Communication

Cyclodextrin Membranes www.advmat.de

Cyclodextrin Films with Fast Solvent Transport

and Shape-Selective Permeability
Luis Francisco Villalobos, Tiefan Huang, and Klaus-Viktor Peinemann*

using block copolymer self-assembly,[7]

This study describes the molecular-level design of a new type of filtration mem- and synthesizing polymers of intrinsic
brane made of crosslinked cyclodextrins—inexpensive macrocycles of glucose, microporosity.[8] Efforts have also been
shaped like hollow truncated cones. The channel-like cavities of cyclodextrins made to control the size distribution of
spawn numerous paths of defined aperture in the separation layer that can the free-volume elements (i.e., cavities) by
inducing thermally driven segment rear-
effectively discriminate between molecules. The transport of molecules through
rangements.[5] We propose an approach
these membranes is highly shape-sensitive. In addition, the presence of hydro- that uses an intrinsically porous macro-
phobic (cavity) and hydrophilic (ester-crosslinked outer part) domains in these cycle as the main building block of the
films results in high permeances for both polar and nonpolar solvents. membrane’s selective layer to increase
the free-volume and to achieve certain
control over the size and shape of its free-
Energy-intensive molecular separations often impose the heav- volume elements. Each macrocycle supplies the matrix with a
iest economical burden and the biggest environmental foot- free-volume element of fixed dimensions. Among the available
print of industrial processes. However, the energetic demands intrinsically porous building blocks to choose from,[9] cyclodex-
of these separations can be reduced considerably by using trins (CDs) are particularly attractive because their large-scale
high-performance membranes rather than a phase change pro- production from starch is sustainable,[10] they are nontoxic[11]
cess.[1] Energy-efficient separations require highly permeable and they have already proven to be useful agents for a wide
membranes with rigorous selective properties in a targeted size variety of separations.[12]
range.[2] The performance of these membranes can be maxi- The preparation of cyclodextrin-based crosslinked polymers
mized by choosing the appropriate material and controlling the has been going on for more than 50 years.[13,14] Surprisingly,
porosity and thickness of the selective layer. the fabrication of a membrane made entirely of cyclodextrins
Unlike rigid microporous materials, such as metal-organic has remained elusive. Rölling and co-workers attempt to do
frameworks, zeolites, porous coordination polymers, and cova- this by radical polymerization of acrylated cyclodextrins yielded
lent organic frameworks that have defined cavity shapes and membranes that were too brittle for use.[15] The flexibility of the
sizes, amorphous polymers contain a wide range of cavity sizes membranes was rescued by copolymerizing cyclodextrins with
and topologies.[3] Nevertheless, the ease of processing and eco- flexible polymers, but this substantially decreased the concen-
nomical cost of these polymers has positioned them as the tration of cyclodextrin in the active layer, diluting their effect.
workhorse materials of the membrane industry. Their perfor- Yue et al. developed an ingenious approach with α-CD as a
mance in small-molecule separations is governed by the size, directing agent in the formation of single-layer ionic organic–
distribution, and connectivity of the disordered voids—free inorganic 2D frameworks.[16] The cavity of the α-CDs was used
volume—that occur in them naturally. Ideally, the free volume as a cap for the “stick” cationic molecules used to build the
of polymeric materials should be permanent and intercon- frameworks. Such layers, when assembled into membranes by
nected and have a narrow aperture-size distribution; however, vacuum filtration, showed a precise size-selectiveness in the
controlling these properties has proven to be challenging.[4,5] ultrafiltration range. To date, cyclodextrins have been incor-
Successful strategies to increase the free volume of and create porated into membranes as fillers, as part of a membrane-
more interconnected voids within these membranes have forming polymer, and as surface modifiers (see Table S1 in the
focused on preventing the efficient packing of polymer chains Supporting Information). This letter highlights how the careful
in the solid state by adding filler particles as “nanospacers,”[6] choice of fabrication conditions enables the manufacture of
thin, defect-free, solvent and mechanically stable membranes
made entirely of crosslinked cyclodextrin molecules.
L. F. Villalobos, Dr. T. Huang, Prof. K.-V. Peinemann We formed thin films of cyclodextrin by interfacial poly­
King Abdullah University of Science and Technology (KAUST) merization, a technique used widely in the membrane industry
Advanced Membranes and Porous Materials Center to make thin-film composite polyamide membranes for water
Thuwal 23955-6900, Kingdom of Saudi Arabia desalination and nanofiltration. The commercially most suc-
E-mail: [email protected]
cessful reverse osmosis membrane for water desalination is
The ORCID identification number(s) for the author(s) of this article
can be found under http://dx.doi.org/10.1002/adma.201606641.
made by an interfacial reaction using an aqueous amine solu-
tion and a solution of trimesoyl chloride in a hydrocarbon.
DOI: 10.1002/adma.201606641 Here, we used a hydrocarbon-based terephtaloyl chloride (TC)

Adv. Mater. 2017, 29, 1606641 1606641 (1 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 1. β-CD films made by interfacial polymerization on top of PAN porous supports. a) Synthesis of the films from β-CD (dissolved in the aqueous
phase) and terephthaloyl chloride (dissolved in the organic phase). b) SEM image of a β-CD film made with 6.5% (w/v) β-CD in a 1 m NaOH solution
(aqueous phase), 1% (w/v) TC in hexane (organic phase), for 10 min reaction time.

solution and cyclodextrins in the aqueous phase instead of essential to use NaOH concentrations of 0.1 m or more to form
the conventional diamines (e.g., m-phenylenediamine, p-phe- the β-CD films.
nylenediamine, piperazine). This approach presents several Surprisingly, the morphology of the film was considerably
challenges: (i) cyclodextrins are much larger and bulkier than affected by the concentration of NaOH used: 0.1–0.8 m pro-
diamines,[17] (ii) the hydroxyl (OH) groups of cyclodextrins are duced flat, smooth films, and 1.0 m or higher produced crum-
less reactive than amines, and have different reactivities among pled films (Figure S3, Supporting Information). We believe that
them,[18] and (iii) cyclodextrins tend to agglomerate in aqueous two processes are occurring at the same time—formation and
solutions.[19,20] Nevertheless, by optimizing preparation condi- hydrolysis of ester bonds—and that the balance between them
tions, we produced continuous β-CD films on top of a commer- determines the morphology of the resulting film. The ester link-
cial polyacrylonitrile (PAN) porous support (Figure 1). ages, formed during interfacial polymerization, can be readily
The chemical reaction rate and the transfer of reactants to hydrolyzed by hydroxide ions. For example, an aqueous solu-
the polymerization zone are important factors during interfacial tion of pH 13.0 or higher was sufficient to immediately destroy
polymerization,[21] as they can substantially influence the mor- a fully formed β-CD film, while a solution of pH 11.0 decom-
phology of the resulting film.[22,23] When dealing with nanom- posed it slowly. Although the alkaline conditions used in the
eter-sized molecules, such as cyclodextrins, diffusion is slow aqueous phase of the interfacial polymerization have the poten-
and could be the limiting step if precautions are not taken. To tial to destroy the ester links that hold the film together, the low
obviate this limitation and to promote a constant and homoge- aqueous solubility of the acid chloride prevents the film from
neous availability of cyclodextrins in the reaction zone, we used forming in the aqueous phase. Hydrolysis of ester bonds can
high concentrations of cyclodextrins and a hydrophilic support only occur along the bottom of the film and the rate is propor-
with high surface porosity and small pore-size distribution. We tional to the concentration of hydroxyl ions in the aqueous solu-
found that the use of lower concentrations of cyclodextrins or tion stored inside the pores of the PAN support. Slow hydrolysis
hydrophobic supports with more surface pore heterogeneity led rates allow for the formation of a continuous highly crosslinked
to defective films (see Section B in the Supporting Information). β-CD film along the top of the support. Once formed, the film
A key parameter in the fabrication of cyclodextrin mem- hinders further diffusion of cyclodextrins to the reaction zone.
branes was the pH of the aqueous phase, which was controlled Alternatively, high hydrolysis rates prevent the formation of a
by addition of sodium hydroxide (NaOH). Hydroxyl groups of continuous β-CD film on the support, and instead it forms away
β-CD start to slowly deprotonate above pH 12.0 and do so more from the surface, in some parts of the membrane, generating
rapidly above pH 13.0.[24] Ionization of OH groups at high pHs a crumpled structure. The top part of crumpled films suffers
avoids aggregation of cyclodextrins,[20] forms the more reactive less hydrolysis of ester bonds because it is further away of the
alkoxide ions,[18,21] and increases their solubility in water. Each aqueous phase during their fabrication. Consequently, crumpled
glucopyranose unit in a cyclodextrin molecule has two sec- films must have a higher crosslinking degree at their top part.
ondary hydroxyl groups (wide face) and one primary hydroxyl X-ray photoelectron spectroscopy (XPS), and Fourier transform
group (narrow face). The primary OH groups are more reac- infrared (FTIR) analysis of the top and bottom part of an iso-
tive, however the secondary ones deprotonate easier. At suffi- lated crumple β-CD film supported this hypothesis. Both, XPS
ciently high pHs, the acid chloride attacks the hydroxyl groups (Figure 2c–f) and FTIR (Figure S11, Supporting Information)
of the cyclodextrins indiscriminately.[18] For these reasons it was analysis, showed clear differences between the bottom and top

Adv. Mater. 2017, 29, 1606641 1606641 (2 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 2. Cross-section and internal structure of a β-CD membrane. The membrane was prepared by interfacial polymerization of β-CD (6.5%, w/v)
in a 1 m NaOH aqueous solution with TC (1%, w/v) in hexane. a) Cross-sectional SEM. Inset: enlargement of an area encompassing the interface of
the β-CD film with the PAN support. b) TEM image of the β-CD membrane’s cross-section. Inset: enlargement of an area encompassing the active
layer of the β-CD film (thickness of 180 ± 55 nm); the higher contrast of the active layer is because the material is denser in that area. Photoelectron
spectra of the crumpled β-CD isolated from the PAN support and placed on top a silicon wafer: c) C1s of bottom part; d) C1s of top part; e) O1s of
bottom part; f) O1s of top part. The groups that could be attributed to each C peak are CC, CC, and CH for C1, CO, and CN for C2, OCO
for C3, and OCO for C4.

Adv. Mater. 2017, 29, 1606641 1606641 (3 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

part of the same β-CD film indicating an asymmetry in its struc- of molecules through them was very fast—water and methanol
ture. Using data from the O1s spectra (Figure 2d,e) it is pos- permeances of up to 20 and 9.4 L m−2 h−1 bar−1, respectively,
sible to estimate the number of TC molecules per β-CD (refer were obtained. Further analysis of a cross-section of the β-CD
to Section C.2 in the Supporting Information for details on the film using transmission electron microscopy (TEM) showed
estimation). For the bottom part it ranges from 1.6 to 4.1, and that it comprises a dense layer of 180 ± 55 nm and a less dense
for the top part from 2.9 to 7.4. A higher number of TC units substructure (Figure 2b). The lower density substructure may
per β-CD are an indication of a higher crosslinking degree. be the result of a greater degree of hydrolysis of ester crosslinks
Before interfacial polymerization occurs, some of the hydroxyl during its formation. On the contrary, flat β-CD films presented
ions present in the aqueous phase are consumed to ionize β-CD a uniform density throughout; no difference in contrast was
molecules and to partially hydrolyze the PAN support.[25] For observed in their TEM images (Figure S4b, Supporting Infor-
this reason NaOH concentrations of 1 m or more were needed mation). Crumpled films showed up to 47 times higher solvent
for sufficient presence of hydroxyl ions inside the pores of the permeances than did flat films (Figure 3a) partly because of the
support to hydrolyze the β-CD film at the rate necessary to larger surface area of the crumpled films but also because of
achieve the desired crumpled morphology. the greater number of hydrolyzed ester linkages in the film.
The internal structure of the β-CD crumpled films is very Molecules can travel either through the cyclodextrin cavity
interesting. Scanning electron microscopy (SEM) revealed a or through the space in between cyclodextrin molecules com-
cyclodextrin film with variable thickness in the micrometer prising crosslinker molecules covalently attached to one or two
range (Figures 2a and 3d). Contrary to what was expected cyclodextrins, carboxylic acid groups produced as a result of
from membranes with such a thick active layer, the transport ester hydrolysis, and unreacted primary and secondary hydroxyl

Figure 3. Performance of β-CD membranes. a) Pure solvent (water and methanol) permeances of five β-CD membranes prepared using different NaOH
concentrations in the aqueous phase. The rest of the preparation conditions were the same: 6.5% (w/v) β-CD in aqueous phase, 1% (w/v) TC in hexane,
for 10 min reaction time. Membranes prepared with 0.1 and 0.8 m NaOH were flat and smooth, while higher NaOH concentrations produced crumpled
films. b) Photograph of an isolated β-CD film floating in water. c) Pure solvent permeances for eight solvents through the crumpled β-CD membrane
prepared using 1 m NaOH aqueous solution. d) Schematic showing how a β-CD membrane separates molecules based on their shape. Cross-section
SEM image corresponds to a β-CD membrane prepared using 2 m NaOH aqueous solution. e) UV–vis absorption spectra of a methanol solution with
PPIX (orange molecule) and RB (pink molecule) to evidence the separation performance of the β-CD membrane.

Adv. Mater. 2017, 29, 1606641 1606641 (4 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

groups. A greater degree of hydrolysis benefits the transport of approximation using ferrocene methanol as a model adsorbate
molecules through the β-CD film in two ways: more open path- revealed 98% ± 8% of cavities were accessible in the film
ways in between cyclodextrin molecules and easier access to (Figure S12, Supporting Information). Moreover, CO2 adsorp-
cyclodextrin’s cavity. Using conditions that favor ester hydrolysis tion experiments at 273 K revealed a characteristically steep
during fabrication can considerably increase permeance without uptake at low relative pressures, indicating a high-degree of
changing the morphology of the β-CD film. For example, two flat microporosity (Figure 4b). Nonlocal density functional theory
films prepared with either 0.1 or 0.8 m NaOH solution caused (NLDFT) calculations applied to the isotherm indicated that
permeances differing by an order of magnitude (Figure 3a). An the majority of micropores in the β-CD film had an aperture
XPS analysis performed to estimate the number of TC units of 0.5–0.6 and 0.8 nm (Figure 4c), which is consistent with
per β-CD and the degree of ester hydrolysis of each of the β-CD that reported for the narrow (0.61 nm) and wide (0.79 nm)
films is available in Section C.2 in the Supporting Information. cavity apertures of β-CD[17] (Figure S14, Supporting Informa-
Flat β-CD membranes were outperformed by crumpled mem- tion). The surface area of the isolated β-CD film was calcu-
branes, and among them, the one prepared using a 2 m NaOH lated at 160 ± 4 m2 g−1 with the Langmuir model. This value
solution showed the highest permeance (Figure 3a). However, is comparable to results from Jimenez-Solomon et al., who
this membrane contained extensive hydrolyzed ester linkages, prepared microporous poly­ arylate nanofilms and reported a
rendering it unstable at high pressures: It collapsed at pressures Langmuir surface area of 79–90 m2 g−1 for polymers made by
above 5 bar. Decreasing the NaOH concentration to 1 m pro- interfacial polymerization and 150–160 m2 g−1 for those made
duced a high-permeance crumpled β-CD film that could with- by monophasic reaction[26] (calculated from CO2 adsorption
stand pressures of more than 20 bar. SEM images of a β-CD film experiments).
after operating for more than 10 h at 20 bar showed no change The transport of solvent molecules through the β-CD mem-
in its structure, neither collapse of the crumples (Figure S8c,d, branes is determined by the size of the solvent molecules and to
Supporting Information). However, a decrease in permeance a lesser extent by the viscosity and polarity of the solvent. Both
was observed during long-term operation. The methanol perme- polar and nonpolar solvents permeate through the β-CD film at
ance dropped around 60% when using either 4 bar (membrane high rates due to the presence of both hydrophilic and hydro-
tested for 50 h, Figure S8a in the Supporting Information), or phobic domains within it. Data for the successive permeation
20 bar (membrane tested for 13 h, Figure S8b in the Supporting of several solvents through a crumpled β-CD membrane are
Information) as the driving force. available in Figure 3c. Integrity of the membrane throughout
The β-CD film prepared using 1m NaOH was isolated by the experiment was confirmed by testing its rejection of methyl
dissolving the PAN support (Figure 3b) and studied it in more orange (molecular weight 327.3 g mol−1, size 0.86 nm3, nega-
detail. The isolated film expanded less than 10%, remained in tive charge) in different solvents. We measured a 90% or
one piece, and maintained its crumpled structure (Figure S2, greater rejection for all solvents in which methyl orange was
Supporting Information). It was robust and flexible, showing soluble (water, acetonitrile, methanol, ethanol, and tetrahydro-
no visible tearing or fragmentation after several successive furan). This translates to permeances at least an order of mag-
aspirations through a pipette (Video S1, Supporting Informa- nitude higher than that of commercially available membranes
tion). Moreover, it was thermally stable up to more than 250 °C (see Table S2 in the Supporting Information) and comparable
in a nitrogen atmosphere (Figure 4a). The number of acces- to recently published results of state-of-the-art thin-film com-
sible β-CD cavities in this film is closely related to the perfor- posite membranes used for solvent resistant nanofiltration (see
mance of the membrane because molecules can only transport Table S3 in the Supporting Information). Unlike these mem-
across cyclodextrins with an uncovered cavity. A conservative branes, however, that rely on a very thin active layer to achieve

Figure 4. Characterization of the isolated β-CD film. The analyzed β-CD film was made with 6.5% (w/v) β-CD in a 1 m NaOH solution (aqueous phase),
1% (w/v) TC in hexane (organic phase), for 10 min reaction time. To obtain enough material for the analysis, we isolated β-CD films from 20 membranes
prepared with the same conditions. a) Thermogravimetric analysis (TGA) curves for β-CD crosslinked film and pure β-CD. b) CO2 sorption at 273 K
for the isolated β-CD film. c) NLDFT-analyzed pore-size distribution based on carbon slit-pore geometry. The cumulative surface area of all the pores
up to 1 nm approximated by NLDFT was 75 m2 g−1.

Adv. Mater. 2017, 29, 1606641 1606641 (5 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

high permeances, β-CD membranes perform similarly despite (Sigma-Aldrich) solution for 10 min. Finally, the membrane was washed
having an active layer of at least an order of magnitude thicker with hexane to remove the unreacted TC, dried for at least 12 h at room
temperature, and then rinsed and stored in Milli-Q water. The Teflon
(detailed comparison available in Section D in the Supporting
frames used for this work produced square membranes of 7 cm × 7 cm
Information). β-CD membranes have permeances in the same that were cut into circles of 1 or 2 in. to fit either the Millipore plastic cell
range of thin polyamide,[23] polyarylate,[26] or polyelectrolyte or the Sterlitech stainless steel cell used in the permeation experiments.
bilayer[27] films. And rejections of methyl orange of 91%–95%, Control experiments were made with no monomer in both aqueous and
compared to 96%–99% of methyl orange for polyamide thin organic phases, no monomer in the aqueous phase and no monomer
films, 97%–99% of crystal violet for polyarylate thin films, and in the organic phase. Detailed information of the optimization of
95%–96% of acid fuchsin for polyelectrolyte bilayer thin films. the preparation conditions is available in Section B in the Supporting
Information.
The most distinctive feature of the prepared β-CD mem- α-CD and γ-CD membranes were made with cyclodextrins obtained
branes is their ability to discriminate small molecules based from Tokyo Chemical Industry.
on their shape. Only molecules with a kinetic diameter smaller Characterization Methods: SEM images were obtained using a FEI
than the narrow aperture of β-cyclodextrin’s cavity (≈0.61 nm) Nova Nano or a FEI Quanta 200. Cross-section samples were prepared
were able to transport across it. Additionally, carboxylic acid by freeze fracturing the membranes in liquid nitrogen. Samples were
groups, produced by the hydrolysis of ester linkages or acid coated with a thin film of Ir or Au/Pd to reduce charging during imaging.
TEM images were obtained using a FEI Titan 80-300 CT operated at
chloride groups, hinder negatively charged molecules from
300 kV. To prepare cross-section samples for TEM, the membranes
entering the cavities. Small, negatively charged molecules that were embedded in a low-viscosity epoxy resin (Agar R1165) and
fit inside the cavity of β-CD were rejected, while molecules of sectioned with a Leica EM UC6 ultramicrotome. TEM samples of an
similar size but neutral passed freely (Figure S9, Supporting isolated cyclodextrin film were prepared by dissolving the PAN support
Information). Molecules that cannot enter the cavity have to in dimethyl sulfoxide (DMSO) and recovering the remaining β-CD film
find possible pathways through the available spaces in between with a copper grid. XPS measurements were carried out in a Kratos Axis
Ultra DLD spectrometer equipped with a monochromatic Al Kα X-ray
the cyclodextrin molecules. For example, two neutral mol-
source (hν = 1486.6 eV) operating at 150 W. The data were analyzed with
ecules of similar molecular weight, protoporphyrin IX (PPIX; CasaXPS software.
molecular weight 562.7 g mol−1, approximate molecular width To characterize properties of the β-CD film alone, the authors isolated
of 1.5 nm[28]) and rhodamine B base (RB; molecular weight sufficient material by dissolving the PAN support of over 20 membranes
442.6 g mol−1, approximate molecular width of 0.6 nm,[29]) can with DMSO, transferred the remaining β-CD films to a water bath,
be effectively separated by a β-CD membrane (Figure 3d,e). washed them with methanol, and dried them at 60 °C under a vacuum.
Permeating a mixture of these two molecules in methanol Thermal analysis of the isolated β-CD films was performed with a
thermogravimetric analyzer (TGA) Q5000 from TA Instruments under
resulted in a rejection of only 3% of the cylindrically shaped a N2 atmosphere from room temperature to 700 °C at a heating rate
RB molecules while 90% of the spherically shaped PPIX were of 5 °C min−1. Low-pressure CO2 sorption experiments on the isolated
retained. This discriminate shape selectivity has also been veri- β-CD films were performed in a Micromeritics ASAP 2020. Samples were
fied for larger molecules: A molecule known for its ability to dried at 60 °C under a vacuum for 48 h and degassed at 120 °C for 20 h
thread cyclodextrin molecules[30] (polyethylene oxide, average before measuring the CO2 sorption isotherm at 273 K. Quantification
molecular weight 1500 g mol−1) was able to freely cross the of ferrocene methanol (Sigma-Aldrich) adsorbed by an isolated β-CD
film was done by immersing 5 cm2 of the film in a 200 ppm ferrocene
β-CD film while a molecule of similar molecular weight but
methanol in MeOH solution. This experiment was repeated three
a shape unconducive with that of the cavity (γ-CD, molecular times. The ferrocene methanol solution was prepared by first dissolving
weight 1297.1 g mol−1) was substantially rejected (Figure S7c,d, ferrocene methanol in 1 mL of acetone (Sigma-Aldrich) and then
Supporting Information). diluting the resulting solution with 199 mL of MeOH. The concentration
Although we focused on β-CD, similar films could also be of ferrocene methanol was monitored by ultraviolet–visible (UV–vis)
fabricated with other common cyclodextrins, such as α-CD or spectroscopy using a NanoDrop 2000c spectrophotometer.
Nanofiltration Performance: It was evaluated in terms of permeance of
γ-CD (Figure S13, Supporting Information). By following a
water or organic solvents and rejection of Na2SO4 (Sigma-Aldrich) or dye
well-established process of fabrication combined with the low molecules (methyl orange from Acros Organics, amido black 10B from
cost and commercial availability of all the materials used, we MP Biomedicals, congo red from Sigma-Aldrich and brilliant blue
demonstrated the feasibility of up-scaling this type of high- R250 from Fisher Scientific), respectively. Nanofiltration experiments
performance membrane. We hope this work will inspire the were carried out in a dead-end configuration using either Millipore stirred
use of cyclodextrins and other cavitands[31] as central building cells (model 8010, effective membrane area of 4.1 cm2) or high-pressure
blocks for novel membranes required for tackling challenging Sterlitech stainless steel cells (model HP4750, effective membrane
area of 14.6 cm2). All experiments were done at room temperature
molecular separations. and repeated three times. Data for consecutive solvent runs presented
in Figure 3c were obtained using Sterlitech stainless steel cells with a
driving force of 10 bar; the rest of the experiments were performed in
Millipore stirred cells with a driving force of 4 bar. All membranes were
Experimental Section preconditioned in the solvent to be measured by permeating it for at
Synthesis of Cyclodextrin Films: Membranes were made using the least 1 h before taking any measurements. The solute concentration
following conditions unless otherwise is specified: First, a commercial used in the tested solutions was 1000 ppm for Na2SO4 and 10 ppm for
porous PAN membrane (GMT GmbH) was immersed for 30 min in a the dye molecules. Analysis of the solute concentration was done with a
6.5% (w/v) dry β-cyclodextrin (β-CD hydrate from Acros Organics) 1 m NanoDrop UV–vis spectrophotometer for the dyes and with an Oakton
NaOH aqueous solution. Next, the cyclodextrin loaded PAN support pH/CON 510 conductivity meter for the salt.
was pressed with a rubber roller to remove excess solution from the Shape selective separations were tested with two pairs of molecules
surface and mounted in a Teflon frame. Then, the surface of the PAN with similar molecular weights: PPIX (Sigma-Aldrich) and RB (Sigma-
support was exposed to a 1% (w/v) TC (Sigma-Aldrich) in hexane Aldrich); γ-CD and PEO 1500. For the first pair of molecules, PPIX

Adv. Mater. 2017, 29, 1606641 1606641 (6 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

was first dissolved in 1 mL of DMSO and then diluted with 199 mL of [3] P. M. Budd, N. B. McKeown, D. Fritsch, J. Mater. Chem. 2005, 15,
MeOH to a final concentration of 10 ppm to prevent agglomeration 1977.
of the dye molecules. This solution was then mixed in a 1:1 ratio with [4] D. L. Gin, R. D. Noble, Science 2011, 332, 674.
10 ppm RB in MeOH solution to obtain the solution to be permeated [5] H. B. Park, C. H. Jung, Y. M. Lee, A. J. Hill, S. J. Pas, S. T. Mudie,
through the β-CD membrane. The concentration of both molecules was E. Van Wagner, B. D. Freeman, D. J. Cookson, Science 2007, 318,
monitored simultaneously via UV–vis spectroscopy. The other pair of 254.
molecules had to be tested separately because their signals overlapped [6] T. C. Merkel, B. D. Freeman, R. J. Spontak, Z. He, I. Pinnau,
in the gel-permeation chromatography analysis used to monitor their P. Meakin, A. J. Hill, Science 2002, 296, 519.
concentration. The first solution tested was 1000 ppm PEO 1500 in
[7] N. Petzetakis, C. M. Doherty, A. W. Thornton, X. C. Chen,
methanol, followed by pure MeOH, and then 1000 ppm γ-CD in MeOH
P. Cotanda, A. J. Hill, N. P. Balsara, Nat. Commun. 2015, 6, 7529.
solution. Concentrations of γ-CD and PEO 1500 were measured with an
[8] N. B. McKeown, P. M. Budd, Chem. Soc. Rev. 2006, 35, 675.
Agilent system equipped with dual columns in series (PL Aquagel-OH
30 8 µm and PL Aquagel-OH 40 8 µm) and a G1362A RID detector. [9] J. R. Holst, A. Trewin, A. I. Cooper, Nat. Chem. 2010, 2, 915.
Ultrapure Milli-Q water was used as the mobile phase with a flow rate [10] H. Dodziuk, Cyclodextrins and Their Complexes, Wiley-VCH Verlag
of 1 mL min−1. GmbH & Co. KGaA, Weinheim, Germany 2006, Ch. 1.
[11] E. M. M. Del Valle, Process Biochem. 2004, 39, 1033.
[12] E. Schneiderman, A. M. Stalcup, J. Chromatogr. B: Biomed. Sci. Appl.
2000, 745, 83.
Supporting Information [13] N. Morin-Crini, G. Crini, Prog. Polym. Sci. 2013, 38, 344.
Supporting Information is available from the Wiley Online Library or [14] A. Alsbaiee, B. J. Smith, L. Xiao, Y. Ling, D. E. Helbling,
from the author. W. R. Dichtel, Nature 2016, 529, 190.
[15] P. Rölling, M. Lamers, C. Staudt, J. Membr. Sci. 2010, 362, 154.
[16] L. Yue, S. Wang, D. Zhou, H. Zhang, B. Li, L. Wu, Nat. Commun.
2016, 7, 10742.
Acknowledgements [17] F. van de Manakker, T. Vermonden, C. F. van Nostrum,
This research was supported by King Abdullah University of Science and W. E. Hennink, Biomacromolecules 2009, 10, 3157.
Technology (KAUST). The authors are grateful to Kexin Yao for assisting [18] A. R. Khan, P. Forgo, K. J. Stine, V. T. D’Souza, Chem. Rev. 1998, 98,
with the CO2 adsorption experiments and to Mohamed Nejib Hedhili 1977.
for performing the XPS analysis. The authors thank Rachid Sougrat, [19] J. Hernandez-Pascacio, Á. Piñeiro, J. M. Ruso, N. Hassan,
Stefan Chisca, Mahendra Kumar, Rahul Shevate, and Federico Pacheco R. A. Campbell, J. Campos-Terán, M. Costas, Langmuir 2016, 32, 6682.
for discussions. [20] G. González-Gaitano, P. Rodríguez, J. R. Isasi, M. Fuentes,
G. Tardajos, M. Sánchez, J. Inclusion Phenom. Macrocyclic Chem.
2002, 44, 101.
Conflict of Interest [21] P. W. Morgan, S. L. Kwolek, J. Polym. Sci., Part A: Polym. Chem.
1996, 34, 531.
The authors declare no conflict of interest. [22] F. Pacheco, R. Sougrat, M. Reinhard, J. O. Leckie, I. Pinnau,
J. Membr. Sci. 2016, 501, 33.
[23] S. Karan, Z. Jiang, A. G. Livingston, Science 2015, 348, 1347.
Keywords [24] E. Gaidamauskas, E. Norkus, E. Butkus, D. C. Crans, G. Grincienė,
Carbohydr. Res. 2009, 344, 250.
cyclodextrins, interfacial polymerization, membranes, nanofiltration [25] T. Godjevargova, A. Dimov, J. Membr. Sci. 1992, 67, 283.
[26] M. F. Jimenez-Solomon, Q. Song, K. E. Jelfs, M. Munoz-Ibanez,
Received: December 8, 2016 A. G. Livingston, Nat. Mater. 2016, 15, 760.
Revised: March 16, 2017 [27] N. Joseph, J. Thomas, P. Ahmadiannamini, H. Van Gorp,
Published online: April 24, 2017 R. Bernstein, S. De Feyter, M. Smet, W. Dehaen, R. Hoogenboom,
I. F. J. Vankelecom, Adv. Funct. Mater. 2017, 27, 1605068.
[28] X. Peng, J. Jin, Y. Nakamura, T. Ohno, I. Ichinose, Nat. Nanotechnol.
2009, 4, 353.
[1] P. Marchetti, M. F. Jimenez Solomon, G. Szekely, A. G. Livingston, [29] D. Wang, L.-J. Wan, C. Wang, C.-L. Bai, J. Phys. Chem. B 2002, 106,
Chem. Rev. 2014, 114, 10735. 4223.
[2] D. F. Sanders, Z. P. Smith, R. Guo, L. M. Robeson, J. E. McGrath, [30] A. Harada, J. Li, M. Kamachi, Nature 1992, 356, 325.
D. R. Paul, B. D. Freeman, Polymer 2013, 54, 4729. [31] D. J. Cram, Science 1983, 219, 1177.

Adv. Mater. 2017, 29, 1606641 1606641 (7 of 7) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim