1174166

Charge Transfer and Catalysis at the Metal–Support Interface
By
Lawrence Robert Baker
A dissertation submitted in partial satisfaction of the

requirements for the degree of
Doctor of Philosophy
in
Chemistry
in the
Graduate Division
of the
University of California, Berkeley
Committee in charge:
Professor Gabor A. Somorjai, Chair
Professor Martin Head–Gordon
Professor Peidong Yang
Professor Jeffrey Bokor
Fall 2012
The U.S. Department of Energy has the right to use this document for any purpose whatsoever including the right to
reproduce all or any part thereof.
DISCLAIMER
This document was prepared as an account of work sponsored by the United States Government. While this
document is believed to contain correct information, neither the United States Government, nor any agency thereof,
nor the Regents of the University of California, nor any of their employees, makes any warranty, express or implied,
or assumes any legal responsibility for the accuracy, completeness, or usefulness of any information, apparatus,
product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein
to any specific commercial product, process, or service by its trade name, trademark, manufacturer, or otherwise,
does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States
Government or any agency thereof, or the Regents of the University of California. The views and opinions of
authors expressed herein do not necessarily state or reflect those of the United States Government or any agency
thereof or the Regents of the University of California.
1
Abstract
Charge Transfer and Catalysis at the Metal–Support Interface
by
Lawrence Robert Baker
Doctor of Philosophy in Chemistry
University of California, Berkeley
Professor Gabor A. Somorjai, Chair
Kinetic, electronic, and spectroscopic characterization of model Pt–support systems are

used to demonstrate the relationship between charge transfer and catalytic activity and
selectivity. The results show that charge flow controls the activity and selectivity of supported
metal catalysts.
Using a Pt/n-Si Schottky junction, it is possible to externally control the electronic
properties at the Pt/Si interface by applying bias or by exciting charge carriers with visible light.
It is found that this device can control the rate of CO oxidation on Pt. Results show that negative
charge on the Pt increases the reaction rate while positive charge on the Pt decreases the reaction
rate. This is the first time that a solid-state device has been used to externally control the rate of
a chemical reaction as determined by directly measuring the product yield. Similar experiments
were performed for the H2 oxidation reaction and analogous results were obtained.
Similarly, electronic control of catalyst performance can be achieved by substrate doping.
It is shown that F doping can tune the electronic structure of TiO2. When F is incorporated into
TiO2, the electrical conductivity is dramatically enhanced because F acts as an n-type dopant. It
is shown that this highly n-type TiO2 acts as an electronically active support for Pt catalysts,
increasing the rate of CO oxidation by electronically activating surface O. It is further shown
that for a multipath reaction, the electronic structure of TiO2 also controls the reaction selectivity.
The electronic activity of highly n-type TiO2 serving as a Pt support for methanol oxidation
selectively enhances the production of formaldehyde relative to CO2.
Finally, sum frequency generation (SFG) vibrational spectroscopy is used to probe the
molecular nature of strong metal–support interactions (SMSI). This is the first time that SFG has
been used to probe the highly selective oxide–metal interface during catalytic reaction, and the
results demonstrate that charge transfer from TiO2 on a Pt/TiO2 catalyst controls the product
distribution of furfuraldehyde hydrogenation by an acid–base mechanism. SFG spectra reveal
that a furfuryl-oxy intermediate forms on TiO2 as a result of a charge transfer interaction. This
furfuryl-oxy intermediate is a highly active and selective precursor to furfuryl alcohol, and
spectral analysis shows that the Pt/TiO2 interface is required primarily for H spillover. Density
functional calculations predict that O-vacancies on the TiO2 surface activate the formation of the
furfuryl-oxy intermediate via an electron transfer to furfuraldehyde, drawing a strong analogy
between SMSI and acid–base catalysis.
2
This dissertation builds on extensive existing knowledge of metal–support interactions in
heterogeneous catalysis. The results show the prominent role of charge transfer at catalytic
interfaces to determine catalytic activity and selectivity. Further, this research demonstrates the
possibility of selectively driving catalytic chemistry by controlling charge flow and presents
solid-state devices and doped supports as novel methods for obtaining electronic control over
catalytic reaction kinetics.
i
Table of Contents
Acknowledgements ...................................................................................................................... iii
Chapter 1 Introduction ............................................................................................................... 1
1.1 Background and Motivation .................................................................................................. 1
1.2 Nanoscience and Heterogeneous Catalysis ........................................................................... 2
1.3 Strong Metal–Support Interactions (SMSI) .......................................................................... 3
1.4 Acid–Base Catalysis .............................................................................................................. 5
1.5 Summary of Dissertation ....................................................................................................... 7
1.6 References ............................................................................................................................. 9
Chapter 2 Experimental Methods ............................................................................................ 13
2.1 Catalyst Fabrication and Synthesis ..................................................................................... 13
2.2 Catalyst Characterization .................................................................................................... 15
2.3 Catalytic Reaction Studies .................................................................................................. 20
Chapter 3 Solid-State Charge-Based Device for Control of Catalytic Carbon Monoxide
Oxidation on Platinum Nanofilms Using External Bias and Light ........................................ 22
3.1 Abstract ............................................................................................................................... 22
3.2 Introduction ......................................................................................................................... 22
3.3 Experimental ....................................................................................................................... 23
3.4 Results and Discussion ........................................................................................................ 24
3.5 Conclusions ......................................................................................................................... 30
3.6 References ........................................................................................................................... 31
Chapter 4 Controlling the Rate of Hydrogen Oxidation on Platinum/Silicon Catalytic
Nanodiodes by Photocurrent Generation with Visible Light ................................................. 33
4.1 Abstract ............................................................................................................................... 33
4.2 Introduction ......................................................................................................................... 33
4.3 Experimental ....................................................................................................................... 34
4.5 Conclusions ......................................................................................................................... 38
4.6 References ........................................................................................................................... 38
Chapter 5 Generation of Highly n-Type Titanium Oxide Using Plasma Fluorine Insertion
.............................................................................................................................................. 40
5.1 Abstract ............................................................................................................................... 40
5.2 Introduction ......................................................................................................................... 40
5.3 Experimental ....................................................................................................................... 41
5.5 Conclusion ........................................................................................................................... 52
5.6 References ........................................................................................................................... 52
Chapter 6 Highly n-Type Titanium Oxide as an Electronically Active Support for
Platinum in the Catalytic Oxidation of Carbon Monoxide ..................................................... 54
6.1 Abstract ............................................................................................................................... 54
6.2 Introduction ......................................................................................................................... 54
6.3 Experimental ....................................................................................................................... 55
6.5 Conclusions ......................................................................................................................... 62
ii
6.6 References ........................................................................................................................... 63
Chapter 7 Controlling the Selectivity of Methanol Oxidation on Platinum Catalysts by
Tuning the Electronic Structure of a Titanium Oxide Support by Fluorine Doping ........... 65
7.1 Abstract ............................................................................................................................... 65
7.2 Introduction ......................................................................................................................... 65
7.3 Experimental ....................................................................................................................... 67
7.5 Conclusions ......................................................................................................................... 71
7.6 References ........................................................................................................................... 72
Chapter 8 Furfuraldehyde Hydrogenation on Titanium Oxide-Supported Platinum
Nanoparticles Studied by Sum Frequency Generation Vibrational Spectroscopy: Acid–
Base Catalysis Explains the Molecular Origin of Strong Metal–Support Interactions ....... 74
8.1 Abstract ............................................................................................................................... 74
8.2 Introduction ......................................................................................................................... 74
8.3 Experimental ....................................................................................................................... 76
8.5 Conclusions ......................................................................................................................... 86
8.6 References ........................................................................................................................... 87
Chapter 9 The Role of an Organic Cap in Nanoparticle Catalysis: Restructuring of
Carbonaceous Material Controls Catalytic Activity of Platinum Nanoparticles ................. 89
9.1 Abstract ............................................................................................................................... 89
9.2 Introduction ......................................................................................................................... 89
9.3 Experimental ....................................................................................................................... 91
9.5 Conclusions ....................................................................................................................... 100
9.6 References ......................................................................................................................... 100
Chapter 10 Removing Poly(vinylpyrrolidone) from Size-Controlled Platinum
Nanoparticles for Catalytic and Sum Frequency Generation Spectroscopic Studies:
Solvent Cleaning as an Alternative to UV Cleaning .............................................................. 102
10.1 Abstract ........................................................................................................................... 102
10.2 Introduction ..................................................................................................................... 102
10.3 Experimental ................................................................................................................... 103
10.4 Results and Discussion .................................................................................................... 105
10.5 Conclusions ..................................................................................................................... 112
10.6 References ....................................................................................................................... 113
Chapter 11 Conclusions and Future Work ........................................................................... 115
12.1 Electronic Oxygen Activation by Charge Transfer from the Catalyst Support .............. 115
12.2 Strong Metal–Support Interactions and Acid–Base Chemistry ...................................... 116
12.3 The Role of an Organic Cap to Mediate Nanoparticle Catalysis .................................... 117
12.4 References ....................................................................................................................... 118
iii
Acknowledgements
This research was supported by the U.S. Department of Energy, Office of Basic Energy
Sciences, under Contract No. DE-AC02-05CH11231. I am grateful for this funding source and
hope that continued scientific funding will be available for the important challenges of the future
pertaining to renewable energy and energy conversion chemistry.
I am deeply grateful to the many individuals who have contributed to my education and
who have supported me during an amazing experience in Berkeley. My experiences have far
exceeded my most optimistic expectations when I moved here four years ago with my family,
and it is a result of the very good people that have surrounded me.
I am grateful to Gabor Somorjai. In my first interactions with him, he radiated an
influence that enthused me with a love for the research that we have shared, and this has
motivated every choice and accomplishment since those early conversations. Following my first
meeting with him during my site visit to Berkeley, I was so excited to join his research group
that I kept a copy of C&EN News showing his picture on the cover with me constantly. Since
that time he has taught me much about how to be a good chemist, a good mentor, and a good
husband and father, and my life will be much better because of what I have learned from him.
He once told me that in life, security multiplied by opportunity equals a constant, meaning that
we can only increase our opportunities by sacrificing our security. However, he has raised the
value of this constant significantly in my life and in the lives of numerous other students, and he
is one of the great mentors that I aspire to emulate.
I am grateful to Inger who has taken good care of me. If Gabor is my scientific father,
then Inger is my scientific mother. I remember shortly after coming to Berkeley, my wife,
Catherine, was expecting our second child. I tried to obtain a parking permit at LBNL so that I
could come home quickly when Catherine started labor. However, the badge office would not
issue the permit, and I found myself in trouble when I responded disputatiously. Of course Inger
calmed this storm and with her and Gabor’s help, I received a permit. I have always been
pleased to think that she is proud of my accomplishments, and I know that it was on the basis of
her nomination that I received the Benjamin Boussert Award.
I am grateful to my lab partners, and I have learned much from my interactions with each
one. Almost my entire Ph.D. experience has been spent in close work with Antoine Hervier. To
me he is much more than a bright chemist; he is an honest friend. Our early experiences working
together to use a catalytic photodiode to control the rate of H2 oxidation is one of the fun
memories of my Ph.D. I will always remember sitting with him at the computer anxiously
watching as each data point confirmed or disproved our anticipated breakthrough. Griffin
Kennedy is another lab partner whose contributions to our project have been invaluable. I will
always be glad that he chose to join our project following our acquaintance in Chem 125. He is a
talented student with a bright future and a good friend, and I am grateful for my close association
with him. There have been many others with whom it has been a privilege to work. Getting to
know each member of the Somorjai Group and learning their personal motivations for studying
science has broadened my perspective and given me many worthy examples to follow.
iv
I am most grateful to my family. When Catherine, Eliza, and I moved to Berkeley, we
could not have anticipated the experiences we have had. During our time here, we have been
joined by Stephen and Beth. I have loved each moment with this family, and they are my
treasures. I hope that my hard work in this program will represent a good example to my
children and that they will also be blessed to love learning as I have loved it. I am also grateful
to my Mom and Dad and my siblings, Lyndsay, Ryan, Quinn, Luke, Spencer, and Brooke. My
desire to honor them has been my motivation.
1
Chapter 1
Introduction
1.1 Background and Motivation
The first known commercialization of a catalyst came in the form of the Dobereiner
lamp. This early household device consisted of a bottle of zinc metal in sulfuric acid to produce
hydrogen. A platinum catalyst was located at the bottle’s outlet. When the bottle was opened,
the platinum would ignite the pressurized hydrogen to produce a small flame. Used as a kitchen
lighter, this lamp was invented by Johann Wolfgang Dobereiner in 1823 at the University of Jena
in Germany and remained in commercialization for about 60 years. The term “catalysis” was not
used until 13 years after Dobereiner’s invention, when the Swedish chemist, Jons Jacob
Berzelius, observed that certain substances accelerate the rate of a reaction while themselves,
remaining unchanged. To describe this phenomenon, Berzelius coined the term catalysis from
the Greek word “καταλύειν,” meaning to untie.
Since Dobereiner’s invention almost two centuries ago, catalysis has become a $900
billion per year industry. Catalysis is used in processes as remote as the catalytic converter that
reduces harmful emissions from automobiles to the synthesis of ammonia using the Haber
process. It is clear that the quality of living would be dramatically different without catalysis.
For example, the world’s present population is largely sustained by agriculture enhanced by
fertilizers synthesized using the Haber process. It is estimated that without the ability to catalyze
ammonia synthesis, the present global population would be cut in half. Currently 90% of all
commercially produced chemical products rely on catalysis in some form.
There currently exists an economic and environmental need to develop catalysts for
chemical energy conversion toward the goal of alternative fuels and renewable energy.
Development of catalytic processes for chemical energy conversion from renewable sources is
proving to be one of the most pressing and challenging issues for modern science. At the heart
of energy conversion is the correlation between charge flow and chemistry. At present, several
processes exist that can convert a chemical reaction to a net charge flow (e.g., fuel cells and
batteries) or that can use a charge flow to drive a chemical reaction (e.g., electrochemistry,
photochemistry, and acid–base catalysis). However, it is clear that the efficiency and stability of
these processes require improvement to meet the existing energy demand or that novel energy
conversion processes will need to be developed. This has been the goal of the Helios Solar
Energy Research Center which originally funded my research. Recently, Helios has been
replaced by the Joint Center for Artificial Photosynthesis. Both of these initiatives are supported
by the Department of Energy, Office of Basic Energy Sciences. A fundamental understanding of
the correlation between charge flow and catalysis, as well as the ability to control charge flow to
drive a chemical reaction, is key to the success of these projects, and this has been the motivation
of my Ph.D. research.
2
1.2 Nanoscience and Heterogeneous Catalysis
Heterogeneous catalysis describes chemical reactions that occur on surfaces and, together
with homogeneous and enzyme catalysis, composes one of the three major branches of catalysis
science. The study of heterogeneous catalysis is closely linked with the evolution of surface
science, because the correlation of surface structure with catalytic activity originally formed the
foundation of this field.1 Although some catalytic reactions are structure insensitive, many
reactions depend strongly on the surface structure of the catalyst.2 By studying catalytic
reactions on single crystal model catalysts, it was discovered that low coordination sites
occurring at step edges, kinks, and defects, are often more active than other highly coordinated
surface sites.3 A few examples of structure sensitive reactions include H2–D2 exchange on Pt,4-6
furan and crotonaldehyde hydrogenation on Pt,7,8 CO oxidation on Pt,9 ammonia synthesis on
Fe,10 and CO hydrogenation on Ni.11
For single crystal model catalysts, the crystal face determines the surface structure.
However, for nanoparticle catalysts, the relative concentration of step edges and kink sites
depends on particle size and shape. Traditional methods for catalyst preparation (i.e. incipient
wetness and ion exchange) use the reduction of a metal salt inside of a mesoporous oxide.12,13
The result is a catalyst consisting of metal particles with a broad distribution of sizes and
morphologies. This type of polydisperse heterogeneous catalyst masks the structure sensitivity
inherent in heterogeneous catalysis because of the impossibility of selecting nanoparticles of a
single size or shape. Recent advances in nanoscience have shown that colloidal synthetic
methods can produce monodisperse nanoparticles with well-defined sizes and shapes.14-18 This
advance has marked a new era in heterogeneous catalysis where monodisperse nanoparticles are
used as model catalysts.19-21 The results of these studies show that nanoparticle size and shape
control the catalytic activity and selectivity for many reactions.22-27
In colloidal synthetic methods, nanoparticles are necessarily encapsulated by an organic
polymer or surfactant. This capping agent lowers the surface energy of the nanoparticle to
prevent aggregation of the particles;18 the cap may also help to control the size and shape of the
nanoparticles.28 This gives rise to an important question regarding the effect of the capping
agent on the catalytic properties of the encapsulated nanoparticles. It is traditionally thought that
the cap acts as a site blocking agent and lowers the metal surface area available for catalytic
reaction.29 In this light, it has been assumed that colloidal preparation methods are impractical
for catalytic applications because the presence of the organic cap decreases the apparent metal
dispersion. However, this is an incomplete assumption that considers the cap as a passive
coating rather than a dynamic shell that restructures depending on reaction conditions. This
work demonstrates that for certain reactions a capped nanoparticle is more active than its
uncapped analog owing to reversible restructuring of carbonaceous material in various gas
environments. When the capping layer is removed from a nanoparticle, the cap is replaced by a
stable carbonaceous shell. This shell reversibly adopts an open, porous structure in H2 and a
closed, impermeable structure in O2. Consequently, a capped nanoparticle is a more active
catalyst for oxidation chemistry because the capping layer prevents formation of a deactivating
carbonaceous deposit. This use of capping agents to control catalytic activity of a nanoparticle is
reminiscent of homogeneous catalysis where ligands are used to tune the catalytic properties of
single metal ions.30
3
1.3 Strong Metal–Support Interactions (SMSI)
1.3.1 Definitions of SMSI
Although metals alone are often catalytically active, most catalysts consist of metal
particles supported on a porous oxide. This not only provides a high surface area for the
heterogeneous catalyst, but as shown by many studies, the oxide support plays an important role
in determining the activity and selectivity of the catalyst.31-34 This phenomenon, known as the
strong metal–support interaction (SMSI), has been widely studied and is an important topic in
both science and industry. The term SMSI was first used by Tauster and Fung to refer to the
dramatic reduction in chemisorption sites observed for noble metal catalysts supported on
titanium oxide.32 In their studies, Tauster and Fung showed evidence for a strong bonding
interaction between metal nanoparticles and a reduced titanium oxide support, and they used the
term SMSI to refer directly to this bonding interaction.34 However, because this metal–support
interaction is closely linked to the catalytic behavior of the metal involved, the definition of
SMSI has since expanded to broadly refer to support-induced changes in the catalytic activity
and selectivity of metal nanoparticles. In this work, the term SMSI refers to the ability of an
inert support to mediate catalytic behavior of a metal without specific attention to bonding at the
metal–support interface.
SMSI is known to play an important role in many catalytic reactions, including CO
oxidation,35-39 CO and CO2 hydrogenation,40,41 hydroformulation,42 and partial hydrogenation
reactions.43-47 The catalyst support also plays an important role in activating molecular oxygen
for selective partial oxidation reactions. Industrially relevant examples include the synthesis of
aldehydes from primary alcohols,48 the production of hydrogen peroxide from hydrogen,49-53 and
the conversion of methane to synthesis gas.54-56 In these cases, molecular oxygen is preferred
over other oxygen donors due to cost, energy efficiency, and environmental concerns.48,56
1.3.2 Mechanisms of SMSI
Several proposed mechanism exist to explain the role of an oxide to mediate the catalytic
behavior of a supported metal nanoparticle. Because SMSI broadly refers a wide variety of
experimental observations, these mechanisms are not mutually exclusive and each focuses on
explaining a slightly different observation related to supported metal catalysts. Together, these
observations describe a complex series of interactions showing how the metal–support interface
can strongly influence surface chemistry and selectively amplify specific reaction pathways. The
majority of proposed mechanisms for metal–support interactions fall into one of the following
categories: 1) formation of active sites when the support wets (or decorates) the metal
particle,57,58 2) activation and spillover from chemically active sites in the support,59 and
3) electronic mediation through various forms of charge transfer.37,60
The concept that an oxide support could diffuse onto and eventually encapsulate a metal
nanoparticle was first suggested to explain the loss of chemisorption sites when metal catalysts
supported on reducible oxides were heated in H2. Tauster and Fung showed that after reduction
of a metal catalyst supported on titanium oxide at 773 K, H2 and CO chemisorption decreased to
nearly zero for Ru, Rh, Pd, Os, Ir, and Pt.32 This was not a result of a structural collapse of the
porous oxide support as the total surface area remained unchanged and the effect was reversible
by oxidizing treatments. Baker et al. first showed by transmission electron microscopy (TEM)
4
that following reduction treatments, titanium oxide migrated over the metal surface in a way that
would block metal sites from chemisorbing H2 or CO.61 Komaya et al. showed by TEM that as
the reduction temperature increased, the coverage of titanium oxide on the metal catalyst
increased resulting in total encapsulation by 773 K.62 With continuously improving TEM
resolution, evidence for metal encapsulation by a reduced support has become unquestionable.63
Although metal decoration by an oxide support can explain the loss of apparent metal
surface area as measured by chemisorption experiments, it falls short of explaining how the
oxide mediates the catalytic activity of the supported or encapsulated metal catalyst. A series of
experiments by Boffa et al. are insightful for understanding how metal decoration by an oxide
affects the surface catalytic reaction.40,41 In these experiments, various oxides were deposited
onto a Rh foil. This geometry with oxide islands on a planar metal support is referred to as an
inverse catalyst and allows a high degree of control over the oxide–metal interface. Because the
surface energy of the oxide is lower than the Rh, the oxide layer grows as a single monolayer
islands on the metal without any multilayer formation until 95% surface coverage. This
preparation method allows for precise control of the relative concentration of oxide and metal
surface sites compared with previous SMSI catalysts prepared by high temperature reduction.
Assuming that the metal sites are catalytically active and that the oxide sites are not,
linear deactivation of the Rh is expected with increasing oxide coverage. This is the case for the
ethylene hydrogenation reaction. However, for CO and CO2 hydrogenation reactions, a >10-fold
rate enhancement occurs as the Rh metal is decorated with oxide islands. This rate enhancement
reaches a maximum at 50% oxide coverage, suggesting that oxide–metal interface sites occurring
at the perimeter of the oxide islands are orders of magnitude more active than the metal alone.
By correlating the maximum rate enhancement with the Lewis acidity of the deposited oxide,
Boffa hypothesized that a Lewis acid–base interaction of CO and CO2 with the oxide is
responsible for carbonyl bond activation. This idea of bond activation at the oxide–metal
interface, coupled with the concept of metal decoration by a reduced oxide, shows one method
by which an oxide support can mediate the catalytic properties of a metal nanoparticle.
Because adsorption energies of molecules can differ substantially between metals and
oxides, the oxide–metal interface can also bring together reactants that would not co-adsorb on
either an oxide or a metal alone. An example of this is found in the CO oxidation reaction on Pt.
On Pt single crystals the activation energy of CO oxidation is as high as 42 kcal/mol,
corresponding with the high heat of CO adsorption.64,65 This indicates that the Pt surface is
predominately CO covered and that the rate limiting step is desorption of CO to allow O 2
adsorption and dissociation. However, O2 adsorption is much more competitive on oxides than
on Pt. Consequently, for Pt supported on certain oxides, the Pt may be predominately CO
covered while the oxide is predominately O covered. This means that at the metal–support
interface the reaction proceeds without the need for CO desorption, and the activation energy is
much lower than on a bulk Pt catalyst.66,67
Additionally, once adsorbed on one site, many reactants can migrate across the oxide–
metal interface.68 For example, H2 will not dissociate on most oxide surfaces but will actively
dissociate on many metals, including Pt.69 Consequently, in the presence of small amounts of Pt,
oxides can be permeated with atomic H as a result of dissociation on the Pt followed by
migration across the Pt–support interface.62 This allows the reaction of atomic H with any
molecule preferentially adsorbed on the oxide support, and gives rise to reaction pathways that
5
do not exist on either the pure Pt or the pure oxide. Results presented in this work show that
specific adsorption modes of aldehyde molecules on titanium oxide activate the carbonyl bond
for selective hydrogenation which occurs via the spillover of atomic H from supported Pt.
Consequently, Pt supported on titanium oxide is highly selective for formation of unsaturated
alcohols from aldehydes, while Pt supported on silicon oxide is not active at all for this reaction
pathway.
Previous studies have indicated that, at least in certain instances, catalytic oxidation
reactions proceed by an electronically activated pathway.70-75 Bonn et al.70 showed that for CO
oxidation on Ru, the activation of chemisorbed O occurs by charge transfer from the metal to
form an active O− intermediate. Because the active O− species cannot form thermally until well
above the desorption temperature of CO, a temperature ramp of a CO/O2 co-adsorbed Ru surface
produces CO and O2, but not CO2. However, a femtosecond laser pulse can produce very high
electronic temperatures on the short time scale without significant lattice heating. This serves to
electronically activate O without thermally desorbing CO, resulting in CO2 formation. Although
fundamentally insightful, this type of experiment employing a femtosecond laser cannot be
scaled for high turnover applications. The research presented here shows that similar electronic
O activation also occurs by electron spillover from a highly n-type catalyst support.39,76,77 These
results demonstrate the electronic nature of SMSI as well as the utility of metal–support
interactions to electronically control the activity and selectivity of oxidation catalysts. In this
work, it is also shown that non-thermal control of catalytic surface chemistry is possible by
externally controlling the electronic structure at a metal–support interface using a solid-state
charge-based device.78
1.4 Acid–Base Catalysis
Like SMSI, acid–base catalysis is often characterized by the selective enhancement of a

single reaction pathway. This is because in acid–base catalysis, the generation of ionic reaction
intermediates determines the reaction rate and selectivity, and the charged intermediate is often
highly active for a specific pathway.79-81 In the case of acid catalysis, charge is transferred to an
organic molecule, usually by H+ addition or H− abstraction.82 Bronsted acids work by H+
addition and result in formation of a carbonium ion (i.e., a carbocation with a five-fold
conjugation), while Lewis acids work by H− abstraction and result in the formation of a
carbenium ion (i.e., a carbocation with a three-fold conjugation). Scheme 1 depicts CH5+ and
CH3+, the simplest carbonium and carbenium ions, respectively. Typically carbonium ions
quickly convert to carbenium ions by the evolution of molecular hydrogen. Consequently, H 2
evolution has been taken as a signature of Bronsted acidity, and this results in a stoichiometric,
rather than catalytic, reaction unless H2 is present to replenish the acidic sites. Because of the
active nature of the resulting carbocations, strong acids are catalysts for isomerization, 83
alkylation,84,85 polymerization,86,87 and carbonylation88 reactions.
The term superacid was first coined to describe the properties of SbF5 in HFSO3, and a
superacid is defined as a material whose acidic strength is greater than 100% sulfuric acid. 89
Typically a pKa value is used to describe the strength of an acid, where pKa is simply
proportional to the Gibbs free energy of change of a proton transfer. However, fundamentally
pKa does not exactly describe the strength of an acid; instead it describes the strength of an acid
relative to a specific proton acceptor, meaning it quantifies the energetics of an acid–base
6
interaction. Because the base in the acid–base pair is always water for aqueous acids, the pKa
scale is useful for comparing the strength of typical aqueous acids. However, in acid catalysis,
the reactant molecule acts as the proton acceptor. This means that acid catalysis requires an acid
with sufficient strength to protonate the hydrocarbon reactant. Consequently, the pKa scale
breaks down for superacids because water will necessarily limit the activity of a superacid by
replacing the reactant as the proton acceptor. Instead of pKa, the Hammett acidity function (Ho)
represents a measure of acidic strength for superacids, and this method relies on the use of
aromatic nitro compounds as indicators.90,91 However, there is still no way to directly measure
the energetics of charge transfer in acid catalysis.
H H
H 3C + +
C
H H H
Carbonium Ion Carbenium Ion
Scheme 1
The first superacids studied were anhydrous liquids. Addition of a strong Lewis acid
such as SbF5 to Bronsted acids like H2SO4 or HFSO3 increases their acidic strength dramatically.
This is because the SbF5 destabilizes the proton on the Bronsted acid by an electron withdrawing
interaction. The acidic strength of these mixed Lewis–Bronsted acids is sufficiently high to
protonate hydrocarbons, and in these acids, carbocations are stable and have been observed by
infrared, Raman, nuclear magnetic resonance, and x-ray photoelectron spectroscopies.92 Because
of the industrial advantages of heterogeneous catalysis compared to homogeneous catalysis,
preparation of solid superacids has become an important area of research. 82 Several classes of
solid superacids exist including liquid acids mounted on a solid support,93 sulfate-activated metal
oxides,94-96 mixed salts,97,98 and zeolites.99
Acid catalysis represents a unique chemistry that relies on charge transfer via protons and
molecular ions, and many similarities appear to exist between SMSI and acid–base catalysis.
The evolution of this work that began by measuring chemically induced charge flow at the
oxide–metal interface100-102 is quickly leading to a molecular level study of acid catalyzed
chemistry with a focus on understanding the close relationship between charge flow and catalytic
selectivity. Specifically, it is clear that a fundamental understanding of catalyst selectivity is
closely related to understanding the flow of charge between various active sites on a catalyst.
This charge flow occurs both by the migration of molecular ions and by electrical conduction.
Additionally, by controlling the flow of charge at a catalyst interface, novel reaction pathways
will become accessible, and these processes are central to the future of chemical energy
conversion.
7
1.5 Summary of Dissertation
1.5.1 Experimental Methods
Experimental methods used in this research are discussed in Chapter 2 with a focus on
catalyst preparation, characterization, and reaction studies. Catalyst preparation includes
fabrication of solid-state catalytic devices, synthesis of colloidal nanoparticles, and deposition
and processing of oxide thin-films used for metal catalyst supports. A range of techniques were
used for catalyst characterization. Ex situ methods include electron microscopy, x-ray
photoelectron spectroscopy (XPS), and electrical probe measurements. In situ methods include
bias and photocurrent measurements and sum frequency generation (SFG) vibrational
spectroscopy. Reaction studies were performed in gas phase batch mode reactors with gas
chromatography (GC) detectors.
1.5.2 Solid-State Device for External Control of Catalytic Chemistry
Chapters 3 and 4 describe the development of a novel solid-state charge-based device for
external electronic control of catalytic chemistry. Chapter 3 describes results for the CO
oxidation reaction and Chapter 4 describes results for the H2 oxidation reaction. Together these
findings show that a bias-induced negative charge at the Pt surface enhances the rate of catalytic
oxidation while a flux of photo-induced positive charges to the Pt surface decreases the rate of
oxidation chemistry. This is the first time that a charge-based device has been shown to provide
external control of a surface catalytic reaction as determined by directly measuring the product
yield, and the results for CO oxidation have been published in Nano Letters.78 The results for H2
oxidation are yet unpublished owing to experimental difficulties affecting the signal quality.
1.5.3 Fluorine-Doped Titanium Oxide as an Electronically Active Platinum Support
Similar electronic effects in catalytic oxidation are possible by chemical doping, and
Chapters 5–7 discuss the effects of F in titanium oxide films used as supports for Pt catalysts.
Chapter 5 describes a method for n-type doping titanium oxide using plasma F insertion. F has
two effects on the doped oxide: First, F passivates O-vacancies. In titanium oxide O-vacancies
give rise to partially occupied midgap defect states that act as the primary conduction channel;
these states also pin the Fermi energy well below the conduction band edge. When these O-
vacancies are passivated by F doping, electrical conductivity decreases, and the Fermi energy
approaches the conduction band edge. Second, F acts as an n-type dopant by donating electrons
to the conduction band of titanium oxide. By these combined effects, F doping allows
fabrication of titanium oxide films with a high Fermi energy and an electron transport channel in
the true conduction band, and these effects on the electronic structure of titanium oxide have
been published in Nano Letters.76
Chapters 6 and 7 show that supporting Pt catalysts on F-doped titanium oxide results in
electronically mediated catalytic activity for CO oxidation and methanol oxidation. In the case
of CO oxidation, electron density from the highly n-type support appears to activate surface O,
resulting in an approximately two-fold increase in catalytic activity. This result is exactly
analogous to the electronic effect observed by external bias in a catalytic solid-state device. In
the case of methanol oxidation, a change in reaction selectivity occurs as a result of substrate
doping. Methanol oxidation on Pt produces both formaldehyde and CO2. It appears that
8
different surface O species give rise to the different reaction products, and the electronically
activated surface O on the highly n-type support favors partial oxidation of methanol to
formaldehyde. These results on the utility of F-doped titanium oxide as an electronically active
support for Pt catalysts have been published in Journal of Physical Chemistry C.39,77 In addition,
two patents are pending based on the reported work with F-doped titanium oxide.103,104
1.5.4 Strong Metal–Support Interactions Probed by Sum Frequency Generation Vibrational
Spectroscopy
Chapter 8 describes a molecular level investigation of strong metal–support interactions

in furfuraldehyde hydrogenation on supported Pt nanoparticles. In this study, no support effect
is observed on the reaction kinetics of Pt nanoparticles capped with poly(vinylpyrrolidone)
(PVP) because PVP inhibits contact between Pt and the support. However, when the PVP cap is
removed by UV cleaning, a 50-fold enhancement is observed in the selective production of
furfuryl alcohol by Pt nanoparticles supported on TiO2, while no effect is observed for Pt
nanoparticles supported on SiO2. Using SFG vibrational spectroscopy, the highly selective
reaction pathway on the Pt/TiO2 catalyst is correlated with the formation of a furfuryl-oxy
intermediate on the TiO2 support. It appears that the role of the Pt/TiO2 interface is to allow H
spillover to active sites on the support, and H spillover is directly observed during reaction as a
change in the nonresonant intensity in the SFG spectrum. Density functional calculations predict
that O-vacancies in the TiO2 are the catalytically active sites and that selective furfuryl alcohol
formation is activated by an electron transfer from a Ti3+ cation to the adsorbed furfuraldehyde
molecule. This work draws a strong analogy between SMSI and acid–base catalysis and has
been submitted for publication to Journal of American Chemical Society. A continuation of this
work is currently in progress to compare the reaction kinetics and SFG spectra for
furfuraldehyde, crotonaldehyde, and acetaldehyde hydrogenation on supported Pt catalysts.
1.5.5 The Role of an Organic Cap in Nanoparticle Catalysis
The organic cap used to stabilize colloidal nanoparticles also has an important role in
mediating the surface chemistry of nanoparticle catalysts. Chapters 9 and 10 investigate this
effect using SFG vibrational spectroscopy to probe the structure of the organic capping layer on
Pt nanoparticles under reaction conditions. Chapter 9 shows that when UV cleaning is used to
remove PVP from Pt nanoparticles, carbonaceous fragments remain on the surface. These
fragments dynamically restructure in H2 and O2 and control the activity of Pt nanoparticles for
ethylene hydrogenation and methanol oxidation across several orders of magnitude. These
carbonaceous fragments form a tightly closed shell in O2 that prevents access to the Pt surface,
but this shell opens in H2. From kinetic measurements on thermally cleaned PVP- and oleic
acid-capped Pt nanoparticles, it appears that formation of this dynamic carbonaceous layer
occurs regardless of the capping agent or cleaning method. This study contradicts conventional
assumptions that the capping agent necessarily inhibits nanoparticle activity and shows that for
methanol oxidation, capped nanoparticles are orders of magnitude more active than their cleaned
analogues. A manuscript describing this discovery will be submitted soon for publication.
Chapter 10 investigates the thermal stability of PVP-capped Pt nanoparticles before and
after UV cleaning. It is found that while the capped nanoparticles are thermally stable in
reaction conditions up to 473 K, the UV-cleaned particles melt below 373 K. This loss of
thermal stability after cap removal precludes UV cleaning for studies of size- and shape-
9
controlled Pt nanoparticles. However, SFG studies of capped nanoparticles during reaction are
challenging because vibrational modes of the organic cap often overlap with vibrational modes
of the reaction intermediates of interest. Solvent cleaning is suggested as an alternative method
for PVP removal from Pt nanoparticles. It is shown that while solvent cleaning does not remove
the cap in its entirety, the remaining PVP disorders in H2 allowing clear SFG identification of
reaction intermediates for cyclohexene hydrogenation. It is further shown that solvent cleaning
preserves the thermal stability of the Pt nanoparticles, making this method a viable alternative to
UV cleaning for SFG studies of size- and shape-controlled nanoparticles. This work is published
in Journal of Physical Chemistry C.
1.6 References
(1) Somorjai, G. A.; Li, Y. Introduction to surface chemistry and catalysis; 2nd Edition ed.;
Johns Wiley & Sons, Inc.: Hoboken, NJ, 2010.
(2) Somorjai, G. A.; Carrazza, J. Industrial & Engineering Chemistry Fundamentals 1986,
25, 63.
(3) Boudart, M. In Advances in Catalysis; D.D. Eley, H. P., Paul, B. W., Eds.; Academic
Press: 1969; Vol. Volume 20, p 153.
(4) Ferrer, S.; Rojo, J. M.; Salmerón, M.; Somorjai, G. A. Philosophical Magazine A 1982,
45, 261.
(5) Salmeron, M.; Gale, R. J.; Somorjai, G. A. The Journal of Chemical Physics 1977, 67,
5324.
(6) Salmerona, M.; Gale, R. J.; Somorjai, G. A. The Journal of Chemical Physics 1979, 70,
2807.
(7) Kliewer, C. J.; Bieri, M.; Somorjai, G. A. Journal of the American Chemical Society
2009, 131, 9958.
(8) Kliewer, C. J.; Aliaga, C.; Bieri, M.; Huang, W.; Tsung, C.-K.; Wood, J. B.;
Komvopoulos, K.; Somorjai, G. A. Journal of the American Chemical Society 2010, 132, 13088.
(9) McCrea, K. R.; Parker, J. S.; Somorjai, G. A. The Journal of Physical Chemistry B 2002,
106, 10854.
(10) Strongin, D. R.; Carrazza, J.; Bare, S. R.; Somorjai, G. A. Journal of Catalysis 1987, 103,
213.
(11) Andersson, M. P.; Abild-Pedersen, E.; Remediakis, I. N.; Bligaard, T.; Jones, G.;
Engbwk, J.; Lytken, O.; Horch, S.; Nielsen, J. H.; Sehested, J.; Rostrup-Nielsen, J. R.; Norskov,
J. K.; Chorkendorff, I. Journal of Catalysis 2008, 255, 6.
(12) Preparation of Solid Catalysts; Ertl, G.; Knozinger, H.; Weitkamp, J., Eds.; Wiley-VCH:
Weinheim, Germany, 1999.
(13) Tsoncheva, T.; Dal Santo, V.; Gallo, A.; Scotti, N.; Dimitrov, M.; Kovacheva, D. Appl
Catal a-Gen 2011, 406, 13.
(14) Ahmadi, T. S.; Wang, Z. L.; Green, T. C.; Henglein, A.; El-Sayed, M. A. Science 1996,
272, 1924.
(15) Peng, X.; Wickham, J.; Alivisatos, A. P. Journal of the American Chemical Society 1998,
120, 5343.
(16) Puntes, V. F.; Krishnan, K. M.; Alivisatos, A. P. Science 2001, 291, 2115.
(17) Oh, M.; Mirkin, C. A. Nature 2005, 438, 651.
(18) Yin, Y.; Alivisatos, A. P. Nature 2005, 437, 664.
10
(19) Song, H.; Kim, F.; Connor, S.; Somorjai, G. A.; Yang, P. The Journal of Physical
Chemistry B 2004, 109, 188.
(20) Rioux, R. M.; Song, H.; Hoefelmeyer, J. D.; Yang, P.; Somorjai, G. A. The Journal of
Physical Chemistry B 2004, 109, 2192.
(21) Song, H.; Rioux, R. M.; Hoefelmeyer, J. D.; Komor, R.; Niesz, K.; Grass, M.; Yang, P.;
Somorjai, G. A. Journal of the American Chemical Society 2006, 128, 3027.
(22) Bratlie, K. M.; Lee, H.; Komvopoulos, K.; Yang, P.; Somorjai, G. A. Nano Letters 2007,
7, 3097.
(23) Kuhn, J. N.; Huang, W.; Tsung, C.-K.; Zhang, Y.; Somorjai, G. A. Journal of the
American Chemical Society 2008, 130, 14026.
(24) Grass, M.; Rioux, R.; Somorjai, G. Catalysis Letters 2009, 128, 1.
(25) Grass, M. E.; Joo, S. H.; Zhang, Y.; Somorjai, G. A. The Journal of Physical Chemistry
C 2009, 113, 8616.
(26) Witham, C. A.; Huang, W.; Tsung, C.-K.; Kuhn, J. N.; Somorjai, G. A.; Toste, F. D. Nat
Chem 2010, 2, 36.
(27) Alayoglu, S.; Aliaga, C.; Sprung, C.; Somorjai, G. Catalysis Letters 2011, 141, 914.
(28) Zhang, Y.; Grass, M. E.; Kuhn, J. N.; Tao, F.; Habas, S. E.; Huang, W.; Yang, P.;
(29) Kuhn, J. N.; Tsung, C.-K.; Huang, W.; Somorjai, G. A. Journal of Catalysis 2009, 265,
209.
(30) Gorin, D. J.; Sherry, B. D.; Toste, F. D. Chemical Reviews 2008, 108, 3351.
(31) Schwab, G. M. Transactions of the Faraday Society 1946, 42, 689.
(32) Tauster, S. J.; Fung, S. C.; Garten, R. L. Journal of the American Chemical Society 1978,
100, 170.
(33) Tauster, S. J.; Fung, S. C.; Baker, R. T. K.; Horsley, J. A. Science 1981, 211, 1121.
(34) Tauster, S. J. Accounts of Chemical Research 1987, 20, 389.
(35) Oh, S. H.; Eickel, C. C. Journal of Catalysis 1988, 112, 543.
(36) Zhu, H.; Qin, Z.; Shan, W.; Shen, W.; Wang, J. Journal of Catalysis 2004, 225, 267.
(37) Chen, M. S.; Goodman, D. W. Science 2004, 306, 252.
(38) Goodman, D. Catalysis Letters 2005, 99, 1.
(39) Baker, L. R.; Hervier, A.; Seo, H.; Kennedy, G.; Komvopoulos, K.; Somorjai, G. A. The
Journal of Physical Chemistry C 2011, 115, 16006.
(40) Boffa, A. B.; Bell, A. T.; Somorjai, G. A. Journal of Catalysis 1993, 139, 602.
(41) Boffa, A.; Lin, C.; Bell, A. T.; Somorjai, G. A. Journal of Catalysis 1994, 149, 149.
(42) Yamada, Y.; Tsung, C.-K.; Huang, W.; Huo, Z.; Habas, S. E.; Soejima, T.; Aliaga, C. E.;
Somorjai, G. A.; Yang, P. Nat Chem 2011, 3, 372.
(43) Vannice, M. A.; Sen, B. Journal of Catalysis 1989, 115, 65.
(44) Lin, S. D.; Sanders, D. K.; Albert Vannice, M. Applied Catalysis A: General 1994, 113,
59.
(45) Kijeński, J.; Winiarek, P. Applied Catalysis A: General 2000, 193, L1.
(46) Malathi, R.; Viswanath, R. P. Applied Catalysis A: General 2001, 208, 323.
(47) Kijeński, J.; Winiarek, P.; Paryjczak, T.; Lewicki, A.; Mikołajska, A. Applied Catalysis
A: General 2002, 233, 171.
(48) Enache, D. I.; Edwards, J. K.; Landon, P.; Solsona-Espriu, B.; Carley, A. F.; Herzing, A.
A.; Watanabe, M.; Kiely, C. J.; Knight, D. W.; Hutchings, G. J. Science 2006, 311, 362.
11
(49) Landon, P.; Collier, P. J.; Papworth, A. J.; Kiely, C. J.; Hutchings, G. J. Chemical
Communications 2002, 2058.
(50) Landon, P.; Collier, P. J.; Carley, A. F.; Chadwick, D.; Papworth, A. J.; Burrows, A.;
Kiely, C. J.; Hutchings, G. J. Physical Chemistry Chemical Physics 2003, 5, 1917.
(51) Fu, Q.; Saltsburg, H.; Flytzani-Stephanopoulos, M. Science 2003, 301, 935.
(52) Edwards, J. K.; Solsona, B. E.; Landon, P.; Carley, A. F.; Herzing, A.; Kiely, C. J.;
Hutchings, G. J. Journal of Catalysis 2005, 236, 69.
(53) Fan, S.; Yi, J.; Wang, L.; Mi, Z. Reaction Kinetics and Catalysis Letters 2007, 92, 175.
(54) Vernon, P. D. F.; Green, M. L. H.; Cheetham, A. K.; Ashcroft, A. T. Catalysis Letters
1990, 6, 181.
(55) Ashcroft, A. T.; Cheetham, A. K.; Green, M. L. H.; Vernon, P. D. F. Nature 1991, 352,
225.
(56) Nakagawa, K.; Ikenaga, N.; Suzuki, T.; Kobayashi, T.; Haruta, M. Applied Catalysis A:
General 1998, 169, 281.
(57) Haruta, M. Cattech 2002, 6, 102.
(58) Pietron, J. J.; Stroud, R. M.; Rolison, D. R. Nano Letters 2002, 2, 545.
(59) Molina, L. M.; Rasmussen, M. D.; Hammer, B. Journal of Chemical Physics 2004, 120,
7673.
(60) Akubuiro, E. C.; Verykios, X. E. Journal of Catalysis 1988, 113, 106.
(61) Baker, R. T. K.; Prestridge, E. B.; McVicker, G. B. Journal of Catalysis 1984, 89, 422.
(62) Komaya, T.; Bell, A. T.; Wengsieh, Z.; Gronsky, R.; Engelke, F.; King, T. S.; Pruski, M.
Journal of Catalysis 1994, 149, 142.
(63) Bernal, S.; Calvino, J. J.; Cauqui, M. A.; Gatica, J. M.; López Cartes, C.; Pérez Omil, J.
A.; Pintado, J. M. Catalysis Today 2003, 77, 385.
(64) Berlowitz, P. J.; Peden, C. H. F.; Goodman, D. W. The Journal of Physical Chemistry
1988, 92, 5213.
(65) Su, X.; Cremer, P. S.; Shen, Y. R.; Somorjai, G. A. Journal of the American Chemical
Society 1997, 119, 3994.
(66) Liu, L.; Zhou, F.; Wang, L.; Qi, X.; Shi, F.; Deng, Y. Journal of Catalysis 2010, 274, 1.
(67) Vayssilov, G. N.; Lykhach, Y.; Migani, A.; Staudt, T.; Petrova, G. P.; Tsud, N.; Skála,
T.; Bruix, A.; Illas, F.; Prince, K. C.; Matolı´n, V. r.; Neyman, K. M.; Libuda, J. Nat Mater 2011,
10, 310.
(68) Takakusagi, S.; Fukui, K.-i.; Tero, R.; Asakura, K.; Iwasawa, Y. Langmuir 2010, 26,
16392.
(69) Conner, W. C.; Falconer, J. L. Chemical Reviews 1995, 95, 759.
(70) Bonn, M.; Funk, S.; Hess, C.; Denzler, D. N.; Stampfl, C.; Scheffler, M.; Wolf, M.; Ertl,
G. Science 1999, 285, 1042.
(71) Lambert, R. M.; Cropley, R. L.; Husain, A.; Tikhov, M. S. Chemical Communications
2003, 1184.
(72) Zhang, Y.; Kolmakov, A.; Chretien, S.; Metiu, H.; Moskovits, M. Nano Letters 2004, 4,
403.
(73) Zhang, J.; Liu, X.; Blume, R.; Zhang, A. H.; Schlogl, R.; Su, D. S. Science 2008, 322, 73.
(74) Kaden, W. E.; Wu, T. P.; Kunkel, W. A.; Anderson, S. L. Science 2009, 326, 826.
(75) Zhang, Z.; Yates, J. T. Journal of the American Chemical Society 2010, 132, 12804.
(76) Seo, H.; Baker, L. R.; Hervier, A.; Kim, J.; Whitten, J. L.; Somorjai, G. A. Nano Letters
2010, 11, 751.
12
(77) Hervier, A.; Baker, L. R.; Komvopoulos, K.; Somorjai, G. A. The Journal of Physical
Chemistry C 2011.
(78) Baker, L. R.; Hervier, A.; Kennedy, G.; Somorjai, G. A. Nano Letters 2012, 12, 2554.
(79) Jencks, W. P. Accounts of Chemical Research 1980, 13, 161.
(80) Greeley, J.; Norskov, J. K.; Mavrikakis, M. Annu. Rev. Phys. Chem. 2002, 53, 319.
(81) Olah, G. A.; Molnar, A. Hydrocarbon Chemistry; 2nd Edition ed.; John Wiley & Sons,
Inc.: Hoboken, NJ, 2003.
(82) Yamaguchi, T. Applied Catalysis 1990, 61, 1.
(83) Takahashi, O.; Hattori, H. Journal of Catalysis 1981, 68, 144.
(84) Schmerling, L.; Vesely, J. A. The Journal of Organic Chemistry 1973, 38, 312.
(85) Guo, C.; Yao, S.; Cao, J.; Qian, Z. Applied Catalysis A: General 1994, 107, 229.
(86) Kovacic, P.; Kyriakis, A. Journal of the American Chemical Society 1963, 85, 454.
(87) Li, W.; Shen, Z.; Zhang, Y. European Polymer Journal 2001, 37, 1185.
(88) Volkova, G. G.; Plyasova, L. M.; Shkuratova, L. N.; Budneva, A. A.; Paukshtis, E. A.;
Timofeeva, M. N.; Likholobov, V. A. In Studies in Surface Science and Catalysis; Xinhe, B.,
Yide, X., Eds.; Elsevier: 2004; Vol. Volume 147, p 403.
(89) Gillespie, R. J. Accounts of Chemical Research 1968, 1, 202.
(90) Gillespie, R. J.; Peel, T. E.; Robinson, E. A. Journal of the American Chemical Society
1971, 93, 5083.
(91) Gillespie, R. J.; Peel, T. E. Journal of the American Chemical Society 1973, 95, 5173.
(92) Olah, G. A. Angewandte Chemie International Edition in English 1973, 12, 173.
(93) Heinerman, J. J. L.; Gaaf, J. Journal of Molecular Catalysis 1981, 11, 215.
(94) Kayo, A.; Yamaguchi, T.; Tanabe, K. Journal of Catalysis 1983, 83, 99.
(95) Yamaguchi, T.; Jin, T.; Tanabe, K. The Journal of Physical Chemistry 1986, 90, 3148.
(96) Jin, T.; Yamaguchi, T.; Tanabe, K. The Journal of Physical Chemistry 1986, 90, 4794.
(97) Ono, Y.; Tanabe, T.; Kitajima, N. Journal of Catalysis 1979, 56, 47.
(98) Ono, Y.; Yamaguchi, K.; Kitajima, N. Journal of Catalysis 1980, 64, 13.
(99) Maxwell, I. E. Catalysis Today 1987, 1, 385.
(100) Park, J. Y.; Somorjai, G. A. Chemphyschem 2006, 7, 1409.
(101) Hervier, A.; Renzas, J. R.; Park, J. Y.; Somorjai, G. A. Nano Letters 2009, 9, 3930.
(102) Maximoff, S. N.; Head-Gordon, M. P. Proceedings of the National Academy of Sciences
2009, 106, 11460.
(103) Baker, L. R.; Seo, H.; Hervier, A.; Somorjai, G. 2011.
13
Chapter 2
Experimental Methods
2.1 Catalyst Fabrication and Synthesis
2.1.1 Thin Film Deposition and Processing
2.1.1.1 Electron Beam Evaporation—In this work thin film deposition was an important
step in the fabrication of catalytic solid-state charge-based devices; oxide thin films also served
as Pt supports for studying metal–support interactions. Two methods were used for thin film
deposition: electron beam evaporation and magnetron sputtering. Electron beam evaporation
uses an electron beam to melt and vaporize a target material. This vapor then expands into a
vacuum and condenses on a desired substrate. Because the expansion of the target vapor into
vacuum is directional, shadow masks can control the pattern of the thin film deposited on the
substrate. This method is capable of depositing both metallic and insulating target materials.
The target material is placed inside a crucible liner that rests in a water-cooled hearth. A high
voltage (5 keV) is then applied between a W filament and the hearth. Subsequent ohmic heating
of the filament results in electron emission, and the electrons are accelerated across the applied
bias toward the hearth. The hearth is covered with a set of permanent magnets that direct the
electron beam from the filament to the target. The high energy electron beam heats the target
which first melts and then vaporizes into the vacuum. The pressure is maintained at 10−5 Torr or
less to allow the vapor to ascend directionally to the top of the chamber. The substrate is
suspended at the top of the chamber, and as the target vapor ascends, it deposits as a thin film on
the substrate.
2.1.1.2 Magnetron Sputtering—Magnetron sputtering uses the collision of ions on a target
material to eject atoms from the target to the gas phase, and these atoms subsequently deposit on
a substrate. This technique requires a backpressure of gas (usually Ar) in the tens of mTorr
range to generate ions for bombarding the target. The magnetron consists of a cathode and
anode, and a high voltage serves to ionize the Ar gas and accelerate the positive ions toward the
target. A magnetic field is typically present to control the trajectory of gas phase ions to further
enhance ionization by secondary collisions. The collision of Ar ions on the target ejects neutral
target atoms into the gas phase, and these atoms deposit on a substrate that is suspended in the
chamber above the magnetron. Magnetron sputtering can deposit both metallic and insulating
thin films. For metal deposition, a DC magnetron is typically used. For insulator deposition, an
AC magnetron operating at an RF frequency prevents charging of the insulating target. To
control the oxidation state of a deposited film, it is possible to leak in a backpressure of O 2
during sputtering.
2.1.1.3 Quartz Crystal Microbalance—In both electron beam evaporation and magnetron
sputtering, the film thickness can be monitored during growth using a quartz crystal
microbalance. A quartz crystal microbalance uses quartz as a piezoelectric material. An
oscillating voltage applied to the crystal causes it to oscillate; a resonant oscillation frequency
exists that is a function of adsorbed mass on the crystal surface according to the Sauerbrey
14
equation. By measuring the resonant frequency of oscillation, it is possible to extract the mass
on the surface of the crystal monitor. This mass is then converted to an average film thickness
on the substrate using the density and a geometric scaling factor (Z factor). The Z factor depends
on the relative positions of the crystal monitor and the substrate inside the deposition chamber.
2.1.1.4 Rapid Thermal Annealing—The crystallinity of the oxide thin films can be
enhanced after deposition by rapid thermal annealing (RTA). RTA uses high intensity lamps to
achieve fast substrate heating; temperatures in excess of 1,000 K are possible with ramp times
less than 1 min. The flow of gasses over the sample during annealing also provides a degree of
control over the oxidation state. Starting from an O deficient oxide film, the O stoichiometry can
be increased by annealing in O2. Alternately, RTA in an inert gas can increase the film
crystallinity without changing the oxidation state.
2.1.1.5 Plasma Fluorine Doping—F doping to titanium oxide thin films was achieved in a
parallel plate reactive ion etch (RIE) chamber. Gasses are introduced in the chamber by mass
flow controllers and the chamber is differentially pumped to maintain a pressure in the tens of
mTorr range. An electromagnetic field oscillating at RF frequencies generates an inductively
coupled plasma inside the chamber; this plasma serves as a supply of reactive ions. F ions are
generated at low concentration in the plasma by introducing a trace amount of SF6 in a
background of N2. A DC bias between the top and bottom plates directs the negative ions from
the plasma downward toward the sample. This bias is in the range of 150 V and serves to drive
negatively charged F ions into the titanium oxide thin film.
2.1.2 Preparation of 2-Dimensional Model Catalysts from Colloidal Nanoparticles
2.1.2.1 Colloidal Synthesis of Nanoparticles—Two methods are used for the colloidal
synthesis of Pt nanoparticles. Both methods rely on the reduction of chloroplatinic acid in
ethylene glycol. In one case chloroplatinic acid hexahydrate and PVP are combined in ethylene
glycol in a 1:4 mass ratio and heated to 438 K under a flow of Ar for 1 h. This results in 4.5 nm
PVP-capped Pt nanoparticles. These nanoparticles are washed by precipitation in hexanes
followed by redispersion in ethanol, and the number of washes determines the amount of PVP
left on the nanoparticles. The other method does not use a capping agent during synthesis.
Instead chloroplatinic acid is reduced in a solution of NaOH in ethylene glycol. Chloroplatinic
acid hexahydrate and NaOH are combined in ethylene glycol in a 1:1 mass ratio and heated to
433 K for 3 h. This results in 1.7 nm uncapped Pt nanoparticles. These nanoparticles are
precipitated by neutralizing the solution with 2 M HCl. The nanoparticles are then re-dispersed
in ethanol containing a desired capping agent. This synthesis allows for the preparation of
identical Pt nanoparticles where the only variable between samples is the capping agent.
2.1.2.2 Langmuir Blodgett Deposition—Formation of 2-dimensional films of

monodisperse nanoparticles to serve as model catalysts was achieved by Langmuir Blodgett (LB)
deposition. The nanoparticles are suspended in a hydrophobic solvent such as chloroform and
dispersed onto a water surface. Time is given for the solvent to evaporate, leaving a 2-
dimensional dispersion of nanoparticles. The nanoparticle film is then compressed with a mobile
barrier, and the surface pressure is monitored as a function of decreasing surface area. The
surface pressure corresponds to the density of nanoparticles on the water. When the desired
surface pressure is reached, a substrate is pulled out from under the surface of the water, and the
15
film of nanoparticles is deposited onto the substrate. The final density of nanoparticles on the
substrate can be controlled by controlling the surface pressure during deposition.
2.1.2.3 Organic Cap Removal by UV Cleaning—UV cleaning was used to remove the
organic cap from the Pt nanoparticles. This increases the Pt surface area available for reaction.
This also brings the Pt into close contact with the support for studying metal–support
interactions. Two low-pressure mercury (Hg) lamps (Lights Sources Inc., model number
GPH357T5VH/4P) are used as the UV source; the Hg lamps emit two lines at 184 and 254 nm.
The two lamps are aligned parallel to each other 2.5 cm apart in a clean Al box. The sample sits
1.2 cm below the lamps. By varying the time of UV exposure, it is possible to control the
amount of organic cap removed from the Pt nanoparticles. This cleaning is believed to be the
combined effect of direct photodecomposition of the organic cap and oxidation of the cap by
ozone produced from the 184 nm Hg line.
2.2 Catalyst Characterization
2.2.1 Electrical Probe Measurements
Electrical probe measurements were used to characterize the performance of solid-state

devices and as well as the surface conductivity of titanium oxide thin films. All electrical probe
measurements were performed using Keithley 2400 Sourcemeter. To evaluate the performance
of Schottky diode and heterojunction devices, current–voltage (IV) curves were measured by
scanning a probe voltage and measuring the current passed through the sample. From an IV
curve it is possible to determine many of the device properties including the rectification factor,
built-in potential, and photovoltaic efficiency. Rectification factor is taken as the ratio of current
passed at equal forward and reverse bias voltages. This value characterizes the efficiency of a
diode which is designed to pass current only in one direction. The built-in potential of a device
describes the difference in Fermi energy between the two active layers in a diode structure. It is
given as the voltage showing current onset in the forward bias direction of an IV curve. The
determination of rectification factor and built-in potential from an IV curve is depicted in
Fig. 1A.
Photovoltaic efficiency is determined by the IV curve of a photodiode during

illumination. The power output of the device is given as the area above the IV curve during
illumination integrated along the voltage axis from zero bias (i.e., closed circuit current) to the
intercept on the current axis (i.e., open circuit voltage) as shown in Fig. 1B. Normalizing this
value to the known power of the incident light gives the power conversion efficiency of the
photovoltaic device. Alternately, the quantum efficiency can be calculated by normalizing the
current flow during illumination to the number of incident photons. Quantum efficiency is
usually reported for closed circuit, zero bias conditions, but an applied bias can tune the quantum
efficiency of a device.
Conductivity of oxide thin films was also determined by IV measurements. In this case,
the slope of an IV curve represents the sample conductance. Conductivity is simply given by
normalizing the conductance to the dimensions of the conduction channel as shown in Eq. 1.
Eq. 1
16
50 4
Forward Bias
(A) Current (I+1.5V) (B)
40 Dark
2
Current Density (mA/cm2)
Current Density (mA/cm2)

Light
30
Rectification Factor = |I+1.5V / I−1.5V|
(1.4 x 104)
0
20
Open Circuit Voltage
10 (VO.C. = 0.35 V)
-2
0
Reverse Bias
Current (I−1.5V) -4
-10 Built-In Potential Closed Circuit Current
(VB = 0.5 V) (IC.C. = 4.5 mA/cm2)
-20 -6
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Bias (V) Bias (V)
Fig. 1 (A) IV curve of Pt/n-Si catalytic diode. From the IV curve the built-in potential and
rectification factor of the device are easily determined. (B) IV curves of Pt/n-Si catalytic diode
in dark and during illumination. The power output of the device is determined by integrating the
shaded area of the curve between the closed circuit current and the open circuit voltage.
where σ is the material conductivity, G is the sample conductance, L is the length of the
conduction channel, and A is the area of the conduction channel.
2.2.2 X-Ray Photoelectron Spectroscopy
X-ray photoelectron spectroscopy (XPS) provides elemental analysis of surfaces. XPS

can also resolve oxidation states of the various elements. The technique works using a
monochromatic x-ray source (usually from an Al or Mg anode) to excite core electrons from a
sample into the vacuum. The kinetic energy of these photoelectrons is measured using an
electron energy analyzer. The difference between the x-ray energy and the kinetic energy of the
photoelectron is equal to the binding energy of the electron in the sample. Because core
electrons in each element have a characteristic binding energy, this method provides elemental
surface analysis. Shifts in this binding energy are characteristic of the oxidation state of the
element. The surface sensitivity of XPS is determined by the mean free path of photoelectrons
escaping from the sample. Depending on the material and the kinetic energy of the
photoelectron, the mean free path is on the order of 1–2 nm, so XPS typically probes ~10 atomic
layers deep.
The monochromatic x-ray source is usually generated from an Al or Mg anode. A high

voltage (30–50 keV) between a W filament and the anode leads to electron emission from the
filament and collision on the anode. This collision of high energy electrons results in excitation
of core electrons. Subsequent electron relaxation leads to emission of well-defined x-ray lines
depending on the anode material. A filter or grating then selects a single emission line (usually
K-alpha) for use as the x-ray source.
A hemispherical analyzer is used to analyze the kinetic energy of photoelectrons ejected

from the sample. A hemispherical analyzer uses a magnetic field to bend the photoelectrons in
17
an arc. The radial trajectory of an electron in the analyzer depends on the strength of the
magnetic field and the kinetic energy of the electron. If a slit is placed in front of an electron
detector located inside the analyzer, only electrons of a specified kinetic energy will pass through
the slit and impinge on the detector for a given magnetic field strength. When the magnetic field
is scanned, electrons across a range of energies are detected resulting in a photoemission
spectrum.
XPS usually requires ultra-high vacuum to allow transmission of photoelectrons from the
sample to the energy analyzer. However, a differential pumping system allows for collection of
photoelectron spectra at pressures up to hundreds of mTorr. The differential pumping system
consists of three differentially pumped chambers. A nose cone that opens to the high pressure
sample cell is placed less than 1 mm from the sample surface. A series of magnetic lenses focus
the photoelectron beam through the apertures of each differential pumping stage and into the
electron energy analyzer. An outline of this detection system is shown in Scheme 1. This
technique known as ambient pressure XPS allows for in situ characterization of a catalyst at
moderate reaction pressures, and these experiments are performed using synchrotron radiation
from the Advanced Light Source at Lawrence Berkeley National Lab. Ex situ XPS used in this
study was performed at ultra-high vacuum on a commercial system (Physical Electronics, PHI
5400 ESCA/XPS) with an Al anode source at 1486.6 eV.
Sample
Electron Analyzer
X-Ray
To Pump To Pump To Pump
Scheme 1
2.2.3 Electron Microscopy
Catalyst morphology was examined before and after catalytic reactions using
transmission electron microscopy (TEM) and scanning electron microscopy (SEM). TEM
imaging was performed on a JOEL 2100 microscope operating at 200 kV. SEM imaging was
performed on a Zeiss Gemini Ultra-55 microscope operating at 5 kV.
2.2.4 Sum Frequency Generation Vibrational Spectroscopy
Sum frequency generation (SFG) vibrational spectroscopy is a nonlinear optical

spectroscopy that, by selection rules, is inherently surface or interface specific. In this work,
SFG was used to observe the vibrational spectrum of reactant molecules on a catalyst surface
during reaction; it is also used to probe the vibrational spectrum of organic capping agents on
18
catalytic nanoparticles in various reaction conditions. In SFG, an IR beam and a visible beam
overlap in time and space on the sample surface and a third beam is generated in reflection with a
frequency equal to the sum of the two input frequencies (see Scheme 2).
IR 532 nm SFG
Sapphire Window
Metal Oxide
Platinum
Scheme 2
The SFG signal intensity is proportional to the factor of the two input intensities as
shown in Eq. 2.
Eq. 2
Χ(2) is a tensor representing the second order polarizability of the sample, IIR is the IR beam
intensity and IVIS is the visible beam intensity. Χ(2) represents a spatial average of molecular
hyperpolarizabilities, Β, weighted by the number of molecules, N, on the surface as shown in
Eq. 3.
〈 〉 Eq. 3
An energy diagram for an SFG process is shown in Scheme 3. This shows SFG as the
coherent combination of an IR absorption event and a Stokes Raman scattering event.
Accordingly, the hyperpolarizability, Β, of a molecule is given as the dot product of the dipole
transition moment, μ (i.e., IR cross section), and the first order polarizability, α (i.e., Raman
cross section), as shown in Eq. 4.
Eq. 4
Molecular Cartesian coordinates are denoted as subscripts a, b, and c. SFG signal is enhanced
when μ diverges at a vibrational resonance, so fixing the frequency of the visible beam and
scanning the IR beam frequency across a range of interest produces the vibrational spectrum of a
sample.
The surface sensitivity of SFG is given by the following relationships involving Χ(2). The
first relationship given in Eq. 5 is an intrinsic property of Χ(2).
Eq. 5
Here I, J, and K represent lab coordinates, and this relationship can be derived from a coordinate
transformation of Eq. 3 and Eq. 4. In fact this relationship is true of all even order terms in the
19
VIS SFG
ν2
IR ν1
ν0
Scheme 3
polarizability expansion. The second relationship of Χ(2) is given in Eq. 6 and is true only of
centrosymmetric media.
Eq. 6
Substituting Eq. 5 into Eq. 6 shows that in centrosymmetric media
Eq. 7
Because Χ(2) is zero for centrosymmetric media, SFG is only sensitive to a break in inversion
symmetry which usually occurs at a surface or interface, and this relationship defines the surface
sensitivity of SFG.
Nonlinear spectroscopy means that the signal intensity does not depend linearly on the
intensity of a single probe beam such as in Raman or IR spectroscopy. In IR or Raman
spectroscopy, if the intensity of the input beam doubles, then the signal intensity also doubles
(i.e., the Raman scattering intensity in the case of Raman, or the absorption intensity in the case
of IR). In SFG if the beam input intensity doubles (i.e., IIR and IVIS each double) then the signal
intensity increases by a factor of 4 as shown in Eq. 2. That is why SFG is referred to as a
nonlinear spectroscopy. This dictates certain experimental requirements for SFG, specifically
high intensity pump lasers with ps or fs pulse widths. If the pulse width of the pump beams is
increased to the ns time-scale, the signal intensity decreases by the square of the intensity drop.
This loss in signal intensity is only compensated by a linear increase in the integration time, so
the signal loss would be approximately 2–3 orders of magnitude. This relationship coupled with
the relatively low cross-section for an SFG process (~10−6) precludes any source other than ps or
fs lasers for this experiment.
For SFG experiments, an active/passive mode-locked Nd:YAG laser (Leopard D-20,

Continuum) produces a 20 ps pulse at a 20 Hz repetition rate. The fundamental output at 1064
nm is passed through an optical parametric generator/amplifier to generate a tunable infrared
(IR) beam (2700–3600 cm−1) and a second harmonic visible (VIS) beam (532 nm). The IR
(100 μJ) and VIS (100 μJ) beams are spatially and temporally overlapped on the back surface of
20
a sapphire window containing the Pt nanoparticles supported on SiO2 or TiO2 thin films. The
VIS and IR beam are incident on the sample at 40° and 50° degrees, respectively, relative to
surface normal. The generated SFG signal is then collected and sent to a photomultiplier tube.
A gated integrator enhances the signal-to-noise. To collect a spectrum, the IR beam is scanned
across the spectral range of interest.
Fig. 2 Reactor cell for in situ SFG measurements. The catalyst temperature can be controlled up
to 523 K, and the cell can hold pressures up to 20 bar.
The beams are directed onto the sample using a sapphire prism as shown above in
Scheme 2. A solution of deuterated polystyrene (d8) in deuterated decalin (d18) serves as an
index matching liquid that does not interfere with transmission of the IR beam at the C–H stretch
frequency. The catalyst surface iss pressed into thermal contact with an aluminum heating block
to heat the catalyst to reaction temperature. A recess in the heating block allows for the flow of
reaction gasses across the catalyst surface. A metal bellows circulation pump provides gas
mixing. A gas tight seal is made between the sapphire window and the heating block using a
Kalrez O-ring. A picture of this system is shown in Fig. 2.
2.3 Catalytic Reaction Studies
Three identical stainless steel batch mode reactors were used for gas phase reaction
studies. Batch mode reactors allow products to accumulate over time during reaction and were
necessary to measure reaction kinetics on 2-dimensional catalysts because of their relatively low
surface area. Compared to high surface area 3-dimensional catalysts, 2-dimensional samples
were preferable for these studies because their planar geometry makes them compatible with the
characterization techniques described above. Batch reactor volumes were each 1 L. Prior to
each reaction, the chamber was evacuated to 10−5 Torr with a turbomolecular pump. The
catalysts were heated during reaction using a resistive boron nitride substrate heater with a
graphite filament (Momentive Performance Materials), and gas mixing was achieved using a
metal bellows circulation pump (Senior Flexonics). Catalysts were typically tested over the
course of several hours and reaction products were monitored as a function of time using gas
chromatography (GC) with thermal conductivity and flame ionization detectors. Conversion was
kept below approximately 10% to avoid the observation of secondary reaction pathways. An
internal gas sampling loop in the GC was placed in line of the recirculation pump to allow
automatic gas sampling at specified intervals during reaction. A diagram of the reactor system is
shown in Scheme 4, and a picture is shown in Fig. 3.
21
Recirculation
Pump
Gas Manifold
Sample Loop Reactor
GC
Turbo Pump
Ready
Gas Chromatograph
Mechanical Pump
Scheme 4
(A) (B)
(C)
Fig. 3 (A) Arial view of batch reactor with Pt/Si catalyst on the boron nitride substrate heater.
(B) Turbomolecular pump used evacuate the reactor to 10−5 Torr prior to reaction. (C) Gas
chromatograph used to detect reaction products.
22
Chapter 3
Solid-State Charge-Based Device for Control of

Catalytic Carbon Monoxide Oxidation on Platinum
Nanofilms Using External Bias and Light
3.1 Abstract
Using a Pt/Si catalytic nanodiode, we externally control the rate of CO oxidation on a Pt

nanofilm. The catalytic reaction can be turned on and off by alternating between bias states of
the device. Additionally, the reaction rate is sensitive to photocurrent induced by visible light.
The effects of both bias and light show that negative charge on the Pt increases catalytic activity,
while positive charge on the Pt decreases catalytic activity for CO oxidation.
3.2 Introduction
The development of more efficient catalysts depends on the ability to control electronic
interactions for the selective enhancement of desired reaction pathways.1-3 Driving catalytic
chemistry with a charge flow already forms the basis of electrochemistry, photochemistry, and
acid-base catalysis.4-9 However, an unexplored approach to charge-mediated catalysis is to use a
solid state device to externally control the catalyst electronic structure with an applied bias or a
photo-induced current. In this work, we demonstrate the viability of using a Pt/Si nanodiode to
control the rate of CO oxidation.
Many industrial catalysts consist of metal nanoparticles dispersed on a porous support.

Because the support is often chemically inert without the metal, the metal is usually considered
the active catalyst. However, numerous studies have demonstrated that the electronic properties
of the support play a major role in determining the activity and selectivity of the metal/support
system.10-15 This complex interaction between a metal and its support has been widely studied
and is commonly called the strong metal-support interaction (SMSI).16-20 We are interested in
SMSI because these interactions often give rise to highly active and selective reaction
pathways.3,21
SMSI affects a wide range of catalytic reactions, including CO and CO2 hydrogenation19-
21
, CO oxidation1,22-24, selective hydrogenation25-27, and selective partial oxidation reactions.2,28-30
We have recently demonstrated that for CO oxidation and methanol oxidation, the Pt/TiO2
interaction is controlled by electron transfer from the TiO2, which can be modified by chemical
doping.24,30
The ability to control the electric field and charge flow at a metal/support interface using
a solid state charge based device would provide nonthermal, electronic control of a catalytic
reaction.31-34 From SMSI studies, it is clear that charge flow at the metal/support interface has a
major influence on the rate and selectivity of a catalytic reaction. If this approach is viable, solid
23
state device technology could soon find major applications in the field of catalysis for tunable
control of selective surface chemistry.
The goal of activating nonthermal chemical processes electronically using a solid state
device is an active area of research, and attempts to drive nonthermal surface chemistry have
focused primarily on using a tunneling junction to emit hot electrons into a metal thin film.31,35-41
When the thickness of the metal thin film is less than the mean free path of the hot electron, it is
assumed that the electron reaches the surface before scattering to low energy electron hole pairs
and subsequent lattice vibrations (a process that occurs in ~1 ps).42 Several studies have
demonstrated the ability of these hot electron emitters to induce a chemical reaction involving a
single pre-adsorbed monolayer of reactant.35-37 However, no reports have been made that
demonstrate the ability to electronically activate a catalyst operating in steady state turnover by
directly measuring the product yield.
We have previously reported the production of hot electron flow during CO oxidation
using a catalytic nanodiode.43,44 In that work a catalytic reaction was used to drive a current flow
across a nanodiode. We demonstrate in this paper that the opposite is also possible: Using a
catalytic nanodiode, an externally applied bias and a photo-induced current flow are used to
control the rate of the catalytic reaction. We find that applying a negative charge to the Pt
increases the reaction rate, while a flux of photo-induced positive charges to the Pt decreases the
reaction rate. In both cases, the rate change is reversible under conditions of steady state
turnover.
3.3 Experimental
To fabricate the catalytic Schottky diode, Pt was deposited on n-type Si (100) by electron
beam evaporation. The n-type Si substrate was phosphorus doped to achieve a conductivity of
1–10 Ω cm. The Pt film was 4 nm thick as measured by quartz crystal microbalance. The
chamber base pressure for electron beam deposition was >1 × 10−5 Torr. No etching of the Si
native oxide was performed. We found that the native oxide did not act as a significant tunneling
barrier to charge flow. This was verified by two observations: 1) linear ohmic contact to the n-
Si with Al metal and 2) high photocurrent efficiency showing easy charge flow between the n-Si
substrate and the Pt active area.
Using a shadow mask, the Pt was deposited as a 4 × 6 mm rectangle which served as the
device active area. An aluminum (Al) pad provided ohmic contact to the n-type Si, and a gold
(Au) pad provided contact to the Pt active area. Both the Al and Au contact pads were 100 nm
thick. A layer of SiO2 (150 nm) insulated the Au pad from the Si substrate. These Al, Au and
SiO2 layers were each deposited by electron beam evaporation using shadow masks.
A halogen lamp was used to illuminate the catalytic diode through a sapphire window on
the reaction chamber. Lamp power was 1.35 W, the numerical aperture was 0.3, and the distance
from the lamp to the diode was 8 cm. This resulted in approximately 60 mW/cm2 of radiation
over the active area of the diode. The spectral profile of the lamp closely followed the solar
spectrum.
24
A batch mode reactor with a boron nitride substrate heater was used to determine activity
of the catalytic diode for CO oxidation. A 2:5 ratio of CO to O2 was used in the chamber in a
background of He. A metal bellows circulation pump provided gas mixing. Reactions were
performed at several temperatures between 423 and 443 K as measured by a type-K
thermocouple. CO2 production was monitored as a function of time using a gas chromatograph
with a thermal conductivity detector.
Electrical connections were made between the contact pads of the device and BNC
feedthroughs on the chamber using Au wire. Depending on the activity of the catalytic diode for
a given reaction condition, up to three diodes were connected in parallel to increase the catalytic
device area. A Keithley 2400 Sourcemeter was used to control the electrical bias across the
device, measure photocurrent, and obtain IV curves of the device during reaction.
Reaction rates are reported in turnover frequency (TOF) as CO2 molecules produced per
Pt site per minute. All error bars represent the 95% confidence interval based on the rate of CO 2
production normalized to the estimated number of Pt sites. The number of Pt sites was estimated
by assuming a uniform (111) surface structure over the entire catalyst area. Although this
calculation is approximate, it provides a consistent normalization to the catalyst area and yields a
reasonable estimate of the absolute TOF.
3.4 Results and Discussion
3.4.1 Electrical and Photo Properties of Catalytic Diode
The Pt served as the catalyst, and a Schottky barrier formed at the Pt/Si interface. The Pt
film was 4 nm thick. We found this was thin enough to transmit visible light, so the device was
also an efficient photodiode showing power conversion efficiency (PCE) of 1.5% and a
spectrally averaged incident photon to current conversion efficiency (IPCE) of 15%. Fig. 1A
shows the device architecture; Fig. 1B shows the current-voltage (IV) curves of the device in the
dark and during illumination by a halogen lamp (~60 mW/cm2) at room temperature. The
spectral profile of the lamp closely followed the solar spectrum.
3.4.2 Bias Effect on Catalytic Activity
Fig. 2A shows the effect of applied bias on the catalytic activity of the device for CO
oxidation. The experiment was performed in 40 Torr CO and 100 Torr O2 at 443 K. Exposing
the device to reaction conditions for approximately 10 h serves as a pretreatment necessary to
observe a rate enhancement by bias. During the pretreatment, the device which is initially active
in open circuit (i.e. no applied bias) deactivates as a result of Pt oxidation. Previous work has
shown that the activity of the oxidized nanofilm can be reactivated by heating to 523 K in CO to
reduce the Pt catalyst.24 However, we find that although the oxidized Pt nanofilm is not active in
open circuit, it responds to an applied bias. Following the pretreatment, we find that the catalyst
activity can be enhanced by applying a reverse bias across the device. As shown in Fig. 2A this
effect is reversible, so that the reaction rate can be turned on and off simply by switching the
applied bias. In the experiment shown, a 1 V reverse bias is used to enhance the catalytic
reaction rate, and the device is switched back and forth between an applied bias and an open
circuit (i.e. no applied bias) state corresponding to high catalyst activity and no activity,
25
(A) (B)
Fig. 1 (A) Schematic of the catalytic nanodiode (B) IV curves of the catalytic nanodiode in the
dark (red curve) and during illumination with visible light (blue curve) at room temperature.
The light source is a halogen lamp emitting 60 mW/cm2 with a spectral profile similar to the
solar spectrum.
respectively. This effect was observed under steady-state turnover conditions for approximately
8 h.
We attempted to perform similar experiments to observe how the magnitude of applied

bias affects the catalyst activity. For example, we may assume that applying a 0.5 V bias rather
than a 1 V bias would result in a smaller rate enhancement, while a 2 V bias would yield a
greater rate enhancement. The major challenge of this experiment is that the bias effect is
sensitive to the oxidation state of the Pt nanofilm which is changing during reaction, and we find
that the rate enhancement is more sensitive to the oxidation state of the catalyst than it is to the
magnitude of the applied bias. Consequently, we could not quantify the magnitude of rate
enhancement as a function of applied bias in a way that is statistically significant.
The direction of the applied bias is important because of the rectifying nature of the
diode. A forward bias induces a current flow, and we find that under conditions of high current
flow induced by a forward bias, the device is unstable in reaction. However, because of the
Schottky barrier between the Pt and n-type Si, a reverse bias does not induce a current flow.
Rather an electric field is generated at the Pt/Si interface with a negative charge build-up on the
Pt and a positive charge build-up on the Si as shown in Fig. 2B. It appears that the negative
charge on the Pt enhances catalytic activity. Our studies of hot electron flow during CO
oxidation using a Pt/TiO2 catalytic nanodiode could be explained by density functional theory
showing that the reaction proceeds via a negatively charged CO2− transition state.45 This finding
implicates the important role of negatively charged reaction intermediates leading to a
correlation of turnover rates with electron flow.43,44 Accordingly, it is not surprising that
generating negatively charged intermediates by external bias leads to an enhancement of the
turnover rate.
26
(A)
(B)
Fig. 2 (A) Effect of bias on the catalytic activity of the nanodiode for CO oxidation. Red squares
show the reverse bias that was externally applied to the device during CO oxidation. Blue
circles show the corresponding turnover frequencies (TOF). The error bars represent 95%
confidence intervals for the TOF measurements. (B) Band diagrams of the nanodiode in open
circuit and during reverse bias. During reverse bias, positive charge builds-up on the Si and
negative charge builds-up on the Pt. This increases band bending in the Si and results in a high
electric field at the Pt/Si interface.
3.4.3 Light Effect on Catalytic Activity
We also explored the possibility of using visible light to control the reaction rate by
generating a photocurrent across the device. Fig. 3A shows the reaction rate for CO oxidation on
the catalytic device first in the dark, then under illumination, then in the dark again. Fig. 3A also
shows the corresponding photocurrent measurements made during the reaction. The results
show that the photocurrent decreases the reaction rate by a factor of ~3. As with the bias, this
effect is reversible, and the rate increases again when the light is turned off.
The photochemistry experiment was performed in 8 Torr CO and 20 Torr O2 at 423 K.

During this experiment, the reverse bias was kept fixed at 0.6 V. The role of the bias was to
increase the reaction rate as well as the efficiency of photocurrent generation which scales with
reverse bias. It is important to note that prior to the reaction, the device was subjected to an
oxidizing pretreatment at 473 K. We found that the oxidizing pretreatment greatly increased the
reaction rate at reverse bias (the same as in the above case showing a turnover rate increase by
27
external bias without light), and this pretreatment was necessary to observe an effect by light on
the reaction rate. Again, this implicates the important role of Pt oxidation in observing an effect
of charge flow on the catalytic reaction rate.
(A)
(B)
Fig. 3 (A) Effect of photocurrent on the catalytic activity of the nanodiode for CO oxidation.
Red squares show the current flow across the device in the dark and during illumination under
reaction conditions. Blue circles show the corresponding TOF. The error bars represent 95%
confidence intervals for the TOF measurements. During this experiment, the reverse bias was
kept fixed at 0.6 V. (B) Band diagram of the device showing the mechanism of photocurrent
generation. The 4 nm Pt film is thin enough to transmit visible which is absorbed in the Si,
generating electron-hole pairs. Because of the interface potential, the holes move across the
Pt/Si interface resulting in a flow of positive charge to the Pt surface.
The mechanism of photocurrent flux is generation of an electron-hole pair in the Si

followed by hole injection to the Pt as shown in Fig. 3B. Accordingly, photocurrent represents a
flux of positive charge to the Pt catalyst and negative charge to the Si support. By bias we
showed that a negative charge on the Pt increases the reaction rate; now we show that a flux of
positive charge to the Pt decreases the reaction rate.
We found the chemical changes of the device under reaction conditions played an
important role in determining both the bias effect and the photon effect. The bias effect shown in
Fig. 2A was reversible for about 8 h prior to losing control of the device. The photon effect
shown in Fig. 3A was shorter, lasting for only about 1 h. However, in both cases, the device was
28
effective long enough to control the chemistry for at least several hundred turnover events per
active site. There appears to be a window in the oxidation state of Pt that shows a dramatic
effect by bias and light on the catalytic activity for CO oxidation. At a lower oxidation state of
the Pt nanofilm, the device is less sensitive to the applied bias and the photocurrent flux, while at
a higher oxidation state, the catalytic activity to too low to measure. In both the bias and the
light experiments, the device pretreatments were empirically optimized to yield the greatest
effect by bias and light, and a detailed description of device hysteresis is discussed below.
3.4.4 Device Stability and Hysteresis
We found that the history of the device in various gas conditions played a major role in
determining both the bias effect and the photon effect. Specifically, we observed switching
behavior by the device induced by treatment in oxidizing conditions, especially during forward
bias. Fig. 4 shows the IV curve of the device under reaction conditions initially and again
following extended time in reaction conditions. We find that the same switching occurs
following extended treatment in pure O2 at reaction temperature (443 K). The initial state of the
device, which we call state A, is characterized by a high forward bias current and good
rectification. The final state of the device, which we call state B, is characterized by a low
forward bias current and an almost symmetrical IV curve with respect bias. We found that the
longer the device was exposed to oxidizing conditions, the more likely it was to fall into state B,
although, there were several instances that we were able to induce a switch back to state A for a
short amount of time by applied bias.
Fig. 4 Current-voltage curves of the device initially and after extended time in reaction
conditions.
Forward bias seems to promote the switching behavior. Fig. 5 shows a current-time (IT)
curve of the device under reaction condition at a fixed forward bias. The sudden drop in current
at 14 s indicates the switch from state A to B. Fig. 6 shows the same type of switch occur during
an IV curve measurement where at 0.7 V forward bias, the device suddenly switches from state
A to B. The switch from state A to B corresponds to the incorporation of subsurface O into the
29
Pt catalyst. This explains why extended treatment in oxidizing conditions is necessary to induce
the switch and explains the drop in forward bias conductivity after the switch.
We find that the switch also corresponds with a major change in the catalytic behavior of
the device. Fig. 7 shows the reaction rate measured over time while the device is alternated
between open circuit and a 1 V reverse bias. Each time the bias was changed, the IV curve of the
device was measured to determine its state, and the switch from state A to B occurred at 596 min
of time in reaction. Before the switch, we find that the device is catalytically active in open
circuit, and the activity is decreased when a reverse bias is applied. However, after the switch,
the device is inactive in open circuit, and the activity is increased when a reverse bias is applied.
Fig. 5 Current-time curve at 1 V forward bias showing a sudden switch of the device from its
initial conductive state to an insulating state.
We suggest the following explanation: Initially the Pt is metallic and the CO oxidation
reaction is rate limited by activation of the Pt-C bond of the adsorbed CO. It was recently
demonstrated by Wolf and coworkers that in a similar Pt/TiO2 Schottky diode system, a reverse
bias results in a red shift of the vibration frequency of adsorbed CO.34 This red shift corresponds
to electron back donation to the π* orbital of the CO. This orbital is anti-bonding with respect to
C-O, but is bonding with respect to Pt-C. Consequently, we can assume that the effect of reverse
bias is to weaken the C-O bond but to strengthen the Pt-C bond. On metallic Pt where the
reaction rate is limited by activating adsorbed CO, strengthening the Pt-C bond by reverse bias
results in a decreased reaction rate.
However, the reaction kinetics change upon incorporation of subsurface O in the Pt. In
this case, the reaction proceeds via a Mars-van Krevelen mechanism in which adsorbed CO
reacts with O in the Pt lattice, and Pt-O bond activation is rate limiting. Our group and others
have demonstrated that metal-O bond activation occurs by negative charge.24,42,46 This explains
why in state B, the device is only active when a negative bias is applied to the Pt, and the activity
decreases during illumination when a flux of positive charge moves from the Si substrate to the
Pt surface.
30
Fig. 6 Current-voltage curve of the device showing a sudden switch from a conductive state to an
insulating state at 0.7 V forward bias.
Fig. 7 Turnover frequency (TOF) for CO oxidation on the catalytic diode as the bias is switched
between open circuit and 1 V reverse bias. Between each TOF measurement a current-voltage
curve is measured to determine the state of the device. A switch occurs from state A (conductive)
to state B (insulating) during the experiment. We find that the effect of bias on catalytic activity
depends strongly on the state of the device.
3.5 Conclusions
This work demonstrates direct control of catalytic CO oxidation using a catalytic

nanodiode. This is the first time that a charge-based device has been shown to provide external
control of a surface catalytic reaction as determined by measuring the product yield, and this was
achieved using both applied bias and visible light. Based on these results, we believe that
31
charge-based device technology will be further developed for catalytic applications in energy
conversion and selective chemistry.
3.6 References

(3) Yamada, Y.; Tsung, C.-K.; Huang, W.; Huo, Z.; Habas, S. E.; Soejima, T.; Aliaga, C. E.;
Somorjai, G. A.; Yang, P. Nat Chem 2011, 3, 372.
(4) Fujishima, A.; Honda, K. Nature 1972, 238, 37.
(5) Ni, M.; Leung, M. K. H.; Leung, D. Y. C.; Sumathy, K. Renewable and Sustainable
Energy Reviews 2007, 11, 401.
(6) Osterloh, F. E. Chemistry of Materials 2007, 20, 35.
(8) Greeley, J.; Norskov, J. K.; Mavrikakis, M. Annu. Rev. Phys. Chem. 2002, 53, 319.
(9) Olah, G. A.; Molnar, A. Hydrocarbon Chemistry; 2nd Edition ed.; John Wiley & Sons,
Inc.: Hoboken, NJ, 2003.
(10) Herrmann, J. M.; Pichat, P. Journal of Catalysis 1982, 78, 425.
(11) Meriaudeau, P.; Ellestad, O. H.; Dufaux, M.; Naccache, C. Journal of Catalysis 1982, 75,
243.
(12) Sexton, B. A.; Hughes, A. E.; Foger, K. Journal of Catalysis 1982, 77, 85.
(13) Resasco, D. E.; Haller, G. L. Journal of Catalysis 1983, 82, 279.
(14) Benvenutti, E. V.; Franken, L.; Moro, C. C.; Davanzo, C. U. Langmuir 1999, 15, 8140.
(15) Jochum, W.; Eder, D.; Kaltenhauser, G.; Kramer, R. Top Catal 2007, 46, 49.
100, 170.
(24) Baker, L. R.; Hervier, A.; Seo, H.; Kennedy, G.; Komvopoulos, K.; Somorjai, G. A. The
Journal of Physical Chemistry C 2011, 115, 16006.
59.
General 1998, 169, 281.
(30) Hervier, A.; Baker, L. R.; Komvopoulos, K.; Somorjai, G. A. The Journal of Physical
Chemistry C 2011.
32
(31) Gadzuk, J. W. Phys Rev Lett 1996, 76, 4234.
(32) Gergen, B.; Nienhaus, H.; Weinberg, W. H.; McFarland, E. W. Science 2001, 294, 2521.
403.
(34) Deshlahra, P.; Schneider, W. F.; Bernstein, G. H.; Wolf, E. E. Journal of the American
Chemical Society 2011, 133, 16459.
(35) Sharpe, R. G.; Dixon-Warren, S. J.; Durston, P. J.; Palmer, R. E. Chemical Physics
Letters 1995, 234, 354.
(36) Wadayama, T.; Kojim, A.; Hatta, A. Applied Physics A: Materials Science & Processing
2004, 79, 1891.
(37) Wadayama, T.; Yokawa, M. Chemical Physics Letters 2006, 428, 348.
(38) Diesing, D.; Janssen, H.; Otto, A. Surface Science 1995, 331–333, Part A, 289.
(39) Diesing, D.; Kritzler, G.; Stermann, M.; Nolting, D.; Otto, A. Journal of Solid State
Electrochemistry 2003, 7, 389.
(40) Thomsen, L. B.; Nielsen, G.; Vendelbo, S. B.; Johansson, M.; Hansen, O.; Chorkendorff,
I. Physical Review B 2007, 76, 155315.
(41) Ray, M. P.; Lake, R. E.; Thomsen, L. B.; Nielson, G.; Hansen, O.; Chorkendorff, I.;
Sosolik, C. E. Journal of Physics: Condensed Matter 2010, 22, 084010.
G. Science 1999, 285, 1042.
(44) Hervier, A.; Renzas, J. R.; Park, J. Y.; Somorjai, G. A. Nano Letters 2009, 9, 3930.
(45) Maximoff, S. N.; Head-Gordon, M. P. Proceedings of the National Academy of Sciences
2009, 106, 11460.
33
Chapter 4
Controlling the Rate of Hydrogen Oxidation on

Platinum/Silicon Catalytic Nanodiodes by
Photocurrent Generation with Visible Light
4.1 Abstract
We report here that turnover frequency of hydrogen oxidation on a Pt/TiO2 nanodiode

can be decreased by exposing the device to visible light. The magnitude of this effect could not
be measured accurately, since the heater used for the experiment desorbed a measurable amount
of hydrogen under illumination, which interfered with the measurement. However, applying a
reverse bias to the device enhances the effect of light, indicating that the light is in fact
modifying catalysis on the surface of the diode.
4.2 Introduction
Solar energy is an increasingly attractive alternative to fossil fuels, but it presents the
major disadvantage of only being available during the day. One of the solutions to this problem
is to use solar energy to drive a chemical reaction on a photocatalytic device. The reaction
product can then be used to regenerate electricity when solar energy is unavailable, or as
transportation or heating fuel.1 The device must be able to generate electron–hole pairs from
visible light, separate those charges in order to avoid their recombination, and finally use the
electrons, the holes, or both to drive a chemical reaction. This is often achieved by placing
catalytic particles in contact with a photovoltaic assembly, where it can collect electrons and
holes, all of this on the nanoscale.2-5
A typical solar cell uses a p–n junction to separate charges, but nanodiodes have been
proposed as an alternative architecture.6 The Schottky barrier at a metal–oxide interface serves
the role of the charge separator. Depending on the direction of the band bending, electrons or
holes are injected into the metal, where they can affect surface chemistry. These charges have
overcome the Schottky barrier, and have energies in significant excess of the Fermi level. If the
metal is thinner than their mean free path, these charges may interact with adsorbates before they
relax, providing additional energy to do chemistry. This idea of collecting charge carriers while
they are still hot has already been suggested for purely photovoltaic solar cells as a strategy for
increasing efficiency.7
We report here that a 4 nm Pt film deposited on Si acts as a photodiode and that its
catalytic activity can be affected by light. Visible light excites electron-hole pairs in Si, and the
holes are injected into Pt. We show that under hydrogen oxidation conditions, the rate of
formation of water is decreased under illumination, indicating that the flow of charges to the
surface interferes with the reaction mechanism. Unfortunately, illumination of the diode also
caused the boron nitride sample heater to desorb H2 into the reactor. Becuase the reaction rate
34
was measured by following the decrease in the amount of H2 by gas chromatography, this artifact
interfered with the observation of the effect of light on the diode.
Nonetheless, the effect of light is enhanced by applying a reverse bias to the Pt/Si diode,
which shows that some of the change in reactivity can be attributed to the effect of light.
4.3 Experimental
To fabricate the catalytic Schottky diode, Pt was deposited on n-type Si (100) by electron
beam evaporation. The n-type Si substrate was phosphorus doped to achieve a conductivity of 1–
10 Ω cm. The Pt film was 4 nm thick as measured by quartz crystal microbalance. The chamber
base pressure for electron beam deposition was >1 × 10−5 Torr. No etching of the Si native oxide
was done prior to Pt deposition, because the native oxide acted as a tunneling barrier to increase
the rectification of the diode. Using a shadow mask, the Pt was deposited as a 4 × 6 mm
rectangle which served as the device active area.
An aluminum (Al) pad provided ohmic contact to the n-type Si, and a gold (Au) pad
provided contact to the Pt active area. Both the Al and Au contact pads were 100 nm thick. A
layer of SiO2 (150 nm) insulated the Au pad from the Si substrate. These Al, Au and SiO2 layers
were each deposited by electron beam evaporation using shadow masks.
The nanodiode is placed on a ceramic heater, and the temperature is monitored by a

thermocouple, fluctuating by less than 0.5 K. The turnover of water molecules was obtained by
monitoring the amount of H2 during the reaction, using a gas chromatograph connected to the
chamber by a sample loop equipped with a metal bellows circulation pump. Hydrogen oxidation
was carried out in 10 Torr H2, 30 Torr O2, and 720 Torr He, at 423 K. Prior to reaction, samples
were heated to 423 K in 100 Torr O2 for one hour. This pretreatment led to increased turnover,
facilitating the measurements.
The nanodiode’s electrodes are connected to an electrical circuit with gold wires, making
it possible to characterize the device by measuring its IV curve. This was done with a Keithley
Sourcemeter, which was also used to measure chemicurrent passing through the diode while the
chemical reaction is taking place. No voltage bias is applied to the diode.
A halogen lamp was used to illuminate the catalytic diode through a sapphire window on
the reaction chamber. Lamp power was 1.35 W, the numerical aperture was 0.3, and the distance
from the lamp to the diode was 8 cm. This resulted in approximately 60 mW/cm2 of radiation
over the active area of the diode. The spectral profile of the lamp closely followed the solar
spectrum. On a typical diode, current at 443 K was on the order of 0.5 mA in dark conditions,
and 1.5 mA under illumination, for an active area of 24 mm2. Current in dark conditions is likely
due to thermionic emission from Pt to Si.
Fig. 1 shows the rate of hydrogen oxidation on a Pt/Si diode, measured by gas
chromatography, in the dark and under illumination at 423 K. Before each on/off cycle, the
reactor was evacuated down to a pressure of 10−5 Torr, and the mixture introduced was
35
comprised of 10 Torr H2, 30 Torr O2 and 740 Torr N2. The rate under illumination is
consistently lower than in the dark.
16
Light On
Turnover Frequency
12
Light Off
(H2O/Pt site/s)
Fig. 1 Turnover frequency of hydrogen oxidation in 10 Torr H2, 30 Torr O2 and 740 Torr He at
423 K. Turnover is calculated based on the rate of decrease of the hydrogen peak height in the
chromatogram. In between each on/off cycle, the chamber was pumped down to 10−5 Torr.
Error bars represent 95% confidence intervals.
As a control experiment, the same procedure was carried out in the absence of O2. In
these conditions, no reaction is occurring on the surface of the diode, and yet fluctuations in the
H2 concentration in the reactor still correlate with light exposure, as shown in Fig. 2. These
fluctuations disappeared when the boron nitride sample heater was covered in aluminum foil,
suggesting that the heater takes up H2, and desorbs it continuously under illumination.
Any effect on the reaction from illuminating the diode cannot be separated from this
artifact. However, since the diode is connected into a circuit through electrical feedthroughs on
the reactor, it can be subjected to a voltage bias. In another set of experiments, H2 oxidation was
performed in the same conditions while different voltage biases were applied to the diode. Fig. 3
shows the rates under illumination and in the dark at 0, −0.6, −0.3 and −0.9 V. Again, the rates
under illumination are lower than in the dark, but the ratio between the two decreases with
applied bias. The ratio of the light on rate to the light off rate is plotted versus voltage in the
graph in Fig. 4.
Turnover rates tend to change over time: the active surface area and the surface coverage
may change as a poison takes up surface sites, and the film may anneal or oxidize under the
harsh conditions used for the reaction. To avoid confusing any of these effects with the effect
from bias, the voltages were not applied in decreasing order, but instead in the following order:
0, −0.6, −0.3 and −0.9 V. The decrease in turnover by light is enhanced with greater reverse
bias. A final measurement was then performed at 0 V again, and showed a value similar to the
initial 0 V measurement.
36
8
Turnover Frequency
4
(H2O/Pt site/s)
Light On
-4
Light Off
-8
Fig. 2 Turnover frequency of hydrogen oxidation in 10 Torr H2, 770 Torr He and no O2 at 423
K. In the absence of O2, there should be no turnover, and the non-zero values observed are the
result of H2 leaking from the reactor, reacting with traces of O2, and desorbing from the heater
under illumination.
30
Turnover Frequency
Light Off
25
(H2O/Pt site/s)
Light On
20
15
10
5
0
0.0 0.3 0.6 0.9
Reverse Bias (V)
Fig. 3 Turnover frequency for hydrogen oxidation in 10 Torr H2, 30 Torr O2 and 740 Torr He at
423 K, over a series of on / off cycles with different bias voltages applied to the diode. In order
to avoid artifacts in device history, the chronological order of the voltages applied was 0 V, −0.6
V, −0.3 V and −0.9 V.
The effect from bias shows that the light does affect the surface chemistry of the diode.
The Pt film is 4 nm thick, and transmits most of the light through to the Si substrate. Si, with its
bandgap of 1.1 eV, absorbs visible light, resulting in the formation of electron hole pairs.8 The
direction of band bending in a Pt/n-Si is such that excited electrons in the conduction band
remain in the silicon, while excited holes are accelerated towards the metal.
37
Fig. 4 Ratio of turnover frequency under illumination to turnover frequency in the dark at
different bias voltages, calculated from the values plotted in Fig. 3 (full circles). Empty circle
represents a final on/off cycle at 0 V carried out after the four others.
Fig. 5 Schematic band energy diagrams for a Pt/Si diode at 0 V and −1V voltage biases,
respectively.
The Si wafers used were covered in an insulating layer of SiO2 several nanometers thick.
The bandgap of SiO2 is 9 eV, so excited holes in Si must tunnel through this native oxide to
reach the Pt film.9 This decreases the efficiency of the diode. The SiO2 film can be etched away
with HF. However, this film grows back to 1 nm thickness in air in only 1 hour.10 Fabricating a
diode with no tunneling barrier for the holes is therefore quite difficult. Because photocurrent
38
obtained with the native oxide present is substantial enough to affect the catalysis, no attempt
was made to remove it.
Illuminating the diode would also increase its temperature, from photon absorption, and
from Joule heating due to the photocurrent. This can safely be ruled out as a cause of the change
in turnover, since an increase in temperature under illumination would increase turnover, not
decrease it.
It would be useful to know the efficiency of the process by which light hinders turnover
in order to understand the mechanism at play. The efficiency for generation of photocurrent in
the diode is known, as is the number of reaction events occurring on the surface. It would
therefore be possible to know how many reaction events are prevented for each hole that is
injected from Si into Pt. Desorption of H2 from the sample heater unfortunately prevents us from
making this calculation, and more experiments are needed to better understand this phenomenon.
Chapter 6 covers similar, more succesful experiments performed with CO oxidation.
4.5 Conclusions
Despite the sample heater interfering with the results, we have shown that turnover for
hydrogen oxidation can be significantly reduced on a Pt/TiO2 photodiode by exposing the diode
to visible light. This is not due to heating of the catalyst under illumination, but instead to a
steady-state flow of hot holes to the surface. The mechanism for this process is not well
understood, and further experiments are required. Without the effect of H2 desorbing from the
heater under illumination, it would be possible to calculate the efficiency for this process, which
would be useful in understanding the mechanism. A comparison of D2 and H2 oxidation would
also be particularly useful in this regard.
This work shows that a catalytic device can be designed that allows us to tune surface
chemistry with light. Using different diode materials, and with a different reaction, it may be
possible to use light to increase activity, or achieve higher selectivity in the case of a multi-path
reaction.
4.6 References
(1) Lewis, N. S.; Nocera, D. G. Proceedings of the National Academy of Sciences 2006, 103,
15729.
(2) Yan, H.; Yang, J.; Ma, G.; Wu, G.; Zong, X.; Lei, Z.; Shi, J.; Li, C. Journal of Catalysis
2009, 266, 165.
(3) Amirav, L.; Alivisatos, A. P. J. Phys. Chem. Lett. 2010, 1, 1051.
(4) Maeda, K.; Xiong, A.; Yoshinaga, T.; Ikeda, T.; Sakamoto, N.; Hisatomi, T.; Takashima,
M.; Lu, D.; Kanehara, M.; Setoyama, T.; Teranishi, T.; Domen, K. Angewandte Chemie
International Edition 2010, 49, 4096.
(5) Zong, X.; Han, J.; Ma, G.; Yan, H.; Wu, G.; Li, C. J. Phys. Chem. C 2011, 115, 12202.
(6) McFarland, E. W.; Tang, J. Nature 2003, 421, 616.
(7) Ross, R. T. Journal of Applied Physics 1982, 53, 3813.
(8) Kittel, C. Introduction to solid state physics; Wiley, 2005.
39
(9) DiStefano, T. H.; Eastman, D. E. Solid State Communications 1971, 9, 2259.
(10) Raider, S. I.; Flitsch, R.; Palmer, M. J. Journal of The Electrochemical Society 1975, 122,
413.
40
Chapter 5
Generation of Highly n-Type Titanium Oxide Using

Plasma Fluorine Insertion
5.1 Abstract
True n-type doping of titanium oxide without formation of mid-gap states would expand
the use of metal oxides for charge-based devices. We demonstrate that plasma-assisted F
insertion passivates defect states and that fluorine acts as an n-type donor in titanium oxide. This
enabled us to modify the Fermi level and transport properties of titanium oxide outside the limits
of O vacancy doping. The origin of the electronic structure modification is explained by ab
initio calculation.
5.2 Introduction
The application of transition metal oxide (TMO) semiconductors for charge-based

devices is becoming prevalent for chemical, electrical, and energy applications.1–3 Among
TMOs, titanium oxide has been widely studied for photocatalysis,4 light harvesting,5 resistive
switching memory,6 and chemical sensing,7 as an intrinsic n-type semiconductor. The charge
transport/transfer in titanium oxide is highly sensitive to O vacancy defects that exist as the
lattice bonding defects in single-crystals and as surface and grain boundary defects in
polycrystalline films.8 In previous studies, O vacancies, equivalent to Ti3+ states, have been used
as intrinsic donors to create a transport channel for increased conductivity.9,10
O vacancies give rise to partially occupied Ti3+ states below the conduction band (CB)
edge and act as n-type donors. Consequently, limited control of titanium oxide conductivity and
Fermi level (EF) is possible by the manipulation of O vacancies. However, the intrinsic doping
of titanium oxide through Ti3+ defect sites is equivalent to the formation of a reduced oxide.
Although, the reduced oxide is often highly conductive, Ti3+ mid-gap states act as a barrier to CB
transport and result in EF pinning.11 Additionally, the electric activity (e.g. charge trapping) and
chemical reactivity of O vacancies create problems for material reliability. Consequently, highly
conductive titanium oxide is seldom useful for device architecture, and fabrication of highly
conductive, defect-free titanium oxide remains an important challenge.
The extrinsic chemical doping of titanium oxide using impurity dopants (e.g. N, S, and C)
by colloidal synthesis methods to modify the electronic structure and achieve visible light
absorption has been actively studied for photocatalytic applications.12,13 Additionally, metal
impurity (e.g. V, Ni, and Nb) doping to titanium oxide has been used to modify E F and increase
the conductivity.14,15 However, in each of these cases, the semiconductor properties of titanium
oxide are compromised by the formation of metallic mid-gap states. To the best of our
knowledge, no one has reported n-type doping of a metal oxide without simultaneously
decreasing the bandgap. This ability would be analogous to phosphorus or arsenic doping of
silicon where carrier concentration is changed without significant band structure modification.
41
Here we report the use of plasma-assisted F insertion to passivate defects and act as an
extrinsic n-type donor in titanium oxide. Electrical analysis shows that F insertion is an excellent
method to achieve highly conductive, low defect titanium oxide suitable for TMO device
applications. Theoretical calculations also investigate the concept of n-type doping to metal
oxides by F insertion and suggest the mechanism by which F generates a highly conductive
surface channel.
5.3 Experimental
Titanium oxide films were deposited by direct current magnetron sputtering. During
sputtering, plasma power was 400 W, bias voltage was 450 V, Ar flow was 50 SCCM, and O 2
flow was 3 SCCM. Two different film thicknesses and substrates were used in this work: 100
nm films deposited on an insulating substrate and 10 nm films deposited on highly doped (0.01–
0.02 Ω∙cm) p-type Si (100) with native oxide. As deposited, the titanium oxide films were
highly O deficient.
Rapid thermal annealing (RTA) was then used to tune the O vacancy concentration.
Three pre-conditions were prepared for F insertion with sub-oxide concentrations ranging from
41 to < 3% (measured as the ratio of Ti4+ to Ti3+ peak areas in the Ti 2p XPS spectrum). To
achieve this distribution of sub-oxide concentrations, all samples were first annealed in N2 at
500 °C to increase crystallinity. Following RTA in N2, annealing in O2 at different temperatures
served to increase the O stoichiometry in the film. No O2 annealing produced films with 41%
sub-oxide; RTA in O2 at 350 °C produced films with 8% sub-oxide; and RTA in O2 at 500 °C
produced films with <3% sub-oxide. It should be noted that the actual stoichiometries varied
from one deposition to another. Consequently, all results reported here are taken from a single
deposition. Although the exact values may vary, the trends reported are consistent between
depositions.
F insertion was achieved by plasma treatment in N2 gas with trace SF6. Trace SF6 was
introduced into the chamber by flowing a 9:1 mixture of SF6 and O2 followed by pumping to
chamber base pressure (5 mTorr). After pumping, a small background pressure of SF6 remained
in the chamber. N2 gas was then introduced at 80 SCCM for plasma treatment. Plasma power
was 20 W with 130 V DC substrate bias. F concentration is tunable with the length of the
plasma treatment.
X-ray photoelectron spectroscopy (XPS) was used to analyze the chemical binding states
and valence band edge of the titanium oxide films (Physical Electronics, PHI 5400 ESCA/XPS
system with an Al anode source at 1486.6 eV). The energy resolution for each point is 0.05 eV.
The analyzer was positioned at 50 degrees relative to sample normal. All binding energies were
calibrated to the C 1s peak state.
The real and imaginary parts of the complex dielectric constant, εc = ε1+iε2, and
absorption coefficients for titanium oxide samples were determined by visible-ultra violet
spectroscopic ellipsometry in a rotating compensator enhanced spectrometer. The
monochromatic light source from a Xenon lamp at photon energies of 1.5–6 eV was used with
spectral resolution of 15 meV.
42
Two types of electrical measurements were made. First, current-voltage (IV) curves were
measured for 10 nm titanium oxide films on highly doped p-Si. Second, surface resistivity was
measured for 100 nm titanium oxide films on insulating supports. Ohmic contact was made to
the titanium oxide using thin film electrodes consisting of 10 nm Ti and 100 nm Au. For the
TiO2/p-Si heterojunction devices, ohmic contact was made to the backside of p-Si with 30 nm Pt.
All metals were deposited by electron beam evaporation. Measurements were made with a
Keithley 2400 sourcemeter.
Based on the relative intensities of the Ti3+ and Ti4+ states in the Ti 2p XPS spectrum, we
refer to the three pre-conditions to F insertion as TiO1.7, TiO1.9, and TiO2 (see Fig. 1).
(A) (B)
Fig. 1 (A) Ti 2p and (B) F 1s XPS spectra of undoped and F-doped TiO1.7, TiO1.9, TiO2. Open
circles show the raw data, and lines show the results of a Gaussian deconvolution.
Fig. 2A shows surface conductivity measurements at room temperature for TiO1.7 and
TiO2 before and after F insertion. In the case of TiO1.7, the high density of suboxide states
associated with the O vacancies leads to a metallic conduction (resistivity ~0.1 Ω·cm ) but the
conductivity decreases slightly with F insertion. This is because F passivation of O vacancies
decreases the density of suboxide states used for transport in these materials. However, in the
case of TiO2, F insertion increased conductivity (i.e., decreased resistivity) by a factor of 40.
Before discussing these results, it is insightful to note that for O vacancy doped TiOX, transport
occurs in the suboxide band structures induced by O vacancy defects. This is clearly evident in
the activation energy of charge transport which is < 0.1 eV for these samples (see Figs. 2B and
3). Polycrystalline TiO2 will always show activation energy for CB transport of ~ 0.3 eV
because of localized electron traps at grain boundaries.16 We cannot measure the activation
43
energy for transport in the TiO2 sample because it is highly insulating. This serves to illustrate
that in oxide materials true CB transport is almost impossible to observe because high
conductivity in oxides is usually only achievable by reducing the O stoichiometry to create a
suboxide band structure. For the F inserted TiO2 (F-TiO2) sample we observe a transport
activation energy of 0.3 eV consistent with CB transport in polycrystalline TiO2 (see Fig. 3).
(A) (B)
(C)
(D)
Fig. 2 (A) Conductivity (S·cm−1) measurements at room temperature for 100 nm TiOX on
insulating substrates. (B) Ln (Conductance, S) as a function of temperature for O vacancy
doped TiO1.7 and F-inserted TiO2 diodes. The Ln (S) values of O vacancy doped TiO1.7 is
normalized to the value of F-inserted TiO2 at 28.7 °C (i.e. 1000/T = 3.31 K−1). The slope is
proportional to the activation energy (Ea) for transport. (C) Imaginary dielectric functions (ε2)
as a function of photon energy for undoped TiO2, O vacancy doped TiO1.7, and F inserted TiO2
taken by spectroscopic ellipsometry measurements. CB absorption (shaded in grey) shows an
onset at 3.2 eV and is nearly identical for the F inserted and undoped TiO2 sample. At photon
energies below 3.2 eV, absorption is due to O vacancy associated mid-gap states. (D) Lowest
excitation energy of TiO2, O vacancy doped TiO1.7, and F inserted TiO2 from absorption
coefficient spectra. The reference absorption onset for each excitation is ~ 105 cm−1.
It is likely that this conduction is highly localized to the surface of the film. Due to the
method for F insertion, we consider that F incorporates in the film primarily in the top 3 ± 1 nm
as confirmed in the XPS depth profile. By acting as an n-type donor (to be discussed later) to
raise EF toward CB edge, F represents a positive space charge in the near surface region resulting
44
in downward band bending toward the surface. If the band bending is such that the CB drops
below EF, then a shallow 2-dimensional electron gas (2DEG) forms. We suggest that the 2DEG
is the mechanism for the dramatic increase in conductivity for the F-TiO2 sample. According to
this theory, the surface channel resistivity is much lower than shown in Fig. 2A, where we
conservatively consider the entire film thickness (100 nm) in determining the area of the
transport channel. If the 2DEG is localized to the top 1 nm of the film as is likely, then true
resistivity calculated from the device channel area could easily be two orders lower (~ 10 2 Ω·cm)
than that determined by considering the entire film thickness (~104 Ω·cm). Note that the F
appears to be stable over long periods of time. We have repeated surface conductivity
measurements after storing the samples in dry N2 for 5 months and observe no measurable loss in
surface conductivity induced by F. The purpose of storing in dry N2 is to prevent water
adsorption that significantly modifies the surface electronic structure of titanium oxide.
Fig. 2C shows the photon energy dependent imaginary dielectric functions obtained by
spectroscopic ellipsometry for undoped titanium oxide (i.e. TiO2), O vacancy doped titanium
oxide (i.e. TiO1.7), and titanium oxide after F insertion (i.e F-TiO2). This allows us to compare
the effects of O vacancy doping and F insertion on the CB structure and minimum excitation
energy. The CB edge shows an onset at 3.2 eV for TiO2 and F-TiO2 samples, indicating that F
insertion does not change the bandgap. Absorption at photon energies below 3.2 eV is the result
of electron excitation from the valence band (VB) to mid-gap states associated with a suboxide
band structure. By looking at the relative absorption coefficients converted from the dielectric
functions, it is possible to compare the density of mid-gap states induced by O vacancy doping
and by F insertion under the joint density of states theory (see below). The relative absorption of
mid-gap states in TiO1.7 (~105 cm−1) is an order of magnitude higher than that (~104 cm−1) of
TiO2. The lowest excitation energy for each film in Fig. 2C is plotted in Fig. 2D. While O
vacancy doped TiO1.7 shows the significant destabilization of intrinsic TiO2 band structure with
the low excitation energy (from top of VB to empty subgap O vacancy states) at 2.2 eV, F
insertion does not induce any of change in CB edge states, stabilizing the bandgap energy at 3.2
eV. As will be discussed later, this stabilized CB edge state of F-TiO2 is due to the higher
excitation energy of molecular orbital states of Ti-F bonds than 3.2 eV bandgap energy of TiO2.
The conduction band (CB) edge states of titanium oxide thin films were analyzed from
the absorption spectra taken by a rotating compensator enhanced spectroscopic ellipsometry
(SE). SE determines the complex reflectance ratio, ρ=rp/rs, where rp and rs are the complex
reflectance of waves in the polarized light that are parallel (p) and perpendicular (s) to the plane
of incidence, respectively. Monochromatic light from a xenon lamp provides visible and
ultraviolet light having an energy range from 1.5 eV (8267 Å) to 6 eV (2066 Å). This light
excites electrons from the top of the valence band to the conduction band in the overlayer (i.e.
titanium oxide) material. From  , optical properties such as index of reflection (n), extinction
coefficient (k), and dielectric function (ε) are determined. To extract the dielectric function of
the titanium oxide overlayer from the pseudodielectric function of the whole stack (air/TiO2/Si) a
three phase model was applied such that the dielectric function of Si was analyzed first, and the
dielectric functions of the titanium oxide overlayer was subsequently extracted. The generalized
equation for the optical three phase model is expressed as follows:
45
Fig. 3 Arrhenius Plot of Ln(1/resistance, Ω-1) against 1000/temperature (K) for undoped and F-
doped TiO1.7, TiO1.9, TiO2. Squares show raw data, and lines show linear fits. The slope of the
fit is proportional to the activation energy for thermally activated charge transport.
1/ 2
4dna  s ( s   o )( o   a )   s 
    s    sin 2  
  o ( s   a ) a  Eq. 1
where <ε> is the measured pseudo dielectric function, εs is the substrate (Si) dielectric function,
εa is the air dielectric function, εo is the overlayer (titanium oxide) dielectric function, d is the
overlayer thickness, na is the index of reflectance for air, λ is the wavelength of incident
polarized light, and φ is the angle of incidence (fixed at 67.08°). Because all parameters except εo
are already known, εo can be obtained by iterative calculation using Eq. 1
The absorption coefficient, α is obtained from the λ and k values by the conversion
equation:
α=4πk/λ Eq. 2
where k is the extinction coefficient and λ is the wave length of light. Fig. 4 shows the resulting
graphs of the extinction coefficients for the different oxide films.
Fig. 5A shows current-voltage (IV) curves for diodes consisting of TiOx on highly doped
p-Si. We present results for TiO2 and TiO1.7 before and after F insertion. Without F insertion,
both TiO2 and TiO1.7 heterojunction devices show almost symmetrical IV curves where the
reverse bias leakage current is nearly as high as the forward bias current. This is the result of
two factors. First, the high density of subgap states in TiO1.7 provide a path for carrier transport
in the reverse bias direction. Second, the low n-type character (i.e., low EF level) of O vacancy
doped TiOX closely aligns with EF in p-Si resulting in minimal interface band bending. For TiO2
and TiO1.7, the reverse bias current decreases and the forward bias current increases with F
insertion. This is the combined effect of defect state passivation by F and increased E F in the F-
TiOX layer to achieve greater band bending at the p-Si interface. In Fig. 5B, the rectifying factor
(taken as the ratio of forward to reverse bias current) increases with F insertion by a factor of 3
46
Fig. 4 Absorption coefficient spectra for undoped and F-doped TiO1.7, TiO1.9, TiO2 taken by SE
measurement and three-phase analysis. Closed circles show undoped TiOX, and open circles
show F doped TiOX.
(B)
(A)
Fig. 5 (A) Current-voltage (IV) curves for ~ 10 nm TiOX on highly doped (0.01 Ω cm) p-Si. Red
lines show TiO1.7 samples, and blue lines show TiO2. The current density values of all TiO2
diodes were multiplied by a factor of 200. (B) Rectifying factor for TiO2 and TiO1.7 before and
after F doping. Rectifying factors are the ratio of forward to reverse bias current. For O
vacancy doped TiOX, low rectifying factors are the result of a high concentration of defect states
and low Fermi level (EF). By passivating defects and raising EF, F doping increases the
rectifying factor for each precondition by more than a factor of 3.
for both cases. A similar trend was observed for TiO1.9 (not shown), except that the rectifying
factor before F insertion was 3.4 and increased to 22.4 after F insertion. This is much higher
47
than for TiO2 and TiO1.7. We do not fully understand this anomaly but believe it is in part a
result of the higher EF in TiO1.9 before F insertion compared to either TiO2 or TiO1.7 as discussed
below. It may seem counterintuitive that surface F doping could affect rectification at the buried
Si-TiOX interface. However, the thickness of the TiOX layer (i.e. 10 nm) is much smaller than
the depletion width. Consequently, n-type doping to the surface will still effect rectification due
to the mean field across the depletion region of this p-n heterojunction.
Fig. 6A shows EF for TiO2, TiO1.9, and TiO1.7 before and after F insertion. We
determined EF by fitting the valence band edge XPS spectra (see Fig. 7). It is insightful to first
consider the effect of O vacancies on the Femi level before considering the effect of F doping.
In the case of stoichiometric TiO2, we measured EF to be 2.4 eV. Doping by O vacancies
initially introduces partially occupied Ti3+ states about 0.3~0.5 eV below the CB edge,17 and
these states serve to pin EF at approximately 2.8 eV in the TiO1.9 sample. However, if the oxide
is further reduced, semi-metal states form that pin EF at a much lower energy. This is evident in
the TiO1.7 sample that shows EF to be 2.1 eV. We note that this value is in excellent agreement
with the spectroscopic ellipsometry spectra that showed a minimum excitation at 2.2 eV (see Fig.
2D). Although this reduced oxide is highly conductive, transport occurs in the metallic band
structure of the suboxide, and it is not possible to use this material as a semiconductor because of
the high density of mid-gap states.
Fig. 6A also shows that F insertion raises EF for each pre-condition by up to 1 eV. In the
case of TiO1.7, where EF is initially pinned at 2.1 eV by the semi-metal states of the suboxide, F
insertion increases EF to 3.1 eV, just below the CB edge. This is partly the result of F
passivation of suboxides states in the bandgap. However, the resulting EF is 0.7 eV greater than
undoped, stoichiometric TiO2. We conclude that in addition to passivating defects, F acts as an
n-type donor. These results demonstrate the ability of F insertion to tune EF outside the limits of
O vacancy doping. Comparing EF after F insertion for each pre-condition shows that EF scales
with the atomic ratio of F to Ti determined by XPS (see Fig. 6B). This indicates that F binding
to an O vacancy gives rise to a free CB electron in titanium oxide. We call this effect n-type
doping by F insertion. Fig. 6B also indicates another binding state of F which we assign as
interstitial insertion. Theoretical calculations discussed below explain the origin of these two
binding states and discuss their respective contributions to modifying the bulk electronic
properties. We note that EF scales with the concentration of a F binding state at 684.9 eV that we
assign as O substitution (or binding to a preexisting O vacancy).
Fig. 6C shows the Ti 2p spectrum for O vacancy doped and F inserted TiO1.7. The peak
binding state at 458.5 eV is characteristic of the Ti4+ states in the stoichiometric oxide, while the
shoulder at lower binding energy (BE, 456.9 eV) represents Ti3+ and Ti2+ states in the suboxides.
Comparison of the Ti 2p spectra of TiO1.7 before and after F insertion shows a decrease in Ti3+
states from 41% to 9%. This is a dramatic reduction in the density of defect states with F
insertion, and is direct evidence that F passivates O vacancies by binding Ti3+. Fig. 6D shows
the absorption ratio of mid-gap to CB states before and after F insertion for each TiOX pre-
condition based on spectroscopic ellipsometry measurements. This provides additional evidence
that F binding is primarily to O vacancies, and that F passivates defect states in non-
stoichiometric oxides. Because the chemical instability of O vacancies is one of the primary
48
(A) (B)
(C) (D)
Fig. 6 (A) Fermi level (EF) before and after F insertion for each TiOX precondition determined
by fits to the valance band edge XPS spectra (see Fig. 8). (B) The atomic ratio of F to Ti after 3
min. plasma treatment for each TiOX precondition. Atomic ratios are shown for two different
binding states identified by XPS (see Fig. 8B) and for total F. We assign these two binding states
as F substituted for O and as interstitial F. Substitutional F concentration depends strongly on
the initial O vacancy concentration before plasma treatment and appears to determine EF after F
insertion. (C) Ti 2p XPS spectrum of TiO1.7 before and after F insertion. Open circles show the
raw data, and lines show the results of a Gaussian deconvolution. The peak state at 458.5 eV is
the Ti4+ state in stoichiometric TiO2. The lower binding energy peak at 456.9 eV is the Ti3+ state
and is a marker for O vacancy induced defect states. The decrease in Ti3+ state concentration
with F insertion indicates that F passivates defects by binding to Ti3+ at O vacancies. (D) Ratio
of defect state to CB absorption measured by spectroscopic ellipsometry before and after F
insertion for TiO1.7, TiO1.9, and TiO2. Defect state/CB absorption was taken as the maximum
absorption coefficient below/above 3.2 eV. The trend shows that F passivates O vacancies that
give rise to defect states.
factors that limit TMO device reliability, we believe that F insertion may greatly improve the
chemical stability of TMO semiconductor devices. However, we have not yet investigated the
effect of F insertion on chemical stability. Note that in the case of stoichiometric TiO2, F
insertion does not significantly change the O vacancy concentration because it is already so low.
We note that SF6 can etch TiO2 by formation of TiF4 which has a vapor pressure at room
temperature. Fracassi et al. showed that during SF6 plasma etching a layer of TiF3 and TiF4
forms on the surface of the oxide.18 TiF4 leaves the surface by sublimation which is likely
49
Fig. 7 VB edge XPS spectra for undoped and F-doped TiO1.7, TiO1.9, TiO2. Open circles show
raw data, and lines show fitting used to determine Fermi level.
assisted by ion sputtering in the plasma. Consequently, the formation of titanium fluorides(TiFX)
and subsequent etching represents an upper limit to F doping in TiO2. For our plasma exposures,
the F concentration stayed beneath the limit for formation of TiF4 as can be seen by atomic
fractions of F to Ti below 0.6. Fracassi et al. also showed that formation of TiF3 and TiF4 during
etching is easily visible in the Ti 2p XPS spectrum by a 3–5 eV shift to higher BE.19 We observe
no significant shift (< 0.2 eV) in the Ti 2p BE with plasma treatment. This suggests true
chemical doping by F insertion rather than formation of a fluoride phase.
Theoretical calculations are carried out using a high level, many-electron theory in order
to investigate fundamental properties of TiO2. This is a well-established method for
investigating the electronic structure of metal oxides.20 The theoretical model considers three
different modifications to TiO2: a) interstitial F, b) F substitution for oxygen or c) an O vacancy.
Of particular interest are effects on the electronic excitation spectrum and induced spin states that
accompany changes in the occupancy of Ti d orbitals. We use a geometry optimized Ti18O36
nanoparticle to investigate these effects. While the nanoparticle differs in detail from bulk TiO 2,
the interior sites involve the same basic electronic structure. Fig. 8A depicts Ti18O36
nanoparticle model with an interstitial F site. In other calculations, a subsurface vacancy is
created by removing a neutral O atom at the interior of the particle, and a F atom is substituted
for O.
In all three systems, the lowest energy excitation is best described as exciton, with a hole
that is largely localized on a single O2p and an electron that is transferred to the d-shell of a
nearest neighbor Ti. In the nanoparticle, the oxygen vacancy and interstitial F alter the excitation
energy or “bandgap” from the ~3eV of the unmodified Ti18O36. Substitutional F increases the
excitation energy while an O vacancy decreases the excitation energy (see Table 1). This
explains why F insertion does not alter the minimum excitation energy observed experimentally.
50
(A) (B)
Fig. 8 (A) Ti18O36 nanoparticle model with interstitial F (yellow ball) used for theoretical
calculation. In this site, F sits in a distorted octahedral formed by six Ti atoms. In this case
charge transfer to the F ion is primarily from a surface O atom (shown as site 50) because the
nearest neighbor Ti atoms are already fully oxidized. The substitutional F site and O vacancy
site (shown as site 43) have three Ti nearest neighbors. (B) F 1s XPS spectrum of TiO2 after 3
min. F insertion. Open circles show the raw data, and lines show the results of a Gaussian
deconvolution. The peak state at 684.9 eV indicates Ti-F bonding as a result of O subsitution.
We assign the state at 686.5 eV as interstitial F. The relative energies of the two binding states
are explained by the qualitatively different charge exchange between F and either Ti or O
depending on the insertion site.
Although F insertion creates new states in the bandgap, the lowest excitation energy is still the
difference between the unmodified valance and conduction band edges of stoichiometric TiO2.
The calculations also reveal major differences in the concentration of free carrriers: an O
vacancy and substitutional F induce unpaired electron spin in the d-shell of a nearest neighbor Ti,
while interstitial F induces unpaired spin on nearby oxygen. One of the key findings that relates
to the present investigation is the difference in behavior of subsurface species in the creation of
surface electronic states involving oxygen atoms that are substantially undercoordinated, e.g., 2-
fold coordinated, on the surface of TiO2. O atoms on the surface can serve as receptors of
unpaired spins generated internally and are strongly affected by nearby subsurface modifications.
To summarize, substitutional F is responsible for the increase of electron density in CB of TiO2
in contrast to the increase in mid-gap states by O vacancy doping. Furthermore, interstitial F
leads to an unpaired spin in the O2p orbital of surface O that draws free carriers to the surface
and acts as a surface conduction.
To emphasize the high level of correlation between the experimental results and the
theoretical study, we briefly discuss the nature of F binding states in TiO 2 as identified by XPS
spectra. Fig. 8B shows the F 1s XPS spectrum for TiO2 after F insertion. The peak is composed
of two binding states centered at 684.9 and 686.4 eV. This spectrum is identical to previous
reports for F insertion of TiO2 by solution phase methods.19 The peak state at 684.9 eV indicates
Ti-F bonds by substitutional F doping. The shoulder at higher BE is a F surface state, and
previous reports have hypothesized as to its exact origin.19 We propose that the higher BE state
51
Table 1
O Vacancy Substitutional F Interstitial F
Increase in number of Ti 3d 1.58 0.89 0.00

electrons*
Increase in number of unpaired 0.00 0.00 0.93

O 2p electrons*
F population — 9.55 9.62
Change in O2p-Ti3d excitation −0.3 +0.8 +0.2

energy (eV)*
represents interstitial F insertion (i.e. F insertion without removal of an O atom). In the case of
O substitution, F coordinates to a partially oxidized Ti atom, electron density is transferred from
the Ti to the F, and the Ti atom assumes a 4+ oxidation state. However, if F is inserted or bound
to the surface without the removal of O, F could not draw electron density from a Ti nearest
neighbor because all nearest neighbor Ti atoms are fully oxidized. In this case, the calculation
suggests that F attracts electron density from a nearby O atom creating a partially unoccupied
O2p state. This is what we term interstitial insertion. The relative energies of the two binding
states are explained by the qualitatively different charge exchange between F and either Ti or O
depending on the insertion site.
In Fig. 1B, two F binding states are resolved at 684.9 eV and 686.4 eV. Based on Ti 2p
XPS data in Fig. 1A, Ti3+ species is reduced by F passivation forming Ti-F bonds. At this bond,
F attracts about one electron from the neighboring Ti because the electronegativity of F is higher
than Ti. This F-induced charge exchange causes the shift of binding energy of Ti higher than Ti3+
and that of F lower than neutral F. This Ti-F formation is equivalent to the substitutional F
doping either replacing the lattice oxygen (O) site or passivating O-vacancy sites. The theoretical
calculation (to be discussed later) predicts the origin of this kind of Ti-F bond as the
substitutional F-doping by replacing (i) the top-surface O site having two near-neighboring Ti
atoms and (ii) the sub-surface O or O-vacancy site having three near-neighboring O sites with F.
For all cases, F attracts electron from Ti, causing the singly occupied 3d electron in Ti. Thus, the
F 1s biding state at 684.9 eV is assigned as the substitutional F doping to Ti. On the other hand,
the F 1s bonding state at 686.4 eV is assigned as the surface F (Fsurface) because it has a higher
concentration at the more surface region. The theoretical calculation considers the interstitial F at
subsurface. The interstitial F occupies the octahedral sites formed by Ti ions. F attracts electron
density because of its electronegativity but since Ti is already depleted by its normal bonding
with O, the electron attracted to F comes from O2p electron density of top-surface
undercoordinated O, which leads to the unpaired electron in O2p orbital. However, the
electronegativity of O is far higher than Ti, the electron density that F attracted from O is a
partial electron density smaller than that for Ti-F case. This partial charge exchange causes the
higher binding energy of interstitial F than that of substitutional F. Therefore, the density of
surface-segregated Fsurface binding state in XPS data and theoretical result strongly suggest that F
1s bonding state at 686.4 eV is due to the interstitial F.
52
Considering the two binding states to be equivalent to substiutional and interstitial F
insertion considered by the theoretical calculations, we find excellent agreement between the
theoretical conclusions and the experimental findings. First, we observe that substitutional F (i.e.
binding state at 684.9 eV) scales with the O vacancy concentration before doping while
interstitial F (i.e. binding state at 686.4 eV) shows little dependence on the precondition to
plasma treatment. Second, we find that EF scales closely with the concentration of substitutional
F which theory predicts to generate a free electron in the Ti d shell, unlike interstitial F. Finally,
the role of interstitial F to increase surface conductivity by drawing free charge carriers to the
surface can explain the enhancement of surface conductivity in stoichiometric TiO2 after F
insertion.
5.5 Conclusion
This study demonstrates that F passivates O vacancy defects and acts as an extrinsic n-
type donor for titanium oxide, resulting in greatly enhanced CB transport. This effect is stable at
ambient temperature with no measureable loss of surface conductivity over several months. It is
likely that F will have the same effect on other TMO semiconductors. The ability to remove
mid-gap states and tune EF while stabilizing the intrinsic TiO2 band structure can result in a
highly conductivity oxide suitable for device architecture and overcomes the limitations of
current O vacancy doping technology for oxide semiconductors. The highly conductive surface
electronic structure of n-type titanium oxide with F insertion, free of EF level pining, is
promising for many metal-oxide applications including solar cells, resistive switching memory,
transparent oxide transistors, sensors, and catalysts.
5.6 References
(1) Wager, J. F. Science 2003, 300, 1245.

(2) Xie, X.; Li, Y.; Liu, Z.-Q.; Haruta, M.; Shen, W. Nature 2009, 458, 746.
(3) Park, J. Y.; Lee, H.; Renzas, J. R.; Zhang, Y.; Somorjai, G. A. Nano Lett. 2008, 8, 2388.
(4) Fujishima, A.; Honda, K. Nature 1972, 238, 37.
(5) Bach, U.; Lupo, D.; Comte, P.; Moser, J. E.; Weissortel, F.; Salbeck, J.; Spreitzer, H.;
Gratzel, M. Nature 1998, 395, 583.
(6) Kwon, D.-H.; Kim, K. M.; Jang, J. H.; Jeon, J. M.; Lee, M. H.; Kim, G. H.; Li, X.-S.;
Park, G.-S.; Lee, B.; Han, S.; Kim, M.; Hwang, C. S. Nat Nano 2010, 5, 148.
(7) Peng, X.; Chen, A. J. Mater. Chem. 2004, 14, 2542.
(8) Di Valentin, C.; Pacchioni, G.; Selloni, A. J. Phys. Chem. C 2009, 113, 20543.
(9) Zhang, Y.; Kolmakov, A.; Chretien, S.; Metiu, H.; Moskovits, M. Nano Lett. 2004, 4,
403.
(10) Frederikse, H. P. R. Journal of Applied Physics 1961, 32, 2211.
(11) Cao, F.; Oskam, G.; Searson, P. C.; Stipkala, J. M.; Heimer, T. A.; Farzad, F.; Meyer, G.
J. J. Phys. Chem. 1995, 99, 11974.
(12) Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, 269 -271.
(13) Chen, X.; Burda, C. J. Am. Chem. Soc. 2008, 130, 5018.
(14) Choi, W.; Termin, A.; Hoffmann, M. R. J. Phys. Chem. 1994, 98, 13669.
(15) Janisch, R.; Gopal, P.; Spaldin, N. A. Journal of Physics: Condensed Matter 2005, 17,
R657.
53
(16) Cox, P. A. Transition Metal Oxides: An Introduction to Their Electronic Structure and
Properties; Oxford University Press, 2010.
(17) Lucovsky, G.; Seo, H.; Lee, S.; Fleming, L. B.; Ulrich, M. D.; Lüning, J.; Lysaght, P.;
Bersuker, G. Japanese Journal of Applied Physics 2007, 46, 1899.
(18) Fracassi, F.; d’ Agostino, R. Pure and Applied Chemistry 1992, 64, 703.
(19) Czoska, A. M.; Livraghi, S.; Chiesa, M.; Giamello, E.; Agnoli, S.; Granozzi, G.; Finazzi,
E.; Valentin, C. D.; Pacchioni, G. J. Phys. Chem. C 2008, 112, 8951.
(20) Lucovsky, G.; Whitten, J. L. Surface Science 2007, 601, 4138.
54
Chapter 6
Highly n-Type Titanium Oxide as an Electronically

Active Support for Platinum in the Catalytic
Oxidation of Carbon Monoxide
6.1 Abstract
The role of the oxide-metal interface in determining the activity and selectivity of
chemical reactions catalyzed by metal particles on an oxide support is an important topic in
science and industry. A proposed mechanism for this strong metal-support interaction is
electronic activation of surface adsorbates by charge carriers. Motivated by the goal of using
electronic activation to drive non-thermal chemistry, we investigated the ability of the oxide
support to mediate charge transfer. We report an approximately twofold increase in the turnover
rate of catalytic carbon monoxide oxidation on platinum nanoparticles supported on
stoichiometric titanium dioxide (TiO2) when the TiO2 is made highly n-type by fluorine (F)
doping. However, for non-stoichiometric titanium oxide (TiOX<2) the effect of F on the turnover
rate is negligible. Studies of the titanium oxide electronic structure show that the energy of free
electrons in the oxide determines the rate of reaction. These results suggest that highly n-type
TiO2 electronically activates adsorbed oxygen (O) by electron spillover to form an active O−
intermediate.
6.2 Introduction
Although metals alone are often catalytically active, most industrial catalysts consist of
metal particles supported on a porous oxide. This not only provides a high surface area for the
heterogeneous catalyst, but as shown by many studies, the oxide support plays an important role
in determining the activity and selectivity of the catalyst.1-4 This phenomenon, known as the
strong metal-support interaction (SMSI), has been widely studied and is an important topic in
both science and industry.5
SMSI affects a wide range of catalytic reactions, including carbon monoxide (CO) and
carbon dioxide (CO2) hydrogenation4-6, selective hydrogenation7-9, and CO oxidation10-13.
Although not traditionally termed SMSI, the catalyst support also plays an important role in
activating molecular oxygen for selective partial oxidation reactions. Industrially relevant
examples include the synthesis of aldehydes from primary alcohols14, the production of hydrogen
peroxide from hydrogen15-19, and the conversion of methane to synthesis gas20-22. In all of these
cases, molecular oxygen is preferred over other oxygen donors due to cost, energy efficiency,
and environmental concerns.14,22
Titanium oxide is perhaps the best known example of an SMSI support.2,4 Additionally,
titanium oxide is a common support for catalysts demonstrating high selectivity toward partial
oxidation.14,18,22 The majority of proposed mechanisms for the role of titanium oxide fall into
55
one of the following categories: 1) formation of active sites when the support wets (or
“decorates”) the metal particle23,24, 2) oxygen (O) activation and/or spillover from chemically
active defects in the support25, and 3) electronic mediation through various forms of charge
transfer12,26. Although each of these mechanisms can play an important role, this study focuses
on the role of charge transfer and electronic mediation due to its importance in energy
conversion reactions.
Previous studies have indicated that, at least in certain instances, catalytic oxidation
reactions proceed by an electronically activated pathway.27-32 Bonn et al.27 showed that for CO
oxidation on Ru, the activation of chemisorbed O occurs by charge transfer from the metal.
Because the active O species cannot form thermally until well above the desorption temperature
of CO, a temperature ramp of a CO/O2 co-adsorbed Ru surface produces CO and O2, but not
CO2. However, a femtosecond laser pulse can produce very high electronic temperatures on the
short time scale without significant lattice heating. This serves to activate the O without CO
desorption, resulting in CO2 formation.
Although fundamentally insightful, this type of experiment employing a femtosecond

laser cannot be scaled for high turnover applications. Another approach to achieve electronic
activation is to use an electronically modified catalyst. Several variables can tune the electronic
structure of a catalyst including size31, support32, and doping30. The objective of this study was
to identify the role of the support in electronic activation of surface chemistry. Here we
demonstrate the ability to modify the electron transport properties of titanium oxide by
controlling the O stoichiometry and by fluorine (F) doping. We find that stoichiometric titanium
dioxide (TiO2) modified by F doping is an electronically active support for CO oxidation on
platinum (Pt) under conditions of high pressure and temperature (760 Torr and 443 K).
However, the highly conductive reduced oxide shows little or no electronic activity toward CO
oxidation, with or without F.
We have previously reported the production of hot electron flow during catalytic CO
oxidation using the Pt/TiO2 nanodiode system.33 In that work we demonstrated that there is an
electron flow from the Pt to the TiO2 in proportion to the catalytic turnover. We now find that
the opposite effect occurs by F doping TiO2. In this case, an electron flow from the titanium
oxide to the surface adsorbates significantly increases the catalytic reaction rate by activating
surface O for reaction with CO.
6.3 Experimental
6.3.1 Catalyst Preparation
Thin films of titanium oxide with variable oxygen (O) vacancy concentrations served as
supports for the platinum (Pt) catalyst. First, titanium oxide films (100 nm) were deposited on
Si(100) wafers with a 500-nm-thick thermal oxide by direct-current magnetron sputtering.
Sputtering conditions were 400 W plasma power, 450 V bias voltage, 50 sccm Ar flow, and 3
sccm O2 flow. As deposited, the titanium oxide films were highly O deficient. Rapid thermal
annealing in O2 at temperatures between 623 and 773 K was used to vary the O concentration in
the films. To avoid grain size variations due to annealing at different temperatures, all samples
were annealed in nitrogen (N2) at 773 K. Titanium oxide films with three stoichiometries were
56
3+ 4+
prepared: TiO2, TiO1.9, and TiO1.7, as determined from the ratio of Ti and Ti in the Ti 2p X-
ray photoelectron spectra (XPS) shown in Chapter 5 Fig. 1A.
Fluorine (F) insertion was achieved by plasma treatment in N2 gas with trace sulfur
hexafluoride (SF6). Trace SF6 was introduced into the chamber by flowing a 9:1 mixture of SF6
and O2 followed by pumping to chamber base pressure (5 mTorr). After pumping, a small
background pressure of SF6 remained in the chamber. N2 gas was then introduced at 80 sccm for
plasma treatment. Plasma power was 20 W and DC substrate bias was 130 V. F concentration
was tunable with the length of the plasma treatment. Two samples of each TiOX stoichiometry
were fabricated, and one of each type was plasma treated. This produced six supports for
subsequent Pt deposition: TiO2, TiO1.9, and TiO1.7, each with and without F. Fig. 1B in Chapter
5 shows the F 1s XPS spectra for each of the F-doped supports. All samples were stored under a
dry N2 atmosphere until the reaction rate measurements.
Pt nanoparticles were deposited onto the samples by electron beam evaporation. Vapor
deposited, rather than colloidally synthesized, nanoparticles were used to avoid the presence of
an insulating polymer capping layer between the metal nanoparticles and the support. For
electron beam evaporation the chamber base pressure was less than 10−5 Torr, and the deposition
rate was 0.02 Å/s. The average film thickness monitored by quartz crystal microbalance was
1 nm. SEM showed that at this average thickness the Pt nanoparticles left ~50% of the support
exposed as shown in Figure S2 (supporting information).
6.3.2 Catalyst Characterization
The electronic structure of the titanium oxide supports was characterized by XPS,
spectroscopic ellipsometry, and measurements of surface conductivity and activation energy for
charge transport.
XPS was used to analyze the chemical binding states of the titanium oxide films and the
supported Pt catalysts both before and after reaction (Physical Electronics, PHI 5400 ESCA/XPS
system with an Al anode source at 1486.6 eV). The analyzer was positioned at 50° relative to
sample normal. All binding energies were calibrated to the Ti 2p peak state.
The real and imaginary parts of the complex dielectric constant, εc = ε1+iε2, and
absorption coefficients for titanium oxide samples were determined by visible-ultra violet
spectroscopic ellipsometry in a rotating compensator enhanced spectrometer. The
monochromatic light source from a Xenon lamp at photon energies of 1.5–6 eV was used with
spectral resolution of 15 meV.
To measure surface conductivity of the titanium oxide films, ohmic contact was made to
the titanium oxide using thin film electrodes consisting of 10 nm Ti and 100 nm Au. All metals
were deposited by electron beam evaporation. Measurements were made with a Keithley 2400
sourcemeter.
6.3.3 Reaction Rate Measurements
A batch mode reactor with a boron nitride substrate heater was used to determine the
reaction rates of CO oxidation on each of the above described catalysts. A metal bellows
57
circulation pump provided gas mixing. Gas pressures were 40 Torr CO and 100 Torr O2 in a
background of He. The catalyst temperature was 443 K. Each catalyst was tested for 112 min
and CO2 production was monitored as a function of time using a gas chromatograph with a
thermal conductivity detector.
Reaction rates are reported in turnover frequency (TOF) as CO2 molecules produced per
Pt site per minute. All error bars represent the 95% confidence interval based on the rate of CO 2
production normalized to the estimated number of Pt sites. The number of Pt sites was estimated
by assuming a uniform (111) surface structure over the entire catalyst area. Although this
calculation is approximate, it provides a consistent normalization to the catalyst area and yields a
reasonable estimate of the absolute TOF.
Fig. 1 shows the effect of F on the reaction rate and on the surface conductivity of each
stoichiometry. In the case of stoichiometric TiO2, F doping increased the reaction rate by 61%,
while F doping to TiO1.7 and TiO1.9 decreased the reaction rate by 21 and 8%, respectively.
There is a surprising similarity between the effect of F on the turnover rate, and the effect of F on
the surface conductivity of the titanium oxide support. As in the turnover rate measurements, F
doping to TiO2 dramatically increased (40-fold) the surface conductivity, while F doping to
TiO1.7 and to TiO1.9 resulted in only a slight decrease (approximately twofold) in surface
conduction. We can explain these results by considering the electronic structure of
stoichiometric and reduced titanium oxide. We characterized the electronic structure using two
measurements: the activation energy of surface conduction and the optical absorption spectrum.
To understand how the activation energy of transport relates to the electronic structure, it
is necessary to consider the role of grain boundary defects in polycrystalline TiO2. Grain
boundaries produce defects in the band structure of titanium oxide with a localized defect state
~0.3 eV below the conduction band edge. Consequently, this is the activation energy for
conduction in polycrystalline TiO2 because electrons thermally trap and de-trap at grain
boundaries during transport. In contrast, O vacancies at high concentrations form a sub-oxide
band structure at a reduced energy. This sub-oxide band shows semi-metal, or nearly
unactivated, electron transport.34
Fig. 2A shows the activation energies for electron transport in TiO1.7 and TiO2. As
expected, the highly reduced oxide shows low activation energy for conduction (0.07 eV)
consistent with semi-metal transport in a low energy sub-band, while F-doped TiO2 shows
activation energy of 0.29 eV, consistent with conduction band transport across grain boundaries.
TiO2 without F was too insulating to accurately measure the activation energy of transport.
However, F insertion to TiO1.7 did not change the transport mechanism.
Fig. 2B compares the optical absorption of TiO1.7 and TiO2. To illustrate the differences
between these samples, we show the minimum excitation energy having a fixed absorption
coefficient of 2 × 104 cm−1. Full spectra are shown in Fig. 4 of Chapter 5. We find that TiO1.7
has a strong absorption located ~0.7 eV below the conduction band edge that arises from
excitation of electron-hole pairs in the sub-oxide band. This measurement reflects the decrease
58
Stoichiometric TiO2 TiOX Sub-Oxide

35
Undoped Undoped
(A)
TOF (CO2/Pt/min)
F-Doped F-Doped
30
25
20
15
Conductivity ( cm )
6
-1
50 x10
(B)
40
-1
30
Conduction Band Sub-Oxide
Transport Transport
20
10
0
TiO2 TiO1.7 TiO1.9
Fig. 1 (A) Turnover frequencies (TOF) for CO oxidation on Pt nanoparticles supported on the
six titanium oxide supports: TiO2, TiO1.9, and TiO1.7, each with and without F insertion.
Reaction occurred in 40 Torr CO, 100 Torr O2, and 620 Torr He at 443 K. TOF data reflect the
stable rate after ~30 min of deactivation. Error bars represent 95% confidence intervals.
(B) Surface conductivity measurements for all six titanium oxide supports before Pt nanoparticle
deposition. Comparison of (A) and (B) shows a surprising similarity between the effect of F on
the TOF and on the surface conductivity.
in bandgap energy for the reduced oxide and shows the difference in the energy of free electrons
in the two supports.
Fig. 3A is a diagram of the two band structures of stoichiometric and reduced titanium
oxide based on the above measurements. In the cases of both TiO1.7 and TiO1.9, F doping
decreased the surface conductivity. The reason for this is that F binds to Ti at the sites of O
vacancies resulting in passivation of defect states. Conduction in these samples occurs primarily
in the sub-oxide band structure induced by the high concentration of O vacancies. Consequently,
a decrease in the O vacancy concentration partially removes the conduction channel in these two
samples. However, F insertion in the TiO2 sample increased conductivity by a factor of 40. This
is because F acts as an extrinsic n-type donor in metal oxide semiconductors, increasing the
concentration of free electrons in the conduction band of TiO2.35
Assuming a reaction mechanism where the rate limiting step is activated by a conduction
band electron from the titanium oxide support, the electronic structure of the titanium oxide
explains the observed trend in reactivity. In the case of TiO1.7 and TiO1.9, the presence of F
59
Minimum Excitation Energy (eV)

4.0
Ln (Conductivity, S) -8 (A)Ea=0.069 eV TiO1.7 (B) Undoped
F-Doped
3.5
-9
Ea=0.073 eV F-TiO1.7 3.0
-20
0.292 eV 2.5
Ea
F-TiO2
2.0
-21
1.5
3.0 3.1 3.2 3.3 TiO2 TiO1.7
-1
1000/T (K )
Fig. 2 (A) Arrhenius plots showing the activation energy for charge transport in TiO1.7 and TiO2.
Because TiO2 without F is highly insulating, the Arrhenius curve could not be accurately
measured for this sample. The different activation energies for conduction in TiO1.7 and TiO2
confirm different transport channels as depicted in Fig. 3A. (B) Minimum excitation energy
having a fixed absorption coefficient of 2 × 104 cm−1. This measurement reflects the decrease in
bandgap energy for the reduced oxide and shows the difference in the energy of free electrons in
the two supports.
(A ) (B )
Fig. 3 (A) Band diagram depicting two types of transport in titanium oxide. Sub-band defect
states in the reduced oxide provide a semi-metal transport channel. In stoichiometric TiO2,
transport occurs in the actual conduction band. Due to grain boundary defects, conduction band
transport is thermally activated as charges trap and de-trap at grain boundaries. (B) Schematic
depicting why electronic activation is possible by F doping and not by O-vacancy doping.
Although a high density of free charge carriers exist in the reduced oxide, these carriers reside
in the sub-band located ~0.5–1.0 eV below the conduction band edge. At this low energy, the
electrons cannot activate surface O for reaction with CO. However, in stoichiometric TiO2,
electrons reside in the true conduction band. Charge transfer from the conduction band results
in electronic activation of surface O for CO2 formation.
60
decreased the surface electron density in the oxide support resulting in a slightly reduced reaction
rate. However, because the majority of electrons in these supports reside in the sub-oxide band
at an energy that is presumably too low to induce chemistry, changes in electron density have a
minimal effect on the reaction rate. In the case of TiO2 where free carriers reside in the
conduction band from which they can spillover to adsorbates, an increase of surface electron
density induced by F correlates with a significant increase in catalytic activity by electronic
activation.
We propose that the mechanism for electronic activation is electron spillover from TiO2
to adsorbed O to form an activated O− intermediate that readily reacts with CO. This is depicted
schematically in Fig. 3B. In the case of O vacancy doping, the energy of the sub-oxide band is
0.5–1.0 eV below the conduction band edge. This energy is much higher than kT (≥13 kT at a
reaction temperature of 443 K). Consequently, a reaction pathway that is electronically activated
by free carriers in the conduction band of TiO2 would not be thermally accessible until a
temperature of ~6,000 K. However, because F insertion to TiO2 increases the surface
concentration of free electrons without formation of any mid-gap states, free electrons reside in
the actual conduction band making this material an electronically active support for CO
oxidation.
XPS and reaction kinetics provide evidence that electronic activation by the F-doped
support occurs by electron spillover to surface O. Fresh catalysts prepared and treated under
reaction conditions for 1 h at 373, 473, or 573 K showed significant changes in the O 1s
spectrum that correlate with deactivation of the undoped catalyst. Fig. 4A shows the O 1s XPS
spectrum for the Pt/TiO2 catalyst following reaction at various temperatures. The surface O peak
at high binding energy (~532.5 eV) increased with reaction temperature. The XPS spectra for
the F-doped analog of this catalyst (not shown here) were identical to those shown in Fig. 4A.
Fig. 4B shows how the formation of surface O affects the activity of the undoped catalyst.
Initially we observe a TOF of ~25 min−1 at 443 K; however, following reaction at high
temperature (543 K), the catalyst deactivated by a factor of four. The initial activity was restored
after removal of surface O by reaction with CO. Treatment in O2 again resulted in catalyst
deactivation by the formation of surface O.
We do not entirely understand the mechanism of catalyst deactivation by surface O, but

at present we suggest the following: Surface O is an inactive species occupying catalytic sites,
so deactivation is simply a result of blocking active sites with strongly bound O. Additionally, O
acts as an electron acceptor, so it is possible that deactivation by surface O is partly an electronic
phenomenon. Surface O may bind and localize free electrons in the titanium oxide support that
are necessary for activating the catalytic reaction. This suggests that O activation is rate limiting
for CO oxidation and is consistent with the idea that electronic activation by the F-doped TiO2
sample occurs by electron spillover to adsorbed O to form an activated O− intermediate. Fig. 4C
provides further evidence for this theory.
Fig. 4C shows time dependent reaction rate measurements for Pt on each TiOX
stoichiometry and on F-doped TiO2. Each undoped catalyst deactivated by 20–36% during 1 h
under reaction conditions. However, Pt supported on F-doped TiO2 showed a 49% rate increase.
We propose that these kinetics correspond to the formation of surface O during the reaction and
61
30
XPS Intensity (Arb. Units)
(A) Surface O (B)
25
TOF (CO2/Pt/min)
O-Ti
573 K 20
473 K 15
10
373 K
5
As-dep
0
534 532 530 528 As-dep CO/O2 CO O2
Binding Energy (eV) Sequential Treatment
40
(C) TiO1.7
35 TiO1.9
F-TiO2
TOF (CO2/Pt/min)
TiO2
30
25
20
15
20 30 40 50 60
Reaction Time (min)
Fig. 4 (A) O 1s XPS spectrum for the Pt/TiO2 catalyst as-deposited (As-dep) and after 1 h under
reaction conditions at 373, 473, and 573 K. The growth of surface O at high binding energy
(~532.5 eV) occurs with increasing reaction temperature. (B) Turnover frequency (TOF) data
revealing the role of surface O to deactivate the undoped catalyst for CO oxidation. Following
reaction at elevated temperature, the catalyst deactivated by a factor of four. Removal of
surface O by 100 Torr CO at 523 K restored the initial activity. Treatment in 100 Torr O2 at 523
K again produced surface O resulting in deactivation. (C) Turnover frequency (TOF) data
showing a change in catalyst activity at 443 K over 1 h. In all cases except for the Pt/TiO2-F, the
catalyst deactivated over this time. However, in the case of Pt supported on F-doped TiO2, the
activity increased by 49, indicating that the highly n-type support can electronically activate
surface O as it forms for reaction with CO.
the ability or inability of electrons from the oxide support to activate it for CO 2 formation. The
high energy of the electrons in the F-doped stoichiometric support is responsible for enhancing
the catalytic activity by electron spillover. However, this effect cannot occur with catalysts
containing high concentrations of O vacancies. Although there is a high density of free carriers
in these supports, they are trapped in the sub-oxide band structure.
62
We also investigated the stability of F in reaction conditions at 373, 473, and 573 K.
XPS showed that F concentration decreased with increasing reaction temperature and was
undetectable after 1 h at 573 K. Fig. 5 shows the F to Ti atomic ratio for F-doped TiO2 as
fabricated and after 1 h in reaction conditions at each of the three treatment temperatures.
Although the surface F concentration began to decreases even at 473 K, it was still present at
significant levels. Consequently, we safely assume that F remained in the support throughout the
duration of the reaction rate measurements shown above which took place at 443 K.
0.35
0.30
As-dep
Ratio
atomic ratio
0.25
0.20
F/Ti Atomic
F/Ti~0.2 at 443K
0.15
0.10
F/Ti
0.05
0.00
300 400 500 600
CO/O2 Reaction Temperature (K)
Fig. 5 F to Ti atomic ratio in F-doped TiO2 as-deposited (As-dep) and after 1 h in reaction
conditions at 373, 473, and 573 K. The atomic ratio is based on the integrated intensity of the F
1s and the Ti 2p XPS spectra and corrected for sensitivity factors. The F concentration
decreases at increasing temperature, but F is still present up to 473 K.
6.5 Conclusions
We achieved approximately twofold increase in the reaction rate of CO oxidation on a

Pt/TiO2 catalyst when the TiO2 is made n-type by F doping. A correlation of this effect with the
electronic structure of titanium oxide indicates that the mechanism responsible for increased
catalytic activity is activation of surface O by electron spillover from the oxide support.
Noting that certain catalytic reactions proceed by charged intermediates, it seems

intuitive that hot carriers can activate non-thermal reaction pathways by spillover to surface
adsorbates. Hot carriers are difficult to generate in a metal catalyst because of their fast
relaxation times. However, n-type doping of oxide semiconductors is a feasible method for
supplying charges of controlled energy to a catalytic interface. This unique utility of oxides also
enlightens the widely studied strong metal-support interaction (SMSI), in which the properties of
an oxide used to support metal clusters strongly influences the activity and selectivity of the
catalyst system.
This result demonstrates that tuning the electronic structure of metal-oxide

semiconductors used as catalyst supports can provide access to reaction pathways that on other
catalysts would not be thermally accessible. This ability implicates applications in many other
63
charge mediated processes, including energy conversion reactions, electrochemical processes,
solid acid/base catalysis, and selective partial oxidation chemistry.
6.6 References

100, 170.
59.
(11) Zhu, H.; Qin, Z.; Shan, W.; Shen, W.; Wang, J. Journal of Catalysis 2004, 225, 267.
(15) Landon, P.; Collier, P. J.; Papworth, A. J.; Kiely, C. J.; Hutchings, G. J. Chemical
Communications 2002, 2058.
(16) Landon, P.; Collier, P. J.; Carley, A. F.; Chadwick, D.; Papworth, A. J.; Burrows, A.;
Kiely, C. J.; Hutchings, G. J. Physical Chemistry Chemical Physics 2003, 5, 1917.
(17) Fu, Q.; Saltsburg, H.; Flytzani-Stephanopoulos, M. Science 2003, 301, 935.
(19) Fan, S.; Yi, J.; Wang, L.; Mi, Z. Reaction Kinetics and Catalysis Letters 2007, 92, 175.
(20) Vernon, P. D. F.; Green, M. L. H.; Cheetham, A. K.; Ashcroft, A. T. Catalysis Letters
1990, 6, 181.
(21) Ashcroft, A. T.; Cheetham, A. K.; Green, M. L. H.; Vernon, P. D. F. Nature 1991, 352,
225.
General 1998, 169, 281.
(23) Haruta, M. Cattech 2002, 6, 102.
(24) Pietron, J. J.; Stroud, R. M.; Rolison, D. R. Nano Letters 2002, 2, 545.
(25) Molina, L. M.; Rasmussen, M. D.; Hammer, B. Journal of Chemical Physics 2004, 120,
7673.
(26) Akubuiro, E. C.; Verykios, X. E. Journal of Catalysis 1988, 113, 106.
G. Science 1999, 285, 1042.
2003, 1184.
64
403.
(30) Zhang, J.; Liu, X.; Blume, R.; Zhang, A. H.; Schlogl, R.; Su, D. S. Science 2008, 322, 73.
(31) Kaden, W. E.; Wu, T. P.; Kunkel, W. A.; Anderson, S. L. Science 2009, 326, 826.
(34) Cox, P. A. Transition Metal Oxides An Introduction to their Electronic Structure and
Properties; Clarendon Press: Oxford, 1992.
(35) Seo, H.; Baker, L. R.; Hervier, A.; Kim, J.; Whitten, J. L.; Somorjai, G. A. Nano Letters
2010, 11, 751.
65
Chapter 7
Controlling the Selectivity of Methanol Oxidation on

Platinum Catalysts by Tuning the Electronic
Structure of a Titanium Oxide Support by Fluorine
Doping
7.1 Abstract
Platinum films of thickness 1 nm were deposited by electron beam evaporation onto 100-
nm-thick titanium oxide films (TiOx) with variable oxygen vacancy concentrations and fluorine
(F) doping. Methanol oxidation on the platinum films produced formaldehyde, methyl formate,
and carbon dioxide. F-doped samples demonstrated significantly higher activity for methanol
oxidation when the TiOx was stoichiometric (TiO2), but lower activity when it was non-
stoichiometric (TiO1.7 and TiO1.9). These results correlate with the chemical behavior of the
same types of catalysts in CO oxidation. Fluorine doping of stoichiometric TiO2 also increased
selectivity toward partial oxidation of methanol to formaldehyde and methyl formate, but had an
opposite effect in the case of non-stoichiometric TiOx. Introduction of oxygen vacancies and
fluorine doping both increased the conductivity of the TiOx film. For oxygen vacancies, this
occurred by the formation of a conduction channel in the bandgap, whereas in the case of
fluorine doping, F acted as an n-type donor, forming a conduction channel at the bottom of the
conduction band, about 0.5-1.0 eV higher in energy. The higher energy electrons in F-doped
stoichiometric TiOx led to higher turnover rates and increased selectivity toward partial oxidation
of methanol. This correlation between electronic structure and turnover rate and selectivity
indicates that the ability of the support to transfer charges to surface species controls in part the
activity and selectivity of the reaction.
7.2 Introduction
Heterogeneous catalysts often consist of metal nanoparticles dispersed over an oxide

support. While the oxide is usually inert by itself, it may affect the catalyst’s behavior in
significant ways, a phenomenon known as strong metal support interaction.1,2 This phenomenon
is in part attributed to charge transfer involving the oxide support.3 For example, it has been
observed that partially covering Rh foil with an oxide enhanced the reaction rates for
hydrogenation of CO and CO2,4 despite the fact that the oxide was not active by itself and
blocked the active Rh sites. The enhancement factor was found to correlate with the Lewis
acidities of the oxides used, i.e., their ability to accept a charge. These results suggest that
modifying the conditions for charge transfer from the oxide support may yield significant
changes in catalytic behavior.
In the present study, methanol oxidation was carried out over Pt/TiO2 catalysts for which
the oxide was modified by introducing oxygen vacancies and fluorine dopant. It is reported that
fluorine doping of stoichiometric TiO2 allows the oxide to transfer an electron to an adsorbate.
66
This activates one or multiple steps in the reaction, leading to a significant increase both in
turnover frequency and selectivity toward the partial oxidation products, formaldehyde and
methyl formate. The present study is the first to correlate a change in selectivity due to doping
of the oxide support with the charge transfer properties of the oxide. Results indicate that
reaction selectivity is controlled, to a significant extent, by charge transfer from the oxide.
The aforementioned results are consistent with a previous study of CO oxidation on

Pt/TiOx catalysts.5 In that study, TiOx films were annealed under different conditions to obtain
various stoichiometries, i.e., TiO1.7, TiO1.9, and TiO2, as determined by X-Ray photoelectron
spectroscopy (XPS). Introducing oxygen vacancies into TiO2 leads to the formation of electronic
states in the bandgap about 0.5–1.0 eV below the bottom of the conduction band.6,7 These
midgap states act as a conduction channel, and the conductivity of the film increases by orders of
magnitude.8,9
One titanium oxide film of each sample was also doped in SF6 plasma, yielding six types
of oxide support: TiO1.7, TiO1.9, and TiO2, both doped and undoped. Fluorine was found to bind
to Ti by filling oxygen vacancies, slightly offsetting the increased conductivity.9 However, F also
acts as an n-type donor, forming donor levels just below the conduction band. For TiO2, this
increases conductivity by forty-fold (Fig. 1B). In the presence of oxygen vacancies, the vacant
midgap states 0.5–1.0 eV below the bottom of the conduction band capture any donor electrons,
and F doping slightly decreases the conductivity of the TiO1.7 and TiO1.9 films.9
While both oxygen vacancies and fluorine doping can increase the film conductivity, the
resulting conduction channels are about 1.0 eV apart in energy. This difference correlates with
the surface chemistry of the Pt/TiOx catalysts. Although the turnover frequency (TOF) increases
by nearly two-fold when stoichiometric TiO2 is F-doped, no such increase is observed with the
non-stoichiometric TiO1.7 and TiO1.9 films (Fig. 1A).
Kinetic measurements of this particular system indicated that O poisons the reaction in
this range of temperature and pressure. As a result, the increase in turnover may be attributed to
electron transfer from the oxide to surface O, activating it for reaction with CO. Non-
stoichiometric TiOx does not show this effect because the conduction channel formed by midgap
states is much lower in energy. Electrons in those states have insufficient energy to transfer to
surface O.
The above information suggests that the turnover rates of a catalyst can be tuned by
modifying the electronic structure of the support. An important contribution of the present
investigation is the proof that the same method can be used to tune selectivity, a metric of great
importance in industrial catalysis. Methanol oxidation (Scheme 1) was chosen for this purpose
because it is relatively simple (only three different products are produced under the conditions of
this study) and turnover can easily be measured at 333 K. At this temperature, the fluorine
concentration in the TiO2 lattice is stable.5 The reaction begins with the adsorption of methanol
that leads to the formation of methoxy, which then dehydrogenates to form formaldehyde that
can either bind with O to form CO2, or with methoxy to form methyl formate.10 Methyl formate
conversion to CO2 was not observed at 333 K. This reaction is of major industrial relevance,
because it is the primary method of producing formaldehyde,11 also forming the basis of
operation of the direct methanol fuel cell.12 In the former, methanol must only be partially
67
(A)
(B)
Fig. 1 (A) Turnover frequencies (TOF) for Pt/TiOx catalysts identical to those of the present
study reported for CO oxidation.5 Fluorine doping only shows a significant effect for
stoichiometric TiO2. Reaction occurred at 443 K under conditions of 40 Torr CO, 100 Torr O2,
and 620 Torr He. TOF data represent stable rates obtained after 30 min in reaction conditions,
when deactivation had stopped. (B) Surface conductivity measurements for the same set of TiOx
supports prior to Pt deposition.5 The numbers for TiO2 and F-doped TiO2 have been magnified
by a factor of 106 for clarity.
oxidized, whereas in the latter, the energy efficiency is maximized if the oxidation is carried out
to form only CO2; hence the need for selectivity control.
7.3 Experimental
Thin films of TiOx with variable O vacancy concentrations were used as supports of the
Pt catalyst. These films were deposited on Si(100) wafers by direct current (DC) magnetron
sputtering under conditions of 400 W plasma power, 450 V bias voltage, and Ar and O2 gas flow
rates of 50 and 3 sccm, respectively.
The Si(100) wafers had a 500-nm-thick thermally grown SiO2 film, while the thickness of
the TiO2 films was equal to 100 nm. As-deposited titanium oxide films were highly O deficient.
The O concentration in the films was controlled by rapid thermal annealing in O2 at temperatures
68
Selectivity toward partial oxidation
+ +
O− * CO2
O CO2 + +
Selectivity toward combustion
Scheme 1
between 623 and 773 K. To avoid variations in grain size due to the different annealing
temperatures, all samples were first annealed in N2 at 773 K. TiOx films of three different
stoichiometries were prepared for this study, i.e., TiO2, TiO1.9, and TiO1.7. These stoichiometries
were determined from the ratio of Ti3+ and Ti4+ in the Ti 2p XPS spectra.9
Fluorine doping was accomplished by plasma treatment in N2 gas and trace amounts of
sulfur hexafluoride (SF6), introduced into the chamber by flowing a 9:1 mixture of SF6 and O2,
followed by pumping to a chamber base pressure of 5 mTorr. A low background pressure of SF6
remained in the chamber after pumping. N2 gas was then introduced at a flow rate of 80 sccm
for plasma treatment under conditions of 20 W power and 130 V DC substrate bias. XPS
analysis of the F-doped samples did not reveal any S traces.9 One of the two samples of each
TiOX stoichiometry was plasma treated, resulting in six supports for subsequent Pt deposition,
i.e., TiO2, TiO1.9, and TiO1.7, each with and without F doping. All substrates were stored in dry
N2 between the time of fabrication and the reaction rate measurements.
Pt nanoparticles were deposited onto TiOx/Si(100) substrates by electron-beam

evaporation. This deposition method was selected instead of colloidal synthesis to avoid the
presence of an insulating polymer capping agent between the Pt nanoparticles and the oxide
support. During electron beam evaporation, the chamber base pressure was less than 10-5 Torr.
The average Pt film thickness, monitored by a quartz crystal microbalance, was equal to about 1
nm.
7.3.2 Reaction Rate Measurements
Reaction rates of methanol oxidation on each of the six catalysts described above were
measured with a batch mode reactor equipped with a boron nitride substrate heater and a metal
bellows circulation pump for gas mixing. In reaction, gas pressures were set at 10 Torr CH3OH
and 50 Torr O2 in He background, while the temperature was fixed at 333 K. All of the samples
showed significant deactivation (up to two orders of magnitude) in the first 60 min of reaction,
thereafter producing stable rates for several hours. Rates reported here refer to the steady-state
rates obtained after initial deactivation.
69
Turnover rates were measured by a gas chromatograph (GC) with a thermal conductivity
detector. Selectivities were estimated after calibrating the detector’s response by injecting
known amounts of CO2 and methyl formate. Because it was difficult to obtain formaldehyde in
pure form, the same calibration could not be performed. Therefore, the thermal conductivity of
formaldehyde was assumed to be the same as that of methyl formate.
Formaldehyde and methyl formate are treated together and referred to as partial oxidation
products. For some samples, the production rate of formaldehyde was less than its consumption
rate, leading to a negative turnover number. That rate was counted as contributing negatively to
selectivity for partial oxidation. Selectivities reported here should not be compared directly to
other results in the literature, and are only used for relative comparisons between samples of this
study.
Reaction rates are presented in the form of turnover frequencies (i.e., number of CO2
molecules per Pt site per minute). Error bars represent a 95% confidence interval based on the
rate of production normalized by the estimated number of Pt sites. Assuming a Pt(111) surface
of uniform structure, the number of Pt sites was determined as the number of Pt atoms over the
catalyst area. Although this is an approximate calculation, it provides consistent normalization
to catalyst area and a reasonable estimate of the absolute turnover rate.
Methanol oxidation was carried out on Pt/TiOx catalysts that differed only in the
preparation of the oxide support. As mentioned earlier, the oxide support was modified by
oxygen vacancy doping and fluorine doping. Fig. 2 shows the effect of these treatments on the
turnover frequencies of the three products, CO2, formaldehyde, and methyl formate. The graph
shows the difference between turnover frequencies of F- .
In the case of stoichiometric TiO2, fluorine doping increased activity ‒ TOF increased by 71%
for CO2, two-fold for formaldehyde, and four-fold methyl formate. An opposite trend was
observed for non-stoichiometric TiO2, i.e., TiO1.9 and TiO1.7. Variation in oxygen vacancy
concentration also exhibited a significant effect on activity, although a clear trend cannot be
established.
A similar fluorine effect was observed when measuring the selectivity of the reaction.
Fig. 3 shows the percentage of partial oxidation products (combining formaldehyde and methyl
formate) obtained in reaction, as opposed to the total oxidation product, CO2. For stoichiometric
TiO2, fluorine doping increased the selectivity toward partial oxidation from 17 to 35%, whereas
for non-stoichiometric TiO2, selectivity toward partial oxidation decreased as a result of fluorine
doping.
The effects of fluorine doping observed here are consistent with a previous study dealing
with CO oxidation.5 In that study, the same catalysts produced turnover frequencies for CO
oxidation that correlated strongly with the conductivities of the TiOx films prior to Pt deposition
(Fig. 1). Selectivities and turnover frequencies reported here for methanol oxidation also
correlate with conductivity measurements (Fig. 1B), suggesting that the same phenomenon
governs the effect of fluorine in both reactions.
70
Fig. 2 Change in turnover frequency Δ(TOF) of CO2, formaldehyde, and methyl formate
products of methanol oxidation over Pt catalysts on six different TiOx supports: TiO1.7, TiO1.9,
and TiO2, both doped and undoped. Δ(TOF) denotes the difference in turnover frequency
between F-doped and undoped TiOx supports. Error bars represent 95% confidence intervals.
Measurements were obtained at 333 K under reaction conditions of 10 Torr CH3OH, 50 Torr O2,
and 720 Torr He.
Fig. 3 Selectivity toward partial methanol oxidation of F-doped and undoped Pt/TiOx catalysts.
Partial oxidation products refer to formaldehyde and methyl formate, as opposed to total
oxidation that produces CO2.
Differences in activities observed in this study can be explained by considering the

charge transfer mechanism in non-stoichiometric and F-doped stoichiometric TiO2, shown
schematically in Fig. 4. TiO2 has a bandgap of 3.2 eV, and the density of electronic states in the
bandgap, by definition, is very low for stoichiometric TiO2. Oxygen vacancies introduced into
the lattice give rise to energy levels in the bandgap (midgap states) between 0.5 and 1.0 eV
71
6
below the bottom of the conduction band, increasing dramatically the conductivity of the
material (Fig. 1B). Fluorine binding to Ti at O vacancy sites partially passivates these midgap
states, resulting in lower conductivity.9 However, in the case of stoichiometric TiO2, where there
are no midgap states, fluorine acts as an n-type dopant, creating donor states just below the
conduction band. This increases conductivity by a factor of 40 (Fig. 1B).
Both oxygen vacancy doping and fluorine doping produce a population of surface
electrons available for conduction. These electrons may also transfer to nearby adsorbate
orbitals, provided they are close in energy. For CO oxidation, this charge transfer occurs in F-
doped stoichiometric TiO2 from the oxide to surface O, which activates the O to react with CO,
leading to increased turnover rate.5 For oxygen-vacancy doped TiO2, midgap electrons produced
from oxygen doping are lower in energy by 0.5-1.0 eV. Because these electrons cannot transfer
to surface O, an increase in turnover is not observed. The effect of fluorine is observed under
steady-state conditions, excluding the possibility of charge accumulation. Therefore, the charged
intermediate must return its additional electron to the oxide support before desorbing as a CO2
molecule. Such a scheme of CO oxidation on platinum has been described in a previous
theoretical study.13 In the last step of the reaction, a charged CO2- intermediate simultaneously
undergoes desorption and loss of an electron to yield CO2.
In the case of methanol oxidation, it is likely that the same activation of a surface species
is responsible for the increased turnover and the shift in selectivity toward formaldehyde and
methyl formate. Further experiments must be carried out to identify which species on the surface
is activated by the support.
A sharp decrease in turnover rate (up to two orders of magnitude) was observed after
approximately 60 min in reaction with each sample, regardless of oxidation state or doping, due
to CO poisoning of the Pt layer. Initial rates obtained before the deactivation did not show a
dependence on the TiO2 substrate, and the turnover frequencies reported here refer to the steady-
state rates obtained after initial deactivation.
Deactivation was found to be reversible. When the sample was cooled down to room
temperature, the gases pumped out, and the reaction restarted, original rates were restored, and
the same deactivation occurred after about 60 min, without any significant change in turnover
rates. CO produced from methanol oxidation is known to poison Pt.14 When 0.5 Torr of CO was
added to the reaction mixture, turnover rates at 333 K decreased below the detection limit. After
the surface was cooled down to room temperature, the reactor evacuated, and a new reaction was
initiated on the same sample without any CO, the turnover rate was restored to its previous value.
This suggests that, although CO was not produced in amounts detectable by GC, it resulted in Pt
poisoning. The reversibility of poisoning indicates that, even at room temperature, a monolayer
of CO is easily removed when the reaction is stopped, in agreement with earlier observations.15
7.5 Conclusions
Methanol oxidation over Pt/TiOx catalysts showed that fluorine doping of stoichiometric
TiO2 increases the selectivity toward the formation of formaldehyde and methyl formate from 17
to 35%. The activities and selectivities of the fabricated Pt/TiOx samples closely followed the
conductivities of F-doped TiOx films, similar to CO oxidation. This effect is likely due to an
72
Fig. 4 Schematic of charge transfer mechanism in non-stoichiometric TiOx and F-doped

stoichiometric TiO2. In the presence of midgap states, electron transfer from the oxide to an
adsorbate species involves an electron with energy between 0.5 and 1.0 eV below the bottom of
the conduction band. In the case of F-doped TiO2, donor electrons occupy states at the bottom
of the conduction band that were previously not thermally accessible. These higher energy
electrons may transfer to an adsorbate species, resulting in higher turnover frequency and
enhanced selectivity toward partial oxidation. (Ec, EF, and Ev denote conduction, Fermi, and
valence energy levels, respectively.)
increased electron population in the conduction band of the oxide, allowing for charge transfer
that activated a surface species. The present study is the first to show that the electronic structure
of the oxide support can be tuned to control the oxidation pathways on catalyst surfaces.
7.6 References
(1) Schwab, G.-M. Discussions of the Faraday Society 1950, 8, 166.

(2) Tauster, S. J.; Fung, S. C.; Garten, R. L. J. Am. Chem. Soc. 1978, 100, 170.
(3) Jochum, W.; Eder, D.; Kaltenhauser, G.; Kramer, R. Topics in Catalysis 2007, 46, 49.
(4) Boffa, A. B.; Lin, C.; Bell, A. T.; Somorjai, G. A. Catalysis Letters 1994, 27, 243.
(5) Baker, L. R.; Hervier, A.; Seo, H.; Kennedy, G.; Komvopoulos, K.; Somorjai, G. A. J.
Phys. Chem. C 2011, 115, 16006.
(6) Cronemeyer, D. C. Phys. Rev. 1959, 113, 1222.
73
(7) Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, 269.
(8) Bilmes, S. A.; Mandelbaum, P.; Alvarez, F.; Victoria, N. M. J. Phys. Chem. B 2011,
104, 9851.
(9) Seo, H.; Baker, L. R.; Hervier, A.; Kim, J.; Whitten, J. L.; Somorjai, G. A. Nano Lett.
2011, 11, 751.
(10) Outka, D. A.; Madix, R. J. J. Am. Chem. Soc. 2011, 109, 1708.
(11) Weissermel, K.; Arpe, H.-J. Industrial organic chemistry; Wiley-VCH, 2003.
(12) Surampudi, S.; Narayanan, S. R.; Vamos, E.; Frank, H.; Halpert, G.; LaConti, A.; Kosek,
J.; Prakash, G. K. S.; Olah, G. A. Journal of Power Sources 1994, 47, 377.
(13) Maximoff, S. N.; Head-Gordon, M. P. Proceedings of the National Academy of
Sciences 2009, 106, 11460.
(14) Beden, B.; Hahn, F.; Leger, J.-M.; Lamy, C.; Lopes, M. I. dos S. Journal of
Electroanalytical Chemistry and Interfacial Electrochemistry 1989, 258, 463.
(15) Zhang, C.; He, H. Catalysis Today 2007, 126, 345.
74
Chapter 8
Furfuraldehyde Hydrogenation on Titanium Oxide-

Supported Platinum Nanoparticles Studied by Sum
Frequency Generation Vibrational Spectroscopy:
Acid–Base Catalysis Explains the Molecular Origin
of Strong Metal–Support Interactions
8.1 Abstract
This work describes a molecular-level investigation of strong metal–support interactions

(SMSI) in Pt/TiO2 catalysts using sum frequency generation (SFG) vibrational spectroscopy.
This is the first time that SFG has been used to probe the highly selective oxide–metal interface
during catalytic reaction, and the results demonstrate that charge transfer from TiO2 on a Pt/TiO2
catalyst controls the product distribution of furfuraldehyde hydrogenation by an acid–base
mechanism. Pt nanoparticles supported on TiO2 and SiO2 are used as catalysts for
furfuraldehyde hydrogenation. As synthesized, the Pt nanoparticles are encapsulated in a layer
of poly(vinylpyrrolidone) (PVP). The presence of PVP prevents interaction of the Pt
nanoparticles with their support, so identical turnover rates and reaction selectivity is observed
regardless of the supporting oxide. However, removal of the PVP with UV light results in a 50-
fold enhancement in the formation of furfuryl alcohol by Pt supported on TiO2, while no change
is observed for the kinetics of Pt supported on SiO2. SFG vibrational spectroscopy reveals that a
furfuryl-oxy intermediate forms on TiO2 as a result of a charge transfer interaction. This
furfuryl-oxy intermediate is a highly active and selective precursor to furfuryl alcohol, and
spectral analysis shows that the Pt/TiO2 interface is required primarily for H spillover. Density
functional calculations predict that O-vacancies on the TiO2 surface activate the formation of the
furfuryl-oxy intermediate via an electron transfer to furfuraldehyde, drawing a strong analogy
between SMSI and acid–base catalysis.
8.2 Introduction
Strong metal–support interactions (SMSI) refer to the ability of a seemingly inert oxide to
have a dominant effect on the catalytic properties of a supported metal nanoparticle.1-4 Tauster
and Fung first used the term SMSI to refer to the dramatic loss of chemisorption sites observed
for noble metal catalysts supported on titanium oxide after reduction.2 In their studies, Tauster
and Fung showed evidence for a strong bonding interaction between metal nanoparticles and a
reduced titanium oxide support, and they used the term SMSI to refer directly to this bonding
interaction.4 However, because this metal–support interaction is closely linked to the catalytic
properties of the metal involved, the definition of SMSI has since expanded to broadly refer to
support-induced changes in the catalytic activity and selectivity of metal nanoparticles. SMSI
plays an important role to enhance many catalytic reactions, including CO oxidation, 5-9 CO and
CO2 hydrogenation,10,11 hydroformylation,12 and partial hydrogenation reactions.13-17 The ability
75
of an oxide support to mediate the catalytic behavior of a supported metal nanoparticle is an
important area in catalysis with both scientific and commercial significance. However, more
than 30 years since its discovery, the mechanism of SMSI catalysis remains an open question.
A number of studies show that the support plays an important role in activating the C=O
bond. A correlation of the catalyst activity for C=O bond hydrogenation with the Lewis acidity
of the oxide empirically shows that charge transfer between the C=O bond and cationic sites in
the oxide controls C=O bond activation.11,18-20 This observation points out a similarity between
SMSI and acid–base catalysis. In acid–base catalysis, the generation of ionic reaction
intermediates determines the reaction rate and selectivity because the charged intermediate is
highly active for a specific reaction pathway.21-23 In SMSI the oxide appears to play a similar
role: electron transfer between a reactant molecule and the oxide support leads to the formation
of a charged reaction intermediate that may be highly selective to a specific reaction pathway.
This hypothesis, which explains why SMSI can dramatically enhance the rate of a single reaction
pathway, is confirmed in this study by direct observation of reaction intermediates on supported
metal catalysts. These results show that SMSI catalysis is a subclass of acid–base chemistry
where the flow of charge determines both reaction rate and selectivity.
In the current study, Pt nanoparticles supported on TiO2 and SiO2 serve as catalysts for
furfuraldehyde hydrogenation. The Pt/SiO2 catalyst serves a reference state, because SiO2 is not
SMSI active.3 We investigate the effects of the support on the catalyst selectivity and correlate
kinetic measurements with the surface intermediates observed by sum frequency generation
(SFG) vibrational spectroscopy. SFG vibrational spectroscopy has been previously used to
observe surface reaction intermediates on metal single crystals and shape controlled metal
nanoparticles.24-27 In this study, SFG is used for the first time to study, at a molecular level, how
an oxide support controls the selectivity of a metal nanoparticle.
We find that initially the support has no effect on the reaction kinetics. This is because a
layer of poly(vinylpyrrolidone) (PVP) encapsulates the Pt nanoparticles and insulates the Pt from
the oxide support. However, by controlled removal of the PVP using UV light,28 direct contact
between the nanoparticles and the oxide support results in a new Pt/support interface. In the case
of Pt/TiO2, this interface leads to a 50-fold enhancement in the formation of furfuryl alcohol,
while the Pt/SiO2 interface has no effect on the reaction kinetics. By SFG we observe that a
furfuryl-oxy intermediate forms on the TiO2 as a result of a Lewis acid–base interaction with the
furfuraldehyde molecule. This furfuryl-oxy intermediate is a highly active and selective
precursor to furfuryl alcohol. Spectral analysis reveals that the Pt/TiO2 interface is required
primarily for H2 spillover from the Pt. Density functional calculations identify O-vacancies on
the TiO2 surface as the catalytically active site and show that the barrier to C=O bond
hydrogenation is dramatically decreased as a result of furfural bonding at the reduced Ti 3+ sites
surrounding an O-vacancy. This represents a detailed molecular understanding of the support-
enhanced reaction mechanism for an SMSI catalyst and demonstrates a strong similarity between
SMSI and acid–base catalysis.
76
8.3 Experimental
To study support effects, 2-dimensional model catalysts were prepared by depositing Pt

nanoparticles onto two different metal-oxide thin films. SiO2 and TiO2 served as the two
supports. As synthesized, the Pt nanoparticles were encapsulated in a layer of PVP. This
capping layer prevented actual contact between the Pt nanoparticle and the oxide support.
However, by photodecomposition of the PVP, the Pt nanoparticles were brought into contact
with the oxide support. By controlling the degree of contact between the particle and the
support, the effect of the oxide-metal interface on catalytic activity and selectivity was studied.
The Pt nanoparticles were synthesized from chloroplatinic acid hexahydrate and PVP in a
1:4 mass ratio. In a small beaker, 110 mg of chlorplatinic acid was dissolved in 10 mL ethylene
glycol. In a separate beaker, 440 mg of PVP was dissolved in 10 mL of ethylene glycol. Once
in solution, the two mixtures were combined into a 50 mL two-neck round bottom flask fitted
with an Ar flow. The solution was purged under vacuum for 15 min. The vessel was then
heated to 438 K for 1 h with vigorous mixing under a flow of Ar. The resulting nanoparticles
were precipitated with acetone and washed three times with ethanol and hexanes. The
nanoparticles were then suspended in chloroform for Langmuir Blodgett (LB) deposition.
Transmission electron microscopy (TEM) showed that the particles were 4.6 ± 2.8 nm.
Two substrates were used to support the nanoparticles: SiO2 and TiO2. For reaction
studies, a Si(100) wafer with a 500 nm thermally grown oxide served as the SiO2 support. The
TiO2 support consisted of a 50 nm TiO2 thin film deposited on the SiO2 substrate by electron
beam evaporation. Following deposition, the TiO2 thin film was annealed at 773 K in O2.
Analogous samples were prepared for sum frequency generation spectroscopic studies. For these
samples, an optically transparent substrate was needed. A sapphire window rather than a Si
wafer served as the substrate. A 50 nm thin film of SiO2 or TiO2 was deposited on the sapphire
window by electron beam evaporation, again followed by annealing at 773 K in O2 to ensure
oxide stoichiometry.
The LB technique was then used to deposit a monolayer of the Pt nanoparticles onto the
metal-oxide supports. This technique has been described previously in detail.25,29 In short, a
suspension of nanoparticles in chloroform is dispersed onto a water surface (18 MΩ). Time is
given for the chloroform to evaporate, leaving a 2-dimensional dispersion of nanoparticles on the
water surface. The film is then compressed with a mobile barrier, and the surface pressure is
monitored as a function of decreasing surface area. The surface pressure corresponds to the
density of nanoparticles on the water. When the desired surface pressure is reached, a substrate
is pulled out from under the surface of the water, and the film of nanoparticles is deposited onto
the substrate. The final density of nanoparticles on the substrate can be controlled by controlling
the surface pressure during deposition. A surface pressure of 14 mN/m was used for these
studies, and films were deposited using a Nima 611 LB trough. Filter paper served as the surface
tension probe. Prior to LB deposition the substrate was cleaned for 1 h in Nochromix solution.
TEM showed that the area coverage for Pt on the substrate following LB was approximately
30%.
77
Immediately prior to reaction, the samples were exposed to UV light in air to
photodecompose the PVP layer that encapsulated the nanoparticles. Two low-pressure mercury
(Hg) lamps (Lights Sources Inc., model number GPH357T5VH/4P) were used as the UV source;
the Hg lamp emitted two lines: 184 and 254 nm. The two lamps were positioned parallel to each
other 2.5 cm apart in a clean Al box. The sample was positioned 1.2 cm below the lamps. By
varying the time of UV exposure, it was possible to control the amount of PVP removed from the
Pt nanoparticles. This cleaning is the combined effect of direct photodecomposition of the PVP
as well as oxidation of the PVP by ozone produced by the 184 nm Hg line.28
8.3.2 Activity and Selectivity Measurements
A stainless steel batch mode reactor was used to determine the reaction rates and
selectivity of furfuraldehyde hydrogenation for Pt supported on SiO2 and TiO2 with varying
degrees of PVP cap removal. The catalyst was heated with a boron nitride substrate heater to a
reaction temperature of 393 K. A metal bellows circulation pump provided gas mixing. Gas
pressures were 1 Torr furfuraldehyde, 100 Torr H2, and 659 Torr He. The furfuraldehyde was
purified by freeze-pump-thaw cycles. Each catalyst was tested for 8 h and reaction products
were monitored as a function of time using a gas chromatograph with a flame ionization detector.
To calculate a turnover frequency (TOF) for each catalyst, it was necessary to determine
the number of Pt active sites. Because ethylene hydrogenation is insensitive to Pt structure30 and
support,11 the number of active sites for each catalyst was determined by measuring the rate of
ethylene hydrogenation and normalizing to a known TOF.31 Ethylene hydrogenation was run on
each catalyst following furfuraldehyde hydrogenation in a separate but identical reactor, and
catalysts were stored under N2 between reactions. The catalyst temperature for the ethylene
reaction was 298 K, and gas pressures were 10 Torr ethylene and 100 Torr H2 in a background of
He.
Turnover frequencies for furfuraldehyde hydrogenation are based on the reaction rate,
measured from a fit to peak area versus time, and normalized to the number of active sites as
determined by ethylene hydrogenation. Selectivity measurements for furfuraldehyde
hydrogenation are calculated as the TOF of a given product normalized to the combined TOF of
all products. Error bars represent 95% confidence intervals for these measurements.
Sum frequency generation (SFG) is a second order, nonlinear process that probes the Χ(2)
tensor. Because Χ(2) is zero for centrosymmetric media, SFG is only sensitive to a break in
inversion symmetry which usually occurs at a surface or interface. Consequently, SFG is useful
for obtaining vibrational spectra of surfaces. In this study, SFG is used to obtain the vibrational
spectra of molecules on the catalyst surface during reaction. Comparison of reaction
intermediates on Pt nanoparticles supported on SiO2 and TiO2 demonstrates that a unique
reaction pathway is active on the Pt/TiO2 catalyst.
Continuum) produced a 20 ps pulse at a 20 Hz repetition rate. The fundamental output at 1064
nm was passed through an optical parametric generator/amplifier to generate a tunable infrared
78
(100 μJ) and VIS (100 μJ) beams were spatially and temporally overlapped on the back surface
of a sapphire window containing the Pt nanoparticles supported on SiO2 or TiO2 thin films. The
VIS and IR beam were incident on the sample at 40° and 50° degrees, respectively, relative to
surface normal. The generated SFG signal was then collected and sent to a photomultiplier tube.
A gated integrator was used to enhance the signal-to-noise. To collect a spectrum, the IR beam
was scanned across the spectral range of interest. All experiments were performed in the ppp
polarization combination.
The beams were directed onto the sample using a sapphire prism as shown in Fig. 1. A
solution of deuterated polystyrene (d8) in deuterated decalin (d18) served as an index matching
liquid that did not interfere with transmission of the IR beam at the C-H stretch frequency. The
catalyst surface was pressed into thermal contact with an aluminum heating block to heat the
catalyst to reaction temperature. A recess in the heating block allowed for the flow of reaction
gasses across the catalyst surface. A metal bellows circulation pump provided gas mixing. A
gas tight seal was made between the sapphire window and the heating block using a Kalrez O-
ring.
Fig. 1 Diagram showing how a sapphire prism directed the VIS and IR beams onto the catalyst
surface for SFG vibrational spectroscopy. The catalyst was prepared on the back side of a
sapphire window and consisted of a thin film of either SiO2 or TiO2 acting as a support for Pt
nanoparticles.
8.3.4 Density Functional Theory Calculations
The calculation is performed using the VASP code based on the density functional
theory. For the exchange-correlation functional, the generalized gradient approximation (GGA)
of Perdew-Burke-Ernzerhof (PBE) is used. The projector augmented-wave pseudopotentials are
used with an energy cutoff of 400 eV for the plane-wave basis functions. To simulate the
molecule absorption on the TiO2 surfaces, a slab model is used with 6 Ti–O double layers (about
20 Angstrom thick) and 10 Angstrom vacuum layer, and a 1×2 supercell (about 10 Angstrom by
8 Angstrom) in the x-y plane. The Brillouin zone integration is carried out using 2×2×1
Monkhorst–Pack k-point meshes.
8.4.1 PVP Cap Blocks Strong Metal–Support Interactions
Although many reaction products exist for furfuraldehyde hydrogenation depending on

the pressure and temperature conditions,17 we observed only four products which we break into
79
two reaction pathways as shown in Scheme 1. The first pathway is decarbonylation to produce
furan. Along this same pathway, furan can further hydrogenate to form tetrahydrofuran (THF) or
further decarbonylate to form propylene.27 The second reaction pathway is selective C=O bond
hydrogenation to produce furfuryl alcohol.
Scheme 1
Fig. 2 demonstrates that the Pt/TiO2 interface plays a major role in the production of
furfuryl alcohol. Each plot shows the formation of reaction products as a function of time. Parts
A and B show the results for PVP-capped Pt nanoparticles supported on TiO2 and SiO2,
respectively. The PVP-capped particles show identical catalytic activity regardless of the
support, indicating that the presence of the PVP cap prevents actual contact between the Pt
nanoparticles and the oxide support. Parts C and D show the results for the same catalysts
following removal of the PVP cap by UV treatment. The cleaned nanoparticles on TiO2 show a
50-fold enhancement in the production of furfuryl alcohol which we attribute to the contact
between the Pt and TiO2 following the removal of PVP. However, the cleaned nanoparticles on
SiO2 show similar activity and selectivity to the capped nanoparticles, indicating that the Pt/SiO2
interface does not play an important role in this reaction.
8.4.2 PVP Cap Removal by UV Cleaning
Fig. 3A shows the C to Pt and the N to Pt atomic ratios of an LB film of nanoparticles as

a function of UV treatment time measured by X-ray photoelectron spectroscopy (XPS). Both C
and N signals decrease relative to the Pt signal with UV exposure, indicating that the PVP cap is
being removed from the Pt nanoparticles. In this experiment the nanoparticles are supported on a
doped Si wafer without any oxide layer to avoid charging during XPS measurements. The
activity of the Pt nanoparticles increases by a factor of ~5 with UV treatment time as determined
by ethylene hydrogenation, indicating that more Pt sites become available as the cap is removed.
The number of active sites after UV cleaning does not vary between TiO2 and SiO2 showing that
the UV cleaning does not depend on the nanoparticle support. Fig. 3B schematically depicts the
effect of UV cleaning on the supported nanoparticles. As the PVP cap is removed, two
important effects are observed: (1) the number of accessible Pt sites increases, so the activity of
the catalyst for ethylene hydrogenation increases, and (2) the Pt nanoparticles come into contact
with the oxide support. Electron microscopy shows that size and shape monodispersity of the
80
Fig. 2 Formation of reaction products as a function of time on four Pt nanoparticle catalysts.

Parts (A) and (B) show results for PVP-capped Pt nanoparticles supported on TiO2 and SiO2,
respectively. Parts (C) and (D) show results for identical catalysts following 3 h UV cleaning to
remove the PVP cap. Furfuryl alcohol is shown in orange, and decarbonylation products (i.e.
furan, THF, and propylene) are shown in blue. The support has no effect on the activity or
selectivity of the capped particles. However, following UV cleaning, the formation of furfuryl
alcohol is selectively enhanced on the Pt/TiO2 catalyst.
UV-cleaned nanoparticles is lost during reaction because the uncapped nanoparticles are not as
thermally stable as the PVP-capped nanoparticles.
8.4.3 Role of Pt/TiO2 Interface on Reaction Kinetics and Selectivity
Fig. 4 shows TOF of furfuraldehyde to decarbonylation products (i.e. furan, THF, and
propylene) and to furfuryl alcohol as a function of UV treatment time for Pt on TiO2 (A) and
SiO2 (B). The activity of each catalyst is normalized to the number of active sites measured by
ethylene hydrogenation. Initially, the PVP-capped Pt does not produce any furfuryl alcohol
regardless of the support. However, as the Pt is brought into close interaction with the TiO2 by
UV cleaning, the activity for furfuryl alcohol production increases. Pt supported on SiO2 does
not produce any furfuryl alcohol outside of measurement error even after extensive UV cleaning.
The data show that Pt activity for decarbonylation does not depend on the support, indicating that
the oxide/metal interface has no effect on this reaction pathway.
Fig. 5 shows the selectivity of Pt supported on TiO2 (A) and SiO2 (B) as a function of UV
treatment time. The selectivity of the Pt/TiO2 catalyst for furfuryl alcohol increases with PVP
removal, reaching ~90% after 180 min UV cleaning. This is a dramatic change in reaction
selectivity changing from ~100% decarbonylation products on the capped nanoparticles to ~90%
81
30 1.8
C:Pt (A)
N:Pt Atomic Fraction

25 1.5
C:Pt Atomic Fraction
N:Pt
20 1.2
15 0.9
(B)
10 0.6 Pt
5 0.3 Metal Oxide
0 0
0 50 100 150 200 UV Treatment Time
UV Treatment Time (min)
Fig. 3 (A) C:Pt (orange) and N:Pt (green) atomic fractions for PVP-capped Pt nanoparticles as
a function of UV cleaning time measured by XPS. Results indicate removal of the PVP cap from
the Pt nanoparticles by UV photodecomposition. (B) Schematic showing the effects of UV
cleaning on PVP-capped Pt nanoparticles. UV cleaning has two effects on the catalyst: (1) the
number of available active sites per nanoparticle increases, and (2) the nanoparticles are
brought into close contact with the support.
Fig. 4 Turnover frequency (TOF) of furfuraldehyde by Pt supported on TiO2 (A) and on SiO2 (B)
as a function of UV treatment time. Orange bars show the formation of furfuryl alcohol,
representing selective C=O bond hydrogenation , and blue bars show the combined formation of
furan, THF, and propylene, representing the decarbonylation pathway. All TOF values are
normalized to the number of Pt actives sites measured by ethylene hydrogenation. Initially, no
furfuryl alcohol is produced by Pt on either support. However, with UV cleaning, the formation
of furfuryl alcohol is selectively enhanced on the Pt/TiO2 catalyst.
furfuryl alcohol on the cleaned nanoparticles. The Pt/SiO2 catalyst shows no effect of PVP
removal on the reaction selectivity and produces only decarbonylation products. These results
suggest that the Pt/TiO2 catalyst generates a unique reaction intermediate that is highly selective
toward the formation of furfuryl alcohol. Below we demonstrate by SFG vibrational
82
spectroscopy that the production of furfuryl alcohol correlates with a furfuryl-oxy intermediate
that forms on TiO2, and this furfuryl-oxy intermediate is the selective precursor to furfuryl
alcohol.
Fig. 5 Selectivity for furfuraldehyde hydrogenation by Pt supported on TiO2 (A) and on SiO2 (B)
as a function of UV treatment time. Furfuryl alcohol is shown in orange, and decarbonylation
products (i.e. furan, THF, and propylene) are shown in blue. The selectivity of the Pt/TiO 2
catalyst changes dramatically with UV cleaning, going from 100% selectivity for
decarbonylation products to ~90% selectivity for furfuryl alcohol. However, the Pt/SiO2 catalyst
is selective only for decarbonylation regardless of UV cleaning time.
8.4.4 Probing Reaction Intermediates by SFG Vibrational Spectroscopy
Fig. 6 shows SFG spectra of Pt/SiO2 (A), TiO2 without Pt (B), and Pt/TiO2 (C) under
reaction conditions following 180 min UV cleaning. The Pt/SiO2 catalyst shows an intense
stretch at 3,030 cm−1. This stretch is significantly lower frequency than the aromatic mode for
the 5-membered furan ring which appears above 3,100 cm−1, and it has been previously assigned
in furan hydrogenation as a vinylic stretch.27 This indicates the presence of an unsaturated furan
ring with broken aromaticity. A loss of aromaticity can occur in two ways: The first is partial
hydrogenation of the ring to dihydrofuran (DHF) as has been previously observed for furan
hydrogenation on Pt single crystals.27 The second is by an interaction of the furan ring with the
catalyst surface via the O atom. In the present study, GC measurements confirm that no DHF
forms on this catalyst which is 60% selective to furan, 20% selective to THF, and 20% selective
to propylene. Consequently, we assign this vinylic stretch to a furan ring bound to the Pt surface
via the O atom as depicted in Fig. 6A.
At lower temperatures we also observe a stretch at 2,765 cm−1 (not shown) indicative of
the aldehyde C–H stretch. However, this stretch disappears upon heating to 393 K due to
decarbonylation of the furfuraldehyde to furan. This process results in CO deposition on the Pt
surface which acts as a poison in the subsequent hydrogenation of furan to DHF, THF, or
butanol. This explains the relatively low activity of this reaction pathway in furfuraldehyde
hydrogenation compared to previous results for furan hydrogenation where at these pressures and
temperatures, THF and butanol would be major products.27 An additional weak stretch appears
83
−1
in Fig. 6A at 2860 cm . We assign this stretch to the symmetric CH3 mode from π-bonded
propylene which is a 20% product on this catalyst.
Fig. 6B shows the spectrum of TiO2 without Pt under reaction conditions. This spectrum
shows two intense features. The first feature appears at 3,130 cm−1 and represents the aromatic
stretch of the furan ring. The second feature is a located near 2,920 cm−1 and represents the CH2
symmetric mode. This spectrum correlates with a furfuryl-oxy intermediate resulting from
furfuraldehdye bonding to the TiO2 via the carbonyl O atom followed by the addition of a single
H atom to the carbonyl C. This furfuryl-oxy surface intermediate is shown schematically next to
the spectrum in Fig. 6B and is the intermediate precursor to furfuryl alcohol formation.
Although TiO2 does not actively dissociate H2, it is not surprising sufficient H atoms are present
to produce a monolayer of furfuryl-oxy surface intermediates. This may be a result of slight H2
dissociation at defect sites in the polycrystalline TiO2 film, or the H atoms may come from
hydroxyl groups that form on the TiO2 surface in ambient or during UV cleaning. However,
without a continual supply of H atoms from supported Pt nanoparticles, the TiO2 substrate does
not turnover.
Fig. 6C shows the spectrum of the Pt/TiO2 catalyst during reaction. In contrast to the
spectra obtained on Pt/SiO2 and TiO2 without Pt, the nonresonant background of the Pt/TiO2
sample is quite high, and the resonant modes appear as negative features against the high
background. This effect, which is common in SFG, is a result of the phase mismatch between
the resonant and nonresonant contributions of the spectrum. In the Pt/SiO2 sample and TiO2
sample without Pt, the nonresonant contribution to the spectrum is low, so any phase mismatch is
not noticeable. However, the nonresonant component of the Pt/TiO2 catalyst is much greater
resulting in destructive interference with the resonant vibrations. The enhanced nonresonant
signal in the Pt/TiO2 catalyst is a result of H spillover from the Pt resulting in a reduced TiO2
support. It is well known that in TiO2, O-vacancies act as electron donors into mid-gap states of
the reduced oxide which results in a greatly enhanced surface conductivity.32 Because the
nonresonant contribution to an SFG spectrum is largely the result of free electron motion at the
surface of the substrate, it is not surprising that increased TiO2 conductivity by H2 reduction
results in a dramatic enhancement of the nonresonant signal. This effect is reversible in O2 as
shown in Fig. 7. Fig. 7 plots the magnitude of the nonresonant signal in alternating H2 and O2
atmospheres for several samples. An effect on the nonresonant signal intensity is only
significant for the Pt/TiO2 catalyst following UV cleaning. The same effect is not observed for
the Pt/SiO2 catalyst and is only weakly observed for the Pt–PVP/TiO2 catalyst before cap
removal. This reflects the non-reducible nature of SiO2 as well as the necessity of removing the
Pt capping agent to enable H spillover to the TiO2.
The resonant components of the Pt/TiO2 spectrum during reaction correlate closely with
the features observed on the Pt/SiO2 and the TiO2 samples, and the Pt/TiO2 spectrum appears to
represent a combination of the spectrum on Pt and the spectrum on TiO2 as shown in Fig 6C.
This indicates that Pt has a similar reactivity when supported on either oxide, but that H spillover
from Pt results in turnover of the furfuryl-oxy intermediate on TiO2 to furfuryl alcohol. As
shown above by kinetic measurements, this reaction pathway is ~10 times faster than the
decarbonylation reaction pathway which occurs only on Pt. This result demonstrates a striking
similarity between SMSI and acid–base catalysis where both processes begin by formation of a
unique reaction intermediate that acts as a highly active intermediate in a selective reaction
84
Fig. 6 SFG spectra of Pt/SiO2 catalyst (A), TiO2 support without Pt (B), and Pt/TiO2 catalyst (C)
during reaction following 180 min UV treatment. Next to each spectrum a schematic depicts the
surface intermediates represented in the spectrum, and red dots on the molecules above show the
C–H bonds responsible for each vibrational mode
pathway. In the case of SMSI, the TiO2 support activates the furfuraldehyde molecule to form
the highly selective furfuryl-oxy intermediate. The role of the crucial Pt/TiO2 interface is simply
to enable H spillover to this active intermediate.
85
Relative Nonresonant Intensity 1.0
0.9
0.8
Pt/TiO2
0.7 Pt/SiO2
0.6 Pt-PVP/TiO2
0.5
0.4
0.3
H2 O2 H2 O2
Fig. 7 Relative nonresonant SFG intensity of a Pt/TiO2 catalyst (blue), Pt/SiO2 catalyst (green),
and Pt–PVP/TiO2 catalyst in alternating H2 and O2 environments (100 Torr) at 333 K. The
Pt/TiO2 and Pt/SiO2 catalysts were UV-cleaned for 10 min (longer UV cleaning treatments do
not change the results shown). The Pt–PVP/TiO2 catalyst was not UV cleaned, so the PVP
encapsulated the Pt nanoparticles and prevented contact between Pt and TiO2. The nonresonant
intensity is determined from the average intensity between 2,700 and 2,800 cm−1 where there are
no resonant features. For the Pt/TiO2 catalyst, the nonresonant background changes
dramatically in H2 and O2. The enhanced nonresonant signal in H2 is a result of H spillover
from the Pt resulting in a reduced TiO2 support, and this effect is reversible in O2. The same
effect is not observed for the Pt/SiO2 catalyst and is only weakly observed for the Pt–PVP/TiO2
catalyst before cap removal. This reflects the non-reducible nature of SiO2 as well as the
necessity of removing the PVP cap to enable H spillover from Pt to TiO2.
8.4.5 Charge Transfer from TiO2 to Furfuraldehyde
Density functional theory (DFT) calculations indicate that O-vacancies on the reduced
TiO2 surface are the catalytically active sites. Furfuraldehyde binding on the [101] surface of
anatase TiO2 is discussed here; binding on other low-energy surfaces was considered with
similar conclusions. When there is no O-vacancy on the surface, the calculations show that the
furfuraldehyde molecule does not bind with either Ti cations or O anions on the surface.
However, when there is an O-vacancy on the surface, furfuraldehyde binds to one of the two Ti
cations closest to the vacancy site, as shown in Fig. 8, and the energy of the system decreases by
1.35 eV. The significant energy decrease associated with the binding can be understood
considering that the Ti cations near the oxygen vacancy are in the reduced Ti3+ state, so one
electron is occupying a high-energy mid-gap state near the conduction band. The binding of
furfuraldehyde molecule oxidizes this cation to a Ti4+ state as one electron transfers to the
furfuraldehyde molecule. The contour in Fig. 8 shows the charge transfer during the binding
process. The electron transfers from the area around the Ti3+ cation (shown by green contour) to
the C=O bond (shown by blue contour). As a new Ti–O bond is formed, the original C=O
double bond is changed into a C–O σ bond.
In the final surface binding state, the O atom in the furfuraldehyde molecule is
coordinated by Ti and C and is in a stable full-shell state. However, the negative charge
86
Fig. 8 The structure plot of a furfuraldehyde molecule binding to a Ti cation at an O-vacancy

site on anatase TiO2, and the charge transfer induced by the surface binding interaction. The
filled cyan, red, brown and white circles show the Ti, O, C and H atoms, respectively. The
hollow green circles show the site of the O-vacancy (without the vacancy, there should be two O
atoms inside the circle). The green contour shows the electron charge loss, and the blue contour
shows the electron charge gain induced by the surface binding. In this charge transfer
interaction, the Ti atom changes from a 3+ to a 4+ oxidation state, and the furfuraldehyde
molecule acquires a negative charge. This negative charge localizes around the carbonyl C,
activating it for H addition.
localized around the C atom represents a dangling bond, which is very reactive for H binding. A
direct calculation shows the energy decreases by 0.2 eV when one H atom leaves the Pt surface
and binds with this reactive C atom. Consequently, this step is fast and results in the steady state
coverage of furfuryl-oxy intermediates on TiO2 observed by SFG. This model of furfuraldehyde
activation at O-vacancy sites in TiO2 is also consistent with the experimental data showing that
SMSI activity correlates with H spillover which occurs only for the UV-cleaned Pt/TiO2 catalyst.
However, in the case of a free furfuraldehyde molecule (not bound to the TiO2 surface), there is
an energy cost of 2.3 eV for a H atom from the Pt surface to bind to the carbonyl C of a
furfuraldehyde molecule. Obviously, this is impossible from a thermodynamic point of view.
Furthermore, the calculations show that furfuraldehyde does not bind with a SiO2 surface. This
is in accordance with the experimental observation that the Pt/SiO2 catalyst is not SMSI active.
8.5 Conclusions
For furfuraldehyde hydrogenation on supported Pt, the oxide support plays a major role
to determine the activity and selectivity of the catalyst. This effect, which is common to many
reactions, is often referred to as a strong metal–support interaction, but a molecular-level
understanding has been lacking notwithstanding the obvious scientific and practical importance.
We demonstrate that when Pt nanoparticles make close contact with a TiO2 support, the Pt/TiO2
interface results in a new reaction pathway that is highly selective toward furfuryl alcohol
formation. SFG vibrational spectroscopy shows that a furfuryl-oxy intermediate forms on TiO2
and is the selective precursor to furfuryl alcohol. The role of the Pt/TiO2 interface is simply to
enable H spillover to this active intermediate, and this reaction pathway is ~10 faster than the
87
reaction rate on Pt alone. DFT calculations suggest that the formation of the active furfuryl-oxy
intermediate is the result of a charge transfer interaction between the furfuraldehyde molecule
and an O-vacancy site on the TiO2 surface. In this charge transfer interaction, the furfuraldehyde
molecule acquires a negative charge that localizes around the carbonyl C, activating it for H
addition. These results provide a detailed picture of the molecular and electronic interactions
that combine to create the SMSI phenomenon and demonstrate that acid–base interactions are the
foundation for highly selective catalysis at the oxide–metal interface.
8.6 References

(2) Tauster, S. J.; Fung, S. C.; Garten, R. L. Journal of the American Chemical
Society 1978, 100, 170.
(3) Tauster, S. J.; Fung, S. C.; Baker, R. T. K.; Horsley, J. A. Science 1981, 211,
1121.
(6) Zhu, H.; Qin, Z.; Shan, W.; Shen, W.; Wang, J. Journal of Catalysis 2004, 225,
267.
(9) Baker, L. R.; Hervier, A.; Seo, H.; Kennedy, G.; Komvopoulos, K.; Somorjai, G.
A. The Journal of Physical Chemistry C 2011, 115, 16006.
(10) Boffa, A. B.; Bell, A. T.; Somorjai, G. A. Journal of Catalysis 1993, 139, 602.
(11) Boffa, A.; Lin, C.; Bell, A. T.; Somorjai, G. A. Journal of Catalysis 1994, 149,
149.
(12) Yamada, Y.; Tsung, C.-K.; Huang, W.; Huo, Z.; Habas, S. E.; Soejima, T.;
Aliaga, C. E.; Somorjai, G. A.; Yang, P. Nat Chem 2011, 3, 372.
(14) Lin, S. D.; Sanders, D. K.; Albert Vannice, M. Applied Catalysis A: General
1994, 113, 59.
(15) Kijeński, J.; Winiarek, P. Applied Catalysis A: General 2000, 193, L1.
(17) Kijeński, J.; Winiarek, P.; Paryjczak, T.; Lewicki, A.; Mikołajska, A. Applied
Catalysis A: General 2002, 233, 171.
(18) Bracey, J. D.; Burch, R. Journal of Catalysis 1984, 86, 384.
(19) Sachtler, W. M. H.; Shriver, D. F.; Hollenberg, W. B.; Lang, A. F. Journal of
Catalysis 1985, 92, 429.
(20) Sachtler, W. M. H.; Ichikawa, M. The Journal of Physical Chemistry 1986, 90,
4752.
(22) Greeley, J.; Norskov, J. K.; Mavrikakis, M. Annu. Rev. Phys. Chem. 2002, 53,
319.
(23) Olah, G. A.; Molnar, A. Hydrocarbon Chemistry; 2nd Edition ed.; John Wiley &
Sons, Inc.: Hoboken, NJ, 2003.
(24) Bratlie, K. M.; Kliewer, C. J.; Somorjai, G. A. The Journal of Physical Chemistry
B 2006, 110, 17925.
88
(25) Bratlie, K. M.; Komvopoulos, K.; Somorjai, G. A. The Journal of Physical
Chemistry C 2008, 112, 11865.
(26) Kliewer, C. J.; Bieri, M.; Somorjai, G. A. Journal of the American Chemical
Society 2009, 131, 9958.
(28) Aliaga, C.; Park, J. Y.; Yamada, Y.; Lee, H. S.; Tsung, C.-K.; Yang, P.; Somorjai,
G. A. The Journal of Physical Chemistry C 2009, 113, 6150.
Chemistry B 2004, 109, 188.
(30) Schlatter, J. C.; Boudart, M. Journal of Catalysis 1972, 24, 482.
(31) Kuhn, J. N.; Tsung, C.-K.; Huang, W.; Somorjai, G. A. Journal of Catalysis 2009,
265, 209.
(32) Seo, H.; Baker, L. R.; Hervier, A.; Kim, J.; Whitten, J. L.; Somorjai, G. A. Nano
Letters 2010, 11, 751.
89
Chapter 9
The Role of an Organic Cap in Nanoparticle

Catalysis: Restructuring of Carbonaceous Material
Controls Catalytic Activity of Platinum
Nanoparticles
9.1 Abstract
Inherent in the colloidal synthesis of nanoparticle catalysts is the presence of an organic

capping agent that encapsulates the nanoparticles to prevent aggregation. However, this capping
agent often remains present on the nanoparticles during catalytic reaction, and the effect of the
cap on catalysis is an important question that will influence the future applications of colloidal
nanoparticles. In this study, the structure of poly(vinylpyrrolidone) (PVP) ligands on Pt
nanoparticles is probed using sum frequency generation (SFG) vibrational spectroscopy. While
the PVP polymer is initially rigid, when the PVP is removed by UV light, carbonaceous
fragments are found on the surface that dynamically restructure in H2 and O2. These fragments
form a porous coating around the Pt in H2 but collapse to a tightly closed shell in O2. Using
ethylene hydrogenation and methanol oxidation as a probe for the catalytic activity of the
nanoparticles in H2 and O2, respectively, it is shown that the structure of these carbonaceous
fragments controls the catalytic activity of the nanoparticles across several orders of magnitude.
Kinetic experiments on thermally-cleaned PVP-capped and oleic acid-capped nanoparticles show
that these findings apply to multiple capping agents and cleaning methods. This work highlights
the dominant role of an organic cap to mediate nanoparticle catalysis and provides one example
where capped nanoparticles are dramatically better catalysts than their uncapped analogues.
9.2 Introduction
Traditional methods for catalyst preparation (i.e. incipient wetness and ion exchange) use
the reduction of a metal salt inside of a mesoporous oxide.1,2 The result is a high surface area
catalyst consisting of metal particles with a broad distribution of sizes and morphologies. This
type of polydisperse heterogeneous catalyst masks the structure sensitivity inherent in
heterogeneous catalysis3-6 because of the impossibility of selecting nanoparticles of a single size
or shape. Recent advances in nanoscience have shown that colloidal synthetic methods can
produce monodisperse nanoparticles with well-defined sizes and shapes.7-11 This advance
marked a new era in heterogeneous catalysis where monodisperse nanoparticles serve as model
catalysts,12-14 and these catalysts have shown that size and shape control the catalytic activity and
selectivity for many reactions.15-21 These results indicate that colloidal nanoscience is an
important tool for the development of new highly selective catalysts necessary to minimize the
environmental impact and improve the economic efficiency of numerous commercial chemical
processes.
90
In colloidal synthetic methods, nanoparticles are necessarily encapsulated in an organic
polymer or surfactant. This organic capping agent lowers the surface energy of the nanoparticle
to prevent aggregation of the particles;11 the cap may also help to control the size and shape of
the nanoparticles.22 This gives rise to an important question regarding the effect of the organic
cap on the catalytic properties of the nanoparticles. It is traditionally thought that the cap acts as
a site blocking agent and lowers the metal surface area available for catalytic reaction. 23 In this
light, it has been assumed that colloidal preparation methods are impractical for commercial
catalytic applications because the presence of the capping agent decreases the apparent metal
dispersion. However, this is an incomplete assumption based on a model that considers the cap
as a passive coating rather than a dynamic shell that can adjust under reaction conditions. It has
been observed in the case of Pt–dendrimer complexes that the dendrimer structure, which is
highly sensitive to gas/liquid conditions, controls access to the Pt surface.24 Although the
dendrimer blocks the Pt surface in air, it adopts an open structure in water that allows ready
access of gasses to the Pt surface. Despite the importance of this effect for nanoparticle
catalysis, little effort has been made to directly probe the role of the nanoparticle cap to mediate
surface adsorption and catalytic activity. The lack of activity in this area is largely due to the
predetermination that an organic cap necessarily hinders nanoparticle catalysis.
Because of the high surface energy of metals, it is impossible to maintain a clean metal
surface, even under conditions of ultra-high vacuum. With this in mind, it is clear that even
catalysts prepared by traditional methods also become “capped” with undefined surface species
as these catalysts quickly become contaminated in ambient conditions or during reaction.25
Accordingly, the capping agents of colloidal nanoparticles may be desirable because they
represent a well-controlled passivation layer which, if carefully selected, may not prevent
catalysis under reaction conditions. This concept is also reminiscent of homogeneous catalysis
where ligands are used to tune the activity and selectivity of single metal ions, 26 and several
examples demonstrate that similar effects are possible on metal clusters and nanoparticles.27,28
The present work investigates the role of poly(vinylpyrrolidone) (PVP) to mediate the
catalytic properties of encapsulated Pt nanoparticles. We probe the molecular structure of the
PVP cap by sum frequency generations (SFG) vibrational spectroscopy. Ethylene hydrogenation
and methanol oxidation serve as model reactions to probe the catalytic activity in reducing and
oxidizing conditions, respectively. The results show that the PVP cap is sterically rigid and
maintains a constant structure regardless of gas environment. However, when the PVP is
removed by photodecomposition using UV light, carbonaceous fragments are found on the
surface that reversibly restructure in H2 and O2. In O2 atmosphere, these carbonaceous fragments
form a tightly closed shell around the nanoparticles that blocks catalytic activity, but this shell
opens in H2. As a result, the PVP-capped nanoparticles are more than 10 times more active for
methanol oxidation than their cleaned analogues, while the UV-cleaned nanoparticles are more
than 10 times more active for ethylene hydrogenation. Kinetic experiments on thermally-cleaned
PVP-capped and oleic acid (OA)-capped nanoparticles show similar results indicating that these
findings apply to multiple capping agents and cleaning methods. This work highlights the
dominant role of organic material (controlled or otherwise) to mediate nanoparticle catalysis and
provides one example where capped nanoparticles are dramatically better catalysts than their
cleaned analogues.
91
9.3 Experimental
9.3.1 Nanoparticle Synthesis
The Pt nanoparticles were synthesized from chloroplatinic acid hexahydrate and PVP in a
1:4 mass ratio. In a small beaker, 110 mg of chlorplatinic acid was dissolved in 10 mL ethylene
glycol. In a separate beaker, 440 mg of PVP was dissolved in 10 mL of ethylene glycol. Once
in solution, the two mixtures were combined into a 50 mL two-neck round bottom flask fitted
with an argon flow. The solution was purged under vacuum for 15 min. The vessel was then
heated to 438 K for 1 h with vigorous mixing under a flow of argon. The resulting nanoparticles
were precipitated with acetone and washed three times with ethanol and hexanes. The
nanoparticles were then suspended in chloroform. Transmission electron microscopy (TEM)
showed that the particles were 4.6 ± 2.8 nm.
To investigate the effect of different capping agents and cleaning methods on reaction
kinetics, Pt nanoparticles were synthesized using an alternate method where the capping agent is
added after synthesis. This allowed us to add different capping agents to identical nanoparticle
aliquots to isolate the cap as the sole variable between two nanoparticle samples.23 In a small
beaker, 350 mg of chloroplatinic acid was dissolved in 17.5 mL ethylene glycol. In a separate
beaker, 350 mg of NaOH was dissolved in 17.5 mL ethylene glycol. Once in solution, the two
mixtures were combined in a 50 mL two-neck flask fitted with an argon flow. The solution was
purged under vacuum for 15 min. The vessel was then heated to 433 K for 3 h with vigorous
mixing under a flow of argon. Aliquots (2 mL) of the resulting nanoparticles were precipitated
with 2 M HCl, and then re-dispersed in 2 mL ethanol containing 10 mg of either PVP or OA.
TEM showed that the particles were 1.7 ± 0.8 nm.
9.3.2 Langmuir Blodgett Deposition
Formation of 2-dimensional films of monodisperse nanoparticles to serve as model

catalysts is routinely achieved by Langmuir Blodgett deposition. This technique has been
described previously in detail.12,29 In short, a suspension of nanoparticles in chloroform is
dispersed onto a water surface (18 MΩ). Time is given for the chloroform to evaporate, leaving
a 2-dimensional dispersion of nanoparticles on the water surface. The film is then compressed
with a mobile barrier, and the surface pressure is monitored as a function of decreasing surface
area. The surface pressure corresponds to the density of nanoparticles on the water. When the
desired surface pressure is reached, a substrate is pulled out from under the surface of the water,
and the film of nanoparticles is deposited onto the substrate. The final density of nanoparticles
on the substrate can be controlled by controlling the surface pressure during deposition. A
surface pressure of 14 mN/m was used for these studies, and films were deposited using a Nima
611 LB trough. Filter paper served as the surface tension probe.
The substrate used to support the nanoparticles was a TiO2 thin film (50 nm) deposited on
a Si(100) wafer with a thermally grown SiO2 layer (500 nm). The TiO2 thin film was deposited
on the SiO2/Si wafer by electron beam evaporation from an oxide source without any substrate
heating. Following deposition, the TiO2 thin film was annealed at 773 K in O2 to increase
crystallinity and ensure a fully oxidized stoichiometry. Analogous samples were prepared for
sum frequency generation spectroscopic studies. For these samples, an optically transparent
92
substrate was needed. A sapphire window rather than a Si wafer served as the substrate. A TiO2
thin film (50 nm) was deposited on the sapphire window by electron beam evaporation, again
followed by annealing at 773 K in O2. The LB technique was then used to deposit a monolayer
of the Pt nanoparticles onto the TiO2 thin films. Electron microscopy showed that the area
coverage for Pt on the substrate following LB was 30–50%. In some cases, TiO2 used as a Pt
support plays an active role in the catalytic chemistry. However, the conclusions reported here
are not substrate dependent, and kinetic results have been obtained for identical nanoparticles
supported on SiO2 substrates, and no significant differences were observed.
LB deposition is challenging for OA-capped Pt nanoparticles owing to the high

hydrophobicity of the OA. Consequently, for investigating the role of different capping agents,
drop cast films of nanoparticles on a TiO2 substrate were used as catalysts for both the PVP- and
OA-capped samples.
9.3.3 UV Cleaning
Immediately prior to reaction, the samples were exposed to UV light in air to remove the
PVP capping layer. Two low-pressure mercury (Hg) lamps (Lights Sources Inc., model number
GPH357T5VH/4P) were used as the UV source; the Hg lamp emitted two lines at 184 and
254 nm. The two lamps were aligned parallel to each other 2.5 cm apart in a clean Al box. The
sample sat 1.2 cm below the lamps. By varying the time of UV exposure, it was possible to
control the amount of PVP removed from the Pt nanoparticles. This cleaning is the combined
effect of direct photodecomposition of the PVP and oxidation of the PVP by ozone produced
from the 184 nm Hg line.30
X-ray photoelectron spectroscopy (XPS) was used to observe the removal of PVP from
the Pt nanoparticles after UV cleaning. Spectra were obtained using a Physical Electronics
system (PHI 5400 ESCA/XPS) with an Al anode source. The analyzer was positioned at 50°
relative to sample normal. The C 1s, N 1s, and Pt 4f peak areas were fit and normalized by the
appropriate sensitivity factors to obtain the surface C:Pt and N:Pt atomic fractions as a function
of UV exposure time. Measurements showed that 90% of the C was removed from the Pt
following 3 h UV exposure.
Thermal cleaning in air was used as an alternative to UV cleaning for drop cast samples.
For thermal cleaning, the samples were placed in a tube furnace and heated to a desired
temperature for 16 h. The PVP- and OA-capped samples were treated at 473 and 573 K,
respectively to remove the two capping agents that show different thermal stabilities.
Sum frequency generation (SFG) is a second order, nonlinear process that probes the Χ (2)
tensor. Because Χ(2) is zero for centrosymmetric media, SFG is only sensitive to a break in
inversion symmetry which usually occurs at a surface or interface.31 Consequently, SFG is
useful for obtaining vibrational spectra of surfaces. In this study, SFG is used to obtain the
vibrational spectrum of the nanoparticle capping layer in H2 and O2 atmospheres.

Continuum) produced a 20 ps pulse at a 20 Hz repetition rate. The fundamental output at 1064
93
nm was passed through an optical parametric generator/amplifier to generate a tunable infrared
(100 μJ) and VIS (100 μJ) beams were spatially and temporally overlapped on the back surface
of a sapphire window containing the Pt nanoparticles. The VIS and IR beams were incident on
the sample at 40° and 50° degrees, respectively, relative to surface normal. The generated SFG
signal was then collected and sent to a photomultiplier tube. A gated integrator was used to
enhance the signal-to-noise. To collect a spectrum, the IR beam was scanned across the spectral
range of interest. All experiments were performed in the ppp polarization combination.
The beams were directed onto the sample using a sapphire prism. A solution of
deuterated polystyrene (d8) in deuterated decalin (d18) served as an index matching liquid
between the prism and substrate that did not interfere with transmission of the IR beam at the C-
H stretch frequency. The catalyst surface was pressed into thermal contact with an aluminum
heating block to heat the catalyst to the desired temperature. A recess in the heating block
allowed for the flow of gasses across the catalyst surface. A metal bellows circulation pump
provided gas mixing. A gas tight seal was made between the sapphire window and the heating
block using a Kalrez O-ring.
9.3.5 Kinetic Measurements
A stainless steel batch mode reactor was used to determine the reaction rates for ethylene
hydrogenation and methanol oxidation on the Pt nanoparticle catalysts before and after UV
cleaning. The catalyst temperature was controlled with a boron nitride substrate heater. A metal
bellows circulation pump provided gas mixing. For ethylene hydrogenation, gas pressures were
10 Torr ethylene and 100 Torr H2 in a background of He, and the catalyst temperature was
298 K. For methanol oxidation, gas pressures were 10 Torr methanol and 50 Torr O2 in a
background of He, and the catalyst temperature was 333 K. The methanol was purified by
freeze-pump-thawing cycles. Each catalyst was tested for 2 h, and reaction products were
monitored as a function of time using a gas chromatograph with a thermal conductivity detector.
Fig. 1A shows the SFG spectra of PVP-capped Pt nanoparticles in sequential gas

environments. As the gas environment is initially cycled between H2 and O2, the SFG signal is
higher in O2 and decreases in H2 corresponding to a structural reconfiguration of PVP on the Pt.
Because of selection rules, SFG is sensitive not only to the individual molecular susceptibilities,
but also to the net susceptibility of the entire ensemble of molecules at an interface. This net
susceptibility depends strongly on the relative orientation of molecules at the interface. It is
usually assumed that systems showing greater SFG signal intensity have a more highly ordered
interface than those showing weaker intensity because a net disorder leads to canceling out of
signal from individual molecular oscillators. Accordingly, we assign the loss of signal intensity
going from O2 to H2 atmosphere as H2-induced disordering of the PVP. However, after the first
H2/O2 cycle, the PVP settles into a relatively stable configuration on the Pt, and minimal
restructuring occurs as the gas environment continues to change between reducing and oxidizing
conditions.
94
7.0 3.0
CH CH2(s) CH2(a) (A) CH3(s) CH2(s) (B)

6.0 2.0
Intensity
H2
5.0 1.0
O2 O2
4.0 0.0
Intensity
H2
2700 2800 2900 3000 3100 3200
3.0 3.0
O2
O2 (C)
H2
2.0 2.0
Intensity
H2
O2
1.0 1.0
O2
H2 H2
0.0 0.0
2700 2800 2900 3000 3100 3200 2700 2800 2900 3000 3100 3200
Frequency (cm−1) Frequency (cm−1)
Fig. 1 SFG spectra of the PVP-capped Pt nanoparticles before UV cleaning (A) and after UV
cleaning for 10 min (B and C). Spectra were obtained in 100 Torr of H2 or O2 in a background
of Ar at 60 °C. A boxcar average over 5 data points improved signal-to-noise. In (A) and (C)
the spectra are arbitrarily offset for clarity and were obtained sequentially from bottom to top.
(B) shows the raw spectra with no offset of UV-cleaned nanoparticles first in H2 then in O2. The
spectrum obtained in H2 shows a much higher nonresonant intensity with resonant features
appearing as dips in the high background. This is a result of the phase mismatch between the
resonant and nonresonant contributions of the spectrum. By subtracting the raw spectra from a
baseline obtained in Ar, the negative features are inverted to appear positive for clarity (shown
in C). Analysis of the resonant features in these spectra shows that before UV cleaning the PVP
has a conformation that does not significantly change depending on the gas environment.
However, following UV cleaning, carbonaceous fragments are found on the surface that are
dynamic and reversibly restructure in H2 and O2.
This is in strong contrast to spectra shown on identical nanoparticles following UV

cleaning to remove PVP from the Pt surface. Fig. 1B shows the raw spectra of UV-cleaned
nanoparticles first in H2 then in O2. The spectrum obtained in H2 shows a much higher
nonresonant intensity with resonant features appearing as negative peaks against the high
background. This effect, which is common in SFG, is a result of the phase mismatch between
the resonant and nonresonant contributions of the spectrum. By subtracting the raw spectra from
a baseline obtained in Ar, the negative features are inverted to appear positive for clarity (see
Fig. 1C). The enhanced nonresonant signal of the UV-cleaned catalyst is a result of H spillover
from the Pt resulting in a reduced TiO2 support, and the effect is completely reversible in O2. In
the PVP-capped sample, the nonresonant contribution to the spectrum is low even in H2 because
the PVP blocks H spillover to the TiO2.
95
Fig. 2 Formation of reaction products as a function of time on the Pt nanoparticles for ethylene
hydrogenation and methanol oxidation before UV cleaning (A and B) and after UV cleaning for
10 min (C and D). These results show that UV cleaning has a diverging effect on the rates of
ethylene hydrogenation which increases by a factor of 10 and methanol oxidation which
decreases by a factor of 3. Only CO2 production is shown for methanol oxidation. However,
formaldehyde was also produced as a minor product before and after UV cleaning.
Fig. 1C shows that as the atmosphere is cycled between H2 and O2, carbonaceous
fragments are present on the Pt surface, and these fragments reversibly restructure, adopting a
structure that is strongly SFG active in H2, and a structure which is almost entirely SFG inactive
in O2. It may be tempting to conclude that O2 reacts with the remaining cap on the Pt surface to
catalyze its complete removal. However, this is not the case as can be seen by the reversible
appearance of the molecular vibrations when H2 is re-introduced to the system. This change
occurs reversibly over multiple cycles with no sign of signal decrease over time that would
suggest eventual removal of the carbonaceous fragments.
It is also evident that the vibrational modes observed following UV cleaning are not the
same as those observed from the intact PVP cap. Initially the vibrational modes observed on
PVP-capped Pt (see Fig. 1A) match closely the assignments previously reported by infrared and
Raman for PVP-Pt complexes.32 Following UV cleaning, the peaks observed are attributable to
CH2 and CH3 symmetric stretching modes outside of a 5-membered ring. It is also important to
note that XPS measurements indicate that the surface C has decreased more than 85% compared
96
to the fully-capped nanoparticles following UV cleaning. It is not definitive whether the
carbonaceous fragments observed by SFG after cleaning are photodecomposition products of
PVP, or if they represent contamination of the UV-cleaned Pt surface, which would be
unavoidable in ambient. However, it is clear that additional UV cleaning does little to remove
these species.
Fig. 2 shows the catalytic activity of the Pt nanoparticles for ethylene hydrogenation and
for methanol oxidation before and after UV cleaning. We selected these two reactions because
they operate at low temperature (i.e., 298 K and 333 K, respectively) and allowed us to observe
the effects of the cap structure on the catalytic activity of the nanoparticles in H2 and O2
atmospheres. The PVP-capped Pt nanoparticles are active for both reactions as shown in Fig. 2A
and B. However, following UV cleaning, the activity of the nanoparticles for ethylene
hydrogenation and methanol oxidation diverge. Fig. 2C and D show that following UV cleaning,
the rate of ethylene hydrogenation increases by a factor of 10 while the rate of methanol
oxidation decreases by a factor of 3. This represents a 30-fold divergence on the effect of UV
cleaning for these two reactions.
10.0
Normalized Reaction Rate
8.0
Methanol Oxidation
Methanol Oxidation
Methanol Oxidation
Ethylene Hydrogenation
6.0
4.0
2.0
0.0
H2 O2 H2 O2 H2 O2
-2.0
Fig. 3 The reaction rates for ethylene hydrogenation and methanol oxidation on a single catalyst
following UV cleaning for 10 min. The reaction rates are normalized to the initial rate for each
reaction before UV cleaning. The results show that the diverging effect of UV cleaning on the
two reactions is reversible and appears to correlate with the restructuring of carbonaceous
fragments on the Pt surface observed by SFG.
Fig. 3 shows the activity of the catalyst following UV cleaning where the catalyst is
cycled several times between ethylene hydrogenation and methanol oxidation. The rates are
normalized to the initial rate for each reaction prior to UV cleaning. It can be seen that the rates
for both reactions are reversibly affected by the removal of the PVP cap. In the case of ethylene
hydrogenation, this is easily understood based on the increased number of Pt sites available
following PVP removal. In the case of methanol oxidation, it is surprising that cap removal
would have a negative effect on the catalytic activity. It appears that the structure of the
97
carbonaceous fragments observed on the Pt following UV cleaning controls the catalytic
properties of the uncapped nanoparticles. We suggest that these carbonaceous fragments which
do little to prevent access to the Pt in H2 atmosphere, collapse to a tightly closed shell around the
Pt in O2 atmosphere.
8.0
1,4-cyclohexadiene
7.0
6.0
5.0 200 Torr H2

Intensity
4.0
10 Torr H2
3.0
1 Torr H2
2.0
0 Torr H2
1.0 10 Torr Cyclohexene
Vacuum
0.0
2700 2800 2900 3000 3100 3200
Frequency (cm−1)
Fig. 4 SFG spectra of 10 Torr cyclohexene in increasing H2 pressure on Pt nanoparticles after

UV cleaning for 60 min. In this experiment, the nanoparticles are supported directly on a SiO2
prism. Cyclohexene dehydrogenates on Pt to form a 1,4-cyclohexadiene surface intermediate
that is characterized by a strong feature in the SFG spectrum at 2,760 cm−1. This feature is
known to form on an atomically clean Pt surface even in the absence of H2. In this case, the
feature is initially absent from the spectrum, but grows in with increasing H2 pressure. This
indicates that H2 is needed to open the carbonaceous shell on the UV-cleaned Pt nanoparticles
and allow cyclohexene to access the Pt.
To confirm this model in which carbonaceous fragments form a porous Pt coating in H2

but collapse to a tightly closed shell in O2, we used cyclohexene to probe the accessibility of
reactants to the Pt surface. Cyclohexene forms a 1,4-cyclohexadiene surface intermediate on
Pt.33 This is a dehydrogenation product of the cyclohexene and forms even in the absence of H2.
The 1,4-cyclohexadiene species has a distinct molecular vibration at 2,760 cm−1 which is
spectrally well resolved from any resonant modes of the capping agent. Consequently,
cyclohexene represents a suitable probe molecule. To avoid changes in the nonresonant
background with changing H2 pressure, this experiment was performed using Pt nanoparticles
supported directly on an SiO2 prism. Fig. 4 shows the results of this experiment. When
cyclohexene is introduced in the gas phase to UV-cleaned Pt nanoparticles, no 1,4-
cyclohexadiene is observed in the SFG spectrum. Only weak features at 2,840 and 2,910 cm −1
98
appear which we attribute to physisorbed cyclohexene. However, as the H2 pressure is increased
from 0 to 200 Torr, a strong feature at 2,760 cm−1 grows in corresponding to 1,4-cyclohexadiene.
Because this feature forms on an atomically clean Pt single crystal surface even in the absence of
H2,33 we conclude that for the UV-cleaned nanoparticles H2 is needed to open the carbonaceous
shell and allow cyclohexene to access the Pt. It is important to note that in this experiment the
nanoparticles were pre-reduced with H2, so that the oxidation state of Pt is not changing.
Fig. 5A shows the effect of UV cleaning time on the activity of the Pt nanoparticles for
ethylene hydrogenation and methanol oxidation. The activity for each reaction is normalized to
the initial activity of the PVP-capped nanoparticles before UV cleaning. For ethylene
hydrogenation the rate increases with increased cleaning time, while for methanol oxidation the
rate decreases with increased cleaning time. Following 3 h there is a 200-fold divergence of the
rates for these two reactions. This suggests that with increased cleaning, the carbonaceous shell
on the Pt nanoparticles becomes tighter and tighter in O2; however, it can continue to open in H2
allowing access to the Pt. In fact, the ethylene hydrogenation rate after UV cleaning is slightly
greater than expected based on the geometric surface area of the Pt as determined by TEM. This
suggests that the carbonaceous fragments have almost no site blocking effect in H2. Fig. 5B
shows the corresponding C and N surface concentrations as a function of UV cleaning time
measured by XPS. After 30 min, the concentration of N is below the limit of detection by XPS.
The C concentration also levels off after 30 min to a steady value representing only 10% of the
initial C. Again it is impossible to distinguish if this C is a decomposition product of the PVP or
if it is contamination of the cleaned Pt surface.
50.00 30 1.5
Normalized Reaction Rate
Ethylene Hydrogenation (A) (B)
N:Pt Atomic Fraction

C:Pt Atomic Fraction
24 1.2
5.00
18 0.9
12 0.6
0.50
6 0.3
Methanol Oxidation
0.05 0 0.0
1 10 100 1000 0 40 80 120 160 200
UV Cleaning Time (min) UV Cleaning Time (min)
Fig. 5 (A) Reaction rates as a function of UV cleaning time for ethylene hydrogenation and
methanol oxidation. The reaction rates are normalized to the initial rate for each reaction
before UV cleaning. The results show that the rates of the two reactions continue to diverge with
increased cleaning time, eventually showing a 200-fold difference in activity. (B) C:Pt and N:Pt
atomic fractions measured by XPS as a function of UV cleaning time. The N concentration on
the surface drops below the limit of detection after 30 min of UV cleaning. The C concentration
on the surface also levels off at a value representing only 10% of the initial C from the PVP
capping layer.
99
To determine if the formation of this carbonaceous shell is a general phenomenon, we
synthesized Pt nanoparticles using an alternate method where the capping agent is added after
synthesis. This allowed us to add different capping agents to identical nanoparticle aliquots to
isolate the cap as the sole variable between two nanoparticle samples. For this experiment, PVP
and oleic acid (OA) were used as capping agents in two separate aliquots, and the activity of
these catalysts were monitored for ethylene hydrogenation and methanol oxidation before and
after cleaning. In this case, thermal oxidation in a tube furnace was used for cap removal rather
than UV cleaning. The OA-capped nanoparticles were cleaned at 473 K, but the PVP-capped
nanoparticles required thermal treatment at 573 K for PVP removal consistent with previous
work on the thermal degradation of Pt–PVP complexes.32 Following cleaning the PVP- and OA-
capped nanoparticles showed a 100-fold and 30-fold increased activity for ethylene
hydrogenation, respectively.
100.00
As synthesized
Methanol Turnover Frequency
11.2 After cleaning

10.00
2.8 3.3
1.00
0.10 0.08
0.05
0.01
0.01
Pt-PVP Pt-PVP Pt-Oleic Acid
(UV-Cleaned) (O2-Cleaned) (O2-Cleaned)
Fig. 6 Methanol oxidation turnover frequency (TOF) for Pt nanoparticle catalysts before and
after cap removal. TOF is given as the number of methanol molecules converted per Pt site per
s. The number of platinum sites was determined from the measured rate of ethylene
hydrogenation using a known TOF. Consequently, each bar shows the methanol oxidation rate
relative to the ethylene hydrogenation rate on the same catalyst. Data is shown for all three
types of nanoparticles studied (i.e., UV-cleaned Pt–PVP, thermally-cleaned Pt–PVP, and
thermally-cleaned Pt–OA). Noting that this graph is on a log-scale, the Pt activity for methanol
oxidation is shown to decrease for each catalyst by a factor of between 50 and 250 following cap
removal.
Fig. 6 shows the effect of cap removal on the activity for methanol oxidation. In this
figure, the methanol oxidation rate for each catalyst is represented as a turnover frequency
normalized to the number of Pt active sites determined by ethylene hydrogenation. 23
Consequently, each bar shows the methanol oxidation rate relative to the ethylene hydrogenation
rate on the same catalyst. Data is shown for all three types of nanoparticles studied (i.e., UV-
cleaned Pt–PVP, thermally-cleaned Pt–PVP, and thermally-cleaned Pt–OA). Noting that this
graph is on a log-scale, the Pt activity for methanol oxidation is shown to decrease for each
100
catalyst by a factor of between 50 and 250 following cap removal. Because these values are
normalized to the catalyst activity for ethylene hydrogenation, this cannot be the result of
nanoparticle sintering. It seems clear that for methanol oxidation, the capped nanoparticles are
dramatically more active than their cleaned analogues, and this finding applies to multiple
capping agents and cleaning methods.
9.5 Conclusions
These results demonstrate the important role of an organic cap to mediate the catalytic properties
of colloidal nanoparticles. We studied the effect of UV cap removal on the catalytic activity of
PVP-capped Pt nanoparticles for ethylene hydrogenation and methanol oxidation. UV cleaning
effectively removes the PVP cap from the Pt, but SFG shows that carbonaceous fragments are
still present on the nanoparticles. These carbonaceous fragments are dynamic and appear to
block Pt active sites in during methanol oxidation but not during ethylene hydrogenation. We
propose a model where a carbonaceous shell forms on the Pt nanoparticle that is tightly closed in
O2 but becomes permeable in H2. To support this model, cyclohexene served as a probe
molecule to show that following UV cleaning, reactant molecules can only access the Pt when H2
is present to open the carbonaceous shell. Kinetic measurements on PVP- and OA-capped
nanoparticles cleaned by thermal oxidation show that this effect applies to multiple capping
agents and cleaning methods. This work highlights the important role of an organic cap in
nanoparticle catalysis and provides one example where capped nanoparticles are dramatically
better catalysts than their cleaned analogues.
9.6 References
(1) Preparation of Solid Catalysts; Ertl, G.; Knozinger, H.; Weitkamp, J., Eds.; Wiley-VCH:
Weinheim, Germany, 1999.
(2) Tsoncheva, T.; Dal Santo, V.; Gallo, A.; Scotti, N.; Dimitrov, M.; Kovacheva, D. Appl
Catal a-Gen 2011, 406, 13.
(3) Strongin, D. R.; Carrazza, J.; Bare, S. R.; Somorjai, G. A. Journal of Catalysis 1987, 103,
213.
(4) McCrea, K. R.; Parker, J. S.; Somorjai, G. A. The Journal of Physical Chemistry B 2002,
106, 10854.
(5) Andersson, M. P.; Abild-Pedersen, E.; Remediakis, I. N.; Bligaard, T.; Jones, G.;
Engbwk, J.; Lytken, O.; Horch, S.; Nielsen, J. H.; Sehested, J.; Rostrup-Nielsen, J. R.; Norskov,
J. K.; Chorkendorff, I. Journal of Catalysis 2008, 255, 6.
2009, 131, 9958.
(7) Ahmadi, T. S.; Wang, Z. L.; Green, T. C.; Henglein, A.; El-Sayed, M. A. Science 1996,
272, 1924.
(8) Peng, X.; Wickham, J.; Alivisatos, A. P. Journal of the American Chemical Society 1998,
120, 5343.
(9) Puntes, V. F.; Krishnan, K. M.; Alivisatos, A. P. Science 2001, 291, 2115.
(10) Oh, M.; Mirkin, C. A. Nature 2005, 438, 651.
(11) Yin, Y.; Alivisatos, A. P. Nature 2005, 437, 664.
Chemistry B 2004, 109, 188.
101
(13) Rioux, R. M.; Song, H.; Hoefelmeyer, J. D.; Yang, P.; Somorjai, G. A. The Journal of
(14) Song, H.; Rioux, R. M.; Hoefelmeyer, J. D.; Komor, R.; Niesz, K.; Grass, M.; Yang, P.;
7, 3097.
C 2009, 113, 8616.
Chem 2010, 2, 36.
(22) Zhang, Y.; Grass, M. E.; Kuhn, J. N.; Tao, F.; Habas, S. E.; Huang, W.; Yang, P.;
(23) Kuhn, J. N.; Tsung, C.-K.; Huang, W.; Somorjai, G. A. Journal of Catalysis 2009, 265,
209.
(24) Albiter, M. A.; Crooks, R. M.; Zaera, F. The Journal of Physical Chemistry Letters 2009,
1, 38.
(25) Lu, J.; Fu, B.; Kung, M. C.; Xiao, G.; Elam, J. W.; Kung, H. H.; Stair, P. C. Science
2012, 335, 1205.
(27) Li, Y.; Liu, J. H.-C.; Witham, C. A.; Huang, W.; Marcus, M. A.; Fakra, S. C.; Alayoglu,
P.; Zhu, Z.; Thompson, C. M.; Arjun, A.; Lee, K.; Gross, E.; Toste, F. D.; Somorjai, G. A.
Journal of the American Chemical Society 2011, 133, 13527.
(28) Mitsudome, T.; Mikami, Y.; Matoba, M.; Mizugaki, T.; Jitsukawa, K.; Kaneda, K.
Angewandte Chemie International Edition 2012, 51, 136.
(29) Bratlie, K. M.; Komvopoulos, K.; Somorjai, G. A. The Journal of Physical Chemistry C
2008, 112, 11865.
(30) Aliaga, C.; Park, J. Y.; Yamada, Y.; Lee, H. S.; Tsung, C.-K.; Yang, P.; Somorjai, G. A.
The Journal of Physical Chemistry C 2009, 113, 6150.
(31) Shen, Y. R. The principles of nonlinear optics; Wiley-Interscience, 2003.
(32) Borodko, Y.; Habas, S. E.; Koebel, M.; Yang, P.; Frei, H.; Somorjai, G. A. The Journal
of Physical Chemistry B 2006, 110, 23052.
(33) Yang, M.; Chou, K. C.; Somorjai, G. A. The Journal of Physical Chemistry B 2003, 107,
5267.
102
Chapter 10
Removing Poly(vinylpyrrolidone) from Size-

Controlled Platinum Nanoparticles for Catalytic and
Sum Frequency Generation Spectroscopic Studies:
Solvent Cleaning as an Alternative to UV Cleaning
10.1 Abstract
Recent work with nanoparticle catalysts reveals that size and shape control on the
nanometer scale allows for control of reaction rate and selectivity. Additionally, it has been
shown that SFG vibrational spectroscopy is a powerful tool for studying heterogeneous catalysis
because it allows for the observation of surface intermediates during catalytic reactions. To
control the size and shape of catalytic nanoparticles, an organic ligand acts as a cap to stabilize
the nanoparticle during synthesis. The presence of an organic cap presents two major challenges
to SFG and catalytic reaction studies: First, the presence of the cap may block active sites on the
nanoparticle reducing the catalytic activity. Second, the cap often shows a strong signal in the
SFG spectrum, making the observation of reaction intermediates on the nanoparticles difficult or
impossible. In this work, we consider two methods for cleaning Pt nanoparticles capped with
poly(vinylpyrrolidone) (PVP): solvent cleaning and UV cleaning. We find that both UV
cleaning and solvent cleaning make it possible to observe reaction intermediates by SFG.
However, we find that the solvent cleaning method produces nanoparticles that are stable in
reaction conditions, while UV cleaning of the nanoparticles leads to particle agglomeration
during reaction. This understanding makes it possible now to study size- and shape-induced
nanoparticle selectivity with SFG vibrational spectroscopy.
10.2 Introduction
Sum frequency generation (SFG)1,2 vibrational spectroscopy is an interface-specific,

second order optical process that measures the vibrational spectrum of molecules on a surface or
at an interface. It has been shown that SFG vibrational spectroscopy is a powerful tool for
studying heterogeneous catalysis because it allows for the observation of surface intermediates
during catalytic reactions. Previously this technique has been applied to studying reactions
occurring on both single crystal catalysts3-5 and 2-dimensional films of nanoparticle catalysts6,7.
Recent work with nanoparticle catalysts reveals that size and shape control on the nanometer
scale allows for control of reaction selectivity.8-10 This is a major advance for heterogeneous
catalysis, and motivates the present effort to observe reaction intermediates on size- and shape-
controlled nanoparticles under reaction conditions.
A major challenge of these studies is to observe the reaction intermediates on the

background of molecular stretches that originate from the organic capping layer of the
nanoparticles. Nanoparticles require a layer of organic ligands to stabilize both its size and
shape. However, the organic cap has two major effects on the catalytic reaction studies. First,
103
the presence of the cap may block active sites on the nanoparticle reducing the catalytic activity.
Second, the cap often shows a strong signal in the SFG spectrum, making the observation of
reaction intermediates on the nanoparticles difficult or impossible. Common capping agents,
including poly(vinylpyrrolidone) (PVP),11 cetyltrimethylammonium bromide (CTAB),12
tetradecyltrimethylammonium bromide (TTAB),13 hexadecylamine (HDA),14 and oleylamine15,
produce signal in the aliphatic range making it difficult, if not impossible, to identify C-H
vibrations of reactive catalytic intermediates.
In previous studies, UV light has been used to clean the organic cap from the
nanoparticles.14,15 The results of these studies show that UV treatment increases the catalytic
activity of the nanoparticles and minimizes C-H vibrations, leaving a clean background for
observing reaction intermediates by SFG. The present study focuses on a comparison between
this UV method for cleaning the nanoparticles and a simpler, yet effective method, namely,
solvent washing to remove excess capping from the nanoparticles. In our studies, we use Pt
nanoparticles capped with PVP. We find that both UV cleaning and solvent cleaning make it
possible to observe reaction intermediates by SFG. However, we find that the solvent cleaning
method produces nanoparticles that are stable in reaction conditions, while UV cleaning of the
nanoparticles leads to particle agglomeration during reaction. This advance makes it possible
now, for the first time, to study size- and shape-induced nanoparticle selectivity with SFG
vibrational spectroscopy.
The cyclohexene (C6H10) hydrogenation reaction is a good candidate for SFG analysis
because its intermediates have unique and well-documented5,16-20 vibrational signatures. Three
different reaction intermediates of cyclohexene hydrogenation exist on Pt and are easily
identifiable from the SFG spectrum: 1,4-cyclohexadiene (1,4-CHD), 1,3-cyclohexadiene (1,3-
CHD) and the π-allyl intermediate. The spectrum of 1,4-CHD has a single peak at 2760 cm-1
corresponding to a strongly red-shifted CH2 asymmetric stretch. 1,3-CHD has three CH2
stretches (2830, 2875, 2900 cm-1)17. The π-allyl has a strong CH2 asymmetric stretch at 2920
cm-1 and a weak symmetric CH2 stretch at 2840 cm-1. Using cyclohexene hydrogenation on Pt
as a model reaction, we investigate the catalytic activity and vibration spectra of solvent-cleaned
Pt nanoparticles under reaction conditions and compare these results with PVP-capped
nanoparticles and UV-cleaned nanoparticles.
10.3 Experimental
10.3.1 Nanoparticle Synthesis
4.2 nm Pt particles were synthesized using 100 mg chloroplatinic acid hexahydrate

(H2Pt(IV)Cl66H2O, 37.5% metal basis, Sigma Aldrich), 440 mg PVP (29 K, Sigma Aldrich),
and 20 mL ethylene glycol (ReagentPlus, Sigma Aldrich). The mixture was combined in a 25
mL round bottom flask, pumped to remove air and flushed with Ar. The flask was then placed in
an oil bath at 165 °C for 1 h under Ar flow.
3.0 nm Pt nm nanoparticles21 were synthesized using 40 mg ammonium

tetrachloroplatinate(II) ((NH4)2Pt(II)Cl4), 220 mg PVP (29 K, Sigma Aldrich), and 20 mL
ethylene glycol (ReagentPlus, Sigma Aldrich). The mixture was combined in a 25 mL round
104
bottom flask, pumped to remove air and flushed with Ar. The flask was then placed in an oil
bath at 165 °C for 30 min under Ar flow.
1.7 nm Pt particles22 were synthesized by combining 250 mg chloroplatinic acid

hexahydrate (H2Pt(IV)Cl66H2O, 37.5% metal basis, Sigma Aldrich), 25 mL ethylene glycol
(ReagentPlus, Sigma Aldrich) and 0.00625 g NaOH under Ar at 433 K for 1 h. After synthesis,
the particles were precipitated with 1.0 M HCl and then combined with PVP.
The nanoparticles of each size were separated from the ethylene glycol synthesis mixture
by precipitation with acetone and centrifugation at 4000 RPM for 5 min. The nanoparticles were
then dissolved by sonication in 20 mL of ethanol and then mixed with 25 mL of hexane and
centrifuged again for precipitation. This process was repeated 3 times to remove ethylene glycol
and excess chloroplatinic acid from the synthesis. In this paper,we refer to these nanoparticles as
fully-capped.
10.3.2 Solvent Cleaning
For the 4.2 and 3.0 nm nanoparticles solvent cleaning consisted of five additional cycles
of ethanol washing followed by precipitation in hexane for a total of eight ethanol washes.
Following these cycles of ethanol washing, the particles were further washed 2 times, once in
chloroform and once in isopropyl alcohol.
1.7 nm Pt retain less PVP and do not require as many washing cycles. These particles
were washed only one additional time in ethanol for a total of four ethanol washes. If washed
more times, the nanoparticles aggregated in solution.
10.3.3 Langmuir-Blodgett (LB) Film Deposition
Particles were dissolved in a 50/50 ethanol/chloroform mixture before being deposited

onto a water surface for LB deposition.23 For LB deposition, a suspension of nanoparticles in the
ethanol/chloroform mixture is dispersed onto a water surface. Time is given for the organic
solvent to evaporate, leaving a 2-dimensional dispersion of nanoparticles on the water. The film
is then compressed, and the surface pressure is monitored. The surface pressure corresponds to
the density of nanoparticles on the water. When the desired surface pressure is reached, the SFG
prism is pulled out from under the surface of the water, and the film of nanoparticles is deposited
onto the substrate. For SFG experiments, a surface pressure of 20 dyn/cm were used for LB
films. Film deposition on TEM grids was achieved by dropping the grid on the compressed LB
film for 10 sec followed by retrieval with tweezers. For TEM experiments, a surface pressure of
only 8 dyn/cm was used. The lower density used in TEM experiments provided better images
and more clear visibility of nanoparticle aggregation that occurred for some samples during
catalytic reaction. The Nima 611 LB trough was used for all experiments with filter paper as the
surface tension probe.
10.3.4 UV Cleaning
Two low-pressure mercury lamps (Lights Sources Inc., model number

GPH357T5VH/4P) were used for UV treatment inside a clean aluminum box. Samples were
105
placed 5 mm below one lamp and exposed to 254 and 185 nm irradiation for the specified period
of time.
10.3.5 Sum Frequency Generation (SFG)
A mode-locked Nd:YAG dye laser (Continuum D-20) with a fundamental output of 1064
nm, 20 Hz repetition rate and 20 picosecond pulse width was used for all experiments described
herein. A frequency doubling crystal is used to generate the fixed visible (532 nm) from the
fundamental beam. An optical parametric generator/amplifier produces tunable infrared (2680 to
3180 cm-1) corresponding to stretching modes of aliphatic and aromatic groups. Visible and
infrared beams are spatially and temporally overlapped at the base of the prism with angles 65
and 42°, respectively, to achieve total internal reflection. Visible and infrared beam powers were
both 130 μJ before entering the prism. All experiments are performed with the ppp polarization
combination. Polished fused silica equilateral (60°) prisms (ISP Optics) were used for all
experiments. SFG photons are detected using a photomultiplier tube with a gated boxcar
integrator. Samples probed by SFG consist of a nanoparticle film deposited on the prism
surface. This film is probed under reaction conditions at pressures up to 1 atm.24
10.3.6 Transmission Electron Microscopy
Transmission electron microscopy (TEM) images of LB films were performed on a JOEL

2100 microscope operated at 200 kV. Silicone (SiO) films on copper grids (Electron
Microscopy Sciences) were used as substrates for all TEM images presented. Scanning Electron
Microscope (SEM) images were taken with a Zeiss Gemini Ultra-55 microscope on thermally
grown silica (SiO2) substrates.
10.3.7 Kinetic Measurements
Turnover rates were measured using a Hewlett-Packard 5890 Series II gas chromatagraph
connected to a circulating batch reactor. The ST-123 column was used to measure ethylene
hydrogenation rate and the XY3928 column was used for cyclohexene hydrogenation.
10.4.1 PVP Removal is Necessary to Detect Reaction Intermediates on Pt Nanoparticles
The capping agent presents a challenge for SFG studies on colloidal nanoparticle
catalysts because signal from the cap often overlaps with signal from the reactive intermediate of
interest. Even if the cap did not interfere with the detection of reaction intermediates, it still
physically blocks much of the available catalyst surface, preventing adsorption of reactants at
high surface coverage. We find that in the presence of H2, PVP disorders dramatically, such that
the background signal produced by the cap is greatly reduced (see Fig. 1A). However, even in
the absence of a background signal from the PVP, no signal is observed from a reaction
intermediate during cyclohexene hydrogenation (see Fig. 1B). This is because very little
adsorption of cyclohexene occurs because the disordered PVP still blocks the majority of Pt
sites.
106
Fig. 1 SFG spectra for 4.2 nm Pt nanoparticles fully-capped with PVP at 295 K. The film is
exposed to an inert atmosphere (760 Torr Ar), H2 (200 Torr H2, 560 Torr Ar) and cyclohexene
hydrogenation conditions (10 Torr C6H10, 200 Torr H2, 560 Torr Ar). Spectra are offset for
clarity in right plot. Unlabeled features are attributed to PVP.
Contrast these results from capped nanoparticles with the results from solvent-cleaned
nanoparticles. In the case of solvent-cleaned nanoparticles, there is still a strong signal from the
PVP (see Fig. 2A). This is because solvent cleaning does not remove all the PVP from the
nanoparticles; if it did, the particles would aggregate in solution. As in the case of the fully-
capped particles, the signal from the PVP diminishes in H2 due to the disordering effect (see Fig.
2A). However, the major difference between the capped particles and the solvent-cleaned
particles is obvious from the spectrum taken under reaction conditions (see Fig. 2B). In the case
of solvent-cleaned nanoparticles, the 1,4-CHD intermediate is clearly visible indicating
significant adsorption on the Pt surface. From this we conclude that although the solvent-
cleaned particles are still lightly capped with PVP, there are sufficient Pt sites available for
catalysis to occur and to observe the reaction intermediates by SFG.
Because the peak of 1,4-CHD occurs at a unique location compared to other aliphatic
vibrations, it is certain that it arises from a chemisorbed reaction intermediate on the surface of
Pt and not from the PVP. Interestingly, cyclohexene is introduced with H2, PVP peaks are
further decreased from the level observed with only H2. This suggests 1,4-CHD may displace
PVP from the Pt surface into an even more disordered geometry. Spectra were also taken during
cyclohexene hydrogenation on 1.7 and 3.0 nm Pt particles. These spectra (not shown) also
indicate the formation of the 1,4-CHD surface intermediate. 1.7 nm Pt particles retain less PVP
than large particles and do not require as much solvent cleaning as the large particles to observe
the surface intermediate (see Experimental Section for details).
The role of solvent cleaning to increase the Pt surface area is intuitive. However, the role
of the H2 to decrease the background signal produced by the PVP is less obvious, so we discuss
it now in detail. H2 dramatically lowers the signal from PVP on both fully-capped and solvent-
107
Fig. 2 SFG spectra for solvent-cleaned 4.2 nm Pt nanoparticles at 295 K. The film is exposed to
an inert atmosphere (760 Torr Ar), H2 (200 Torr H2, 560 Torr Ar) and cyclohexene
hydrogenation conditions (10 Torr C6H10, 200 Torr H2, 560 Torr Ar). Spectra are offset for
clarity in right plot. Unlabeled features are attributed to PVP.
cleaned nanoparticles. Under an inert atmosphere of Ar gas, the full signal of PVP capping is
strong for both types. However, adding H2 reduces the signal of PVP to relatively weak peaks
located at 2850 cm-1 (CH2-s, chain), 2870 cm-1 (CH2-s, ring), 2930 cm-1 (CH2-a, ring) and 2990
cm-1 (CH2-a, ring).25
Selection rules of SFG indicate decreasing concentration or increasing disorder reduce

SFG signal. The effect of concentration on the spectral intensity is intuitive, but it is important
here to emphasize the effect of molecular ordering on the spectral intensity. Because SFG
vibrational spectroscopy is a coherent process, it is not only sensitive to the individual molecular
susceptibilities, but also to the net susceptibility of the entire ensemble of molecular oscillators.
Consequently, if molecules on a surface are highly disordered, the net ensemble will not produce
a strong SFG signal, even if the concentration is high.
The following two examples emphasize this concept: In a recent study,26 the ordering of
self-assembled monolayers of C60 functionalized phosphoric acid was tuned by varying
fluorination of the phosphoric acid alkyl chain. For certain F/C60 ratios, SFG intensity from C60
molecules decreased even as surface density increased. This effect is attributed to disordering of
the self-assembled monolayer. In another case,5 the observed peaks from the π-allyl intermediate
on Pt(111) increased in intensity by eight times when pressure was changed from ambient to
ultra-high vacuum. Although the surface concentration did not increase in vacuum relative to
ambient pressure, an increase in order strongly contributed to the signal enhancement. We
conclude that the addition of H2 disorders PVP on Pt as indicated by a decrease in SFG signal of
~90%. This spectral change is reversible, and if the cell is evacuated back to vacuum, the full
signal of PVP reappears.
108
UV light at 254 and 185 nm is known to photo-decompose organic material, including
PVP, through photolysis and ozonolysis mechanisms.27 Accordingly, UV cleaning is proposed28
as a viable method for removing the PVP cap from Pt nanoparticles, but the effects of UV
cleaning are on the catalytic behavior of the Pt nanoparticles are not fully understood.
Fig. 3 shows SFG spectra of PVP-capped Pt nanoparticles after 180 min UV exposure.
The spectrum of the Pt nanoparticles in Ar (Fig. 3A) shows a significant signal reduction in the
aliphatic range compared to the fully-capped or solvent-cleaned nanoparticles under the same
condition. This is due to the photodecomposition of PVP by the UV light. XPS data
substantiates the idea of PVP removal by UV cleaning; XPS shows that the C content on the Pt
nanoparticles decreased by 90% following UV cleaning. However, when the UV-cleaned
nanoparticles are exposed to H2, the signal in the aliphatic range increases significantly. This
suggests that although UV cleaning removes the majority of the PVP, it leaves a C shell that
reacts with H2. This observation highlights the difficulty of preparing atomically clean Pt
nanoparticles by UV cleaning. We find that even extensive cleaning by UV light, fails to remove
a C shell that subsequently reacts with H2. This shell appears to be a dynamic actor in the
catalytic reaction, and further studies are in progress to probes the effect of the C shell on Pt
catalysis. However, the shell does not prevent adsorption of cyclohexene, and the 1,4-CHD, 1,3-
CHD, and π-allyl intermediates are all readily observed on the UV-cleaned particles under
reaction conditions (see Fig. 3B).
Fig. 3 SFG spectra for UV-cleaned 4.2 nm Pt nanoparticles at 295 K. The film is exposed to an
inert atmosphere (760 Torr Ar), H2 (200 Torr H2, 560 Torr Ar) and cyclohexene hydrogenation
conditions (10 Torr C6H10, 200 Torr H2, 560 Torr Ar). Spectra are offset for clarity in right plot.
Fig. 4A shows the SFG of Pt nanoparticles (4.2 nm) during cyclohexene hydrogenation
reaction conditions as a function of UV cleaning time, and Fig. 4B shows the C:Pt atomic ratio
also as a function of UV cleaning measured by XPS. We find that with UV cleaning the PVP is
removed as evidenced by a decrease in the XPS C signal. Correspondingly, we find that the
109
Fig. 4 SFG spectra (A) collected at 295 K after UV treatment (0 to 180 min) under cyclohexene
hydrogenation reaction conditions (10 Torr C6H10, 200 Torr H2, 550 Torr Ar). The observed
intermediates change from being exclusively 1,4-CHD (0 to 3 min) to 1,4-CHD, 1,3-CHD and the
π-allyl (30 to 180 min). XPS results (B) show the C/Pt ratio as a function of UV exposure. TEM
images (C, D) collected after UV show clustering following long treatment but no aggregation
(scale bar = 10 nm).
110
signal from reaction intermediates steadily increases as the nanoparticles are cleaned. This is
most obvious in the signal from the 1,4-CHD intermediate at 2765 cm-1. However, there are
additional changes to the spectrum that occur with increased UV cleaning. For particles treated
0-3 min, only the 1,4-CHD is observed. After 10 min of UV cleaning, an intense new peak
appears at 2900 cm-1 with a weak counterpart at 2850 cm-1, revealing the π-allyl adsorption
geometry. With UV cleaning times from 30 to 180 min, the π-allyl signal decreases and weak
peaks appear indicating 1,3-CHD. To observe each of these three intermediates on the Pt[111]
single crystal, the temperature must change from 303 to 483 K.20
We have confirmed by TEM (see Fig. 4C and D) that the nanoparticles are
morphologically stable during these experiments that were performed only at low temperature
(298 K). Consequently, we cannot attribute the change in observed intermediates to a size or
shape effect. It is interesting to note that the 1,3-CHD intermediate is much more active for
cyclohexene hydrogenation than the 1,4-CHD intermediate.29 Accordingly, we would expect
that the 180 min UV-cleaned nanoparticles would be more catalytically active than the fully-
capped nanoparticles because the highly active 1,3-CHD intermediate is only observed on the
catalyst after UV cleaning. In fact, this is true. The UV-cleaned catalyst is approximately three
times more active for cyclohexene hydrogenation than the fully-capped catalyst, even after
correcting for the number of Pt active sites. These spectral changes, which correlate with
reaction kinetics, suggest that besides simply blocking Pt sites, the capping agent may also
actively promote or inhibit the formation of certain reaction intermediates. The idea of
considering the cap as an active, dynamic support for nanoparticle catalysis is an interesting
direction which we hope address in the future.
10.4.2 Solvent Cleaning Preserves the Size Monodispersity of Pt Nanoparticles During Reaction
We now compare the stability of the nanoparticles under reaction conditions following
solvent cleaning and UV cleaning. Fig. 5 shows TEM images of fully-capped Pt nanoparticles of
three sizes (4.2 nm, 3.0 nm, and 1.7 nm) as-deposited, following UV cleaning, and following UV
cleaning and exposure to H2 at 373 K. For the 4.2 and 3.0 nm Pt nanoparticles, UV cleaning
leads to aggregation of particles into clusters, but the particles are still discrete and maintain a
monodisperse size distribution. However, following exposure to H2 at 373 K, these
nanoparticles melt causing a loss of size monodispersity. The 1.7 nm Pt particles are even less
stable and melt after UV cleaning even before exposure to elevated temperature. Although the
images shown were taken on silicone TEM grids, we have observed the same phenomenon on
several other supports, including SiO2 and TiO2.
Fig. 5 also shows TEM images demonstrating the stability of solvent-cleaned Pt

nanoparticles. These images show the solvent-cleaned nanoparticles of 4.2, 3.0, and 1.7 nm sizes
following H2 exposure at 473 K (100 K higher temperature than for the UV-cleaned
nanoparticles). We see that even at 474 K the solvent-cleaned nanoparticles are completely
stable. This represents a major advantage of solvent cleaning compared to UV cleaning.
Although UV cleaning is effective for removing the PVP, it leads to instability of the
nanoparticles in reaction conditions so size monodispersity is lost. However, solvent cleaning
produces stable nanoparticle catalysts that, for the first time, will allow for SFG studies of
nanoscale size effects in catalysis.
111
Fig. 5 TEM images (scale bar = 10 nm) comparing UV-cleaned and solvent-cleaned Pt
nanoparticles before and after reaction. For 4.2 and 3.0 nm nanoparticles, 60 min UV causes
clustering where neighboring nanoparticles move close to one another while still maintaining
size. Under hydrogenation reaction conditions (200 Torr H2, 560 Torr Ar, 373 K), clustered
nanoparticles often fuse and make much larger particles. Spherical 1.7 nm nanoparticles
aggregate immediately under 60 min UV. In contrast, solvent-cleaned nanoparticles are
unchanged up to at least 473 K with H2 (200 Torr H2 with 560 Torr Ar).
10.4.3 Both Solvent Cleaning and UV Cleaning Increase the Number of Pt Active Sites
We also investigated the role of solvent cleaning and UV cleaning on the available
surface area of the nanoparticles. For this study we used ethylene hydrogenation at 298 K as a
way to probe the number of Pt active sites on a LB film. Ethylene hydrogenation is a structure-
insensitive reaction with a well characterized turnover frequency (TOF) on Pt.30 Accordingly the
activity of a catalyst for ethane production from ethylene can be used to determine the
approximate number of active Pt sites on the catalyst.
Fig. 6 shows the number of Pt active sites per cm2 for LB films of fully-capped, solvent-
cleaned, and UV-cleaned nanoparticles. The surface coverage of Pt nanoparticles on the LB
films was controlled by the surface pressure of the nanoparticles during film deposition, and
112
should not vary significantly between samples. We find that at a surface pressure of 30 mN/m,
there is a 30–50% surface coverage of Pt nanoparticles deposited on the substrate. Assuming a
4.2 nm spherical particle with a [111] surface, this surface coverage would indicate
approximately 2×1015 Pt sites per cm2 if there were no poisoning by the PVP cap.
Fig. 6 Active sites calculation based on ethylene hydrogenation rate for 4.2 nm Pt-PVP with
three preparations: fully-capped, solvent-cleaned, and UV-cleaned.
We find that the fully-capped nanoparticles show 5×1014 active sites per cm2 as measured
by ethylene hydrogenation which represents approximately 20% of the geometric area. The
solvent-cleaned nanoparticles show 1×1015 active sites per cm2, and the UV-cleaned
nanoparticles show an 4×1015 active sites per cm2 as measured by ethylene hydrogenation,
corresponding to 40% and >100% of the geometric area, respectively. We attribute the high
activity of the UV-cleaned particles to a deviation of the actual number of active sites per particle
from what is expected by assuming a sphere with a perfect [111] surface. However, this
comparison is still insightful and shows that while solvent cleaning increases the available
number of Pt sites by a factor of 2, UV cleaning is much more effective for removing the PVP
cap, increasing the number of available Pt sites by a factor of 10. This likely also explains the
instability of UV-cleaned particles during reaction compared to the solvent-cleaned particles.
The solvent-cleaned particles are stabilized by a significant amount of PVP that still caps more
than half of the nanoparticle surface.
10.5 Conclusions
We have addressed two major challenges to observing catalytic reaction intermediates on

ligand-capped nanoparticles: 1) interference by SFG signal generated by the organic cap and 2)
blocking of catalytic active sites by the organic cap. We demonstrate that the under
hydrogenation reaction conditions, PVP on Pt disorders such that the SFG signal in the C-H
stretch range from the PVP is negligible. This provides a clean background for probing reaction
intermediates by SFG. We also demonstrate that solvent cleaning of the PVP from the
nanoparticles increases the number of Pt active sites. This increases both the catalytic activity of
the nanoparticles and allows for sufficient signal from adsorbed reactants to probe the vibrational
113
spectrum of reaction intermediates by SFG. While a similar increase in catalytic activity can be
obtained by UV cleaning of the Pt nanoparticles, UV cleaning leads to agglomeration of the
nanoparticles during reaction at temperatures as low as 373 K. However, solvent-cleaned
nanoparticles are completely stable, even in reaction conditions in excess of 473 K.
Accordingly, solvent cleaning is desirable when probing nanoscale size and shape effects in
nanoparticle catalysis because it allow for SFG studies without compromising the size
monodispersity of the nanoparticles.
10.6 References
(1) Lambert, A. G.; Davies, P. B.; Neivandt, D. J. Applied Spectroscopy Reviews 2005, 40,
103.
(2) Shen, Y. R. Nature 1989.
2009, 131, 9958.
(4) Yang, M.; Somorjai, G. A. Journal of the American Chemical Society 2004, 126, 7698.
(5) Yang, M.; Chou, K. C.; Somorjai, G. A. Journal of Physical Chemistry B 2003, 107,
5267.
(6) Bratlie, K. M.; Komvopoulos, K.; Somorjai, G. A. The Journal of Physical Chemistry C
2008, 112, 11865.
7, 3097.
(10) Somorjai, G. A.; Park, J. Y. Journal of Chemical Physics 2008, 128.
(11) Borodko, Y.; Humphrey, S. M.; Tilley, T. D.; Frei, H.; Somorjai, G. A. Journal of
Physical Chemistry C 2007, 111, 6288.
(12) Ullah, M. H.; Chung, W. S.; Kim, I. Small 2006.
(13) Kim, J.; Chou, K. C.; Somorjai, G. A. Journal of Physical Chemistry B 2003, 107, 1592.
(14) Ramirez, E.; Erads, L.; Philippot, K. Advanced ‚Ä¶ 2007.
(15) Zheng, H.; Smith, R. K.; Jun, Y.; Kisielowski, C. Science 2009.
(16) Su, X. C.; Kung, K.; Lahtinen, J.; Shen, R. Y.; Somorjai, G. A. Catalysis Letters 1998,
54, 9.
(17) Su, X. C.; Kung, K. Y.; Lahtinen, J.; Shen, Y. R.; Somorjai, G. A. Journal of Molecular
Catalysis a-Chemical 1999, 141, 9.
(18) McCrea, K. R.; Somorjai, G. A. Journal of Molecular Catalysis a-Chemical 2000, 163,
43.
(19) Yang, M.; Somorjai, G. A. Journal of Physical Chemistry B 2004, 108, 4405.
(20) Yang, M. C.; Chou, K. C.; Somorjai, G. A. Journal of Physical Chemistry B 2004, 108,
14766.
(21) Kuhn, J. N.; Tsung, C. K.; Huang, W.; Somorjai, G. A. Journal of Catalysis 2009, 265,
209.
(22) Rioux, R. M.; Song, H.; Hoefelmeyer, J. D.; Yang, P.; Somorjai, G. A. Journal of
114
(23) Song, H.; Kim, F.; Connor, S.; Somorjai, G. A.; Yang, P. D. Journal of Physical
Chemistry B 2005, 109, 188.
(24) Kweskin, S. J.; Rioux, R. M.; Habas, S. E.; Komvopoulos, K.; Yang, P.; Somorjai, G. A.
Journal of Physical Chemistry B 2006, 110, 15920.
(25) Borodko, Y.; Habas, S. E.; Koebel, M.; Yang, P. D.; Frei, H.; Somorjai, G. A. Journal of
(26) Rumpel, A.; Novak, M.; Walter, J.; Braunschweig, B. Langmuir 2011.
(27) Vig, J. R. 1986.
(28) Aliaga, C.; Park, J. Y.; Yamada, Y.; Lee, H. S.; Tsung, C. K.; Yang, P. D.; Somorjai, G.
A. Journal of Physical Chemistry C 2009, 113, 6150.
(29) Yang, M.; Chou, K. C.; Somorjai, G. A. The Journal of Physical Chemistry B 2004, 108,
14766.
(30) Kuhn, J. N.; Tsung, C.-K.; Huang, W.; Somorjai, G. A. Journal of Catalysis 2009,
265, 209.
115
Chapter 11
Conclusions and Future Work

12.1 Electronic Oxygen Activation by Charge Transfer from the Catalyst Support
This work implicates O as an electron acceptor that gives rise to charged reaction
intermediates at the metal–support interface. This appears to be true for CO oxidation and
methanol oxidation where electronic activation of surface O determines both the reaction rate
and the product selectivity, respectively. Chapters 3 and 4 show that O can be electronically
activated by an external bias applied across a metal–support interface in a solid-state device.
Chapters 5–7 demonstrate that an exactly analogous effect is possible by n-type doping the
support. The exact chemical nature of the electronically active O species is not yet clear. Initial
attempts have been made to characterize this species in situ using ambient pressure x-ray
photoelectron spectroscopy (XPS). However, it is possible that this charged species is short
lived owing to its high chemical activity, and it may not be possible to directly observe this state
without time resolved spectroscopy.
Chapter 3 shows that the history of a solid-state device in reaction conditions has an
important effect on the ability of external bias to electronically activate the reaction
intermediates. A correlation of the electronic activity of the catalytic device with the current–
voltage curves obtained during reaction suggests that the CO oxidation reaction operates in
different regimes depending on the nature of the catalyst and the O intermediate. On a CO
covered Pt surface, the reaction follows a Langmuir–Hinshelwood mechanism where activation
of the Pt–C bond in adsorbed CO for reaction with atomic O is rate limiting. In this case,
electronic activation of O has little effect on the reaction kinetics. However, the Pt catalyst
oxidizes during reaction, and that the Pt–O bond in the oxide is activated by electron density
from the support. Electronic activation of O atoms in Pt oxide can be understood by considering
the band structure of transition metal oxides where the valence band primarily represents
bonding states and the conduction band primarily represents anti-bonding states with respect to
the metal–O bond.1 Consequently, electron transfer to the conduction band of Pt oxide will
weaken the Pt–O bond resulting in a catalytic reaction involving an active O− ion from the lattice
in a Mars–van Krevelen type mechanism. Time-dependent kinetics are also observed in the case
of CO oxidation on Pt using F-doped TiO2 as an electronically active support (see Chapter 6, Fig.
4C). However, in this case in situ characterization of the catalyst is more challenging because
there is no circuit for probing the electronic properties of the catalyst during reaction. In both
cases electron spillover to the O occurs as a result of the electric field at the metal–support
interface. This field can be enhanced by an externally applied bias (Chapters 3 and 4) or by n-
type doping of the support to increase the intrinsic built-in potential (Chapters 5–7). In both
cases, the charge transfer results in enhanced catalytic activity for oxidative chemistry.
The methanol oxidation reaction discussed in Chapter 7 is unique in these studies because
it is a multipath reaction where electronic activation controls the catalyst selectivity rather than
activity. These results strongly suggest that two separate (rather than sequential) pathways exist
116
for formaldehyde and CO2 formation from methanol and that electronically activated O favors
the formaldehyde pathway. There are relatively few catalysts capable of the selective oxidation
of primary alcohols to aldehydes using molecular O2, and H2O2 or other stoichiometric O donors
are often used as the oxidant to increase reaction selectivity.2 Of course, molecular O2 is
preferred for cost and environmental concerns. However, it was recently demonstrated that
certain supported catalysts can produce formaldehydes from primary alcohols at high selectivity
using O2 as the oxidant.3 These catalysts are also shown to be highly selective for H2O2
synthesis from H2 and O2.4 In addition, it is known that halogen-induced electronic effects in
Ag/AgO catalysts dramatically increase the activity for epoxidation chemistry.5 Based on
correlation of these studies with the present work on electronic activation of O, it appears that
negatively charged O represents an important reaction intermediate that may be present on many
partial oxidation catalysts.
The concept of electronic activation by control of catalyst electronic structure or by

external control using a catalytic solid-state device is promising for a variety of important
chemical applications. Electrochemical water splitting is an important reaction for energy
conversion that may benefit from these findings. While H2 evolution occurs at high rates near
the thermodynamic efficiency limit, O2 evolution requires a significant overpotential.6,7
Presumably, a kinetic barrier to electron transfer limits the efficiency of this half reaction.
Already studies using F-doped TiO2 as a support for Pt catalysts show that the onset potential of
electrochemical CO, methanol, and water oxidation reactions can be modified by controlling the
electronic structure of the TiO2 support. In addition, the ability to reduce the overpotential using
an externally applied electric field could allow for fast reaction kinetics at near thermodynamic
efficiencies and would represent a significant advance toward efficient chemical energy
conversion. The concept of a catalytic field effect transistor is a promising idea for this
application, and a patent is pending on this process.8
12.2 Strong Metal–Support Interactions and Acid–Base Chemistry
This work used sum frequency generation (SFG) vibrational spectroscopy to probe the
molecular nature of strong metal–support interactions (SMSI). These results from Chapter 8
show that SMSI appears to be a special case of acid–base catalysis occurring at the oxide–metal
interface. Hydrogenation of furfuraldehyde to furfural alcohol requires selective C=O bond
activation that occurs by charge transfer between the furfuraldehyde molecule and O vacancies
in a TiO2 support. This gives rise to a furfuryl-oxy intermediate on the TiO2 support that is
observed by SFG. This active intermediate selectively hydrogenates to furfuryl alcohol by H
spillover from supported Pt, and the Pt/TiO2 catalyst is 90% selective for this reaction. The
active furfuryl-oxy intermediate does not form on a SiO2 support, and accordingly, kinetic
measurements show that the Pt/SiO2 catalyst does not produce any furfuryl alcohol. Recent
results show that Pt/TiO2 catalysts are also SMSI active for crotonaldehyde and acetaldehyde
hydrogenation reactions to produce crotyl alcohol and ethanol, respectively.
In addition to direct observation of the active furfuryl-oxy intermediate, H spillover is

also directly observed by SFG as an increase in the nonresonant signal intensity from a reducible
support. The increase in nonresonant signal intensity on a Pt/TiO2 catalyst is reversible in O2
and appears to represent an in situ marker for the SMSI state of a catalyst as originally predicted
by Tauster based on chemisorption measurements.9 Recent kinetic measurements for
117
furfuraldehdye hydrogenation show that CeO2 behaves similarly to TiO2, while Al2O3, ZrO2,
Nb2O5, and WO3 show no SMSI effect. Current work is in progress to correlate the SMSI
activity of various supports with the nonresonant SFG intensity under reaction conditions and
with the reducibility of the oxide in H2 as probed by ambient pressure XPS.
These studies demonstrate that SMSI is a special case of acid–base catalysis that strongly
affects C=O bond activation. An important application of C=O bond activation is Fischer–
Tropsch synthesis for the conversion of CO to higher molecular weight hydrocarbons. When
synthesis gas derived from steam reforming of methane is used as the feedstock for Fischer–
Tropsch synthesis, the net process represents conversion of natural gas to liquid fuels and this
process promises to have enormous economic and environmental impact in the current energy
market. An even more challenging chemical process is CO2 hydrogenation for renewable energy
conversion. It has been proposed that acid catalysis is capable of achieving this lofty goal, and in
light of this work, this idea is deserving of considerable attention.
12.3 The Role of an Organic Cap to Mediate Nanoparticle Catalysis
Inherent in colloidal nanoparticle synthesis is the presence of an organic capping agent to

stabilize the size and shape of the nanoparticles. Recent research shows that the size and shape
of catalytic nanoparticles provide control of reaction rate and selectivity for numerous catalytic
reactions.10-16 However, the effect of the organic cap on the catalytic behavior of the
nanoparticles has not been well established. The capping agent is usually considered to act as a
site blocking agent that lowers the metal surface area available for catalytic reaction. In this
light, some have assumed that colloidal nanoparticles are incompatible for commercial catalytic
applications because the presence of the organic cap decreases the apparent metal dispersion.
The studies discussed in Chapters 9 and 10 use SFG to probe the structure of
poly(vinylpyrrolidone) (PVP) on Pt nanoparticles under reaction conditions. The results show
that the capping agent is dynamic and disorders in H2. When the PVP is removed, carbonaceous
fragments continue to occupy the catalyst surface. This carbonaceous material is also dynamic
and appears to control the catalytic activity of the Pt nanoparticles across several orders of
magnitude. The carbonaceous fragments on the cleaned nanoparticles are porous in H2 but
collapse in O2 to form a tightly closed shell. Consequently, removing the organic cap increases
the activity for ethylene hydrogenation, but decreases the activity for methanol oxidation. This
result suggests that for oxidation reactions the organic cap is desirable because it acts as a
passivation layer against surface contamination but does not prevent catalysis. These findings
highlight the important role of the organic cap to mediate nanoparticle catalysis and demonstrate
one example where a capped nanoparticle is much more active than its cleaned analogue.
This work is reminiscent of homogenous catalysis where ligands are used to tune the
activity and selectivity of single metal ions.17 Recent work has also shown that active
involvement by a dendrimer ligand allows small Pt clusters to function as homogenous reaction
catalysts.15 These results suggest that in the future the catalytic activity and selectivity of metal
clusters and nanoparticles will be controlled by carefully selected capping agents in much the
same way as homogenous and biological catalysis achieves high selectivity in complex reaction
pathways using a framework of organic ligands. Other applications of this work include the
design of capping agents to selectively block undesired reaction pathways. For example, coke
118
deposition currently limits the active lifetime of many hydrocarbon catalysts. Perhaps an
appropriate capping agent would passivate the metal surface against coke deposition while still
allowing high activity toward hydrocarbon conversion.18
12.4 References
(1) Cox, P. A. Transition Metal Oxides An Introduction to their Electronic Structure and
Properties; Clarendon Press: Oxford, 1992.
(2) Sheldon, R. A.; Kochi, J. K. Metal-catalyzed oxidations of organic compounds:
mechanistic principles and synthetic methodology including biochemical processes; Academic
Press, 1981.
2003, 1184.
(6) Hall, D. E. Journal of The Electrochemical Society 1983, 130, 317.
(7) Lee, J.-W.; Popov, B. Journal of Solid State Electrochemistry 2007, 11, 1355.
7, 3097.
C 2009, 113, 8616.
Chem 2010, 2, 36.
(18) Lu, J.; Fu, B.; Kung, M. C.; Xiao, G.; Elam, J. W.; Kung, H. H.; Stair, P. C. Science
2012, 335, 1205.

1174166

Uploaded by

Copyright:

Available Formats

1174166

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1174166

Uploaded by

Copyright:

Available Formats

Charge Transfer and Catalysis at the Metal–Support Interface

A dissertation submitted in partial satisfaction of the

Charge Transfer and Catalysis at the Metal–Support Interface

Lawrence Robert Baker

Doctor of Philosophy in Chemistry

University of California, Berkeley

Professor Gabor A. Somorjai, Chair

Kinetic, electronic, and spectroscopic characterization of model Pt–support systems are

1.3.1 Definitions of SMSI

Like SMSI, acid–base catalysis is often characterized by the selective enhancement of a

Carbonium Ion Carbenium Ion

1.5.1 Experimental Methods

Chapter 8 describes a molecular level investigation of strong metal–support interactions

2.1.1 Thin Film Deposition and Processing

2.1.2.2 Langmuir Blodgett Deposition—Formation of 2-dimensional films of

2.2 Catalyst Characterization

2.2.1 Electrical Probe Measurements

Electrical probe measurements were used to characterize the performance of solid-state

Photovoltaic efficiency is determined by the IV curve of a photodiode during

Current Density (mA/cm2)

2.2.2 X-Ray Photoelectron Spectroscopy

X-ray photoelectron spectroscopy (XPS) provides elemental analysis of surfaces. XPS

The monochromatic x-ray source is usually generated from an Al or Mg anode. A high

A hemispherical analyzer is used to analyze the kinetic energy of photoelectrons ejected

To Pump To Pump To Pump

2.2.3 Electron Microscopy

2.2.4 Sum Frequency Generation Vibrational Spectroscopy

Sum frequency generation (SFG) vibrational spectroscopy is a nonlinear optical

Substituting Eq. 5 into Eq. 6 shows that in centrosymmetric media

For SFG experiments, an active/passive mode-locked Nd:YAG laser (Leopard D-20,

2.3 Catalytic Reaction Studies

Solid-State Charge-Based Device for Control of

Using a Pt/Si catalytic nanodiode, we externally control the rate of CO oxidation on a Pt

Many industrial catalysts consist of metal nanoparticles dispersed on a porous support.

3.4 Results and Discussion

3.4.1 Electrical and Photo Properties of Catalytic Diode

3.4.2 Bias Effect on Catalytic Activity

We attempted to perform similar experiments to observe how the magnitude of applied

3.4.3 Light Effect on Catalytic Activity

The photochemistry experiment was performed in 8 Torr CO and 20 Torr O2 at 423 K.

The mechanism of photocurrent flux is generation of an electron-hole pair in the Si

3.4.4 Device Stability and Hysteresis

This work demonstrates direct control of catalytic CO oxidation using a catalytic

(1) Chen, M. S.; Goodman, D. W. Science 2004, 306, 252.

Controlling the Rate of Hydrogen Oxidation on

We report here that turnover frequency of hydrogen oxidation on a Pt/TiO2 nanodiode

The nanodiode is placed on a ceramic heater, and the temperature is monitored by a

4.4 Results and Discussion

Generation of Highly n-Type Titanium Oxide Using

The application of transition metal oxide (TMO) semiconductors for charge-based

5.4 Results and Discussion

Increase in number of Ti 3d 1.58 0.89 0.00

Increase in number of unpaired 0.00 0.00 0.93

F population — 9.55 9.62

Change in O2p-Ti3d excitation −0.3 +0.8 +0.2

(1) Wager, J. F. Science 2003, 300, 1245.

Highly n-Type Titanium Oxide as an Electronically

Although fundamentally insightful, this type of experiment employing a femtosecond

6.3.1 Catalyst Preparation

6.3.2 Catalyst Characterization