the
pore the higher will be the interaction. In a previous paper' it was shown that this
statement is also confirmed experimentally, where adsorption enthalpies on zeolites with
different pore diameters were compared. It was observed that on zeolite ZSM-22 the
adsorption enthalpies of branched hydrocarbons were significantly lower than on zeolite
ZSM-5 with a slightly larger pore diameter. This was attributed to a different adsorption
mechanism operating on ZSM-22, namely pore-mouth adsorption mode.
The pre-exponential factors on ZSM-22 and ZSM-23 are given in Table 1. Similar
values are obtained for the C5-CS range linear and branched hydrocarbons. Values of
linear alkanes on ZSM-23 are slightly higher than on ZSM-22, and values of branched
molecules are slightly lower. Pre-exponential factors of the isomers are significantly

567
higher than their parent linear alkane on ZSM-22, whereas comparable values are
measured for linear and branched alkanes with the same number of carbon atoms on
zeolite ZSM-23.

5 Conclusions

Adsorption of linear and branched alkanes in the C5-C8 range was studied on two
1 0-membered ring zeolites, ZSM-22 and ZSM-23, by using the pulse-chromatographic
technique. Henry constants, adsorption enthalpies, pre-exponential factors and separation
factors are determined. Although the pore diameter and framework structure of these two
zeolites are very close to each other, significant differences in adsorption properties are
observed. For ZSM-22, the adsorption properties of branched alkanes are significantly
different than those of the linear alkanes; this is elucidated by a different adsorption
mechanism for the former, namely pore mouth adsorption. With ZSM-23, the differences
in Henry constants, adsorption enthalpies and pre-exponential factors between linear and
branched alkanes are less pronounced. The studied ZSM-23 sample exhibits adsorption
properties between zeolites allowing all linear and branched alkanes into their pores (such
as ZSM-5) and a zeolite where pore mouth adsorption mode is active (such as ZSM-22).
Further work should be directed towards the study of adsorption of double and triple
branched alkanes on ZSM-23. Measurements will be repeated using a ZSM-23 sample
synthesized according to a different method, in order to investigate the effect of the
zeolite surface properties on the adsorption behaviour.

Acknowledgements

We are grateful to the IAP program of the Belgian Federal Government for support
in the project 'Supramolecular Chemistry and Catalysis'. J. Denayer is grateful to the
F. W.0.-Vlaanderen, for a fellowship as postdoctoral researcher.

6 References

1. Derouane, E. G.; Proceedings of thefirst Francqui Colloqium, 1996,7

2. Martens, J. A. ; Parton, R. ;Uytterhoeven, L. ; Jacobs, P. A. Appl. Catal. 1991,
76,95
3. Webb Ill, E. B.; Grest, G. S.Catal. Lett., 1998,56,95
4. Maesen, Th. L. M.; Schenk, M.; Vlugt, T. J. H.; de Jonge, J. P.; Smit, B. J.
Catal. 1999, 188,403
5 . Claude, M. C. ;Vanbutsele, G. ;Martens, J. A. J. Catal. 2001,203,213
6. Souverijns, W.; Martens, J. A. ;Uytterhoeven, L. ; Froment, G. F. ;Jacobs, P. A.
Progress in Zeolite and Microporous Materials 1997, 105, 1285
7. Ocakoglu, A. R., Denayer, J. F. M., Marin, G. B., Martens, J. A., Baron, G. V.;
Journal of Physical Chemistry B, 2002 (in press)
8. Denayer, J. F.; Souverijns, W.; Jacobs, P. A.; Martens, J. A.; Baron, G. V. J.
Phys. Chem. B 1998, 102,4588
9. Ruthven, D. M. Principles of adsorption and adsorption processes; John Wiley
and Sons: New York 1984,220

568
 THE LOW-TEMPERATURE SORPTION BEHAVIOUR OF CRYOSORBENT
MATERIALS

CHRIS DAY AND VOLKER HAUER

Forschungszentrum Karlsruhe GrnbH, Institute of Technical Physics,
PO Box 3640, 76021 Karlsruhe, Germany
E-mail: [email protected]

This paper presents low-temperature sorption data (experimental and correlated values) of gases that
are of particular relevance to cryosorption vacuum pumping, e.g. nitrogen, helium, hydrogen, and
deuterium. The parametric measurements were performed using a granular coconut-charcoal
(CHEMVIRON SC-I]), which was shown to have an excellent vacuum pumping performance.
Moreover, a porous non-evaporable getter material was investigated for comparison. Nitrogen
isotherms at 77 K were used for benchmarking the measurement device. The DFT approach is
employed to derive pore size distributions from the 77 K data for nitrogen on charcoal; Toth's
Y-function is also applied. The Langmuir model is found to satisfactorily represent the hydrogen data.
It is highlighted how the accessiblepore size and volume depend on the adsorptive under investigation.

1 Objectives and experimental aspects

Physisorption of gases in highly porous substances at cryogenic temperatures is a common

technique of vacuum generation in cryogenic accumulation pumps for those gases which
cannot be condensed under vacuum even at very low temperatures. Hydrophobic activated
charcoal is the favoured sorbent material, as it has some advantages as compared to other
porous materials [3]. If large quantities of hydrogen and helium are to be pumped
concurrently, as planned for the exhaust gas system of future nuclear fusion devices, such as
ITER, competitive sorption of the gases must be reliably excluded. Therefore, activated
charcoal species with a high capacity and excellent sorption performance are needed for
fusion-relevant gas species in general. This is why measurements were conducted with
helium, protium (H2), and deuterium (D2) at various temperatures between 77 K and 4 K.
For reasons of comparison, additional measurements were carried out with nitrogen at 77 K
PI.
Unfortunately, there were no solutions available on the market to characterise charcoal
materials in the required temperature range. As a consequence, an own concept was
developed, which evolved in the new experimental facility COOLSORP (see [2, 41 for
technical details). COOLSORP combines a standard apparatus for sorbent characterisation
at 77 K and a two-stage Gifford McMahon refrigerator(the cold head being in direct contact
with the sample cell) so as to extend the operational range down to cryogenic temperatures.
By defined heating of the second stage, each temperature between 4 K and 100 K can be
adjusted. The reference cryosorption material is a coconut-shell-based, highly activated
granular charcoal (CHEMVIRON CARBON, type SC-11, US mesh size 12x30), with the
sample masses being about 2.5 g.

2 Results

2. I Studies with nitrogen

Sorption of nitrogen at 77 K was investigated on several sample batches for benchmarking

569
the new facility, for consistency checks, and because this represents a standard procedure.
The reproducibility of the measurements was better than 5% for samples of different
production years and otherwise constant conditions. Comparison of the isotherms
measured with the literature data available for the same material [I, 51 yielded a good
agreement, cf. Fig. 1. A classical type4 behaviour is clearly evident. However, some
differences become obvious by a more in-depth analysis of the data sets, as will be
described below.
The center of Fig. 1 illustrates the Y-functions as proposed by T6th [6,7] for
apparent type4 isotherms. This helps to identie the appropriate isotherm equation. The Y
maximum denotes the completion of monolayer formation, which, however, is not
consistent for the different samples analysed here. More experience with the T6th
approach will be gathered in the ongoing analysis and help gaining a better understanding
of this result.
The differential pore distribution as obtained by the Density Functional Theory is also
shown in Fig. 1 (bottom). Overall distribution for all measurements was found to be in the
range of 6-30 A, with pores above 16 A making up a very small fraction only. A peak
value that is common to all data sets is 12.5 A. Only the NBC samples, although not all of
them, show a second peak at 10.7 A.

2.2 Studies with hydrogens

Fig. 2 highlights the isotopic effect for H2and Dz by showing the adsorption isotherms
measured under both sub- and supercritical conditions (T&(Hz)=33.19 K,
T&t(D+38.34K). The lower part illustrates the evaluation of the experimental data
according to the Langmuir equation. It becomes clear that the Langmuir fit works very
well for nitrogen, shows an acceptable performance for the hydrogens, but is not
applicable to helium. The variation of the results is shown in Table 1.

Table 1. Rcsults of the Langmoir plot evaluation at n o d boiling temperature.

GaS Temp. [K] V (monolayer) [cd/g] Surface [m2/g]

Protium 20.4 493 1734
Deuterium 23.7 519 1826
Nitrogen 77.4 318 1386

570
 I I

120

100

80
A

&
3
* 60

40

20

0.0 0.2 0.4 0.6 0.8 1.0

Relative Pressure (p/po)

0.14

9-
0.12

0.10

-5 0.08

sg
I

0.04

0.02

0.00
0 5 10 15 20 25 30 35 40
pore width (A)

Figure 1. Comparison of own data measured for NZat 77 K with data by Ldewyckx [5]. The isotherms (top),
the derived T6th Y-function (middle), and the DE;T analysis of these data (bottom)are presented.

571
 0

0.0035

0.003

0.0025
6-
E
-8 0.002

50.0015
\

n
0.001

0.0005

0
0 0.2 0.4 0.6 0.8 1
Relative Pressure (p/Po)

Figure 2. Iuustration of typical measured isotherm curves of Hz and 9(top); intercomparison of different
sorptives by evaluation of experimental data (data For He from 141) at their n o d W i g temperam in a
Langmuir plot, which should yield a straight line when applicable (bottom).

2.3 Studies using getter material

For comparison, a porous non-evaporable getter alloy material (ZrNalFe, trade name
ST707,SAES Getters) which is commonly used in getter vacuum pumps was investigated
as well. The sample was a typical strip with an active surface of 13.5 c d and a specific
getter mass of 38 mg getter/cm2.The measurements, as shown in Fig. 3, indicate a type-III
sorption behaviour with small amounts adsorbed. However, the relative differences of the
three sorptives are more pronounced on the getter than on charcoal, especially for helium
compared to hydrogen.

572
 0 0.2 0.4 0.6 0.8 1
RelativePressure (p/pc)

Figure 3. Adsorption isotherms of ST707getter material.

3 Conclusion and Acknowledgements

Two porous materials relevant to cryosorption vacuum pumping were characterised by

means of adsorption isotherms. Based on the example of the Langmuir isotherm, it is shown
that the use of standard models may produce ambiguous results when applied to
non-standard cases.
The helpful support of Peter Lodewyckx, Belgian Army - NBC Division, who provided us
with a lot of reference data and thus enabled us to benchmark our own facility, is gratehlly
acknowledged.

References

1. Blacher S. Sahouli B., Heinrichs B., Lodewyckx P.,Pirard R. and Pirard J.P., Lungmuir
16 (2000) 16, pp. 6754-6756.
2. Day Chr. and Hauer V., Pore characterisation of cryosorbent carbon materials. In K.
KANEKO (ed.) Fundamentals of Adsorption 7 (International Adsorption Society,
2002) pp. 1093-1100
3. Day Chr., Colloids undSu$aces A 187-188 (2001) pp. 187-206.
4. Hauer V. and Day Chr., Cryosorbent characterization of activated charcoal in the
COOLSORP facility, Report FZKA 6745 (ForschungszentrumKarlsruhe, Karlsruhe,
2002).
5. Lodewyckx P., Measurement data, personal communication, NBC Division Belgian
Army, 200 1.
6. T6th J., J. Coll. Inte$ace Science 225 (2000) pp. 191-195.
7. T6th J., Advances Coll. andInte$ace Science 55 (1995) pp. 1-239.

573
 THERMAL AND SURFACE MECHANISM STUDIES ON
ADSORPTION-TEMPERATUREPROGRAMMED DESORPTION OF
NITROGEN OXIDES OVER CHEMICALLY ACTIVATED CARBON FIBER

HYUN-JIN KIM,YOUNG-WAN LEE, AND DAE-KI CHOI

Korea Institute of Science & Technology, P.0.BOX 131, Cheongryang,Seoul. 130-650. Korea
E-mail: zzini77@st, re.kr

EUNIL LEE
Korea University, IJ-Ka, Anam-dong Sungbuk-ku, Seoul,136-701, Korea

Chemically activated carbon fibers (CACF) manufactured KOH impregnated to ACF and confirmed
selective adsorption conduct of NO and NO2 in fixed bed adsorption column. NO, desorption
studied on He at SC/min up to 800c using temperature programmed desorption (TPD),and
observed surface characteristic at absorption-desorption. CACF was increased adsorptivity with
offers selective adsorptivity by KOH in NO, adsorption, and CACF (K0H:ACF = 1:3) had an
adsorptivity that was four times higher than that of ACF. NOx desorption on ACF was mostly
occurred within 200C. The results of surface characterization were found that NO, was produced
as a KNOX(x=2,3)after adsorption and potassium ions were distributed without loss after desorption.

1 Introduction

Anthropogenic sources of NO, include transport vehicles such as automobiles and

airplanes and stationary sources such as fossil fuel power plants, boilers and incinerators.
Though NO,-inhibiting technologies have continuously been advanced, thanks to the
efforts of many researchers, no omnipotent process is yet in place, because it is naturally
very difficult to control NO, . Active studies are underway to improve the existing
DeNO, process and to employ a new process. A NO,-controlling technology should be
commercially feasible in terms of relative efficiency and applicability. Especially, to be
able to select the most optimum technology, economic evaluation based on cost efficiency
should be made. Adsorption requires little electricity and needs no additional cost for
treatment, because it does not cause secbndary pollution. The adsorption process with
the fixed bed adsorber has been widely used for the abatement of air pollutants. The
fixed bed adsorption process is recognized as efficient in the present situation where there
is no magic formula process for the control of NO,. Adsorbents packed to fixed bed
adsorber are carbonaceous materials such as activated carbon, activated carbon fiber and
activated coke. Currently drawing attention is an on-going study that intends to enhance
the abatement of certain noxious gases by impregnating chemicals with high selective
adsorptivity into the aforementioned adsorbents [ 11. The possibility of removing harmful
gas using carbonaceous materials during the adsorption process has been examined
because the unique characteristics of carbon materials themselves. Especially, activated
carbon fiber (ACF) has been used in treating harmful gas in the air due to its hydrophobic
property, high surface area and intrinsic pore structure. Because activated carbon fiber
has excellent performance, capacity had increased rapidly day after day [2].
Impregnated activated carbon fiber is material that made to modify surface of activated
carbon fiber chemically. The impregnated activated carbon fiber raises adsorption
performance for material that is not adsorbed well to activated carbon fiber, and improves

574
selectivity of specific catalysis and chemical reaction. It can expect very good effect if
condition of most suitable about choice of impregnant and the amount is established.
And it must ponder how expense is cost because impregnated activated carbon fiber is
costlier than activated carbon fiber. When apply in adsorption for specification gas, if
impregnant exists on the surface of activated carbon fiber, removal mechanism by
chemical adsorption that is concerned to chemical reaction/catalysis preferably more than
physical adsorption acts by much bigger factor. Chemical adsorption need activation
process like most chemical reaction, and generate single adsorption. Also, unlike
physical adsorption, as the temperature is higher, endothermic reaction generates and
collision fraction increases. The chemical reaction by lively molecular motion between
impregnant and adsorbate can be generated. Important factor that affect the chemical
property of impregnated activated carbon fiber is interference action between impregnant
and activated carbon fiber. Usually, for other metal or impurities with possibility to be
unrelated chemically with metal substance of surface when activated carbon fiber is
support, their effect to activated carbon fiber is possibility that give toxicity to impregnant.
Therefore, when make impregnated activated carbon fiber, the properties of activated
carbon fiber should be have well-developed pores, high surface area, strong mechanical
strength and no impurities [3,4].
Our study applies an adsorption process using chemically activated carbon fibers
(CACF) to effect NO,, a prevalent atmospheric pollutant difficult to control. A concrete
understanding of adsorption and desorption behavior and surface chemistry on was
progressed.

2 Experimental Methods

1. Sample Preparation. Pitch-based ACF is used as adsorbent. CACF was prepared

by impregnating a KOH. KOH was impregnated in an aqueous solution state in the
ACF via incipient wet impregnation.
2. Apparatus and Method. The fixed-bed column was a 316 stainless steel tube, with
an inside diameter of 1.09cm and a length of 40cm. The temperature of the coIumn
was maintained with an electric &mace located at its outer wall. Concentrations of
NO and NO2 that exhausted from the bypass line and adsorption column were
analyzed by using a chemiluminescent NO, analyzer (Thermo Environmental
Instruments Inc., model 42C).
3. TPR Procedure. The first step is a heat treatment of CACF to 10-100C in the
presence of He. The second step is the NO, adsorption in the presence of 02. In
the third step for cooling down the sample to 10C. Then, after maintaining to the
temperature for 30min, a temperature programmed desorption (TPD) experiment is
done in He at SC/min up to 800 c.
4. Analysis. By using TGAiDSC, SEM, EPMA and ToF-SIMS, we examined the
thermal behavior and chemical characteristics created on CACF through the
adsorption and desorption of NO,.

3 Results and Discussion

We observed the character changes of CACF manufactured with impregnation of

KOH on ACF. It showed results of TGA and DSC analysis on KOH, ACF and CACF.

575
The results may be deduced that the weight of CACF decreases at a lower temperature
when compared to KOH because the bond energy of KOH is ionized after being
impregnated with ACF. However, the fact that TGA curve showed a similar pattern for
CACF and KOH proves that KOH does not exist in the form of crystal on or inside of
ACF, but has been bonded with ACF physically or chemically, or its electronic property
has been changed due to the movement of its electrons on the surface. This trend may be
seen in DSC, which showed the modified properties of surface. The difference in the
TGA and DSC trends of the three samples are attributable to functional groups and
chemical structures existing on surface, or interaction and physical -and-chemical bonds
between KOH and ACF.
In order to confirm the effect of CACF, compared the breakthrough curve of NO, on
CACF and ACF. The result confirms that CACF shows a good selectivity with NO,, and
it can be determined that KOH provided strong basic functional group to the ACF. And
when adsorbing NOz/Air to CACF as the loading amount of impregnant increases, the
removal efficiency relatively increases because KOH given more selective adsorption site
for NO,. When NO2 was adsorbed on CACF at the temperature 2 IOO'C, the
completely-oxidized KN03, the partially-oxidized KNOz and H20 were produced at the
same time. This indicates that the adsorbent loses its function when selective adsorption
sites are all converted into KN03. However, as for the sample having not reached
saturation points, KOH and KN02 co-exist on CACF at the same time. H20, produced at
the same time, may serve as a side effect to produce HN03or HN02 on the surface. As
the degree of the influence to be exerted by H20 may be dependent on temperature, the
reaction occurs more vigorously when temperature becomes lower.

E,
a
4000
.-0
5c! 2000
u
0
E

u o0
- 100 200 300
Temperature ("C)
400 500 600

0 100 200 300 400 500 600

Temperalure ( " C )

Figure 1. NO and NO2 concentration profiles obtained from Temperature Programmed Desorption when the
bed depths ofNO, adsorbed CACF are (a) ACF, (b) KOH:ACF=I:9, (c) KOH:ACF=l:4, (d) KOH:ACF=l:3, (e)
KOH:ACF=l:2 and (0 KOH:ACF=l:I.

Figure 1 shows NO and NO2 concentration profile obtained t?om temperature

-
programmed desorption. (a) is ACF, (b) (f) as CACF, the ratio of KOH to ACF is 1 :9,
1:4, 1:3, 1:2, and 1:1, respectively. CACF was increased adsorptivity with offers
selective adsorptivity by KOH in NO,adsorption, and CACF (K0H:ACF = 1:3) had an
adsorptivity that was four times higher than that of ACF. However, excess impregnated
KOH (1 :2, 1:1) was deactivated by pore blocking. NO, desorption on CACF was mostly

576
occurred within 20072. As the ratio of KOH increases except for 1:l and 1:2,
concentration of NO and NO2desorbed increases.
Potassium distribution on surface after desorption was analysed by using EPMA. A
previous report concerned the result of the K mapping after production of CACF.
According to the mapping result with magnitude x 3500, the white parts showed
distribution of K. K did not show any difference from the result of the test conducted
after the manufxture of CACF against the sample where NO, was adsorbed and desorbed.
Therefore, Figure 2 was found that NO, only was desorbed fiom the surface with the
crystal, comprised of KNO,(x=2,3) having been created after adsorption, upon desorption
at high temperature 5 600C. The results of ToF-SIMS, stemming from polarity of
positive ions detected through the analysis, were seen in the range of 1-100 m/z, which
provides the data on major peaks. As a result of a qualitative analysis of the spectrum,
K+(m/z-39), KO'(dz=55), K;(m/z=78) and K20+(m/z=94) were observed. Such a
result confirms that a large amount of K existed on the surface after desorption. The
result of surface analysis by EPMA, SEM, and ToF-SIMS confirms that K was distributed
with the level of concentration similar to that of K distributed on the manufactured CACF,
which was not included in the present study though.

Figure 2. The ToF-SIMS positive and negative spectra : C / C d . 9 8 , adsorbed for 2hr with near 5OOppm of
inlet NO, concentration at l o o t , (a)-(b) nonadsorbed on CACF (KOH:ACF=I:3) ; (c)-(d) NO, adsorbed on
CACF (KOH:ACF=l:3).

Reference

1. Lee Y. W., Choi D. K., and Park J. W., Surface chemical characterization using
AES/SAM and ToF-SIMS on KOH-impregnated activated carbon by selective
adsorption of NO, Ind Eng. Chem. Res. 40 (2001) pp. 3337-3345.
2. Mochida I., Kawabuchi Y., Kawano S., Matsumura Y. and Yoshikawa M., High
catalytic activity of pitch-based activated carbon fibers of moderate surface area for
oxidation of NO to NO2 at room temperature. Fuel 76 (1997) pp. 543-548.
3. Illhn-Gom6zM.J., Linares-Solano A., Radovic L. R. and Salinas-Martinez de Lecea
C., NO Reduction by Activated Carbons. 2. Catalytic Effet of Potassium. Energy

577
 Fuels 9 (1995) pp. 97-103.
4. Zhu Z. H.and Lu G. O., Catalytic Conversion of N20to N2over Potassium Catalyst
Supported on Activated Carbon. J. Catall87 (I 999) pp. 262-274.

578
 SURFACE PROPERTIES OF ACTIVATED CARBONS CONTAINING BASIC
HYDROXIDE IONS AND NOx ADSORPTION-DESORPTION PROCESS

Y.W.LEE, H. J. KIM, D.K.CHOI, AND B. K.NA

Korea Institute of Science & Technologv
P.O.BOX 131, Cheongryang,Seoul 130-650, Korea
E-mail: dkchoi6iJiiJist.re.b

J. W.PARK AND C . H.LEE

Yonsei Universiw
134, shinchon-dong,Seodaemun-ku. Seoul, 120-749, Korea

Activated carbon modified chemically with potassium hydroxide via low temperature wet
impregnation method for NOx adsorption was used. The present study examined adsorption and
desorption behaviors and the accompanied surface chemistry. In particular, typical desorption
behavior was examined after NOx was adsorbed at 100C while the temperature was increased up to
60OoC. This study found that the presence of a relatively larger amount of adsorbent delays surface
oxidation. NOx has become oxidized while inducing three types of physical and chemical bonds
on the surface of BHAC. The results concerning the surface properties after adsorption and
desorption were analyzed using various instruments. Potassium existed on the surface without
consumption as K-IAC was adsorbed by potassium oxide after desorption.

1 Introduction

Nitrogen oxides (NOx) is regarded as the most notorious toxic gases emitted into the air in
the process of fossil fuels consumption. However, NOx emission remains difficult
control issues despite decades of research [l-31. NOx is not subdued effectively by
existing granular activated carbons. Over the recent several years, development of
surface modified activated carbon served as a springboard for improving the existing
granular activated carbon (GAC). Such surface modified activated carbon proved to
have far superior adsorption capacity against environment -polluting hazardous
inorganic/organic gases [1,4,5,6,7,8]. Despite the advantages described as above,
surface modified activated carbon still needs to undergo significant improvement via more
systematic and advanced technologies in order to reach perfect process. Adsorption onto
the activated carbon modified chemically with potassium hydroxide (BHAC), in particular,
accompanies not only physical adsorption but also chemical adsorption with adsorbate,
which makes desorption process of BHAC more complicated when compared to GAC.
Desorption requires surface chemical information on molecular ions of adsorbate
distributed on the surface. However, the previous researches lacked study of these aspects.
Therefore, this study investigated a typical behavior that appears when BHAC is used to
examine adsorption at different bed depths at the temperature of 100C and the subsequent
desorption by adjusting the temperature upward. Furthermore, this study analyzed various
surface characteristics in order to collect concrete data concerning surface chemistry after
adsorption and desorption.

579
2 Methods

BHAC was prepared by impregnating a potassium hydroxide into GAC at 85C for 3h by
wet impregnation method. The fixed-bed column was a 3 16 stainless steel tube, with an
inside diameter of 1.09cm and a length of 40cm. Concentrations of NO and NO2 that
exhausted from the bypass line and adsorption column were analyzed by using a
chemiluminescent NO, analyzer. The first step of adsorption-desorption is a heat
treatment of BHAC to 10-100C in the presence of He. The second step is the NOx
adsorption in the presence of 02. In the third step for cooling down the sample to 10C.
Then, after maintaining to the temperature for 30min, a temperature programmed
desorption(TPD) experiment is done in He at 5"C/rnin up to 600 "C. For surface
characterization, Differential Scanning Calorimetry(DSC), Thermal Gravimetric
Analysis(TGA), Scanning Electron Microscopy(SEM), Energy Probe
Micro-Analysis(EPMA) and Time of Flight Secondary Ion Mass
Spectroscopy(ToF-SIMS) are applied for the identification of thermal and surface
chemical information.

3 Results and Discussion

The adsorption experiments, when the bed depth(amount) of BHAC was lcm (0.534g),
2cm (1.067g) and 9cm(4.803g), were tested at 403 K of temperature, 500ppm of NOx
concentration, 30cm/s of linear velocity. The strong basic surface of BHAC gets
oxidized in its adsorptiordreaction with NOx gradually. In this case, the rate, at which
impregnant loses its function after the oxidation of the BHAC surface, mostly depends on
the amounts of impregnant and adsorbent, structure of the activated carbon, concentration,
temperature, and linear velocity. The overall surface oxidation rate decreases when the
amount of NOx adsorbed increases comparatively, i.e. the bed depth gets deeper. The
primary reason of adsorptivity increase is that there are many selective adsorption sites
with the K+ and OH' ions, which can react with NOx. The second reason is deemed to be
related to the re-adsorption of the NO produced [ 1,2,3]. For the bed depth of lcm, 2cm
and 9cm, the each concentration means C/Co of 0.9122, 0.6766 and 0.2828 and the
surface oxidation rate of 91%, 68% and 28%, respectively. This result implies the fact
that fictional loss of the selective adsorption sites is directly related to oxidation of
alkaline surface. When NO2 was adsorbed on BHAC, surface oxidation occurred as seen
in the Eq. (1).
2BHAC + K N 0 2 ( a ) + K N 0 3 ( a ) +H,O(g) + 2KN03(a)+NO(g)+ H,O
ZNO, NO2
(1)
As the degree of the influence to be exerted by H20may be dependent on temperature, the
reaction in Eq. (2) occurs more vigorously when temperature becomes lower.
GAC + C ( N O * ) , ( a ) + C ( N O , ) , ( a )2C(NO,),(a)+3NO(g)
,~~ (2)
6N02

Eq. (2) may occur as a minor side reaction. However, adsorption was not smooth owing
to the boiling point (b.p=2 1.15"C) of NO2 at temperature 2 100C. The adsorption via
chemical reaction, as in Eq. (l), shows high adsorptivity at the temperature range fiom
100C to 150C, which is several times higher than that of adsorption of NOx to GAC
itself in Eq. (2) as reported in the previous papers[2,3]. The chemical modification of
the activated carbon uses chemical substances that have high selectivity against each

580
adsorbate and induces chemical bond on the surface.

Figure 1. A. NO and B.NO2 concentration profiles obtained from Temperature Programmed Desorption when
the bed depths of NOx adsorbedBHAC are lcm, 2cm, and 9cm.

Figure 1A shows NO concentration profile obtained from temperature programmed

desorption. For the bed depth of lcm and 2cm, similar desorption peaks were observed
at similar temperatures. For bed depth of lcm, each of the three peaks showed its
maximum NO concentration of 445ppm at 15OoC, 1500ppm at 300C and 1008ppm at
385C. The bed depth of 2cm displays the maximum NO concentration as much as
approximately twice, when compared to bed depth of lcm: it was observed that the
maximum NO concentration reached 1302ppm at 16OoC, 23OOpprn at 295C and
215Oppm at 375C. On the other hand, the bed depth of 9cm shows the maximum NO
concentration of 1279ppm at 280C while displaying a weak NO peak of 133ppm at
360C. Figure 1B shows the NO2 concentration profile obtained from temperature
programmed desorption under the same condition of Figure 1A. For lcm of bed depth,
the maximum concentration was observed to reach 49ppm at 125"C, 13ppm at 270C and
288ppm at 395C. In bed depth of 2cm, the maximum NOtconcentration was observed
to reach 4lppm at 120C and 400ppm at 395 "C. The bed depth of 9cm shows the
maximum concentration of 48ppm at 280C and 4.8ppm at 380C. Accordingly, the
peaks of desorption behavior may be classified into three categories depending on
adsorption strength of NOx. As seen in the graphs of Figure 1A and 1B with surface
characterization using TGA-DSC, SEM, EPMAEDS, ToF-SIMS, three peaks are
classified [8]:
(1) Weak bond (WB at 100-210C): Weak bond is formed on the surface via physical
adsorption which may be attributable the following possibilities: A) incomplete bond
between unstable carbon out of carbon forming the surface of BHAC and N O M 4 (Eq.
(3)) and B) production of HNOz and HN03 (Eq. (4)). The both reactions emit NO in
higher concentration than NO2.

A 3
2HN03(I,a)+ 2HNO,(I,a)+ NO,(g) + 3NO(g)+ 2H,O(g) + TO,(g) (4)
He
(2) Strong bond (SB at 215-350C): potassium nitrite, having been produced on the
surface, seems to depend on decomposition. The sample having not reached complete
saturation / surface oxidation has high intensity. The decomposition of potassium nitrite
also occurs mostly due to production of NO as seen in Eq. (9, where potassium oxide is

581
adsorbed.
A
2rnO2(4 +
He
OW + 2 W g ) + OAg) (5)
(3) Very strong bond (VSB at 355-600C): A peak occurs with decomposition of
potassium nitrate, i.e. a more stable substance than potassium nitrite. As NO and NO2 are
produced at the same time, potassium oxide is produced and adsorbed (Eq.(6)).
A
2 m 0 , (4
+ K 2 0 ( 4 + NO2 ( g )+ N a g ) + 0 2 ( g ) (6)
He

A. B. C.

Figure 2. SlMS depth profiles of A. NOi, B. NO,and C. OH ions in the non adsorbed BHAC and BHAC after
NOx adsorption and desorption.

Figure 2 A and 2 B shows far larger distribution of NO; and N O to cover 462A on the
surface with 68% oxidation of the surface (bed depth =2cm). The overall distribution of
NO; to reach 1232 A of surface depth, was found to be the highest in NOx ahorbed
>BHAC>after desorption in order. It is deemed that the lower distribution of NO; after
adsorption than BHAC may be attributed to the fact that N hnctional group, having
existed on BHAC, was bonded with surface oxygen and came to be partially desorbed
from the surface when temperature increased. Figure 2 C shows sputter depth profile of
OK. Distribution of OH- is found to be the highest in BHAC. On BHAC, OH- ions
provide selective adsorption sites and react with NO2, as seen in Eq. (l), and evaporate
into H20. Accordingly, after adsorption of NO2, a substantial reduction of O K was
observed regardless of surface depths. However, after desorption, O K ions increased on
the surface which may be explained by two possibilities:
A
KOH(s) + K 2 0 ( a )+ H,O(g) (7)
He

1) As KOH in BHAC, non-reacted selective adsorption sites, was decomposed, as seen Eq.
(7), to produce H20 upon desorption at a high temperature, and H20 evaporated to enable
a hydrogen bond between H with surface oxygen; and 2) H bond was broken away from
some deficient carbon due to a high temperature and H was bonded with surface oxygen.

4 References

1. Lee Y. W., Choi D. K., and Park J. W., Surface chemical characterization using
AES/SAM and ToF-SIMS on KOH-impregnated activated carbon by selective
adsorption of NO, Ind Eng. Chem. Rex 40 (2001) pp. 3337-3345.
2. Lee Y. W., Choi D. K., and Park J. W., Characteristics of NOx adsorption and

582
 surface chemistry on impregnated activated carbon, Sep. Sci. Technol. 37 (2002) pp.
937-956.
3. Lee Y. W., Choi D. K., and Park J. W., Performance of fixed-bed KOH impregnated
activated carbon adsorber for NO and NOz removal with oxygen, Carbon 40 (2002)
pp. 1409-1417.
4. Lee Y. W., Choung J. H., Choi D. K., and Park J. W., NOx adsorption on
impregnated activated carbon, Fundamentals of A&orption, IK International Ltd.
(2002) pp. 154-161.
5. Lee Y. W., Park J. W., Choung J. H., and Choi D. K., Adsorption characteristics of
SO2 on activated carbon prepared fkom coconut Shell with potassium hydroxide
activation, Environ. Sci. Technol.36 (2002) pp. 1086-1092.
6. Lee Y.W., Park J. W., Yun J. H., Lee J. H., and Choi D. K., Studies on the surface
chemistry Based on competitive adsorption of NO,-SOz onto a KOH impregnated
activated carbon in excess 02,Environ. Sci. Technol.,36 (2002)pp. 4928-4935.
7. Lee Y.W., Park J. W., Kim H. J., Park J. W., Choi B. U., and Choi D. K.,Adsorption
and reaction behavior for simultaneous adsorption of NOx and SOz over
carbon-supported potassium catalysts, Carbon, in revision.
8. Lee Y. W., Kim H. J., Choi D. K., Yie J. E., and Park J. W., Temperature
programmed reaction and regeneration studies of NOx over potassium
hydroxide-containingactivated carbon, Emiron. Sci. Technol, in revision.

583
ADSORPTION CHARACTERISTICSOF NITROGEN COMPOUNDS ON SILICA
SURFACE

HYUN JONG KIM,CHANG HA LEE AND YONG GUN SHUL

Department of Chemical Engineering, Yonsei Universiw, Seoul, Korea
E-mail: [email protected]

WHA SIK M M
SK R&D Center, Daejeon, Korea
E-mail: [email protected]

The interaction between silica surface and nitrogen compounds was studied by using mainly solid
state NMR. The quinoline as basic nitrogen cornpound and carbazole as non-basic nitrogen
compound were adsorbed on the dry or wet silica Both of them made a hydrogen bonded hydroxyl
proton on the surface of silica. Surface water on the silica might effect on the interaction between
silica surface and nitrogen compounds.

1 Introduction

A strive toward a cleaner environment has led to the global tightening of the sulfur content
in automotive diesel fuel. For example, the sulfur limit of 500ppm has been adopted by
EEC since 1996, and in Japan since 1997[1]. More severe specifications with a sulfur
content of 35Oppm S is now practiced, and a sulfur level as low as 5Oppm or even at least
lOppm is being proposed in Europe for the year 2005[2]. For this reason, there are
considerable efforts being expended to develop new technologies for the production of
clean fuels, like adsorption, extraction, oxidation, alkylation, and bioprocessing[2].
Currently, however, hydrodesulfiuization(HDS) appears to be the technologically
preferred solution.
The most practical method to produce 10 ppm ultra low sulfbr diesel is considered to
be the two-stage HDS process. In this process, the sulfur levels are reduced down to
around 250 ppm in the first stage and are further reduced down, after the hot gas that
contains the inhibiting components are removed, to below 10 ppm in the second stage.
This second stage requires much higher pressure, hydrogen to feed ratio, hydrogen purity
and lower space velocity compared to the conditions used in most of the conventional
HDS processes. It could increase the capital investment[3].
The performance of HDS is lowered by organic hetercompounds and polyaromatic
hydrocarbons, for which the following order of inhibition has been reported Nitrogen
compounds > organic sulfur conpounds > oxygen compounds > monoaromatic
hydrocarbons[4]. SK Corporation in Korea has been developed a unique technology,
called SK HDS Pretreatment Process, to produce the lOppm ultra low sulfur diesel[3].
The technology is based on the adsorption technique, and is designed to remove nitrogen
compounds fiom the feedstock to the HDS process. It improved the performance of
conventional HDS units, by removing effectively the components that inhibit the
desulfiuizationreaction.
The nitrogen compounds have been characterized among the strongest HDS
inhibitors. Many researchers reported a strong inhibiting effect on the thiophene and DBT
HDS reactions[5-8]. It is well known that the basic nitrogen compounds could poison the

584
acidic sites of catalysts and, consequently, retard desulfurization efficiency[5,61. Recently,
it is reported in some papers that the non-basic nitrogen compounds can also strongly
inhibit hydroprocessing reactions through competitive adsorption[7,8]. In this study, the
nitrogen compounds in light gas oil was removed by adsorption technology, similarly with
SK HDS Pretreatment Process. And the local interaction between nitrogen compounds
and adsorbent was explored by solid-state NMR.For the analysis, we used model light gas
oil containing the basic or non-basic nitrogen compounds,

2 Experimentals
The nitrogen compound in the light gas oil was classified with basic nitrogen compound
and non-basic compound. For the practical condition,quinoline was used as basic nitrogen
compound and carbazole was as non-basic nitrogen compound. And normal hexane was
selected as model light gas oil. And, silica was used for the adsorbent of nitrogen
compounds.
The model light gas oil was prepared with 300 ppm nitrogen compounds in hexane.
For the adsorption of nitrogen compounds, the adsorbent was immersed in the model light
gas oil for 6 hours. The filtered silica particle was dried in the room temperature for 4
hours.
The 'H M A S NMR was performed on a JNM-ECP300 JEOL spectrometer with a
spinning rates of ca. 5.5 kHz, for the understanding of the local interaction between
nitrogen compound and adsorbent. 1H NMR spectra was carried out at 300.53 MHz using
single-pulse excitation. The d2 pulse width and pulse delay were 4 p and 10s.

3 Results
Before the adsorption of nitrogen compounds, the silica was exposed in the moistured air
to have different water contents. Figure 1 shows the 'H MAS NMR spectra of silica used
in this study. Figure l(a) is the N M R spectrum of dry silica powder. The narrow peak at 6
= 1.8 ppm is assigned the isolated silanol. It showed general spectrum of dry silica as
reported in many studiesf9-1I]. After the exposure in wet condition, the resonance at 3.5
ppm which means the physically adsorbed water was appeared. The water content of silica
was increased by exposure in wet air up to 0.11 1glg.silica . These dry and wet silica were
used for adsorption of nitrogen compounds.

20 15 10
- 5 0 -6
shin (6. m)
40 -15 -20

Figure 1. 'H MAS N M R spectra of silica (a) before the exposure in moisture and @I) after the'exposure in
mosture.

585
 20 15 10 5 0 -5 -10

Chmicpl shift (S, ppm)

Figure 2. 'HMAS NMR spectra of dry silica (a) before the adsorption of quinolineand (b) after the adsorption
of quinoline.

Quinoline as basic nitrogen compound was adsorbed on dry silica. As shown in figure 2,
the 'H N M R spectrum of the silica with adsorption of quinoline was considerably
changed. The peaks at k 8 . 5 , 7.9 and 7.5 ppm were newly appeared. These peaks could
be assigned to the hydrogen of quinoline. It means that quinoline was adsorbed on the
silica surface. The intensity of peaks was not changed above 130 ppm of quinoline in
hexane. The adsorbed amounts of quinoline was 0.06 g/g.
And, the isolated silanol signal at k1.8 ppm was disappeared and the resonance at k 5 . 0
ppm was highly increased. The broad peak at W . 0 ppm is the hydrogen bonded hydroxyl
proton. With adsorption of quinoline on the silica surface, the hydrogen bonded silanol
was increased. It might be because quinoline could make strong interaction with silica
surface.

20 15 10 5 0 .5 -10
Chemical shirt (6. ppm)

Figure 3. 'H MAS NMR spectra of wet silica (a) before the adsorption of quinoline and (b) after the
adsorptionof quinoline.

586
Quinoline was also adsorbed on the wet silica. In the 'H MAS NMR spectra of figure 3,
the originations of peaks are same to dry silica with quinoline in figure 2. The resonance
at k8.5.7.9 and 7.5 ppm could be due to the adsorbed quinoline. And, isolated silanol
group was disappeared. However, the signal at k5.0 ppm which means hydrogen bonded
hydroxyl proton was more increased than that of figure 2(b). The reason why hydrogen
bonding was increased by surface adsorbed water is not obvious. Water might effect on
the interaction between silica surface and quinoline. The detailed analysis is now in
progress.
In the case of carbazole(non-basic nitrogen compounds), there are no characteristic
signal of carbazole. However, the peak of isolated silanol was disappeared and that of
hydrogen bonded hydroxyl proton was highly increased. It could mean that small amount
of carbazole might be adsorbed on the silica. And, carbazole might also strongly interact
with silica surface.

4 Acknowledgements

This wore was supported by SK corporation and Korea Research Foundation.

References

1. Koltai T., Macaud M., Guevara A., Schulz E., Lemaire M., Bacaud R. and Vrinat M.,
Comparative inhibiting effect of polycondensed aromatics and nitrogen compounds
on the hydrodesulfurization of alkyldibenzothiophenes, Appl. Caral. A: Gen. 231
(2002) pp. 253-261.
2. Wang X.,Clark P. and Oyama S. T., Synthesis characterization and hydrotreating
activity of several iron group transition metal phosphides, J. Caral. 208 (2002) pp.
321-331.
3. Min W. S., A unique way to make ultra low sulfur diesel, KIChE annual meeting
(2002).
4. Kwak C., Lee J. J., Bae J. S. and Moon S. H., Poisoning effect of nitrogen
compounds on the performance of CoMoS/Al203 catalyst in the hydrodesulfurization
of dibenzothiophene 4-methyldibenzothio phene and 4-6-dimethyldibenzothiophene,
Appl. Catal. B: Environ. 35 (2001) pp. 59-68.
5. Laredo G. C., Reyes 3. A., Can0 J. L.and Castillo J. J., Inhibition effect of nitrogen
compounds on the hydrodesulfurization of dibenzothiophene, Appl. Catal. A: Gen.
207 (2001) pp. 103-112.
6. Nagai M.and Kabe T., Selectivity of molybdenum catalyst in hydrodesulfurization
hydrodenitrogenation and hydrodeoxygenation: effect of additives on
dibenzothiophenehydrodesulfurization,J. Catal. 81 (1983) pp. 4 4 0 4 9 .
7. LaVopa V. and Satterfield C. N., Poisoning of thiophene hydrodesulfurization by
nitrogen compounds, J. Catal. 110 (1988) pp. 375-387.
8. Nagai M., Sata T. and Aiba A., Poisoning effect of nitrogen compounds on
dibenzothiophene hydrodesulfurization on sulfide NiMo/Alz03catalysts and relation
to gasTphasebasicity, J. Caral. 87 (1986) pp. 52-58.
9. Liu C. C. and Maciel G. E., The fumed silica surface: A study by NMR,J. Am. Chem.
SOC.118 (1996) pp. 5103-5119.

587
10. Siahkali A. G., Philippou A, Dwyer J. and Anderson M. W., The acidity and catalytic
activity of MCM-41 investigted by MAS NMR FIX2 and catalytic tests, Appl. CuraL
A: Gen. 192 (2000) 57-69.
11. Mariscal R., Lopez-Granados M.,Fierro J. L. G., Sotelo J. L., Martos C. and Van
Grieken R., Morphology and surface properties of titania-silica hydrophobic xerogels,
Langmuir 16 (2000)9460.9467.

588
ADSORPTION CHARACTERISTICSOF VOCS ON MESOPOROUSSORBENTS

W.G. SHIM', M.S. YANG', J.W. LEE', S.H. SUH3AND H. MOON'

'Faculty of Applied Chemisq, Chonnam National University,Gwangiu 500- 757, Korea
E-mail: [email protected]

'Department of Chemical Engineering, Seonam University,Namwon 590-711, Korea

E-mail:[email protected]

'Department of Chemical Engineering, Keimyung University.Daegu 704-701,Korea

E-mail: [email protected]

Adsorption equilibria of VOCs on mesoporous sorbents (MCM-48) synthesized in our laboratory are
measured at several temperaturesusing a quartz spring balance equipped in a high vaccum system.
A new hybrid isotherm model for mesoporous sorbents is proposed by combining both the Langmuir
isotherm at low pressure and the Sips isotherm in multilayer region. Since mesoporous sorbents have
narrow pore size distribution between 20 to 46 A, capillary condensation of VOCs can be observed
at p / po4 . 2 4 3 . Also, the adsorbed amount decreased considerably with increasing pelletizing
pressure. The isosteric heat of adsorption was calculated using experimental data,

1 Introduction

Due to the significanteconomic and environmentalimplications of disposing VOCs, much

attention has been recently directed towards cost-effective pollution - prevention
techniques aimed at reducing VOCs emissions from industrial facilities [l]. There are
many techniques available to control VOCs emissions with different advantages and
limitations [2]. One of the most effective methods for controlling VOCs is an adsorption
process [3]. Among adsorptiodseparation technologies, the activated carbon has been
widely used in adsorption processes due to higher micropore volume, easy operation, low
operating cost and efficient recovery of most VOCs. However, it often encounters some
problems such as combustion, pore blocking, and hygroscopicity. As a result, alternative
adsorbents have been receiving much attention. For example, hydrophobic zeolites are
found to be one of advancement in controlling VOCs because of their advantage [4,5].
On the other hand, recent discovery of a new family of mesoporous molecular sieves
named M41S has been receiving a great deal of attention after introducing by Mobil
researchers [6]. The M4 1S family is classified into several members: MCM-4 1, MCM-48
and other species. Their synthesis and utilization have been investigated by many
researchers because of their peculiar characteristics such as large internal surface area,
uniformity of pore size, easily controlled size of pore, and high thermal stability. These
mesoporous materials may be useful as adsorbents, supports, and catalysts. Moreover,
interwoven and branched pore structures of MCM-48 provide more favorable mass
transfer kinetics than MCM-41. Therefore, MCM-48 seems to be a better candidate as an
adsorbent in separation techniques or as a catalyst support than MCM-41 [7,8].
In the past few years, some works have been done on the synthesis and
characterization, mechanical stability and adsorption characteristicsof MCM-48. Recently,
Hartman and Bischof reported the experimental adsorption isotherm and breakthrough

589
curves of VOCs on MCM-48 without quantitative prediction. A few isotherms have been
developed for the adsorption of condensable vapors and reviewed elsewhere. However,
none of the models yields a complete description of adsorption isotherms over a wide
pressure range with a uniform set of parameter [9,10].
Therefore, in this paper, we used a simple hybrid isotherm for interpreting the
adsorption isotherms of VOCs on MCM-48 (unpressedpressed) as a function of
temperature. The simple isotherm model is constructed by combining both Langmuir
equation at low pressure and Sips equation at regions of multilayer and capillary
condensation. The reason for the selection of Sips equation to explain capillary
condensation is based on an engineering point of view, namely, both flexible fit of
isotherm data and saving calculation time in dynamic simulation although its physical
meaning is somewhat lack. MCM-48 are prepared by conventional hydrothermal
syntheses and characterized by X-ray dihction (XRD) and nitrogen adsorption and
desorption. Also, the isosteric heat of adsorption is calculated using experimental data.

2 Experimental

2. I Synthesis of mesoporous materials

MCM-48 sample was synthesized as follows. 12.4 g of cethyltrimethylammonium bomide
(CTMABr, CI9Hd2BrN,Aldrich), 2.16 g LE-4 (polyoxyethylene lauryl ether, C12H25
(OCH2CH2)40H,Aldrich) were dissolved in Teflon bottle containing 130 g of deionized
water at 333 K. This aqueous solution was added dropwise to a another aqueous solution
in Teflon bottle containing 40 g of Ludox AS-40 (Du Pont, 40 wt% colloidal silica in
water), 5 g of NaOH, and 130 g of deionized water under vigorous stirring. The solution
mixture was preheated in a water bath kept at 3 I3 K and was stirred at 500 rpm for 20 min.
The resultant gel was loaded to autoclaves, and the mixture was hydrothermally treated at
373 K for 78 h. The mixture was then filtered and washed with 500 mL deionized water.
The washing procedure was repeated 4-5 times to assure the complete removal of the
bromide and other free ions. After drying at 333K for overnight, the dried solid was then
calcined in air at 873 K.

2.2 Measurement of mechanical stability

To test the adsorption property of MCM-48, the samples were compressed using a
hand-operated press. The pelletized MCM-48 diameter is 10 mm and the external pressure
applied is 0, 50, 100, 200, 300, 400 and 500 kg/cm2. Subsequently, the obtained pellet
was crushed and sieved to obtain pellets with a diameter of 0.1 to 0.2 mm that were used
for adsorption equilibrium and fixed bed studies.

2.3 Characterization
In designing an adsorption column, the characterization of adsorbents should be done
prior to experiments. In particular, one should know not only the specific area but also the
pore size distribution of the adsorbent in order to confirm that it would be proper for a
given purpose. Nitrogen adsorption and desorption isotherms, BET surface areas, and
BJH (Barren, Joyner and Halenda) pore size distributions of the synthesized sorbents

590
were measured at 77 K using a Micromeritics ASAP 2000 automatic analyzer. Prior to
measurment, the samples were outgassed at 623 K for 10 h. X-ray powder diffhction data
of mesoporous sorbents were collected on Phillips PW3 123 difictometer equipped with
a graphite monochromater and Cu K, radiation of wavelength 0.154 nm. XRD patterns
were obtained between 2 and 50 with a scan speed of l/min.

2.4 Aakorption stu*

The adsorption amounts of VOCs vapor were measured by a quartz spring balance, which
was placed in a closed glass system. A given amount of MCM-48 particles were placed on
the dish, which was attached to the end of quartz spring. This system was vacuumed for
15 hours at 10 Pa and 250C to remove volatile impurities from the mesopoprous
sorbents. The variation of weight was measured by a digital voltmeter that was connected
to the spring sensor. The adsorption equilibrium was usually attained within 30-60min.
Equilibrium experiments were carried out at different temperatures.

3 Results and Discussion

Figure 1 presents the nitrogen adsorption and desorption isotherms with BJH pore size
distribution curves for MCM-48. The isotherms are type IV according to the IUPAC

800
E
UJ
9
Is00

B
0400

-f
s
200

0.0 02 0.4 0.6 0.8 1.o

ppo

Figure 1. Nitrogen adsorption and desorption isotherms on MCM-48 at 77 K

classification. Also, the isotherms exhibit sharp steps in the relative pressure P/Po =
0.2-0.4, which are associated with capillary condensation in channels of MCM-48
structure. The sharpness of the capillary condensation steps indicates uniformity of pore
channels and their narrow size distribution. The isotherm is reversible and does not exhibit
hysteresis between adsorption and desorption. The surface area of MCM-48 was about
1100 m2/g. And the maxima in the pore size distribution curves indicate the uniform
mesopores of approximately 32 A. The BJH data showed approximately20 A distribution

591
from the mean pore diameter for MCM-48, and this matches well with other distribution
curves reported earlier.
Adsorption isotherms play a key role in either the design of the adsorption-based
process for the disposal of wastes containing VOCs or modeling the catalytic oxidation
process. The equilibrium data for mesoporous sorbents are fitted to combined model of
h g m u i r and Sips equations. This hybrid isotherm model with four isotherm parameters
(4,,b, ,b, ,n) is as below:

q=qm
[l+b,P+
blP
1+b2P"
b2P" 1
The isotherm parameters were determined using Ne1der:Mead simplex method by
minimuig the sum of residual, namely, the differences between experimental and
estimated adsorption amount. Figure 2 showed the adsorption isotherms of TCE on
MCM-48 at 303,308,3 13,323 K. As one can be expected, the adsorption capacity was
decreased with increasing temperature. The hybrid isotherm model for a pure adsorbate
was found to fit the individual isotherm data very well. The parameters of the hybrid
equations are listed in Table 1.

Table 1. Hybrid equation parameters for different temperatures

Temperature Temperature Temperature Temperature
303 K 308 K 311 K 323 K
4m 6.260EM0 5.968EMO 6.36084-00 5.978E4-00
bl 1.825E-01 1.764E-01 8.592E-02 7.522E-02
b2 9.303E-40 3.297E-25 8.441E-32 1.993E-49
n 9.848EM 1 4.695E4-01 4.448E4-01 5.544E4-01

15

I I

0 3 6 9 12 0 2 4 6 8 10
Pressure, kPa Amount adsorbed, mmoUg

Figure 2. Adsorption isotherms of Figure 3. lsosteric heat of adsorption with

TCE for different temperatures respect to amount adsorbed

592
 It has been known that condensation pressure depends on the adsorbate, temperature,
pore size, and geometry of sorbent. As increasing temperature, the capillary condensation
pressures is increased and adsorption capacity is decreased. This fact implies that
adsorption and desorption can be easily achieved by only little adjustment of pressure and
temperature. Therefore the effective removal of VOC can be done by pressure swing
adsorption (PSA) or thermal swing adsorption (TSA) processes.
As a useful thermodynamic property, the isosteric heat of adsorption has been
generally applied to characterize the adsorbent surface. The isosteric heat of adsorption is
evaluated simply by applying the Clausius-Clapeyron equation if one has a good set of
adsorption equilibrium data obtained at several temperatures.

where 4,, is the isosteric heat of adsorption, R is the gas constant, and N is the
amount adsorbed.
In Figure 3,, the isosteric heats of adsorption for vapors studied are plotted as a
function of the amount adsorbed. The isosteric heat is approximately 40 kJ/mol between
the adsorption amounts of 0.5 to 2.0 mmol/g, but that is 45 kJ/mol between the adsorption
amount of below 2.0 m o V g and above 8.0 mmoVg. This difference comes from the
capillary condensation. Because of the joint effects of the energetic non-uniformity of the
adsorbent surface and the interaction of adsorbate molecules in the adsorbed film itself,
the heat of adsorption in general varies significantly with the amount adsorbed. The
isosteric heat of adsorption can be divided into two sections, namely, low and capillary

12
0 Pmsun-100 A Pmssure-200
m Pmsun-300 0 Pmssunr-400
...-.... 10

8
2
g 6
P
U
4

0
0.0 1.5 3.0 4.6 6.0 7.5 0 I00 200 300 400 500
Pressure, kPa Pressure, kg/cm2

Figure 4. Adsorption isotherms of Figure 5. Comparison of the amount

benzene on MCM-48 at 300K adsorbed of MCM-48 with
pelletizing pressure

593
condensation regions. Especially, the isosteric heat of adsorption changes rapidly in
capillary condensation range.
To test the mechanical stability, MCM-48 samples were pressed into pellets using six
different pressures. Figure 4 showed the effect of pelletizing pressure on adsorption
capacity. The pelletizing pressure did not affect the capillary condensations although
benzene adsorbed amount changed dramatically. The adsorption amount of sample
pressed at 500 kg/cm*corresponded to 60 % of the origin sample as shown in Figure 5.

4 Acknowledgements

This work was supported by grant No. (R-01-2001-000-00414-0)from the Basic Research
Program of the Korea Science & Engineering Foundation.

References

1. Khan, F.I. and A.K. Ghoshal., Removal of Volatile Organic Compounds from
polluted air. J. Loss. Prevent. Proc. 13 (2000) pp.527-545.
2. Clausse, B.; Garrot, B.; Cornier, C.; Paulin, C.; Simonot-Grange M.-H.; Boutros, F.
Adsorption of Chlorinated Volatile Organic Compounds on Hydrophobic Faujasite:
Correlation between the Thermodynamicand Kinetic Properties and the Prediction of
Air Cleaning. Micro. and Meso Mat 25 (1998) pp.169-177.
3. Kim, D. J., Shim, W.G. and Moon, H., Adsorption Equilibrium of Solvent Vapors on
Activated Carbon. KJChE 18 (2001) pp.518-524.
4. Takeuchi, Y.; Hino M.; Yoshimura, Y.; Otowa, T.; Izuhara H.;Nojima T. Removal of
Single Component Chlorinated Hydrocarbon Vapor by Activated Carbon of Very
High Surface Area. Sep. and Puri! Tech. 15 (1999) pp. 79-90.
5. Chintawar, P. S.; Greene, H. L. Adsorption and Catalytic Destruction of
Trichloroethylene in Hydrophobic Zeolites. AppZ Cat B. 14 (1997) pp. 37-47.
6. Beck, J. S.; Vartuli, J. C.; Roth, W.J.; Leonowicz, M. E.; Kresge, C. T.; Schmitt, K.
D.; Chu, C. T-W.; Olson, D. H.; Sheppard, E. W.; McCullen, S. B.; Higgins, J. B.;
Schlenker, J. L. A New Family of Mesoporous Molecular Sieves Prepared with
Liquid Crystal Templates. J. A n Chem SOC.114 (1992) pp 10834-10843.
7. Zhao, X. S.; Lu, G. Q. Organophilicity of MCM-41 Adsorbents Studied by
Adsorption and Temperature-Programmed Desorption. Colls and SurfA 179 (2001)
pp 26 1-269.
8. Ryoo, R.; Joo, S.H.; Kim, J.M. Energetically Favored Formation of MCM-48 from
Cationic-Neutral SurfactantMixtures. J. Phys. Chem. B 103 (1999) pp 7435-7440.
9. Hartmann, M.and C. Bishof, Mechanical Stability of Mesoporous Molecular Sieve
MCM-48 Studied by Adsorption of Benzene, n-Heptane, and Cyclohexane. J. Phys.
Chem. B. 103 (1999) pp. 6230-6235.
10. Sonwane, C.G. and Bhatia, S.K.Adsorption in mesopores: A molecular-continuum
model with application to MCM-41 Chem Eng Sci 53 (1998) pp. 3143-3156

594
 MOLECULAR SIMULATION FOR ADSORPTION OF HALOCARBONS IN
ZEOLITES

Kazuyuki Chihara*, Tsuyosbi Sasaki, Shingo Miyamoto. Michi Watanabe,

Caroline F.Mellot-Draznieks2,Anthony K.Cheetham3
1 ) Department of Industrid Chemistry,Meiji Universio
1-1-1 Higashi-mita,Tama-ku, Kawasaki, Kanagawa 214-8571 JAPAN
*Correspondingauthor: Far: 044-934-7197. E-mail:chihara0isc.meijiac.jp
2) Institut Lavoisier, Universite de VersailIes,Saint-Quentin-en-Yvelines45,
Avenue &s Etats-Unis, 78035, Versailles Cedex, France
3 ) Material Research Laboratory, University of California, Santa Babara,
CA93106-5130, USA

The Grand Canomcal Monte Carlo (GCMC) method is simulation method for solving a phenomenon from a
microscopic level, and it is turning into the powerful analysis technique in the field of the enginering. However,
information on forcefield parameters and charges are often inadequate, even in systems where the structure is
well known. From the environmental point of view, the adsorption of chlorinated hydrocarbons by the use of
zeolites may have some potential utility in ground water or soil remediation and other areas. It is becoming
possible to interpret the adsorption characteristic in a molecule level rationally, and to predict the macroscopic
characteristicsuch as adsorption isotherms in recent years using Computer modeling
In this study, equilibria and isosteric heat of adsorption for the system of chlorinated hydrocarbons and Y-
type zeolite were obtained with gravimetric method and chromatographic method. By comparing an experiment
result with a molecular simulation result, the validity of forcefield parameters and zeolite model was examined

1. Introduction

Molecular simulation has now become powerful tool for the study of adsorbed
molecules in zeolites, and the Grand Canonical Monte Carlo (GCMC) method is
especially useful for predicting adsorption equilibria. However, information on forcefield
parameters and charges are often inadequate, even in systems where the structure is well
known. From the environmental point of view, the adsorption of chlorinated
hydrocarbons by the use of zeolites may have some potential utility in ground water or
soil remediation and other areas. Mellot et al. [3] recently reported new forcefield
parameters and charges for chlorinated hydrocarbons in the faujasite zeolite: NaX, NaY
and siliceous Y. These yield heats of adsorption that are in good agreement with
calorimetric data [2]. In this study, their forcefield was used to simulate adsorption
isotherms and isosteric heats of adsorption for chloroform, trichloroethylene and
tetrachloroethylene in USY-type zeolite, separately. The results were compared with
gravimetric and chromatographic experiments.

2. Experimental

2.1. Gravirnetric Method

Fig.1 shows experimental apparatus for gravimetric analysis. The zeolite sample
(about Ig) was placed in a quartz basket (G). Then the adsorbate in flask (B) was fed to

595
adsorption tube (N). The whole apparatus was in a constant temperature air bath. The
temperature range was 303-323 K. The amount adsorbed was measured correspondingto
the pressure of the vapor in the tube. The pressure was measured by pressure sensor (P) at
higher pressure range (2 0.013 atm) and baratron (0)at lower pressure range (<0.013
atm). In this way, adsorption isotherms were obtained (Fig. 5).

2.2 Chromatographic Method [ I ]

h

Al-A4 : needle valve

B : adsorbate-flask
C : desplacement meter
D : heater
E : thermo-controller
F : thermo-detector
G :quartzbasket
H : stertch-detector
I : chart recorder
J : mass spectrum meter
K :quartz basket
L : mantle heater
M : vacuum pump
N : adsorbent tube
0 : baratron
P : pressure sensor

Fig.1 Experimental apparatus to measure adsorption equilibrium

In order to study the isotherm at lower pressure range, we used a pulse response
method with a gas chromatograph.
The gas chromatograph (GC 9-A; Shimadzu Co., Ltd.) was used with helium carrier.
Siliceous zeolite pellets (with binder) were crashed and screened to obtain particle size
between 4 . 9 5 ~ 1 0to~8 . 3 3 ~ 1 0m~(an average particle radius of 6 . 6 4 ~ 1 m)
0 ~ and packed
to the column (length, 30cm, diameter, 3mm). The experimental column pressure was
kept 200 kPa, the experimental column temperature was kept at 393, 423 and 453 K,
respectively. Pulse responses of vaporized chloroform with helium were detected by
TCD. Response data were stored and processed by a personal computer.
From retention time of pulse response, adsorption equilibrium constant at zero
coverage was obtained and plotted as vant Hoff plot to get heat of adsorption and
adsorption equilibrium constant extrapolated to 303 K on all the adsorbates. This
extrapolated value was plotted in Fig. 5 as initial slope.

2.3. Simulation
Cerius2 (MSI Inc.) was used
throughout the simulations. Forcefield 0-H 90.535 2.698
parameters obtained by Mellot et all. [3] c-c 25.860 3.753
C-CI 55.650 3.707
are listed in Table 1. The Grand Canonical C-H 26.730 3.358
CIcl 119.80 3.822
Monte Carlo method (under constant CtH 57.530 3.392
chemical potential (p), volume (V), H-H 27.630 2.%3
temperature (T)) was used to get the x-x 0.000 0.000
X-H O.Ol0 6.000
equilibrium amount adsorbed. X-CI 0.010 6.000
x-c 0.010 6.000

596
2.3.1 Models
Three models were considered
here. Pure siliceous faujasite (Y-
type) (Fig. 2) was in the database of S i +2.4
Cenus2. Dummy atoms were put in
sodalite cages to avoid impossible 0 -1.2
occupation of adsorbates. PQ-USY
was dealuminated in preparation. So
it definitely contains silanol nest.
PQ-USY has the silica-alumina ratio
of SiO2/Al2O3=70.So there remains Fig3 Pure SiliceousY Type model
aluminum. Then we put 7 silanol
nest as in Fig. 3 and 4 acid site as in Fig. 4.
Charges for Si and 0 were set to be +2.4
e.u and -1.2 e.u, respectively, as usual. H
and 0 in silanol nest were assumed to have
the charges as M.1 e.u and 4.7 e.u,
respectively. Al, 0 and H in replaced A1
were assumed to have the charges as +1.4
e.u, -0.59 e.u and 0.39 e.u. Si charges were
adjusted after these setting to neutral by Fig3 Silanol nest model
averaging.

3. Results and Discussion

In Fig.5, experimental adsorption

isotherms for chloroform in PQ-USY at 303
K are shown. Initial slopes for 303 K are
also shown. This whole range of adsorption
isotherm at 303 K could be compared with
simulation.
At higher pressure (>O.l am), all the Fig.4 Acid site model
simulations were coincident and almost
correspond to gravimetric data. At lower Table2 Meat ot Adsorption [kJ/moll
pressure, simulation for the acid site model (PQ-USY-chloroform)
~~ ~ ~~~ ~~

was good agreement with chromatographic Exp. Chromato. 44.25

data and baratron data.
Experimental heat of adsorption obtained Pure Siliceous Y 34.05
by chromatography at zero coverage is
compared with simulated heat of adsorption Silanol nest 33.66
for 3 models in Table 2. Here simulation
for the acid site model was closer to the Acid site 39.98
experimental value than the pure siliceous
model and the silanol nest model.
In Fig.6, experimental adsorption isotherms for tetrachloroethylene in PQ-USY at 303
K are shown.

597
 In this system, all simulation results Table3 Heat of Adsorption [kJ/moll
with three models were coincident, (Pa-USY-t etrachloroethylene)
and almost correspond to
chromatographic data and gravimetric Exp. Chromato. 43.64
data. This is thought to be because
tetrachloroethylene is different from Pure Siliceous Y 41.40
chloroform and it i s a non-polar
molecule. Silanol nest 41 30
In Table 3, all simulations show the
heat of adsorption corresponding to Acid site 43.68
chromatographic data for the same
reason, and especially acid site model
show the best agreement in the three models.

1
exp. ad. 303K

0.1
8 exp. ad. 303K
Baramon
0.01
- exp. 303K
chromato.
3
= 0.001
0 PureSiliceousY
model
0.m1
A Silanol nest model

o.woo1 0 Acid site model

0.m1 0.m1 0.001 0.01 0.1 1
Pressure [am]
Fig3 Comparison between exp. and
simulations for adsorption isotherms of PQ-
USY-Chloroform system

4 exp. ad. 303K

exp. ad. 303K

Baratoron
-exp. 303K
chromto.
0 Pure Siliceous Y
&I
A S i b 1 nest model

0 Acid site model

0 . m 1 0.m1 0.001 0.01 0.1 1

pressure [atm]
Fig.6 Comparison between exp. and
simulations for adsorption isotherms of PQ-
USY- Tetrachloroethylene system

598
4. Conclusions

Equilibrium and isosteric heat of adsorption for the system of chloroform-Y-type zeolite
were studied.
The adsorption equilibria were measured using a gravimetric method and were
expressed as isotherms. A chromatographic method was used to get the initial slope of the
isotherms. In the simulation, GCMC method was used to calculate amounts adsorbed for
various conditions. When the experiment result and simulation result of chloroform are
compared, the simulation for the acid site model was most agreement with
chromatographic data and baratron data. The simulation result of tetrachloroethylene with
three models corresponded mostly for the non-polar molecule, and above all the acid site
model was the closest to the experiment result. Therefore, to get better coincidence
between experimental data and simulation, it was found to be necessary to account for
aluminum rather than silanol nest.
As a conclusion, FF parameters were confidently applied. And modified structure model
are effective for simulation.

Nomenclature
K* :adsorption equilibrium constant [cc/g]
M : molecular weight [glmol]
q : amount adsorbed [g/g]
R :gas constant [JK moll
R-Min : Bond Length Equilibrium [A]
T :temperature [K]
Epsilon : potential energy minimum [eV]
K :Boltzmann constant [JK]

References
1. K. Chihara, M. Suzuki, K. Kawazoe, Adsorption rate on Molecular Sieving Carbon
by Chromatography, AIChE J., 24,237 (1978)
2. C. F. Mellot, A. K. Cheetham, S. Harms, S. Savitz, R. J. Gorte, A. L. Myers,
Calorimetric and Computational Studies of Chlorocarbon Adsorption in zeolites, J.
Am. Chem. Soc., 120,5788. (1998a)
3. C. F. Mellot, A. M. Davidson, J. Eckert, Adsorption of Chloroform in NaY Zeolite:
A Computational and Vibrational Spectroscopy Study, J.Phys. Chem. B, 102,2530
(1998b)

599
 ADSORPTION OF BTX ON MSC IN SUPERCRITICAL COz,
A CHROMATOGRAPHIC STUDY

Kazuyuki Chihara, Naoki Omi, Yusuke Inoue,

Takuji Yoshida, Takashi Kaneko

Department of Industrial Chemistry,Me& University

1-1-1. Higashi-Mita, Tama-Ku, Kawasaki, 214-8571, Japan
tel&fa: 044-934-7197,e-mail:[email protected]

Chromatographic measurements were made for the adsorption of benzene, toluene and m-xylene on
molecular sieving carbon (MSC) in supercritical fluid C02 mixed with organics. Supercritical
chromatographpacked with MSC was used to detect pulse responses of organics. Adsorption equilibria and
adsorption dynamics for organics were obtained by moment analysis of the response peaks. Dependences
of adsorption equilibrium constants, K*,and micropore diffusivity,D, on amount adsorbed were examined.

Introduction
Adsorption equilibria and adsorption dynamics in supercritical fluids have been
reported recently and it will be possible to apply the supercritical fluid to some new
adsorptive separation processes. Fundamental informations on adsorption under
supercritical conditions are necessary to design such processes. Supercritical
chromatography has been used for study on the adsorption equilibria and adsorption
dynamics.Adsorption of organics, i.e., benzene, toluene and m-xylene, respectively, on
MSC under supercritical conditions has already been reported in reference (Chihara,
1995). In the previous study, chromatographic measurements were made for the
adsorption of benzene, toluene and m-xylene on MSC in supercritical COz mixed with
benzene, toluene and m-xylene respectively. Moment analysis of the chromatogram was
carried out. In the study, the organics used in the form of pulse were the same as organics
mixed with supercritical COz. The dependencies of adsorption equilibrium and micropore
diffisivities on the amount adsorbed were obtained.
In the present study, supercritical COz chromatograph packed with MSC was again
used to detect the pulse responses of organics, and the moment analysis technique was
used to analyze. Equilibrium and dynamics were studied for benzene, toluene and m-
xylene, respectively, -MSC systems in the supercritical COz mixed with organics which
were different fiom that used in the form of pulse. Furthermore, the dependence of
adsorption of the organics on the amount adsorbed of other organics was discussed.

Experimental procedure and conditions

The experimental apparatus (Super 200-type 3; Japan Spectroscopic Co., LTD) was
shown in Fig.1. The carrier fluid of the chromatograph was supercritical COz (critical
temperature 304K, critical pressure 7.3 MPa) and its mixture with the above-stated

600
organics (benzene, toluene or m-xylene) respectively. The adsorbates used in the form of
pulse were different fiom organics mixed with supercriticalC02. For example, in the case
of C02 mixed with benzene, the organic used in the form of pulse was toluene or m-
xylene. In the previous study, pulse organics were the same as organics in the carrier. The
volumes of the pulse were fixed to be 8 x 10m9m3 as liquid. MSC 5A (Takeda chemicals
Co., HGK882.)was crushed and screened to obtain particle size between 1.49 x lo4 -
1.77 x. 104m (an average particle radius of 8.12 x. lo- m). 4.82 x 104kg of these
particles were packed into the chromatographic column of 5 x 10-*mlong and 4.6 x lOm
in diameter. The void hction, E, of the bed was determined to be 0.355. The properties of
MSC5A are shown in Table 1 in reference (Chihara, 1978). Flow rate of supercritical C 0 2
was 1.33 x 10-m3/s at 268K and at 15.0,20.0 and 25.OMPa respectively and flow rate of
adsorbate (benzene, toluene or m-xylene) was 1.67 x 10-0m3/s,5.00 x lo- m3/s and 1.00
x m3/s as liquid at room temperature (298K). The column pressure was kept at 15.0,
20.0 and 25.0 MPa respectively. The pressure drop across the adsorbent bed was
estimated to be about 0.1Mpa and was assumed to be negligible. The experimental
column temperature was kept at 313, 333 and 353 K respectively. Before experimental
runs started, the adsorbent particles were regenerated and stabilized by feeding pure C02
for 2 hours at the experimental pressure and temperature. Pulse responses were detected
using a multi-wave length UV detector (Multi-340; Japan Spectroscopic Co., LTD.) (195-
350 nm). Response data were processed by a personal computer.
A : Liquid C02 Cylinder
B : Valve
C : Cooler
D1.2: Pump
E : MixingColumn
F : Adsorbate (Liquid)
G : InjectionColumn
H : Six-way Valve
1 : Valve
J : PackedColumn
K : Valve
L : UV Detecter (MULTI-340)
M : Back Pressure Regulator
N : PersonalComputer
0 : Fraction Collector
P : Vent

Fig. 1 Experimentalapparatus

Moment analysis of supercritical fluid chromatogram was tried, and the apparent
adsorption equilibrium constant, K* and time constant of micropore diffusivity, D/a2
obtained from first and second moment of response peak, as in references (Chihara, 1993;
Chihara, 1995).

601
Result and discussion
Figure.2 shows adsorption isotherm of toluene at 313K. According to Fig.2, The
amount of adsorption increased with increases of molarity of toluene, and reached to
saturation. The amounts adsorbed became larger with decreases of column pressure. It
was considered that the situation is competitive adsorption and amount adsorbed of
toluene decreases as COz adsorption increase with increases of column pressure.

+ 15MPa
H 20MPa
p 1.5 A 25MPa

4 1.0 adsorbate : toluene

5 a5
solute : toluene

E
~~

0
Concentration [ml/ma]
20 40 80

Fig. 2 Adsorption isotherm : toluene at 3 13K

toluene
y 0.1 benzene
A m-xylene
1
E
Y
temperature :3 13[K]
2 0.01 1 pressure :2O[MPa]

awt
0 a5 1.0 1.5 20
Amount adsorbated of toluene [ml/kgl
Fig. 3 Dependencies of K* on the
amount adsorbed of toluene

Figure.3 shows dependency of adsorption equilibrium c nstar s, K*, for benzene,

toluene, and m-xylene on amount adsorbed of toluene at 15Mpa. This is reasonably
decreasing, which corresponds to Fig.2.

0 toluene
H benzene
A in-xylene

temperature : 353[K1
pressure : 20CMPal

0 0.5 1.a 1.5 2.0

. Amount adsorbated of toluene [mol/kd
Fig. 4 Dependencies of D&exp(d) on the amount adsorbed of toluene

602
 Figure.4 shows dependency of micropore diffusivity, D/a2exp(o2), for benzene, toluene,
and m-xylene on amount adsorbed of toluene at ISMPa. The increase of D/a^expCo2) for
toluene could be reasonably explained by chemical potential driving force. However, as
for dependency of D/a^expCo2) of benzene and m-xylene on amount adsorbed of toluene,
further discussion would be necessary.

Molecular simulation
A simulation is assuming a molecule on a computer and performing various kinds of
physical chemistry calculation. It was with the molecular design support system Cerius2
(Version4.2) made from MSI. The purpose of performing simulation is to elucidate an
adsorption mechanism on the molecule level. The simulation was performed on the same
conditions as an experiment in order to compare with an experiment. MSC68-RC1 and
MSC84-RC1 model were used as adsorbent. There is 6.8A and 8.4A of distance between
adsorption spaces, respectively.

Adsorption state *"

In the beginning, we will
examine how molecules of
adsorbate is benzene was
used for the adsorbate here. MSC68RC1 model
The results are shown in
Figure 5. We see from Figure
5 that benzene is adsorbing to
the adsorption space
reproduced micro pore. Here,
benzene is adsorbed to layer
in parallel in MSC68 model, MSC84RC1 model
on the other hand, Fig.5 Adsorption state of benzene
It is adsorbed aslant in
MSC84 model 0.0025
p
KS

Adsorption isotherm
- i

The simulation was carried out

01

the single component Benzene is CDO

used for adsorbate. Conditions are
S

313K and ISOatm. The results are OMSO68-RC1

1

shown in Figure 6. DMSC84-RC1

a

As for Fig.6, adsorbed amount

for MSC84 model is larger than 06
MSC68 model. The cause of this 0 0.00002 0.00004 0.00006 0.00008
result is that the amount adsorbed is molar concentration [mol/ni]
dependent on the size of the
adsorption space. According to Fig.6, Fig.6 Adsorption isotherm of benzene
The amount of adsorption increased with increases of molarity of benzene, and reached to
saturation.
Figure 7 shows comparison of adsorption isotherm for a molecular simulation different
force field and an experiment. Conditions are 313.15K. and ISOatm. The simulation was
carried out the two components system. UNIVERSAL 1.02, London, and DREIDING2.11
were used for the force field, respectively. As for the amount adsorbed, in every figure,

603
the simulation was small rather
than the experiment. The difference 0.0014
of an experiment and a simulation - 0.0012
RC
l-8- I
0 MSC84-RC1
is large, so that molarity becomes 2
e, 0.001 Y
A
A Exp:
large. 3 0.0008
In the simulation using the force
field parameter of London,
4
c
0.0006

although morlarity was increasing 5 0.0004

by MSC84 model, the result that P 0.0002

the amount adsorbed decreases

0 0.00002 0.00004 0.00006 0.00008
sharply was shown. In the two
molar concentration [molhl]
components system, not only the UNIVERSALI.
interaction of an adsorbent model 0.0014
and adsorbate model but the
interaction of adsorbate model is -
,M
0.0012
0.001
0 MSC84-RC1
added, calculation becomes more ij

complicated, it is thought that such 0.0008

a result arose. 3I 0.0006
Conclusion
1 0.00041 ~ , j
Adsorption equilibrium and B 0.0002
adsorption dynamics on MSC were 0'
evaluated for each organics in
supercritical C02 fluid mixed with
adsorbate by chromatographic
measurement. The dependencies of 0.0014
adsorption equilibrium constants,
K*, and micropore diffiivity, D,
-
-"
0.0012

of toluene, benzene and m-xylene, i: 0.001 A

Y
0 MSC84-RC1
I
A Exp.
on molarity of toluene with each 4
-0
0.0008
A
parameters of temperature or 4- 0.0006 . ~ _ _ -
I-I
pressure were obtained. It was 3 0.0004 Q
0
found that the values of K* and D 6 0.0002 0 0
for an organic substance depended G A
v

on the amount adsorbed of other - 0

0 0.00002 0.00004 0.00006 0.00008
organics strongly. The state of the molar concentration ImoVml]
molecule could be observed by the
molecular simulation. As for the Londo
amount adsorbed, the simulation is Fig.7Comparison with experiment and
small, in comparison with the molecular simu1ation:Adsorption isotherm
experiment.

Reference
Chihara, K., Kawazoe, K.,Sumki, M., Seisan Kenkyu, AIChE J, 24,237-246. (1978)
Chihara, K., Aoki, K., AIChE, Annual Meeting, (1993)
Chihara, K., Oomori, K., Kaneko, R., Takeuchi, Y., AIChE, Annual Meeting, (1995)

604
 POROUS ALUMINA WITH BIMODAL PORE SIZE DISTRIBUTION As AN
ORGANIC ADSORBENT

YOUNGHUN KIM, CHANGMOOK KIM, PIL KIM,JONG CHUL PARK

AND JONGHEOP YI'
School of Chemical Engineering, Seoul National University,Seoul 151-742, Korea
E-mail: [email protected]

Hierarchical channel or wellconnected smaller and larger pore networks show multiple advantages
for application in catalysis or adsorbent in aqueous condition. Micro- and mesopores may provide
size or shape selectivity for guest molecule, while additional macropores can reduce transport
limitations and enhance the accessibility to the active sites. In this study, we proposed a novel
method to prepare bimodal p o r n aluminas with meso- and macropores with narrow pore size
distribution and well defmed pore channels. The framework of the porous alumina is prepared via a
chemical templating method using alkyl carboxylates. Polystyrene beads are employed as physical
templates for macropore. We examined PDDA treated aluminas as organic adsorbent in aqueous
solution. Most anionic dye is removed within 10 min, and the adsorption rate of PDDNP4 is faster
than that of PDDA/P2 because macropore of P4 may have reduced transport limitation and
enhanced the accessibility to the active site, cationic charge.

1 Introduction

Porosity is one of the important factors that influence the chemical reactivity and the
physical interactions of solids with gases and liquids [l]. Physical properties such as
density, surface area and strength are dependent upon the porosity and the pore structure
of a solid. Especially for industrial applications, the control of porosity with well-ordered
structural pore networks is of great importance, for example, in the design of catalysts,
adsorbents, membranes, structural materials and ceramics. Researches on uniform pore
structure of nanometer to micrometer dimensions have been progressed with great interest
due to a variety of possible application in catalysis, molecular separation, membranes,
structural materials and adsorbent [2]. Recently, ordered inorganic structures with
macropore have been prepared using physical templating method [3], latex sphere and
emulsion droplets, or chemical templating method [4] by post-hydrothermal treatment and
primary particle seeding. Micro- and mesopores provide size or shape selectivity for guest
molecule, while additional macropores can reduce transport limitations and enhance the
accessibility to the active sites [5]. As theoretically proven by Levenspiel, the bimodal
catalyst can guarantee high diffusion efficiency [6]. For example, cobalt catalysts
supported on bimodal silicas show remarkably high activity in liquid-phase
Fischer-Tropsch synthesis [7].
Relatively few studies on the synthesis of mesoporous alumina have been reported to
date [8]. One of the limitations of the reported synthetic strategies is that the rate of
hydrolysis (and condensation) reaction of aluminum alkoxide are much faster than that of
silicon alkoxide. In this study, we proposed a novel method to prepare bimodal porous
aluminas with meso- and macropores with narrow pore size distribution and well-defined
pore channels. The b e w o r k of the porous alumina is prepared via a chemical
ternplating method using a w l carboxylates. Here, self-assemblied micelles of wboxylic
acid were used as a chemical template. Mesoporous aluminas were prepared through
careful control of the reactants pH, while the procedures are reported elsewhere [9].

605
2 Experimental

Polystyrene beads (PS) are employed as physical templates for macropore. The
emulsifier-fiee emulsion polymerization method used here allows for the synthesise of
nearly monodisperse latex beads of PS in the size of ca. 100 nm [101. PS beads were
prepared using 700 ml degassed water, 54 ml styrene monomer, 0.65 g potassium
persulfate as initiator, and 20 ml divinylbenzene as cross-linking agent. PS beads were
obtained at 7OoC and 350 rpm, and dried under ambient condition.Aluminum
sec-butoxide and stearic acid were separately dissolved in parent alcohol, sec-butyl
alcohol, at room tempature, and then the two solutions were mixed. Appropriate amount
of HN03 solution was dropped into the mixture at a rate of I mVmin to acidify and
hydrolyze the aluminum precursor. PS beads were added into aluminum hydroxide
solution after stirring for 10 h. The fmal pH of the reactant was approximately 7. Organic
templates, both stearic acid and PS bead, were easily removed fiom dried aluminum
hydroxide by calcination. The overall synthetic procedure is as shown in Fig. 1.
An adsorption test for dye was performed using model solution, which was prepared
with acid red 44 (crystal scarlet, Aldrich). Prepared aluminas (P2and P4)were added into
2%polydiallyldimethylammonium chloride (PDDA, Aldrich) solution, and stirred for 3 hr.
0.1 g of PDDA treated aluminas were stirred with 10 ml of dye solution (50 ppm).
Depletion of dye was determined by UV (510 nm) spectrometry (HP8453, Hewlett
Packard).

*!
Stearic acid
-b ___,
calcination

macropore
Figure 1. Schematics for the synthesis of porous alumina with bimodal pore size distribution. Templates
removal steps are followed by dotted-arrow for polystyrene beads and solid-arrow for silica gels as physical
templates.

3 Results and discussion

3. I Synthesis of bimodal alumina with meso- and macropore

Mesoporous aluminas prepared using only chemical template (P2)show a narrow
pore size distribution, adjustable pore diameter (2-7 nm) and 300-500 m2g-' of surface
area [9].Based on the XRD, DSC, and TGA analysis, the phase of P2 calcined at 420C
and for 3 h is y-fom, and thermally stable up to %Or. The mesopore size of P2 is
dependent upon the carbon tail length of alkyl carboxylate, the ratio of water to aluminum
precursor, calcination condition, and the pH of reactants. Oxolation reaction of aluminum
alkoxide was minimized at IEP (ca. 8-9), and P2 prepared at pHIEPshowed poorly
organized mesostructure [11,12], while P2 prepared (Table 1 and Fig. 2) at pH 7 had a
relatively well organized mesostructure. Therefore, we adjusted the final pH of reactants
using HN03. The XRD patterns of P2 show only one peak at low angle, which indicates
that the pore system is lacking in a long-range order. It is in good aggreement with the

606
TEM result that less ordered worm-like pore distribution was formed in mesoscale (Fig.
3b).
Table 1. Characteristics of bimodal porous aluminas
Sample Pore size [nm] V [cm3g-'] SA[m2g-l]
Damla1 DBni M DWB-FHH Icl
PI 4.2 0.15 149
P2 6.8 3.4 0.53 485
P3 6.4 51 0.24 216
P4 6.2 3.5 50 0.35 239
[a] Gurvitsch (4VIA) model. [b]BJH model on the desorption. [c] BdB-FHH spherical model on the desorption.

Important trends in N2isotherm when the PS beads are used as a physical template
are shown in Table 1 and Fig. 2. In Table 1, PI is the alumina prepared without any
templates, P2 is prepared without physical template (PS bead), P3 is prepared without
chemical template (stearic acid), and P4 is prepared with all templates. For above 10 nm
of pore size and spherical pore system, the Barrett-Joyner-Halenda (BJH) method
underestimates the characteristics for spherical pores, while the Broekhoff-de
Boer-Fre~el-Halsey-Hill(Bdl3-FHH) model is more accurate than the BJH model at the
range 10-100 nm [13]. Therefore, the pore size distribution between 1 and 10 nm and
between 10 and 100 nm obtained fiom the BJH model and BdB-FHH model on the
desorption branch of nitrogen isotherm, respectively. N2isotherm of P2 has typical type
IV and hysteresis loop, while that of P3 shows reduced hysteresis loop at P/Po cu. 0.5 and
sharp lifting-up hysteresis loop at P/Po> 0.8.This sharp inflection implies a change in the
texture, namely, textural macro-porosity [4,14]. It should be noted that P3 shows only
macropore due to the PS bead-fiee fiom alumina h e w o r k .

--
-Pz
P3
P4

0.0 0.2 0.4 0.6 0.8 1.0

PPO
Figure 2. Nitrogen adsorption-desorption isotherms of the porous alumina prepared using stearic acid and/or
PS bead as a template (curve are shifted for clarity). The inset shows the corresponding pore size distribution
from the desorption branch.

When neither chemical nor physical template is used in preparation of alumina, PI,
the resulting material shows less ordered pore size distribution, while P4 shows bimodal
pore systems with both mesopore (3.5 MI)and macropore (50 nm) after thermal treatment.

607
As shown in Fig. 3% the size of PS bead is approximately 100 nm, and dried PS beads
show an aggregate form. However, the pore sizes of calcined P4 and P3 are
approximately 50 nm. Calcination resulted in polymer-he metal oxide strucures, with
spherical voids smaller than the PS bead diameter, by 50%. This contraction phenomenon
is obviously related to the reorganizaiton of the surrounding walls during the calcination
process as evidenced in Fig. 3c and Fig. 3d. The part of the wall retracts by the burning
PS bead [I I]. This thermal treatment causes the link of metal oxide in the structure (Fig.
3d), and it is also affected by the calcination temperature.

Figure 3. E M micrographs of (a) colloidal PS bead and (b) calcined Pt,and SEM images of (c) as-made P4
and (d) calcined P4.

In SEM micrographs, unimodal mesoporous alumina (P2) shows smooth surface,

while the surfaces of calcined P3 and P4 were coarse. As-made P4 materials show
aggregated form of PS beads surrounded by aluminum hydroxide, and this is similar to
that of pure PS beads aggregates. After calcination, the morphology of the calcined P4
changes into inverse shape of as-made P4. As shown in Fig. 3d, several dark spots of
about 40-70 nm diameter can be observed inside the alumina structure. These dark spots,
macropre, surrounded by mesopore framework and connected with neighboring
macropore. Therefore, multi-pore structure shows interlinked cylindrical mesopore with
spherical macropore, like a spongy shape.

3.2 Organic adsorption test

As a model case, we examined the adsorption capability of organics in aqueous solutions
using the alumina developed here. Wang et al. [151 showed that siliceous materials, which
were treated with cationic polymer and oppositely charged mixed surfactant micelles,
adsorbed organophilic compound. Such organic removal reaction is known to be very fast
and reversible ion-exchange reaction. Cationic charge site (N3 in PDDA reacts with
anionic charge site (SO? of dye through charge matching. The alumina sample of P4 was
prepared from attaching PDDA onto the surface and contacted with the Acid Red 44
dissolved in aqueous solution. Most anionic dye is removed within 10 min, and the
adsorption rate of PDDA/P4 is faster than that of PDDNP2 because macropores of P4
reduced the transport limitation and enhanced the accessibility to the active site, as shown
in Fig. 4. The maximum uptake of dye was dependent upon the amount of attached PDDA
on the surface, while the elusion of adsorbed dye was easy and fast when 0.5 M HN03
solution was applied.

Financial support by the National Research Laboratov WRL) of the Korean

Science and Engineering Foundation (KOSEF) is gratefully acknowledged

608
 1.2,

1.0 -

0.8 .

3 0.6 -

0.4 ~

0.0 J I

0 20 40 BO 80 i m m
time (nin)
Figure 4. The concentrationchange of acid red 44 as a function of time.

References

1. U. Schubert, N. Husing, Synthesis of Inorganic Materials, Wiley-VCH, 2000.

2. A. Stein, B. J. Melde ,R. C . Schroden, A h . Muter. 2000, 12, 1403; B. Lee, Y. Kim,
H. Lee, J. Yi, Micropor. Mesopor. Mat. 2001,50,77.
3. B. Lebeau, C. E. Fowler, S. Mann, C. Farcet, B. Charleux, C. Sanchez, J. Mater.
Chem. 2000, 10, 2105; C. F. Blanford, H.Yan, R. C. Schroden, M. Al-Daous, A.
Stein, Adv. Muter. 2001,13,401.
4. X. Wang, T. Dou, Y. Xiao, Chem. Commun. 1998, 1035; J. Sun, Z. Shan, T.
Maschmeyer, J. A. Moulijn, M.-0. Coppens, Chem. Commun.2001,2670.
5 . - M.-0. Coppens, J. Sun,T. Maschmeyer, Catal. Today, 2001,69,331; Y. S. Cho, J. C.
Park,W. Y. Lee, J. Yi, Catai. Lett. In press, 2002.
6. 0. Levenspiel, Chemical Reaction Engineering, 3rd Ed. John Wiley & Sons, 1999.
7. N. Ysubaki, Y. Zhang, S. Sun, H. Mori, Y.Yoneyama, X. Li, K. Fujimoto, Catal.
Commun.2001,2,3 1 1.
8. F. Vaudry, S. Khodabandeh, M. E. Davis, Chem. Muter. 1996,8, 1451; S . Cabrera, J.
El Haskouri, J. Alamo, A. Beltran, S. Mendioroz, M. D. Marcos, P. Amoros, Adv.
Muter. 1999, 11,379.
9. Y. Kim, B. Lee, J. Yi, The Korean J. of Chem. Eng. 2002, 19(5).
10. S. Vaudreuil, M. Bousmina, S. Kaliaguine, L. Bonneviot, A h . Muter. 2001, 13,
1310.
11. S. Valange, J.-L. Cuth, F. Kolenda, S. Lacombe, Z. Gabelica, Micropor. Mesopor.
Mat. 2000,35-36,597.
12. C. J. Brinker, G. W. Scherer, Sol-Gel Sience, Academic press, 1990.
13. M. Kruk, M. Jaroniec, Langmuir 1997, 13, 6267; W. W. Lukens, P. S. Winkel, D.
Zhao, J. Feng, G. D.Stucky, Langmuir 1999,15,5403.
14. W. Lin, J. Chen, Y. Sun, W. Pang, J. Chem. SOC.Chem. Commun. 1995,2367; K. R.
Kloetstra, H. W. Zandbergen, J. C. Jansen, H. B. Bekkum, Microporous Muter. 1996,
6,287.
15. Y. Wang, J. Banziger, P. L. Dubin, G. Filippelli, N. Nuraje, Environ, Sci. Technol.
2001,35,2608.

609
 STORAGE AND SELECTIVITY OF METHANE AND ETHANE
INTO SINGLE-WALLED CARBON NANTOUBES

YANG GON SEO

Division of Applied Chemical EngineeringIERI, GyeongsangNational Universiy
900 Gajwa-dong,Jinju 660-701, Korea
E-mail:[email protected]

BYEONG HO KIM
Department of Environmental Protection, Gyeongsang National University
900 Gajwa-dong,Jinju 660-701, Korea
E-mail:[email protected]

NIGEL A. SEATON
School of Chemical Engineering, University of Edinburgh
King's Buildings, Edinburgh EH9 3JL, UK
E-mail: [email protected]

The adsorption equilibria of methane, ethane and their mixture into single-walled carbon nanotubes
(SWNTs)were studied by using a Grand Canonical Monte Carlo (GCMC)method. The equilibrium
isotherms of methane and ethane and the selectivity from their equimolar mixture were reported.

1 Introduction

Although carbon nanotubes (CNTs) have been recently discovered [9], they have been
attracting a great deal of scientific interest due to their potential application in areas such
as adsorbents and composite materials. CNTs have the number of graphite sheets in tube
walls can vary fkom 1 for single-walled nanotubes (SWNTs) to over 50 for multi-walled
nanotubes (MWNTs), and inner diameter raging from Inm to 5nm with a definite
diameter [ 1,3]. The possibility of controllable pore size and distribution suggests that
CNTs might be used applications such as gas storage and selective separations from gas
mixtures.
Up to now, numerous studies have been conducted on their synthesis [9,10],
treatment [5,13] and physical properties [4]. However only limited number of studies has
been carried out on the adsorption of gas in CNTs, including experimental works [8,1 I ]
and molecular simulations [3,7,14-151. Adsorption behavior depends strongly on the
microporous structure of the particular adsorbent. In this work the effect of pore size on
the adsorption behavior is of interest. The adsorption equilibria of methane, ethane and
their mixture into SWNTs were studied by using a Grand Canonical Monte Carlo
(GCMC) method. We reported equilibrium isotherms of methane and ethane, and the
selectivity from their equimolar mixture.

2 Model and Simulation

CNTs are cylindrical structures and retain their cylindrical shape when their internal
diameters are less than lnm. CNTs flatten to form a honeycomb structure when the

610
internal diameters exceed 2 . 5 ~ 1[12]. Although the nanotubes typically have their ends
capped, selective oxidation can remove their end caps and reveal their hollow central
cavities [5,13]. A high-resolution transmission electron microscope (TEM) image of
SWNTs (manufactured by Iljin Nanotec, Korea) is shown in Figure I (a).

(4 0
Figure 1. A high-resolution TEM image(a) and a segment of an armchair mode of single-walled carbon
nanotubes(b).

There are two modes of rolling graphite sheet, which give rise to the armchair and
saw-tooth configurations. We construct SWNTs according to the armchair mode of
rolling. SWNTs constructed in this manner have only certain allowed diameters. SWNTs
at different allowed diameters can be produced by the saw-tooth mode of rolling. Figure 1
(b) illustrates a segment of an armchair SWNT of diameter D=l.496nm. D is the center-
to-center distance of two diametrically opposite carbon atoms on the nanotubes walls. The
boundary condition of axial direction was applied. We used a length of 7.87nm for all
simulationsreported in this work. The C-C bond length of 0.142nm corresponding to that
of graphite was used.
The intermolecular interactions between two molecules and fluid-wall interactions in
SWNTs were given by a 12-6 Lennard-Jones (LJ) potential. Methane was modeled as a
spherical LJ molecule and ethane as two LJ sites with the unified methyl group. The
interactions were cut at 2.286nm which corresponding to 5 times the methane o parameter.
In each step, one of these was chosen with equal probability at random. For each
point on the isotherm, the system was allowed to equilibrate for 5 ~ 1 steps 0 ~ before
collecting data. After equilibration, the simulation continued for 2 ~ 1 steps
0 ~ in order to
calculate the average values of the extent of adsorption. Further details of the simulations
are given elsewhere[2,6].

3 Results and Discussion

In Figure 2 the pure-component adsorption isotherms of methane and ethane in SWNTs

are presented as an amount adsorbed per unit volume of the pore. At low pressures the
greatest adsorption occurs in the small pores. This is due to smaller pores having larger
adsorbate-adsorbent interaction potentials. Small pores fill rapidly, even low pressure, due
to the presence of a strong wall potential function. The complex variation between the
isotherms for different pore sizes is caused by a trade-off between the strength of
methane-SWNTs interaction and the ability of SWNTs to accommodate methane

611
 - capacity
molecules. For D=O.678nm storage - of methane was very low due to sieving
effect of small pore.

8
. .- x(2.035~11
10

0.678~11
0.950~11
0.814nm

5
0 0 L
0 10 20 30 40 0 10 20 30 40
Prrasurc [bar] Pressure [bar]

Figure 2. Simulated isotherms for methane and ethane adsorption in single-walled carbon nanotubes of different
diameters.

The most important parameter from the point of view of mixture separation is the
selectivity. Figure 3 shows the binary selectivity at 298.2K with a 50% methane, 50%
ethane bulk-gas mixture over a range of pressures and pore diameters. The curves show
behavior with an initial increase in selectivity due to cooperative interactions between the
ethane molecules as the pressure is increased. As pressure increases further the adsorbate
densifies, which imposes an ordering of the adsorbate. The selectivity is a strong function
of pressure and pore width. The most commonly used approach to predict
multicomponent adsorption is Ideal Adsorbed Solution Theory (IAST) based on a
classical thermodynamic analysis. This approach requires pure component adsorption
isotherms, at the proposed temperature of operations of the adsorption unit, for all of the
components in the mixture. The pure-component isotherms form the inputs to IAST, and
the binary predictions are plotted in Figure 3. The binary simulation results by GCMC are
in good agreement with IAST.

612
 80
GCMC
0
0
-
-
lAST D[m]
1.628

2.035
A - 9 . 2.713

20

Figure 3. Comparison of GCMC simulation and IAST predictions of selectivity derived from simulated single
component isotherms.

4 Conclusions

The grand canonical Monte Car10 (GCMC) method was applied to calculate adsorption
equilibria of methane, ethane and their mixture. At low pressure small pores filled rapidly
due to strong wall potentials. The selectivity strongly depended on pressure and pore
width.

5 Acknowledgments

This work was supported by Korea Research Foundation Grant (KRF-2000-005-D00005).

References

1. Ajayan P. M., Stephan O., Colliex C. and Trauth D., Aligned carbon nanotubes
arrays formed by cutting a polymer resin-nanotube composite. Science 265 (1994)
pp. 1212-1214.
2. Allen M. P. and Tildesley D. J., Computer simulation of liquids (Clarendon, Oxford,
England, 1987).
3. Ayappa K. G., Simulation of binary mixture adsorption in carbon nanotubes:
Transitions in adsorbed fluid composition. Lungmuir 14 (1998) pp. 880-890.
4. Dujardin E. Ebbesen T. W., Hiura H. and Tanigaki K., Capillarity and wetting of
carbon nanotubes. Science 265 (1 994) pp. 1850-1 852.
5. Ebbesen T. W., Ajayan P. M., Hiura H. and Tanigaki K., Purification of nanotubes.
Nature 367 (1994) p. 5 19.
6. Frenkel D. and Smit B., Understanding molecular simulation from algorithms to
applications(Academic Press, San diego, 1996).

613
 7. Gu C., Gao G. H., Yu Y. X.and Nitta T., Simulation for separation of hydrogen and
carbon monoxide by adsorption on single-walled carbon nanotubes. Fluid Phase
Equilibria 194-197 (2002) pp. 297-303.
8. Hilding J., Grulke E. A., Sinnott S. B., Qian D., Andrews R. and Jagtoyen M.,
Sorption of butane on carbon multiwall nanotubes at room temperature. Langmuir 17
(2000) pp. 7540-7544.
9. Iijima S.,Helical microtubulesof graphitic carbon. Nature 354 (1991) pp. 56-58.
10. Iijima S. and Ichihashi T., Single-shell carbon nanotubes of 1-nm diameter. Nature
363 (1993) pp. 603-605.
11. Mackie E. B., Wolfson R. A., Arnold L. M., Lafdi K. and Migone, Adsorption
studies of methane films on catalytic carbon nanotubes and on carbon filaments,
Langmuir 13 (1997) pp. 7 197-7201.
12. Tersoff J. and Ruoff R. S.,Structure properties of a carbon-nanotube crystal. Phys.
Rev.Lett. 73 (1994) pp. 676-679.
13. Tsang S. C., Chen Y. K., Harris P. J. F. and Green M. L. H., A simple chemical
method of opening and filling carbon nanotubes, Nature 372 (1994) pp. 159-162.
14. Yin Y.F., Mays T. and McEnaney B., Molecular simulations of hydrogen storage in
carbon nanotubes arrays. Langmuir 16 (2000) pp. 10521-10527.
15. Zhang X. and Wang W., Methane adsorption in single-walled carbon nanotubes
arrays by molecular simulation and density functional theory. Fluid Phase Equilibria
194-197 (2002) pp 289-295.

614
 HYDROTALCITES FOR CARBON DIOXIDE ADSORBENTS
AT HIGH TEMPERATURE

J.I.YANG,M.H.JUNG,S.H.CHOANDJ.N.KIM
Separation Process Research Center, Korea Institute of Energy Research,
71-2, Jang-don, Yusung-gu,Taejeon, 305343, Korea
E-mail:[email protected]

Carbon dioxide adsorbing capacity was investigated using hydrotalcites as high temperahue
adsorbents. A gravimetric method was used to determine the Co2 adsorbing capacities of the
hydrotalcites, and the temperature was 450 "C. Hydrotalcite took higher adsorbing capacity
compared to other basic materials such as MgO, Alza. To increase the Co2 adsorbing capacity of
the hydrotalcite, MglAl ratios, preparation methods, and K2C03 impregnation were. checked. As a
result, the hydrotalcite prepared by high supersaturationwith MglAl=2 showed desirable adsorption-
desorption pattern and a higher C@ adsorbing capacity. Furthermore, KzC03 impregnation on the
hydrotalcite increased the C a adsorbing amount and the optimum value of &Ca impregnation
was 20 wt%. The hydrotalcite prepared by high supersaturation with Mg/Al=2 and 20 wt% K z C a
impregnation took the highest C@ adsorbing capacity of 0.77 mrnol/g at 450 "C and 800 mmHg.

1 Introduction

Hydrogen is commonly produced by the endothermic steam methane reforming (SMR)

reaction that is generally carried out in a catalytic (Nily-AlzO3) reaction at a temperature
of 750-900 "Cand a pressure of 50-600 psig. According to Hufton et al. [l], the hydrogen
production was improved by sorption-enhanced reaction process (SEW) that used an
admixture of a S M R catalyst and an adsorbent. The key requirement for practical use of
the SEW for HZproduction is development of the adsorbent that has a high COz working
capacity at moderately high temperatures (300-500 "C). Although zeolite A, zeolite X and
activated carbon are frequently used as carbon dioxide adsorbents at room temperature, it
is very difficult to use them at high temperature. Recently hydrotalcite-like compounds
were reported to have good features for carbon dioxide adsorbent at high temperatures
PI.
Hydrotalcite is a double-layered material that is composed of a positively charged
brucite-like layer and a negatively charged interlayer. Hytrotalcites are represented by the
general molecular formula, [M(II)I.xM(III)x(OH)~"'A~n~yH~O wherein, M(II)=Mg, Cu,
Ni, Co, Zn; M(III)=Al, Fe, Cr,V; A" is any interlayer anion such as CG", Cl-, NO',
S
O
:
- and x=0.1-0.33 [3]. The basicity of hydrotalcite depends on chemical composition
(type of cation, ratio of Mz'/Mh, type of anion existing in the interlayer) and activation
conditions. For example, calcination of the hydrotalcite above 450 "C results in mixed
metal oxides with strong basicity and high surface area. Thus, they can be used as base
catalysts or catalyst supports. In this research we prepared several hydrotalcites with
different preparation methods, varying Mg/Al ratios and adding different amounts of
K2C03for the purpose of using them as high temperature COzadsorbents.

615
2 Methods

2.1 Preparation of hydrotalcites.

Two different processes were used for preparing hydrotalcites; low supersaturation and
high supersaturationaccording to the methods described in the literature [4].
2.2 Evaluation of C02 adsorbing capacity.
A gravimetric method was used to determine the C02adsorbing capacities of the prepared
hydrotalcites. The amount of sample loaded into the TGA unit (C-1100,CAHN
INSTRUMENTS.INC.)was 65 mg. The sample was pretreated in flowing He at 500 "C
for 3 h. After the temperature was lowered to 450 "C,C02 was introduced to the balance
and the weight change was measured. The range of adsorption pressure was between 0 to
lo00 mmHg.

3 Results

3.I Comparison of hydrotalcite with other C02 adsorbentsat high temperature.

C02 adsorbing capacities of a pure MgO, a pure A1203 and a hydrotalcite (Mg/Al=2,
prepared by high supersaturationmethod) were measured at 450 "C,and their adsorptive
isotherms are shown in Figure 1. Both of individual components of hydrotalcite, MgO and
Al2O3, showed appreciable C02 adsorbing abilities even at high temperatures due to their
basicity. However, the hydrotalcite adsorbed the significantly more amount of C02than
MgO and A 1 2 0 3
3.2 Effect of aluminum contents.
Figure 2 shows COz adsorbing capacities of hydrotalcites prepared by the low
supersaturation method with different Mg/Al ratios. The hydrotalcite with Mg/AI ratio of
2 adsorbed the largest amount of C02 among hydrotalcites with Mg/Al ratios higher than
2. As the ratio of MglAl increased to 5 , C02adsorbing capacity decreased. However, as
the Mg/Al ratio increased further, hydrotalcites adsorbed more C02and the amount of
C02adsorption by the hydrotalcite with Mg/N ratio of 10 became almost equal to that of
hydrotalcite with Mg/Al ratio of 2. Yong et al. [2] also observed the similar valley shaped
results with the effect of aluminum content on C02 adsorbing capacities of Mg/Al
hydrotalcites. When the ratio of Mg/Al equaled two, the high layer charge density of the
hydrotalcite due to larger Al contents enabled higher C02 adsorbing capacity. However,
as the content of Al decreased with the increasing Mg content, the C02adsorbing capacity
decreased due to the loss of the layer charge density. On the other hand, the number of
basic sites increased with the increase in Mg content, resulting the increased C02
adsorption on the hydrotalcite with Mg/Al=lO.
3.3 Effect of preparation methods.
Hydrotalcites were prepared by two different methods of the differing precipitation rate
during the preparation of hydrotalcite precursor. Although both hydrotalcites showed
almost same adsorbed amount of C02, the hydrotalcite prepared by the high
supersaturation nethod had the higher desorption capacity. The hydrotalcite of high
supersaturation is known to have a low crystallinity, many crystalline nuclei and high
surface area due to its small particle size. And the higher surface area of hydrotalcites

616
prepared by the high supersaturation method may yield the desirable adsorption-
desorption behavior.
0.50,

..
.
0 100 2w 300 4M) 6w 800 7w 800 ow 1000 1100
Peq. mmHg

Figure 1. COzadsorption isotherms of various adsorbents ( T 4 5 0 "C,

+ Hydrotalcite(Mg/AI=2.high), + MgO , -A- AIzO3).

0.30
.
0

2 0.25
E
g 0.20
a
5 0.15
U
3 0.10
U
' 0.05
d
0
0.09
2 3 5 7 10
MglA ratio

Figure 2. Adsorbed amount of COz on hydrotalcites with varying ratios of Mg/Al (T=450
"C,prepared by low supersaturation method).

3.4 Effecr of &COj content.

Hoffman et al. [5] reported that the COz adsorbing capacity of the hydrotalcite could be
increased by KzCO3 impregnation in the presence of steam according to the reaction
(K2C03+COz+HzO=2KHC03).Furthermore, the increased basicity due to the
impregnated alkali metal carbonate may provide positive effect of adsorbing more COz to
the hydrotalcite even without steam. Based on that assumption, different amounts of
KzCO3 were impregnated on the hydrotalcite and their COZ adsorbing capacities were

617
investigated. It should be pointed out that COPadsorptionsin this experiment were carried
out without s t e m

l'O

02t
0.1

0.0 I
0 5 1 0 1 5 2 0 2 5 r ) 3 5 4 0 4 5 !
h C 0 , loading amount, wt %

Figure 3. Effect of
the KzCO3 loading on COz adsorption ( T 4 5 0 "C, P=800 mmHg,
MglAl=2, prepared by high supersaturation).
As shown in Figure 3, the amounts of KzCO3 up to 48 wt% were impregnated on the
hydrotalcite with Mg/Al=2 prepared by the high supersaturation method, which was
proved to be more effective in the COz desorption. As a result, the COZadsorbing capacity
was dramatically increased with the loading of KzCO3 compared to the hydrotalcite
without KzC03 impregnation. And there was an optimum amount of K&03 loading.
Therefore, the COz adsorbing capacity was 0.77 mmoVg in 20 wt96 KzCO3 impregnation.
If the loadiig amount increased further, the COz adsorbing capacity decreased. Therefore,
it was c o n f i i that high loading of KzCO3 could be favorable to COZadsorption due to
its increased basic sites, but the pores for COz adsorption were blocked with the loading
of KzCO3.

4 Conclusion

The hydrotalcite with the composition of Mg/Al=2 was a suitable COZadsorbent at high
temperature, viz. 450 "C, and its C02 adsorbing capacity was 0.28 mmoVg. A high
supersaturation method for making a hydrotalcite was desirable because the structure of
the hydrotalcite prepared by the method was proper for regeneration by pressure swing
operation. The KzC03 impregnation on the hydrotalcite was favorable to COz adsorbing
capacity because it made chemical properties of the hydrotalcite more basic than that of
the hydrotalcite without KzC03 impregnation. The 20 wt% K2CO3 impregnation in the
hydrotalcite (Mg/Al=2, high supersaturation) took the highest COZadsorbing capacity of

618
0.77 mmoVg at 450 "C and 800 d g . Above the amount of K2Ca impregnation, the
Cot adsorbing capacity decreased.

References

1. Hufton J. R., Mayorga S. and Sircar S., Sorption-enhanced reaction process for
hydrogen production. AIChE J. 45 (1999)pp. 248-256.
2. Yong Z.,Meta V. and Rodrigues A. E., Adsorption of carbon dioxide onto
hydrotalcite-like compounds (HTls) at high temperatures. I n d Eng. Chem Res. 40
(2001)pp. 204-209.
3. Mckenzie A. L., Fishel C. T. and Davis R. J., Investigation of the surface structure
and basic properties of calcined hydrotalcites. J. Cutul. 138 (1992)pp. 547-561.
4. Narayann S.and Krishna K..Hydrotalcite-supportedpalladium catalysts. Appl. Cutul.
A 174 (1998)pp. 221-229.
5. Hoffman J. S. and Pennline H. W., Investigation of C02 capture using regenerable
sorbents. Proc. of the I ThAnnual Intemutional Pittsburgh Coal Conference. 2000.

619
 EFFECT O F POLARITY OF POLYMERIC ADSORBENTS O N DESORPTION
OF VOCs UNDER MICROWAVE FIELD

XIANG LI ZHONG LI* HONG XIA XI HUAN WANG

College of Chem. Eng., South China University of Technology
Guangzhou, 510640 I! R.China. Email:[email protected]

In this work, desorption of volatile organic compounds (VOCs)from polar, weak polar and non-polar
polymeric adsorbents using microwave was investigated experimentally. Benzene and toluene were
separately used as adsorbates. Results showed that the application of microwave to regenerate the
polymeric adsorbents not only got higher regeneration efficiency in comparison with the use of heat
regeneration, but also made the temperatures of the fixed beds much lower than that when using the
heat regeneration. The weaker the polarity of a polymeric adsorbent,the easier its regenerationwas.

1. Introduction
Recent years, the pollution of volatile organic compounds (VOCs) has attracted
much attention. The VOCs have become important pollutants of air"]. The leakage of the
VOCs from chemical and pharmaceutical manufacturing, printing processes, paint and
adhesive manufacturing and applications, composites and fiberglass molding etc. is the
source of the pollutants. A number of adsorbents are capable of capturing a wide range of
VOCs. However, the conventional process of regenerating the adsorbents by using
organic solvents or thermal fluid still poses a major challenge in this field, notably
because of the high expense or secondary pollution. Thermal regeneration involves
higher temperature, which results in excessive burnout of the carbon (Grant 1990) It].
Chemical regeneration usually requires the use of organic solvent and involves inevitably
a secondary separation or pollution. Recently, some efforts, including supercritical
regeneration 13], ultrasound regeneration [4751 and bioregeneration of an adsorbent, have
been underway. During the 1980s a number of researchers investigated the otential for
using microwave heating to regenerate adsorbents. Burkholder (1986) [6 found that P
applying microwave energy enhanced the desorption of ethanol without heating silicalite
because it selectively heated the ethanol without heating adsorbent. Schmidt (1993) ['I
used microwave to desorb water from activated alumina and from Type 4A and 13X
zeolites, molecular sieves, as well as methanol from 13X zeolite. Initial results indicated
that for particular adsorbentladsorbate systems, microwave heating could dramatically
reduce re eneration time. Further, heating was uniform within a fixed bed. Opperman and
Brown (1999) firstly compared the microwave regeneration and the thermal
regeneration of activated carbons. Their experimental results showed that the temperature
required to regenerate completely the activated carbon when the microwave method was
used was much lower than that when the thermal regeneration method. Turner (2000)
proved that the use of microwave can make the adsorbate with the greater microwave
absorptivity be desorbed selectively, and the surface and adsorbed species can be heated
selectively. The objective of this work is to use microwave energy to regenerate
polymeric adsorbents saturated with benzene and toluene respectively.

2. Principle of microwave desorption

Different substances have different abilities of adsorbing microwave energy and

620
converting into heat. The property that describes how well a material or an adsorbate
adsorbs microwave energy and convert into heat is the effective dielectric loss factor
(c> ). Generally, the heat -up rate of an actuating medium in an applied electric field is
in proportion to its dielectric loss fator, and the frequency and the intensity of the
electric field, and on the other hand, its heating rate under microwave field is in inverse
proportion to its density and its specific heat capacity, as indicated in equation (1) [91.
Where M is mass of the substance or actuating medium, c i is an effective
dielectric loss factor of the actuating medium under microwave field, P is density of the
actuating medium, C, is specific heat capacity of the actuating medium, f is fiequency of
microwave, and E is intensity of the electric field. It means that the smaller the
dielectric loss factor of the actuating medium, the less the amount of microwave energy
adsorbed by the medium is, and thus the slower the its heating rate is.

t
If the dielectric loss factor c> of a material is equal to zero or very small, then the
material is transparent or semi-transparent to microwave, which means that microwave
can easily penetrate through the material without being adsorbed. For the fixed bed in
which the adsorbent has saturated with the adsorbate, when the use of microwave heats or
regenerates it, if the effective dielectric loss factor of the adsorbate is much larger than
that of the adsorbent, the adsorbate would be rapidly and selectively heated and then
desorbed, while the main body of the fixed bed, the adsorbent, would not be obviously
heated. It implies that in this case the most of microwave energy is used to heat the
adsorbate instead of the adsorbent It not only gets high efficiency of regeneration but also
decreases consume of the energy required to regenerate the adsorbent. As a result, it is an
important advantage of the application of microwave to regenerate the fixed bed.
Usually, the dielectric loss factors of polymeric adsorbents are relative small, i.e. the
polymeric adsorbents are transparent or semi-transparent to microwave. So they are
suitable to be used to adsorb VOCs and then be regenerated by microwave.

3. Experimental section

3.1. Reagent and Materials

Benzene (AR), Toluene (AR), N2 (purity: 99.5%). Polymeric resins: NKA 11 resin (a
polar resin), AB-8 (a weak polar resin) and D4006 (a non-polar resin), which were
purchased fiom Nankai Chemicals plant, Tianjin. Diameter of the resins ranged from
0 . 3 m to 1.0mm.
3.2. Device Instrument
WD8OO microwave oven whose power intensity was 0.734W/cm2, which was made
in Tianjin; FNJA electronic scale with 0.0001g of accuracy, which was made in Shanghai;
Adsorption column made of Teflon, which was 7cm long, and 1.3cm in diameter.

3.3. Adsorption of VOC on Polymeric Resins

Firstly, the resin was filled in the adsorption column, and then was dried by N2 under
microwave field for 10-15 minutes. Secondly, gassy mixture of dried air and VOC was
introduced into the column at constant flow rate 0.16 m3/h until the resin packed in the

621
column was saturatedby the VOC.The adsorption column was put on the electronic scale
with 0.0001g of accuracy. Its weight increased gradually as the VOC adsorbed on the
resin, and then got constant when the resin was saturated. The weights of the column
before and after the adsorption of the VOC can be measured using the electronic scale.
Finally, knowing the weights as well as the weight of the resin packed in the column,
ones can calculate the amount adsorbed on the resin, qo(g/g).

3.4. Microwave and Thermal Desorption

For the microwave regeneration, firstly, put the adsorption column packed with the
polymeric adsorbent on which the saturated amount adsorbed of VOCs was 40 in
microwave oven, and subsequently, use synchronously microwave to heat the column and
N2to sweep the VOCs desorbed from the adsorbent out. Regeneration efficiency of the
adsorbent was measured every thirty seconds by weighing method until regeneration
operation ended.
For the thermal regeneration, firstly, put the fixed bed packed with the polymeric
adsorbent on which the saturated amount adsorbed of the VOCs was qo in
temperature-maintaining container. Then, heat the column and maintain its temperature at
9O'C. After that, use N2 to sweep the VOCs desorbed from the adsorbent out. The
regeneration efficiency of the adsorbent can be measured using previously stated method.
Regeneration efficiency of the polymeric resin or the desorption efficiency of the
adsorbate, %%, can be found out according to equation (2). Where qo (g/g) was initial
amount adsorbed of the adsorbate on the resin before regeneration, and qt (glg) was
transient amount adsorbed of the adsorbate on the resin.

4. Results and discussion

4.1 Comparison between Microwave and Thermal Regeneration

Figures 1-3 showed the comparison

between the microwave and thermal
regeneration. It can be seen that the
application of microwave to regenerate
the polymeric adsorbents can get higher
regeneration efficiency than the
application of the thermal regeneration
method. For three kinds of the polymeric
resins that had adsorbed benzene, the 0 L . n . . . . . . . . . I . .* J
0 50 100 150 200 250 300 350 400
regeneration efficiency obtained by
t (6)
using microwave was up to go%, while
the regeneration efficiency obtained by Fig. I Comparison between Microwave and Thermal
using microwave was merely about Regeneration of NKA-11 Resin. Adsorbate: Benzene,
60-70%. NZFlow Rate: 0.16 m3/h

Table 1 showed the time dependence of temperatures, Tb, of three kinds of the fixed
beds respectively packing with the NKA I1 resin, the AB-8 resin and the D4006 resin. It
can be seen that the temperatures of the fixed beds when using the microwave
regeneration were relative low in comparison with the temperature (90%) of the fixed
bed when using the thermal regeneration. A interesting result was that not only the

622
 temperatures of the fixed beds when using the microwave regeneration were much lower
than that when using the heat regeneration, but also the regeneration rate of the polymeric
adsorbents obtained by using the microwave regeneration were much higher than that by
using the heat regeneration method. It implied that these polymeric adsorbents were
semi-transparent to microwaves, allowing the microwave energy to be applied efficiently
throughout the adsorbent bed, and meanwhile, under the microwave field the activated
energy of desorption of these organic compounds from the resins became decreased,
which made the regeneration of the adsorbents easy in comparison with the heat
regeneration.

- 0
0 50 100 150 200 250 300

Fig. 2 Comparison between Microwave and Thermal Fig. 3 Comparison between Microwave and Thermal
Regeneration0fAB-8 Resin. Adsorbate: Benzene, N2 Regenerationof MOO6 Resin. Adsorbate: Benzene, N2
Flow Rate: 0.16 m'h Flow Rate: 0.16 m'h

On the other hand, it should be noted that, the temperature of the fixed bed packing
with the D4006 resin was the lowest one, while the temperature of the fixed bed packing
with the NKA I1 resin was in the highest one. That is to say, the weaker the polarity of the
polymeric resin was, the lower the temperature of the fixed bed corresponding to the
resin was. It meant that the dielectric loss factor of the polymeric resin was in proportion
to its polarity. It should be mentioned that since a thermocouple or thermometer can not
be used to measure directly the temperature of the fixed bed under microwave field. The
temperatures of the fixed beds was measured by means of a special thermometer and
Microwave workstation Mar5 purchased from CEM company, US.

Table 1 Temperature Variance of the Fixed Beds Packed with Different Resins under Microwave Field
Microwave radiation time (s) d 60 240 300 400
D4006 Fixed bed (C) 28 31.5 37 38 40.5
AB-8 Fixed bed (TI 28 41 55 58 61
NKA-I1 Fixed bed ( C ) 28 40 69 75 80.3

4.2 Effect of different polymeric adsorbents on microwave desorption efficiency

In this work, three kinds of the polymeric adsorbents were used, of which the NKA
I1 resin was polar, the AB-8 resin was weak polar and the D4006 resin was non-polar.
Figures 4-5 showed the regeneration efficiency of the polymeric adsorbents obtained
when using the microwave regeneration. It can be seen that whatever the adsorbate used
was, benzene or toluene, the regeneration efficiency of the D4006 resin was the highest,
while the regeneration efficiency of the NKA I1 resin was the lowest. The regeneration

623
efficiency of the AB-8 resin was intervenient. It implied that adsorption affinity between
these VOCs and the polymeric adsorbents decreased with the decrease of adsorbent
polarity. The weaker the polarity of the adsorbent, the lower the corresponding desorption
activated energy of the VOCs on the adsorbent was, which would make the VOCs more
easier to desorb from the adsorbent. For example, compared with the regeneration
efficiency of the NKA I1 resin, not only the regeneration efficiency of the D4005 and the
AD-8 resin was more than 90%, but also the regeneration rates were higher.

100-

80 - A

d
-0-Adsorbent:

-0-
D4000
-A- Adsorbent: AB-0
Adsorbent: NKA-II
2obdsp/ --o-Adsorben(:
-0-Adsorbent: AB-0
4 -
D4000

Adsorbent: NKA-II
--
0 100 200 300 400 500
t (S)

Fig. 4 Effect of Different Resins on the Microwave Fig. 5 Effect of Different Resins on the Microwave
RegenerationEficiency. Adsorbate: Benzene, NZFlow Regeneration Eficiency. Adsorbate: Toluene, NZ
Rate: 0.12 mh. Flow Rate: 0.12 m3/h

5. Conclusion
The application of microwave to regenerate the polymeric adsorbents can not only get
higher regeneration efficiency in comparison with the use of conventional heat
regeneration, but also make the temperatures of the fixed beds be much lower than that
when using the heat regeneration. The weaker the polarity of a polymeric adsorbent, the
easier its regeneration was.
Acknowledgements
This work was supported by the National Natural Science Foundation of China (No.
29936100) and the Natural Science Foundation of Guangdong Province.
References
1. Opperman S.H. and Brown R. C., Pollution Engineering, 3 l(1) (1999) p58
2. Grant T. M. and King C. J., Ind. Eng. Chem. Res., 29 (1 990) p264
3. Xie Lanying, Xi Hongxia, LI Xiangbin, Li Zhong, Ion Exchange and Adsorption,
16(5) (2000) p413
4. Rege S.U. Yang R.T. and Cai C.A., Desorption by Ultrasound: Phenol on Activated
Carbon and Polymeric Resin, AIChE Journal, 44(7) (1998), pp1519-1528
5. LI Zhong, LI Xiangbin, XI Hongxia, Effects of Ultrasound on Adsorption
equilibrium of Phenol on Polymeric Resin, Chemical Engineering Journal, Vo1.86
(2002) pp375-379.
6. Burkholder H. R., Fanslow G. E., and Bluhm D. D., Ind. Eng. Chem. Fund , 25:
(1986) p414
7. Schmidt Philip S. and James R. Fair, Waste Management, 13(5-7) (1993) p25
8. Turner Michael D., Laurence R. L. and Yngvesson K. S., AIChE Journal, 46(4)
(2000) p758
9. JIN Qinhan, Microwave Chemistry, Beijing: Science Press, 1999

624
 MIXED-GAS ADSORPTION ON HETEROGENEOUS SUBSTRATES IN THE
PRESENCE OF LATERAL AD-AD INTERACTIONS

A. J. RAMIREZ-PASTOR, F. M.BULNES AND J. L. RICCARDO

Laboratorio de Ciencias de Superficiesy Medios Porosos. Dpto. ak Fisica. UniversidadNacional de
San Luis - CONICET. Chacabucoy Pedernera. 5700- San Luis. Argentina.
E-mail: [email protected]

Adsorption of binary gas mixtures in the presence of ad-ad interactions is studied through grand
canonical Monte Carlo simulation in the framework of the lattice-gas model. The disordered surface
has been characterized by patches of shallow and deep sites, arranged in a chessboard-like topography.
The adsorption process is monitored through partial and total isotherms, and differential heats of
adsorption. Interesting behaviors have been observed depending on ad-ad interactions and energetic
disorder.

1 Introduction

Adsorption of gas mixtures on solid heterogeneous substrates has received an increasing

interest in the last decades, due to its importance in relation with new technological
developments, like gas-separation and purification [ 1,2].
The description of real adsorption requires to take into account two main effects on the
calculation of the thermodynamic quantities: lateral ad-ad interactions, and characteristics
of the energy surface. In addition, in the case of multicomponent adsorption, different
species could "see" different disordered topographies.
Previous works dealing with disordered surfaceshave been dedicated mainly to random,
or correlated topographies. In the latter case, the combination of heterogeneity and ad-ad
interactions effects produce complex behaviors on the equilibrium properties. An exact
statistical mechanical treatment is unfortunately not yet available and, therefore, the
theoretical description of adsorption has relied on simplified models. One way of
circumventingthis complication is the Monte Carlo (MC) method, which has demonstrated
to be a valuable tool to study surface processes [3,4].
In this work, the adsorption of binary mixtures is studied through MC simulation in the
context of the lattice-gas model. The topography has been characterized by shallow and
deep sites, arranged in a chessboard-like structure. The disorder has been associated to one
of the species, while the other component interact with an homogeneous substrate. The
process is monitored through partial and total isotherms, and differential heats of adsorption,
which appear as sensitive to both lateral interactionsand energetic disorder.

2 Model and Monte Carlo Simulation

The substrate has been represented by a square lattice of M=LxL adsorbing sites with
periodical boundary conditions. The heterogeneityhas been introduced by considering the
two referred kinds of sites. This is a so-called bivariate surface, in equal concentration,
forming hd patches distributed in a chessboard-likeordered topography.
Let us introduce the site occupation variable ci: ci=O if the site i is empty, and ~ = - l
[%=I] if the site i is occupied by a B[A]-atom. The variable a labels the different kind of

625
sites; a 4 [a=1] represents a shallow [deep] site. Each component occupies only one
adsorption site. The energies involved in the process are:
&AD [EBD]: adsorption energy for an A p]particle on a deep site.
EAS[EM]:adsorption energy for an A [B] particle on a shallow site.
W M [WBB] nearest neighbor (NN) energy interaction between an ad-pair AA [BB].
WAB, NN energy interaction between an ad-pair AB (or BA).

Under these considerations, the Harniltonian H of the system, is given by

where 6 is the Kronecker delta function, pA [ p ~ is] the chemical potential of specie A [B],
and I",i means that for a given site i, the sum runs over its four NN sites.
The binary mixture adsorption is simulated by assuming an ideal AB-gas at fixed T, pA
and p ~In. equilibriumthere are two ways to perform a change of the system state: adsorbing
(desorbing) one molecule onto (fiom) the surfkce.
For a given topography, an elementary MC simulation step (MCS) is as follows:
1) Set PA, p ~T, , and an initial state by placing randomly N molecules on the lattice.
2) Choose randomly one of the components of the mixture -*X ( X=A or B).
3) Choose randomly one of the M sites 4 i; generate a random number 5 E [0,1]:
i) if the site i is empty, and 5 5 W,, ,then an X particle is adsorbed on i. Otherwise,
the transition is rejected. Wadsis the transition probability fiom a state with N
particles to a new state with N+1 particles
ii) if the site i is occupied by an X particle, and 5 S wd, , then the X particle is
desorbed fiom i. Otherwise, the transition is rejected. wda is the transition
probability fiom a state with N particles to a new state with N-1 particles.
4) Repeat fiom step 2)M times.
W* and Wda were obtained in the Metropolis scheme [5]. Then, the total and partial
isotherms are obtained as simple averages over m successive configurations: @LA,~B)=
=(N)/M, eA(pA,pB)=(NA)/M, and ~PA,CIB)=(NB)/M,where (...) means the time average
throughout the simulation, and Nx (X=A or B) is the number of adsorbed X-particles
(N=NA+NB).The differentialheat of adsorption qi for the i-specie is [6]:

where U is the energy of the adsorbate and pil/kBT.

The simulations were developed for square LxL lattices, with L=200,and periodic

626
boundary conditions. Finite size effects are negligible. The first mo=2x105MCS were
discarded to allow equilibrium. The next m=2x105 MCS were used to compute averages.
The chemical potential of one of the components is fixed through the process, p~=0,while
the other one is variable.

3 Results and discussion

Figure 1 shows the effect of WBB on adsorption isotherms and differential heats of
adsorption, for a strongly heterogeneous substrate (A.EA=EAD-EAS=-32hT) and repulsive
interactions W M ~ C B T WAB=O. For pA+ -00 (being p~=0),0 ~ and
and 4 0~ is a function of
WBB. 8B(pA+ -00 ,p~=o) diminishes with WBB, since the B species behaves as a repulsive
single gas on an homogeneous system. In fact, the range of variation of 8B(pA+ -00 ,pB=O)
lies between 0.5 and 0.226, for WBB=O and WBB + 00, respectively.
40

30
4,
20

10

-10
*.
-4.00
-20
-40 -20 0 -40 -20 0 20 0.0 0.5 I ,O
P A P A QA

Figure I: W M ~ WA&
, and differentWBBS. a) total and b) partial isotherms; c) differentialheats of adsorption.

As pAincreases, some A particles are adsorbed at expenses of B particles which results

in a decreasing of the B coverage. Both A and B particles tend to form 42x2)
interpenetrating ordered structures as WBB increases. Owing to WAB=O,8A(pA) and qA(pA),
do not depend on WBB. As eA=O.75,the B isotherms (for different WBBS) collapse to a limit
curve. In this conditions, the A particles have completed the deep patches and are forming a
42x2)phase in the shallow patches. On the other hand, the B particles occupy the holes in
the shallow patches, surrounded by four NN A particles. It is interesting to note that total
and B partial curves are contained between two limit ones, corresponding to wBB=O and
WBB+~O.
In order to analyze the effect of the A-B interaction, Fig. 2 shows the behavior of
isothermsand differential heats of adsorption for wM=4kBT,W B B and ~ different WABS. As
pA+ -00, the surface is half covered by B particles, distributed at random. With increasing
PA, NA increases, which produces a decreasing h 6 B owing to the WAB repulsions. In fact,
the probability of fmding a pair AB (or BA) adsorbed on NN sites diminishes with wAB.In
the limit wAB+ w every A particle on the lattice excludes five sites for the adsorption of B
particles (the occupied site plus its four NN sites).
For 0<0~<0.25, the A species is adsorbed on deep patches; [OBI increases
[decreases]. During this regime the number of desorbed particles is greater than the number

627
of adsorbed particles which results in a decreasing of 0, which exhibits a minimum at
eA=0.25(see Fig. 2 (a)). At this coverage, the B species covers 15% of the lattice, occupying
inner sites (plus some border sites) of shallow patches. This novel phenomenon, due to the
AB interactions, provides a theoretical basis for the evaluation of experimental findings
involving adsorption of interacting gas mixtures.
I I I 40

, qd

20

-20

-40 -20 0 20 -20 0 20 0.0

F A FA
Figure f:W M ~ ,w s d and different WM'S. a) total and b) partial isotherms;c) differential heats of adsorption.
For 0.25<0A<0.5,the A particles fill border sites of deep patches. [OBI increases
[decreases] resulting in a decreasing of the total coverage 0 [see Fig. 2 (a)-(b)]. This
adsorption regime is observed until eBm0.06 where half of the inner shallow sites are
occupied (A particles adsorbed on border shallow sites would increase the energy of the
system due to the A-B interactions). The adsorption process follows until the deep patches
become completely covered (0,,=0.5); in this limit, the A isotherm exhibits the well known
plateau due to the surface heterogeneity, which induces a plateau in the B isotherm.
For increasing pA,beyond 0,=0.5, the A particles cover the shallow patches giving
place to the formation of ordered structures inside them while the B particles are desorbed
due to the repulsions. This behavior appears as a fast drop to zero in 0 ~while, 0~ and 0
increase The total coverage climbs from 0.5 to 1 with an intermediate plateau at 0 m0.75
because of the A-A repulsions.
Figure 2 (c) shows q ~ ( 0and~ )@A) for the same cases depicted in Fig. 2 (a)-(b). The
influence of WAB on the differential heats of adsorption can be easily understood by
following the same reasoning like for the corresponding adsorption isotherms. For
0<0~<0.25,48'0, even for ~ ~ 3because 0 , A and B particles are adsorbed avoiding A-B
repulsions. In this regime, q A measures the adsorption energy of deep sites. For
0.25<0A<0.5, q ~ = 0while q A shows an abrupt jump due to the formation of ordered
structures in the adsorbate. In O.5GA<0.75,qAand q B slowly diminish with eA;finally, for
eA>O.75, both of them remain constant.
In Figure 3 we have plotted the results corresponding to W M = W A B ~ CAQ= B T ,-32k~T,
and different values of wBB. Isotherms and differential heats of adsorption can be
understood by following the analysis of Figs. 1-2, because the behavior observed in Fig. 3 is
the superpositionof those effects.
In summary, when B-Brepulsions are introduced the initial coverage decreases. In this
case, both total and B isotherms are strongly dependent on wBB, unlike the A isotherm which
do not depend on wBB; for wBB+ 00 there exists a limit curve for QA) and %(PA). On the

628
other hand, a novel behavior is evidenced in the total isotherm at low coverage when A-B
interactions are considered. During this regime the number of desorbed particles is greater
than the number of adsorbed particles which results in a decrease of 0, which shows a
minimum at OA=O.2S.

40

qd
20

-20

1 - 1 . 1 . 1 . 1 . 1 . 1

-40 -20 0 -40 -20 0 20 0,O 0.5 1 ,O

FA FA OA

Figure 3: W M ~ WA& , and differentWBB'S. a) total and b) partial isotherms; c) differential heats of adsorption.
Solid circles represent the case W M ~ ~ B WT - B, ~ .

MC simulations have proven to be an adequate and powerful tool to study

multicomponent adsorbates with repulsive interactions. The model can be applied in
principle to any kind of topography and seems to be useful for analyzing experimental data
without any special requirements, or large time-consuming computations.

References

1. Ruthven D., Principles of Adsorption and Adsorption Processes (Wiley, N.Y. 1984).
2. Rudzinski W., Steele W. A. and Zgrablich G., Equilibria and Dynamics of Gas
Adsorption on Heterogenous Solid Surfaces (Elsevier, Amsterdam 1997).
3. Bulnes F., Ramirez-Pastor A. J., Riccardo J. L. and Pereyra V., Adsorption of Binary
Mixtures on Heterogeneous Surfaces. In Adrorption Science and Technology ed. By
Do D. D. (Word1 Scientific Publishing, Singapore2000) pp. 5 17-52 1.
4. Bulnes F., Ramirez-Pastor A. and Pereyra V.,Study of Adsorption of Binary Mixtu-res
on Disordered Substrates. J. of Mol. Cat. A: Chem 167 (2001) pp. 129-139.
5. Nicholson D. and Parsonage N. D., Computer Simulation and the Statistical Mechanics
of Adsorption (Academic Press, London 1982).
6. Bakaev V. A. and Steele W. A., Grand Canonical Ensemble Computer Simulation of
Adsorption of Argon on a Heterogeneous Surface. Langmuir 8 (1992) 148.

629
 ADSORPTION ON CORRELATED DISORDERED SUBSTRATES

R. H. LOPEZ, F. M. BULNES, J. L. RICCARDO AND G. ZGRABLICH

Lab. de Ciencias ak Superjicies y Medios Porosos. Univ. Nacional ak San Luis - CONICET
Chacabuco 91 7, 5700- San Luis. Argentina.
E-mail: [email protected]

F. ROJAS
Dpto.de Quimica,Univ.Autdnoma Metropolitana-Iztapalapa,C.P. 55534, MpXico D.F. 09340.
Mtxico

Adsorption of interacting gases on correlated heterogeneous surfaces is studied through Monte Carlo
simulations in the framework of the lattice-gas model. The substrate is generated from the Dual-Site
Bond Model, which consider both entities sites and bonds for the description of the correlated
disordered medium. The process is monitored by following adsorption isotherms and coverage
fluctuations. Different behaviors have been observed depending on lateral interactions and energy
correlations.

1 Introduction

The role of the adsorptive surface characteristicsin many processes of practical importance
is a topic of increasing interest in surface science [ 11. The energetictopography can strongly
affect some phenomena related to the solid-gas interface, like surface diffusion, reactions,
reconstruction, etc. It is well known that in the case of adsorption, lateral interactions
usually compete with topography and obscure its effects on adsorption isotherm, entropy,
adsorption heats, etc 121. However, much less attention has been paid to analyze the
dependence on heterogeneity of the thermodynamic factor. To study the effects of energy
correlationsand energy interactions on the thermodynamic factor is of central importance in
order to analyze the behavior of the chemical diffusion coefficient of adsorbed species by
using the Kubo-Green approximation [3].
We have performed Monte Carlo simulations of adsorption on heterogeneous surfaces,
in the context of the lattice-gas model [4], where the energetic topography has been
generated from the Dual Site-Bond Model (DSBM) [5]. The process is monitored through
adsorption isotherms and coverage fluctuations [3]. Our scope here is to discuss how energy
correlations and lateral ad-ad interactions affect the main characteristics of adsorption
properties, particularly the coverage fluctuations which appears as very sensitive to both
disorder and interactions.

2 The heterogeneous surface

To model the substrate we have used the Dual Site-Bond Model (DSBM) [ 5 ] , which has
proved to be useful to analyze topography effects on several molecular processes on
heterogeneous surfaces. DSBM provide a statistical description of the disordered media
based on two elements: sites, and the corresponding saddle points (bonds); the adsorptive
energy surface is described by site and bond probability density functions Fs(Es) and
FB(EB).The distribution hnctions Sand B associated to Fsand FBare defined by

630
 ES
S(E,)= IF,(E)dE B(E,) = JFB(E)dE (1 1
0. 0

The joint probability density function of finding a connected pair of a site with energy
ES and a bond with energy Esis given by

The correlation function 4 carries information on how sites and bonds are assigned to
each other. The simplest picture is to consider the maximum randomness allowed by the
Construction Principle (the energy of a site cannot be less than the energies of the
connected bonds). Finally Fs(Es) and FB(EB) are, in general, arbitrary functions fulfilling
the condition B Q S(0,which ensures that there will be sufficient amount of bonds to be
connected to sites with sizes equal or smaller than E. The overlapping degree f2,defmed as
the intersection area under Fs and FB is a natural measure of site-bond correlation.
Non-overlapping distributions represent a completely random heterogeneous surface with
zero correlation length, and increasing overlapping generates random patches topographies
of increasing sizes (or correlationlengths). Intermediate values of represent more realistic
situations where correlations between neighboring entities are neither null nor infinitely
extended but rather short ranged. The greater CI the larger the correlation Cdr)between two
elements separated by a distancer. Cdr) has an exponential decay for intermediate and high
values of f2, Cdr) oc exp[-r/ro] where ro is a typical correlation length depending on n:ro
=2f22/(1-n)2 [5].
Different kinds of heterogeneous surfaces can be generated by choosing adequate site
and bond distributions. The procedure by which correlated heterogeneous topographies are
simulated in the DSBM is described in detail elsewhere [6].

3 Basic definitions and Monte Carlo simulation scheme

In the m e w o r k ofthe lattice gas model, the adsorption process is simulated by assuming a
square lattice ofM=LxL adsorption sites, with periodic boundary conditions, in equilibrium
with an ideal gas characterizedby chemical potential ,uand temperature T. The surface as
well as the adsorbate are inert upon adsorption. Then, for a given configuration of
adparticles, the hamiltonian H of the system is given by
M M
H = W C c i c j +cC,Ei - p C c i (4)
2 1i.j) i=l i=1

where w is the nearest neighbor (NN) interaction constant, { i j )denote pairs of NN sites,
and Eiis the adsorption energy of the site i. The spin variable cj take the value c f l [ ~ i l if]
the site i is empty [occupied] (multiple site occupation is excluded).
At a fvted T and for a given value of f i the adsorption process has been simulated by
using the grand canonicalMonte Carlo method [5]. At any elementary step, a site chosen at
random is tested to change its occupancy state according to the Metropolis scheme of
probabilities: W(XpX,)=min{ 1,exp[-(Hf-Hi)/kBTI},where Hf and Hi are the hamiltonians
of the final (X,)and initial states (Xi), and RB the BoltPnann constant. A Monte Carlo Step
(MCs) consist of testing at random Msites to change its occupancy state. Then, mean values
of thermodynamic quantities are obtained by simple averages over m succesive
configurations, like the surface coverage, 8, and the internal energy of the system, U,given

631
by f+(N)/M and U9(H)-p(N)where N is the number of adparticles,and the brackets denote
averages over uncorrelated configurations.
The thermodynamic factor Th and its inverse,f, are defined as:

4 Results and discussion

Without loss of generality, we can consider that all energies are measured in units of kBT.
Then, all results will be independent of the temperature.
We used the Monte Car10 method described in [5,6] for generating heterogeneous
substrates fiom DSBM. We have considered uniform distributions for sites and bonds
(A&=MB=1) with mean site energy Eh=2.5; the site distribution is fixed while the bond
distribution is shifted toward lower energies to increase the correlation degree. We used
square lattices of size LxL, with L=400 [L=700] sites for fk0.5 [R20.5]. The
approximation to thermodynamic equilibrium usually required -10 MCs. Then, the next
4x10 MCs were used to evaluate the equilibrium properties.
Fig. 1 shows the MC results for the coverage fluctuationsf(f=Ti) correspondingto
the random topography (W), and for different values of the NN interaction, w. As it was
expected, the adsorption process is qualitatively similar to the corresponding to the
homogeneous substrate. For w=O, f exhibits the linear dependencef+l-O); in this case,
particles occupy firstly the strongest sites, which are distributed at random on the surface
and the process follows with the filling of weaker and weaker sites.
For the repulsive case (w>o), and for low values of w, f monotonically decrease with 8
while for a fixed value of efdecreases with w, because for a given value of pthe number of
adparticles N decreases with w. The minimum infat M . 5 for large values of w indicates
the tendency to ordered structures in the adsorbate, since the particles adsorb avoiding NN
interactions. in this case, lateral interactions become more important than heterogeneity
(wc=1.763668, 8,=0.5 are the critical values for a orderdisorder transition on an
homogeneous substrate [3,7]). In this conditions fluctuations will strongly promote the
collective diffusion in the adlayer. Contrarily attractive interactions inhibit the collective
diffusionas f increases with w for all 8 (fig. 1(b)).
The effect of the energy correlations onffor repulsive and attractive interactions, is
shown in figs. 2 and 3, respectively.
In the case of repulsive interactions, neither for R+O (random heterogeneous
substrate) nor Q+ 1 (strongly correlated patches of sites) the energy correlation influences
importantly the thermodynamicfactor (and accordingly the collective diffusion).
As shown in fig. 2 the adsorption isotherms are poorly sensitiveto the correlations. The
adsorption proceeds in such a way to avoid the occupation of NN sites, because NN
repulsions would increase the energy of the system.
The effects of the energy correlation on adsorption isotherms and coverage fluctuations
in the presence of attractive interactions is shown in Fig.3. For k-0, there exists a
competence between the convenience of occupying strong adsorbingsites and the attractive
interactions. Thus, the adsorption is more favorable at intermediate 8, and Q) increases

632
faster with 0 at 830.5. Accordingly f reaches a maximum around half coverage. For
increasing a,the topography of quasi-homotaticpatches favors an increase in the number of
adparticles at low f i Then, the slope of)@t increases more rapidly with Q at low pressure,
and the maximum offshifts to lower values of coverage. At intermediate 0, the slope of the
isotherm changes due to the strong adsorptive patches are almost completely covered,
producing a remarkable drop in$ At high coverage, f develops another local maximum
owing to the adsorption occurs on less adsorptive patches, and for increasingp, the lateral
interactionsenhances the isotherm slope. The local maximum off (at 8 0 . 5 ) which shifts to
higher 8 with a.

1,o

0,8
- ~=-0.25

0,6
f0,4 f
02 OS

0,o 0,o
0,O 0,2 0,4 0,6 0,8 1,0 0,2 0,4 0,6 0,8 1,0
e e
Figure 1: coverage fluctuations for M.(a) repulsive interactions; (b) attractive interactions.

1,o 1.o

03

0,6
f0,4 -&-&2=0.75
4 = 0 . 8 5

0,2
lines: Q=0.85

0,o
0,O 0,2 0,4
e 0,6 0,8 1,0 0 4
P
8 12

Figure 2: (a) coverage fluctuations, and (b) adsorption isotherms for w>o and different values of n.
Unlike the repulsive case, this important dependence off on the energy correlation for
attractive interactions is also expected to affect the collective diffiion coefficient. It is
worth noting that diffusion will be significantly inhibited at low and high coverage except at
intermediate 8.
In summary, within the thanework of a general model of heterogeneous substrates with
energy correlations we have shown, through the analysis of the thermodynamicfactor, that

633
energy correlations are expected to influence collective d i m i o n only in presence of
attractive interactions. For repulsive interactions heterogeneity and correlations would not
affect much collective diffusion.

f
4

n-0,O 0,2 0,4 0,6 0,8 1,0 -1 0 1

0,o

0 P
Figure 3: (a) coverage fluctuations, and (b) adsorption isotherms, for 1 6 0and different values of n.

References

1. W. Rudzinski and D. Everett, Adsorption of gases on heterogeneous surfaces, Acad.

Press, N. York (1992); W. Rudziisb, W. Steele and G. Zgrablich, Equilibria and
dynamics of gas adsorption on heterogeneous surfaces. Elsevier,Amsterdam (1996)
2. G. Zgrablich, C. Zuppa, M. Ciacera, J.L. Riccardo and W. Steele, The effect of
energetic topography on the structure of the adsorbate. Surface Sci. 356 (1996).
3. F. Bulnes and A.J. Ramirez-Pastor, Monte Carlo simulation of collective diffusionon
heterogeneous media: Dual Site-Bond model Granular Mutter 3, 1-2 (200 1).
4. D. Nicholson and N. Parsonage, Computer simulation and the statistical mechanics of
adsorption Acad. Press, London (1 982).
5 . V. Mayagoitia, F. Rojas, V. Pereyra and G. Zgrablich, Surf: Sci. 221 (1989); R. H.
Ldpez, A. M. Vidales and G. Zgrablich, Correlated site-bond ensembles: statistical
equilibrium and finite-size effects, Lungmuir 16,7 (2000).
6. J.L. Riccardo, W. Steele, A. Ramirez-Cuesta and G. Zgrablich, Pure Monte Carlo
simulation of model heterogeneous substrates Langmuir 13 (1997).
7. A. J. Ramirez-Pastor and F. Bulnes, Differential heat of adsorption in presence of an
order-disorderphase transition. Physicu A 283 (2000).

634
 TEMPERATURE EFFECTS ON THE SCALING PROPERTIES OF
ADSORPTION ON BIVARIATE HETEROGENEOUSSURFACES

F. ROMdl, F. BULNES,A. J. RAMIREZ-PASTOR AND G . ZGRABLICHt

Laboratorio de Ciencias de Superficies y Medios Porosos, Departamento de Fisica, Universidad
Nacional de San Luis, CONICEX Chacabuco 91 7, 5700, San Luis, Argentina
t Present address: UniversidadAutdnoma Metropolitana,Iztapaiapa, Departamento de Quimica,
P.O. Box 55-534, Mixico D.F..M6xico.
E-mail:[email protected]

Adsorption of particles with nearest-neighbor repulsive interactions is studied through Monte Carlo
simulation on bivariate surfaces characterizedby patches of weak and strong adsorbing sites of size 1.
Patches are considered to have square geometry and they can be either arranged in a deterministic
ordered structure or in a random way. Quantities are identified which scale obeying power laws as a
function of the scale length 1. Consequences of this finding are discussed for the determinationof the
energetic topographyof the surface from adsorption measurements.

1 Introduction

Bivariate surfaces may mimic, to a rough approximation, more general heterogeneous

adsorbents [1,2]. Just to give a few examples, we may mention the surfaces with energetic
topography arising fiom a continuous distribution of adsorptive energy with spatial
correlations, like those described by the Dual Site-Bond Model, or that arising from a
solid where a small amount of randomly distributed impurity (strongly adsorptive) atoms
are added [3]. In both cases the energetic topography could be represented by a random
distribution of irregular patches (with a characteristicsize) of weak and strong sites.
Accordingly, the scope of the present work is to continue previous Monte Carlo
simulation studies on the general properties of the adsorption of interacting particles on
model bivariate surfaces with a characteristic correlation length, 1, and fmd out to what
extent this length scale could be determined from adsorption measurements. The behavior
of relevant quantities, like adsorption isotherms and isosteric heats of adsorption, is shown
to lead to a power law with a universal exponent a. The temperature variation of this
exponent is studied in particular, extending previous results obtained at kBT=l [4]. The
effect of temperature T on the behavior of a as a function of the relative strength of ad-ad
interactions and energy gap between strong and weak sites is determined, showing that at
high temperatures a tends to a constant. The possibility of determining I from experiments
is discussed.

2 Model

We assume that the substrate is represented by a two-dimensional square lattice of

M=LxL adsorption sites, with periodic boundary conditions. Weak and strong sites, with
adsorptive energies EI and cZ (E,<E~) respectively, form square patches of size 1 (I=
1,2,3,..),which are spatially distributed either in a deterministic alternate way (chessboard
topography) or in a non overlapping random way.
The substrate is exposed to an ideal gas phase at temperature T and chemical

635
potential p Particles can be adsorbed with the restriction of at most one adsorbed particle
per site and we consider a nearest neighbor (NN) repulsive interaction energy w among
them. Then the adsorbed phase is characterized by the hamiltonian:

where t?h9,+& is the total s h c e coverage (summing the coverage on weak and strong
sites), ni is the site occupation number (4 if empty or =I if occupied) and the sum runs
over all pairs of NN sites (ij9. Without loss of generality, we can consider that all energies
are measured in units of b T , and that sI=O and s2=sI+AE, in such a way that the
adsorptive energy is characterizedby the single parameter AE.
The adsorption process is simulated through a Grand Canonical Ensemble Monte
Carlo (GCEMC) method [5]. Given T and p, an initial configuration with N=M/2 ad-
particles at random positions is generated. Then an adsorption-desorption process is
started, where a site is chosen at random and an attempt is made to change its occupancy
state with probability given by the Metropolis rule -in{ 1,exp[-AWkBT]},where AH=
=Hf-Hi is the difference between the hamiltonians of the final and initial states. A Monte
Carlo Step (MCS) is achieved when M sites have been tested to change its occupancy
state. The approximation to thermodynamic equilibrium is monitored through the
fluctuations in the number N of adsorbed particles; this is usually reached in -I05MCS.
After that, mean values of thermodynamic quantities, like the coverage 0 and the internal
energy U,are obtained by simple averages over M configurations: &(N)/M and U=(H)-p
(N) where the brackets denote averages over uncorrelated configurations.
By changing the value of p, the adsorption isotherm at a given temperature can be
obtained. Furthermore, from the simulation results, the differential heat of adsorption qd as
a function of the coverage is calculated as qd@=[a(U)/iWJ~.
We have used M=104 and w 1 0 5 .With this size of the lattice (L=lOO,in such a.way2
that it is a multiple of r) we verified that finite size effects, which affect the isotherms in
the case of repulsive interactionsat much smaller sizes, are negligible.

3 Results and discussion

It is known that, at low temperatures and depending on w/AE, the adsorption process
follows two different regimes [4].
For w/AE I1/4, which corresponds to the so-called Regime 1 : i ) strong patches are
filled first up to W . 2 5 , where a 42x2) structure is formed on them; ii) the filling of
strong patches is completed up to e0.5. Processes corresponding to the regions
0.5-0.75 and 0.75-1, respectively, are equivalent to processes i ) and ii) for weak
patches.
For w/AE 2 1/3, denoted as Regime I1 : 11 the strong patches are filled until the
42x2) phase is formed on them; ii) idem i) for the weak patches; iii) the filling of the
strong patches is completed; iv) the filling of the weak patches is completed. These
characteristics can be easily followed through the behavior of the differential heat of
adsorption. Regimes I and I1 are disconnected; in between (1/4< w/AE<1/3), the system
behaves in a mixed transition regime changing continuously from one to another.
Figures 1 and 2 show the behavior of adsorption isotherms, (a), and 4d@, (b), for
different topographies in regimes I and 11, respectively. We have identified the different

636
topographies as: 1, [l,] for chessboard [random] patches of size 1, and bp for the case of a
surface with two big patches. It can be seen that all curves are contained between two
limit ones: the one corresponding to lc and the one corresponding to bp. The fact that
both adsorption isotherm and heat of adsorption curves for different topographies,
characterized by a length scale 1, vary between two extreme cwes, suggests that we
should search for some appropriate quantity to measure the deviation among these curves
and study the behavior of such quantity as the length scale is varied. The quantity we
found most suitable is the area between a given curve and a reference curve. For
adsorption isotherms, this quantity, ;~b,is defined as

where @@) is the reference adsorption isotherm. A similar quantity, x,,, can be defined
for the heat of adsorption curves. By taking as a reference curve the one corresponding to
the bp topography, we obtain Figure 3 (a), where we can see that x,, behaves as a power
law in 1 with two different values of a, for regime I, and regime 11. Exactly the same
behavior is also found for n.
1 .o

0.2

L . I - . I . I . I . 11-20
-30 -20 -10 0 I0 20 0,O 0,2 0,4 0,6 0,s I,O
P e
Figure 1: Adsorption isotherm (a), and differential heat of adsorption (b) correspondingto regime 1.

-20 -10 0 10 20 0,O 0,2 0,4 0,6 0,s I,O

P 6
Figure 2: Idem Figure 1, for Regime 11.
For a given regime, a is the same for chessboard and random topographies. This
reinforce the idea that a random topography characterized by a scale length I behaves like
a chessboard topography with a larger scale length. In ref. [4] we have demonstrated that

637
chessboard and random topography curves for x (either xa or a)should become the same
curve as a h c t i o n of an effective length scale (representing an effective patch size), I,,
given by Ic-s.~, where s=l [s=2] for a chessboard [random] topography. MC results are
in agreement with this theoretical prediction (see Fig. 3).
, . I : I . -1. I . I

REGIME1 i i REGIME11

AE=4w ; AE=3w
(b) , ,\ , //, , , --7
A
1 10 0,l 0,2 0,3 0,4 0,5
kn w/AE
Figure 3: (a) Power-law behavior of showing the collapse of data for chessboard and random topographies
on a single curve for each adsorption regime. Filled [empty] symbols indicate chessboard [random]
topographies. (b) Universal behavior of a vs. w/AE for MC simulation.
The value of a is different for different adsorption regimes. By changing w and AE
we have also found a power law for intermediate regimes, obtaining for exponent a the
general behavior in the variable wlABrepresented in Figure 3 (b). We found that this can
be expressed as:
a=a,=-1.952f0.053; f o r w l A E 1 1 1 4
a = a 2=-3.049&0.065; forwIAE2113 (5)
a =az+ [12(113 - w l hE)lS(a,
- a , ) ; for 1 / 4 5 w l hE 1113; p = 0.42 f 0.04
Then, we can establish with great generality that x (mo or a),calculated by using any
suitable reference curve, behaves as a power law in the effective length scale le
Inx =consr-aInf,rr (6)
where the exponent a has a universal behavior given by equation (5). We stress that the
same universal behavior is obtained by using as a reference isotherm a suitable theoretical
approximation, for example, mean field.
Even if the scaling law, eq. (6), is valid in general, the values of the exponent given in
eq. (5) are valid only for b T = l . For different temperatures, we fmd the following results:
1) the value of a for regime I, al, does not change with T [hollow squares in Fig. 4 (a)];
2) the value of a for regime 11, a2,increases steadily as T increases approaching
asymptotically the value of al[hollow circles in Fig. 4 (a)]. The full line represents a
fitting curve to the Monte Carlo results. When the same variation is applied to the
transition region, the general behavior represented in Fig. 4 (b) is obtained.

4 Conclusions

The results suggest a method to solve the problem of the characterization of the
energetic topography of heterogeneous substrates which can be approximated by bivariate

638
surfaces, through adsorption measurements of particles experimenting repulsive
interactions. Given the surface, adsorption measurements that are strictly necessary are: i)
449, which can be obtained by using microcalorimetric techniques, and ii) the ad-ad
interaction energy, w,which can be obtained by LEED or STM measurements at different
temperatures to determine the critical temperature for the formation of the ordered ~(2x2)
structure. By using this information, the method can be summarized as: 1) since 4,40)=~~
and q,41)=~,+4w, it is possible to determine E], 2 and AE 2) Then, given w / U , the
value of Q can be obtained from figure 4. 3) Finally, by choosing an appropriate
theoretical approximation as a reference curve for @) [or 4,,(9], the value of x. [or a]
can be calculated allowing 1, to be obtained 60m eq.(6).

-1-
I
kT-2

AE=4w i AE=3w

-4,O
0,l 0,2 0,3 0,4 0,5
w/AE
Figure 4: Variation of universal exponent a with T. (a) Hollow squares, al(T) at fixed d E l 2 and w=2
(regime 1); hollow circles, e ( T ) at fixed e l 2 and w=5 (regime 11); full line fitting to the later. (b) General
diagram showing the behavior of exponent a for different regimes and different temperatures.
This method of characterization is being tested against adsorption isotherms on well
defined correlated substrates in computer experiments, before to extent it to real systems.

References

1. Steele W.A., The interaction of gases with solid surfaces (Pergamon Press, New York,
1974).
2. Rudziiski W. and Everett D., Adsorption of Gases on Heterogeneous Surfaces
(Academic Press, New York, 1992).
3. Zgrablich G., Zuppa C.,Ciacera M, Riccardo J. L. and Steele W. A., The Effect of
Energetic Topography on the Structure of the Adsorbate. Surface Sci. 356 (1 996)
257; F. Bulnes, F. Nieto, V. Pereyra, G. Zgrablich and C. Uebing, Energetic
Topography Effects on Surface Diffusion. Langmuir 15 (1999) 5990.
4.Bulnes F., Ramirez-Pastor A.J. and Zgrablich G., Scaling Behavior in Adsorption on
Bivariate Surfaces and the Determination of Energetic Topography. J. Chem. Phys.
115 (2001) 1513; Power Laws in Adsorption and the Characterization of
Heterogeneous Substrates. Acisorprion Sci. and Tech. 19 (2001) 229; Scaling Laws in
Adsorption on Bivariate Surfaces P&s. Rev. E, 65 (2002) 3 1603.
5. Nicholson D. and Parsonage N. D., Computer Simulation and the Statistical
Mechanics of Adsorption (Academic Press, London, 1982).

639
 ADSORPTION OF POLYATOMIC SPECIES: AN APPROACH FROM
QUANTUM FRACTIONAL STATISTICS

J. L. RICCARDO, A. J. RAMIREZ-PASTOR AND F. ROMA

Laboratorio ak Ciencias de Superjiciesy Medios Porosos, Departamento de Fkica, Universidad
Nacional ak San Luis. CONICET, Chacabuco 91 7,5700,San Luis, Argentina
E-mail:jlr@unsl. edu.ar

Multisite-occupancy adsorption is described as a Quantum Fractional Statistic problem. Site

exclusion is characterized by statistical exclusion parameter g which relates to the molecular size and
lattice geometry. The adsorption isotherm is obtained and comparisons with computer experiments
indicate that adsorption configuration and lateral interactions may accurately be assessed from this
theory.

1 Introduction

A major bottleneck in dealing with lattice gases of polyatomic species within the frame
work of classical statistical mechanics is to properly calculate the configurational entropic
contributions to the thermodynamic potential; this means the degeneracy of the energy
spectrum compatible with given number of particles and adsorption sites. In single-site
adsorption, this is usually done through the mean-field assuming that the particles are
distributed randomly over the sites regardless their mutual interactions. However, the
entropic effects in multisite-occupancy adsorption are by no means negligible. For
instance, the coverage dependence of configurational entropy of linear k-mers in one
dimension is strongly dependent on k [l] even for noninteracting particles. It is the fact
that a polyatomic particle excludes more that one state of the states available to the
remaining particles upon adsorption that makes the problem extremely difficult the
counting of number of configurations.
Quantum Fractional Statistics (QFS, also called General Exclusion Statistics)
initiated by Haldane [2], based upon a generalization of the Paulis exclusion principle
(generalized exclusion principle), has recently deserved much attention in quantum
systems since particle interactions can be regarded as statistical interactions
characterized by a parameter g which essentially accounts for the number of states (out of
the states available to a single particle) that a particle excludes when added to the system.
Fermions, bosons, and particles with fractional statistics correspond to g=1, g=O and
O<g<l, respectively.
In the present work we propose the extension of these ideas to classical polyatomic
lattice gases as to describe the thermodynamic properties of multisite adsorption and show
that its treatment can be significantly oversimplified from this perspective. In this case will
be shown that g lies in the range g>l. It is also shown that the adsorption isotherm of k-
mers on a one-dimensional lattice (recently presented in [3]) can be rigorously obtained if
the statistical parameter g equals the k-mers size (g=k). Application to adsorption of k-
mers of various geometries on two-dimensional lattices is carried out and compared with
MC simulations (adsorption isotherm and entropy). Results show that the approach
appears as a promising way of gaining insight into the thermodynamics of multisite-
occupancy adsorption and in the determination of gas-solid potential and the configuration
of adsorbed polyatomics from adsorption experiments.

640
2 The Monolayer Adsorption Theory of Polyatomic Lattice Gases from QFS

Quantum Fractional Statistics relies on the fundamental assumption [2] that AdP-gAN,
where Ad is the change in the Hilbert-space dimension of a single particle when a number
AN of particles is added to the system confined into a finite region of the space, and g
(exclusion statistical parameter1 accounts for the number of states excluded by a single
particle added to the system. If 6 denotes the total number of states available to a single
particle, then, I

d=G-g(N- 1) (1)
where d is the number of remaining states available to a single particle when N-1 particles
have been added (adsorbed) to the system (adlayer).
It is assumed here that g is not density-dependent. Although this is arbitrary, it is
necessary to make the problem analytically handable hereafter.
The grand partition function of identical polyatomic molecules is

where

and
(d+N-l)! - [G+l)W-l)]!
(4)
w(N)=
N!(d-N)! N![D-g(N-l>l]

Particularly, for monolayer adsorption on an homogeneous su-strate wiwut lateral

interactions, for which
E,(N) = &,N (5)
where E, denotes the adsorption energy per molecule, the distribution which maximizes eq.
(1) constrained to the above condition is [4]

where a3 must fhlfil w(1)8(1 +@(A))'-" = A = exp(&- p ) 1k,T .

For a linear k-mer adsorbed on an one-dimensional lattice of M sites with periodic
boundary conditions, it is straightforward that G=M, and g=k, yields the adsorption
isotherm obtained independently in Ref [3].
We analize now, the case of linear adsorbates of size k on a two-dimensional lattice
of connectivity c=4. Thus G=cM/2, n=2N/cM=N/2M. From eq. (4) a=l/n-g=2W0-g, if
0=lr", E,+. Accordingly

641
representing the adsorption isotherm of linear rigid species on a square lattice, where g
has been let as force parameter. In order to account for lateral interactions present in
experiments, we can, in first approximation, include them through a mean-field
contribution. Hence

w being the interacting energy between NN adsorbate's units.

3 Monte Carlo Simulation in Grand Canonical Ensemble

The adsorption process is simulated through a Grand Canonical Ensemble Monte Carlo
(GCEMC) method [4]. For a given value of the temperature T and chemical potential p,
an initial configuration with N dimers adsorbed at random positions (on 2N sites) is
generated. Then an adsorption-desorption process is started, where a pair of nearest
neighbor sites is chosen at random and an attempt is made to change its occupancy state
with probability:
P = rnin(1, exp[- AH / k,TB (9)

where AH=&-Hi is the difference between the Hamiltonians of the final and initial states.
A Monte Carlo Step (MCS) is achieved when M pair of sites have been tested to change
its occupancy state. The equilibrium state can be well reproduced after discarding the first
rn'=105-106MCS. Then, averages are taken over m=l@-106configurations.
Thermodynamic quantities, such as mean coverage 8 and average internal energy U
are obtained as simple averages:

where <N> is the mean number of adsorbed particles, and <...> means the time average
over the Monte Carlo simulation runs.
The computational simulations have been developed for square and honeycomb LxL
lattices, with L = 144 and periodic boundary conditions. With this lattice size we verified
that finite-sizeeffects are negligible.

4 Results and Discussion

Monte Carlo isotherms of interacting and noninteracting dimers on a square lattice have
been interpreted in terms of the new isotherm proposed in eq. (8). w and g were assumed
as fitting parameters of our analysis; and k=2 since data correspond to dimer adsorption.

642
 1.0-

0.8- / / 0 Attractive Dimers -

0.6- k
1 ' *''
p MC Simulation
e ' w/kj=0.00
k ," . w/k B T = -0.50
0.4- .' A w/k B T = -1.02
fa /
* ' . QF Statistics
j *! ..!// U T A AA -I- A A*>- o 1 1-(-AA1
0.2-
A w / kgT = -0.51 * 0.01; g ~ 3.74 0.01
j
n n . ..->'
1' ... / 1- T ft O"7 I ft ftl n ^ 7*> 1^ ft ftl

-5 10

Figure 1. Adsorption isotherms for homonuclear dimers adsorbed on a square lattice with attractive nearest
neighbor interactions. The symbols represent the results from MC simulations and the lines represent the results
from QF Statistics.
Figure 1 and 2 correspond to attractive and repulsive dimers respectively. For the
attractive case, which is the most interesting for experimental systems, it is observed that
the value of g obtained comes very close to 4, and w agrees very well with the one set in
the computer experiment, within a small statistical uncertainty, g = 4 is also consistent
with the number of states that an isolated dimer on a square lattice excludes to others.
1.0-

0.8-
.'A Repulsive
A
' Dimers
fi 0.6- MC Simulation
w/k B T = 0
w/k B T = 2
0.4- A w/k B T = 4
QF Statistics
0.2- -0.06 0.02; g = 3.83 0.01-
1.940.08; g = 4.010.01
w/k B T= 4.270.21; g = 3.960.07
0.0
10 20 30 40

Figure 2. As figure 1 for repulsive nearest neighbor interactions.

This results means that the adsorption isotherm within the framework of the present
theory may provide us of a reliable tool to obtain and interpret the adsorption
configuration and lateral interactions of polyatomic adsorbates.

643
 With regards to the repulsive case, the model adsorption isotherm reproduces the MC
data fairly well for w / k ~ T d The
. discrepancy in the case w/kT=4 links to the order-
disorder phase transitions that dimers develop on a square lattice at w k T , ' d , 8,'=1/2
and w/k~T:nr5, l:=20 (see ref. [5]). These transitions can not be reproduced by the
approximation (mean field) used in eq. (8). As it is clear from Figure 2, the simulated
isotherm for w/k~T=4display a plateau (characteristicof an ordered phase) at l=1/2 since
w/k~T,'<rl. However, only a knee around 8-20 is visible at this temperature because
w/k~T,34.A deeper discussion of the referred phase transitions can be found in ref. [ 5 ] .
The fact that the values of g and W ~ Bobtained
T from fitting of isotherms match very
well the theoretical values used in the simulations of the strong repulsive cases is
attributed to compensation effects and should not be considered as having much physical
meaning, provided the theoretical isotherm differs significantly from the MC one.
In summary, we have presented a new theoretical perspective of adsorption of
polyatomics. The approach simplifies the inherent complexity of this kind of adsorption
systems in a statistical exclusion parameter which hass a physical meaning traced to the
adsorption configuration on the lattice sites.
In addition, mean field approximation works unuasully well in the frame of the
present model, at least for the case of small polyatomics. Both adsorption configuration
and ad-ad interactions could be better known from the present theory.
Future studies already in progress address the analysis of experimental results and the
relationship between g and the configuration of the adsorbed particle on lattice sites.

References

I . Ramirez-Pastor A. J., Eggarter T. P., Pereyra V. and Riccardo J. L., Statistical Ther-
modynamics and Transport of Linear Adsorbates. Phys. Rev.B, 59 (1999) pp. 1 1027-
1 1036.
2. Haldane F.D.M., Fractional Statistics in Arbitrary Dimensions: A Generalization of
the Pauli Principle. Phys. Rev. Lett, 67 (1991) pp. 937-940.
3. Ramirez-Pastor A. J., Rom6 F., Aligia A. and Riccardo J. L., Multisite-occupancy
adsorption and surface diffusion of linear adsorbates in low dimensions: rigurous
results for a lattice gas model. Lungmuir, 16, (2000) pp.5100-5105; Rom6 F.,
Ramuez-Pastor A. J. and Riccardo J. L., Configurational Entropy in k-mer
Adsorption. Lungmuir, 16 (2000) pp. 9406-9409; Romh F., Ramirez-Pastor A. J. and
Riccardo J. L., Configurational Entropy for Adsorbed Linear Species (k-men). J.
Chem. Phys,114 (2001) pp. 10932.
4. Wu Y. S., Phys. Rev. Lett, 13 (1994) pp. 922.
5 . Ramirez-Pastor A. J., Riccardo J. L. and Pereyra V. D., Monte Carlo study of dimer
adsorption at monolayer on square lattices. Surjiuce Science, 41 1, (1 998) pp.294-302.

644
 MULTILAYER ADSORPTION WITH MULTISITEOCCUPANCY

F. R O I d , A. J. RAMIREZ-PASTOR AND J. L. RICCARDO

Laboratorio de Ciencias de Superjiciesy Medios Porosos, Departamento de Fisica, Universidad
Nacional & San Luis, CONtCET, Chacabuco91 7,5700, San Luis, Argentina
E-mail:[email protected]

A closed-form analytic solution for the thermodynamicsof multilayer adsorption of linear species on
one and two-dimensional surfaces is presented. The theoretical description is obtained by using a
new formalism, which provides the isotherm of multilayer adsorption from the adsorption isotherm
at monolayer. The validity of the approach is tested by comparison with Monte Carlo simulations.

1 Introduction

Among the various theories and models that have been proposed to describe multilayer
adsorption in equilibrium, the ones of Brunauer-Emmet-Teller (BET) [l] is the most
widely used and practically applicable. The great popularity of the BET equation in
experimental studies of adsorption led some authors to extend the original theory of
multilayer adsorption. Thus, numerous generalizations of the BET model have been
reported in the literature, including surface heterogeneity, lateral interaction between the
admolecules, differences between the adsorption energy and structure between the first
and upper layers, etc, [2,3].These leading models, along with much recent contributions,
have played a central role in the characterization of solid surfaces by means of gas
adsorption [4]. One fundamental feature of BETS model is preserved in all these theories.
This is the assumption that an adsorbed molecule occupies one adsorption site. In a recent
work [ 5 ] , an improved solution for multilayer adsorption with multisite-occupancy on
one-dimensional substrates was presented. The particular case of multilayer dimer
adsorption was dealt with in detail and a new adsorption isotherm was obtained for
determining of surface area and adsorption energy from experiments.
In the present paper, the calculations have been extended to k-mers in one and two-
dimensional surfaces, based upon the Occupation Balance Approximation [6]. The
theoretical results are compared with simulated data. The proposed model is simple, easy
to apply in practice, and leads to new values of surface area and adsorption heats.
Physically, these advantages are a consequence of properly considering the
configurational entropy of the adsorbate.

2 Model and Numerical Procedure

In order to maintain the simplest model that accounts for multisite-occupancy in

multilayers we define it in the spirit of the BETS original formulation. The adsorbent is a
homogeneous lattice of sites. The adsorbate is assumed as linear molecules having k
identical units (k-mers) each of which occupies an adsorption site. Furthermore, i) a k-mer
can adsorb on empty sites in the first layer or exactly onto an already adsorbed one; ii) no
lateral interactions are considered; iii) the adsorption heat in all layers, except the first one,
equals the molar heat of condensation of the adsorbate in bulk liquid phase. Thus,
c==I/qi=qr/q with qi=q (i=2,..,o) denotes the ratio between the single-molecule partition

645
functions in the first and higher layers. The fact that k-mers can arrange in the first layer
leaving sequences of 1 empty sites with l<k, where no fiuther adsorption of a k-mer can
occur in such a configuration, makes the calculation of entropy much elaborated than the
one for monomer adsorption.
For a lattice having M adsorption sites, the maximum number of columns that can be
grown up onto it is n-=M/k. The grand partition function, &,, of the multilayer in
equilibrium with a gas phase at chemical potential p and temperature T,is given by

n =O

where I(Lk(n,M) is the total number of distinguishable configurations of n columns on M

sites and 6 is the grand partition function of a single column of k-mers having at least one
k-mer in the fmt layer. Then,
m m
{=Cq,q'-'& =cCqi'L cR,q
=-=- cx
i=l i=l l-',q I-x
where &=t?xp(CJkBT) is the fugacity and kB is the Boltzmann constant. In addition, it is
possible to demonstrate that x=&q =p/po is the relative pressure [4,5].
On the other hand, the grand partition function of monolayer, El, can be written as

n=O

in this case, n represents the number of k-mers and X1 is the monolayer fugacity.
The substitution (=Ilshows that Z,,, can be transformed to El, then

From the condition in eq. (4), we can write the monolayer coverage, el,as

where T and fi are the temperature and the mean number of k-mers (columns) at
monolayer (multilayer), respectively. From eq. (2) and (9,the total coverage, 8, results

where R is the total number of adsorbed k-mers at multilayer.

In summary, the numerical procedure can be descripted in two steps: 2) by using el as
a parameter (0 5 el S l),the relative pressure is obtained by using eq. (4). This calculation
requires the knowledge of an analytical expression for the adsorption isotherm of k-mers
in the first layer at monolayer regime; and 2) the values of el and p/po are introduced in eq.
(6)and the total coverage is obtained.

646
 Following this scheme, the exact solution for multilayer adsorption of k-mers on a
lattice in 1-D can be obtained. We start h m the equation

which represents the one-dimensional exact isotherm of k-men at monolayer [7].

Substitutingeq.(7) into eq.(4), one obtains the expression of the relative pressure,

k-1
Po ck(l-Bl)k + B l [ I - ( ~ ) B 1 ]

The eqs. (6) and (8) represent the exact solution describing the adsorption of k-mers
at multilayer regime on a homogeneous surface in I -D. In the case of monomer adsorption
(when k=l), the equations (6) and (8) reduce to the well-known BET isotherm [l]. For
k=2, the dimer isotherm can be written in a simple form:

3 Monte Carlo Simulation in Grand Canonical Ensemble

Adsorption of k-mers in multilayer regime was simulated following the Metropolis

scheme. In this framework, the transition probability fiom a state i to a new state j, W( i
+ j), is defined by W( i + j)=min{l,exp[-~(AU-~)J},where B=l/kBT and (AN)AU
represents the variation in the (number of particles) total energy, when the system changes
fiom the state i to the state j.
In adsorption-desorption equilibrium there are four ways to perform a change of the
State, namely, adsorbing one molecule onto the surface wad?", desorbing one molecule
from the surface Wdegud,adsorbing one molecule in the adsorbed liquid phase (2ndand
upper layers) Wadrbulk, and desorbing one molecule fiom the liquid phase wdwbu'k.
In the first cases, W.~'"d=min{l,exp~-B(Ul-p)]~and W~,""''=min{1,exp[~(U1-p)]},
where U1 is adsorption energy of one molecule in first layer. In addition, exp[-B(U1-
p)]=qlA, q l s q and qA=p/p,. Then, W,~8ud=min{l,cp/po}and Wdusud=min{l,pJ(cp)}.
on the other hand, Wadrbuik =min{l,exp[-p(~-p)1} and WdabUik =min{l,expIp(U-p)j),
where U is adsorption energy of one molecule on the i-th layer with i22. In addition, exp[-
B(U-p)]=qh=p/p,. Then, Wadrbulk =min{l,p/p,} and WdabUlk =min{l,pJp}.
The elementary step in Monte Carlo simulation (MCS) is analogous to the standard
insertion deletion procedure, constrained in this case, to that k-mers can only adsorb on
empty sites of the first layer or onto an already adsorbed one.
The equilibrium state is well reproduced after discarding the fmt lo4 MCS. Then,

647
averages were taken over lo4 MCS succesive configurations. The total coverage was
obtained as simple averages, B=k(N)/M, where (N) is the mean number of adsorbed
particles, and (...)means the time average over the Monte Carlo simulation runs.

4 Results and Discussion

In figure 1 a) we address the comparison between the analytical adsorption isotherm in

one dimension and Monte Carlo simulation. The simulations have been performed for
monomers, dimers and 10-mers adsorbed on chains of M/k =I000 sites with periodic
boundary conditions. Different values of the parameter c have been considered. In all
cases, the computational data fully agree with the theoretical predictions, which reinforce
the robustness of the two methodologies employed here.

e
. -
0.005- Arlnonporocuskka(77K)
c (23.4t)1.6

-
N, I non porour sklka (77 K)
c (1063 t)17.8
v. = (40.8 t 0.6) em' 0.'

-
,

nO.OO0, . , . , . , . ,
0.0 0.2 0.4 0.6 0.8 1.0 0.00 0.05 0.10 0.15 0.20
P 1 Po P 1 Po
Figure 1. a) Adsorption isotherms for k-mers adsorbed on I-D with different values of the parameter c. Lines
correspond to exact theoretical results and symbols represent data from Monte Carlo simlation. b) Fitting of
experimental adsorption isotherms of At, Nz / nonporous silica, through the linearized dimer isotherm equation
(9). The resulting values of the parameter c and the monolayer volume v, are shown in the figure.
Although a detailed analysis of experimental data is out of the scope of the present
work, fitting of standard data for the systems Ar,N2 / nonporous silica is shown in figure 1
b). We have adjusted the values of the parameter c and the monolayer volume v, (being
k v / v m )from the proposed adsorption isotherm for dimers, which differ appreciably from
the BET values (namely, ~ 1 3 5 . 2 v,=37.7
; for N2 and ~ 4 1 . 5 v,=32.5
; for Ar). The
larger values obtained for c in the BET case trace to the compensation arising in the BET
(monomer) model because of its larger entropy with respect to the dimer (k-mer) case.
The larger values of v, arising from the dimer case may appear more controversial,
although it results straightforwardlyfrom the model.
Regarding with k-mer multilayer adsorption in two dimension, we propose an
approximation to the adsorption isotherm based on the recently developed Occupation
Balance approach (OB) [6].In the simplest case of k=2, the monolayer isotherm for non-
interacting dimers adsorbed on a square lattice can be written as,
4 9 3 2
2;' =- -7+ - 8, + -6,
el 4 4
Analogous adsorption isotherms result for honeycomb and triangular lattices. The relative
pressure for the square lattice is obtained inserting eq. (1 0) into eq. (4).

648
 P
-= 48, (1 1)
Po 16c+(4-28~)8,+9c@ +3c@
The eq. (1 1) and similar for other connectivities provide a theoretical solution to
multilayer of dimers on lattices in 2-D. This treatment, in which the entropic effects of the
adsorbate size are accounted for, bears theoretical interest because it represents a
qualitative advance with respect to the existing models of multilayer adsorption.
As shown in figure 2 the lattice geometry affects appreciably the multilayer regime
for weakly adsorbing surfaces (-1, 10). It is expected that the effect increases with the
adsorbate size. Additionally, MC simulation agrees very well in the whole pressure range
with the proposed approximation.

0 0

- 1rIanp-r W e

0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 0.5 0.6
PIPo PIPo
Figure 2. Adsorption isotherms for dimers adsorbed on two-dimensional lattices with different values of the
parameter c. The theoreticai results for different connectivities are shown at left. A comparison between
theoretical (lines) and simulated (symbols) results for dimers adsorbed on square lattices is presented at right.
In summary, an analytical approach to the multilayer adsorption isotherm of
polyatomic adsorbates (k-mers) in one and two-dimensional lattices has been proposed.
Monte Car10 simulation support the high accuracy of the approximation. It arises that
adsorbate size as well as lattice structure influence multilayering.

References

1. Brunauer S., Emmet P. H. and Teller E., Adsorption of Gases in Multimolecular

Layers. J. Am. Chem. SOC.60 (1938) pp. 309.
2. Gregg S. J. and Sing K. S. W., Adsorption, Surface Area, and Porosity (Academic
Press, New Y ork, 1991).
3. Steele W., The Interaction of Gases with Solid Surfaces (Perg. Press, Oxford, 1974).
4. Rudzinski W. and Everett D. H., Adsorption of Gases on Heterogeneous Surfaces
(Academic Press, New York, 1991).
5. Riccardo J. L., Ramirez-Pastor A. J. and Romh F., Multilayer Adsorption with
Multisite Occupancy: An Improved Isotherm for Surface Characterization. Lungmuir,
18 (2002) pp. 2130.
6. Romh F., Ramirez-Pastor A. J. and Riccardo J. L., Configurational Entropy for
Adsorbed Linear Species (k-mers). J. Chem. Phys, 114 (2001) pp. 10932.
7. Ramirez-Pastor A. J., Eggarter T. P., Pereyra V. and Riccardo J. L., Statistical Ther-
modynamics and Transport of Linear Adsorbates. Phys. Rev. B, 59 (1999)pp.11027.

649