Electrooxidation of small organic molecules on single crystaland bi-metallic electrodesCitation for published version (APA):Housmans, T. H. M. (2005). Electrooxidation of small organic molecules on single crystal and bi-metallicelectrodes. Technische Universiteit Eindhoven. https://doi.org/10.6100/IR598438

DOI:10.6100/IR598438

Document status and date:Published: 01/01/2005

Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne

Take down policyIf you believe that this document breaches copyright please contact us at:openaccess@tue.nlproviding details and we will investigate your claim.

Download date: 30. Oct. 2020

Electrooxidation of small organic moleculeson single crystal and bi-metallic electrodes

T.H.M. Housmans

Electrooxidation of small organic molecules on single crystal and bi-metallic electrodes

Proefschrift

ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de

Rector Magnificus, prof.dr.ir. C.J. van Duijn, voor een commissie aangewezen door het College voor

Promoties in het openbaar te verdedigen op dinsdag 6 december 2005 om 16.00 uur

door

Thomas Hubertus Maria Housmans

geboren te Sittard

Dit proefschrift is goedgekeurd door de promotoren: prof.dr. R.A. van Santen en prof.dr. M.T.M. Koper Printed by the Eindhoven University of Technology Press. CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EINDHOVEN Housmans, Thomas H.M. Electrooxidation of small organic molecules on single crystal and bi- metallic electrodes / by Thomas H.M. Housmans. – Eindhoven : Technische Universiteit Eindhoven, 2005. Proefschrift. – ISBN 90-386-2927-3 NUR 913 Trefwoorden: elektrochemie / elektrochemische oxidatie / adsorptie / methanol / koolmonoxide / gestapte elektroden / rhodium / platina / reactiekinetiek Subject headings: electrochemistry / electrochemical oxidation / adsorption / methanol / carbon monoxide / stepped electrodes / rhodium / platinum / reaction kinetics

To Verena

Table of contents

Chapter 1. Introduction 1 Chapter 2. CO oxidation on stepped Rh[n(111)×(111)] single crystal 21 electrodes: a voltammetric study Chapter 3. CO oxidation on stepped Rh[n(111)×(111)] single crystal 39 electrodes: a chronoamperometric study Chapter 4. CO oxidation on stepped Rh[n(111)×(111)] single crystal 61 electrodes: anion effects on CO surface mobility Chapter 5. CO oxidation on stepped single crystal electrodes: a Monte 75 Carlo study Chapter 6. CO oxidation on Pt modified Rh(111) electrodes 97 Chapter 7. Methanol oxidation on stepped Pt[n(111)×(111)] single 115 crystal electrodes: a chronoamperometric study Chapter 8. Structure sensitivity of methanol electrooxidation pathways 137 on platinum: an On-Line Electrochemical Mass Spectrometry study Chapter 9. The electrooxidation of small organic molecules on platinum 159 nanoparticles supported on gold: particle size vs. particle shape effect Summary 179 Samenvatting 181 List of Publications 183 Curriculum Vitae 184 Acknowledgements 185

Introduction

1.1. History of Fuel Cell Systems

1.1.1. Historic Trends

The history of fuel cells is certainly a subject in its own right, and is an integral part of the history of the science and technology of energy and energy production devices.[1] In the scientific community, Sir William Grove is considered to be the father of the fuel cell. In 1839, he reported in the Philosophical Magazine, that the electrolysis of water could be reversed by passing hydrogen and oxygen gas over platinum electrodes in a dilute solution of sulfuric acid, consequently generating electricity and water.[2] The term “Fuel cell” was first used by Sir Humphry Davy in 1802 [3], but it was not used again until 1889, when Ludwig Mond and Charles Langer, attempting to build the first practical device using air and industrial coal gas, coined the term again.[4]

Between Grove’s invention in the early 19th century and the present day the interest in the field of fuel cell catalysis has waxed and waned many times. A detailed overview of the historic trends in fuel cell systems can be found in refs. [5] and [6]. However, we will only describe the main research activities here.

After many years of research, the Second World War essentially stopped all fuel cell related research. The internal combustion engine (ICE) became the most important power source for transportation devices, followed by the jet engine for airplanes, and gas turbine engines in power plants. Only years after the war did fuel cell research become interesting again. The emergence of space flight demanded a reliable, non-combustion based power source, which was found in the form of an alkaline fuel cell (AFC). The AFC consisted of porous nickel anodes and lithiated porous nickel oxide cathodes separated by a circulating 30% aqueous potassium hydroxide solution. It is essentially this fuel cell, which allowed men to fly to the moon. Unfortunately, as alkaline electrolytes do not reject CO2 due to the formation of carbonate crystals (CO2 + 2OH- → CO3

2- + H2O), the AFC technique is restricted to specialized applications where pure H2

Chapter 1

and O2 are used. Nevertheless, despite these problems the renewed research efforts to create electrical automobiles propelled by fuel cells lead to the development of more hydrogen-air fuel cell prototypes.

In the mid-1970s, research interests shifted from the alkaline fuel cells, having reached an advance stage of development, to the phosphoric acid systems, which seemed better suited as stationary power sources. In parallel, the increasing popularity and availability of fossil fuels steered research efforts in the direction of the development of reformers for the on-board production of hydrogen gas. The desire to develop a reformer was enhanced by the difficulties encountered with the storage of hydrogen and the necessity of high purity gases in original hydrogen-air systems. However, reformer based fuel cell systems, which rely on the in-situ production of hydrogen gas from solid or liquid C-H fuels, are complicated and suffer a considerable weight penalty. Moreover, the issue of gas-feed purity remained. These difficulties inherent to hydrogen-air systems powered by either onboard hydrogen storage, or in-situ hydrogen production, lead to the rise of another fuel cell type: the Direct Alcohol Fuel Cell (DAFC), with methanol as best candidate for the alcohol. The benefits of such a fuel cell, which directly oxidizes methanol at low temperatures, were (and in fact, still are) alluring. The theoretically high power/weight ratio and relatively low operation temperature make it a (potentially) more versatile power source than other fuel cells. It can, in principle, be used as power source for small mobile appliances, such as laptops, mobile phones, and other devices, but also in vehicles or as stationary power supply.

The direct electrochemical oxidation of methanol was first investigated by E. Müller in 1920, but the actual pioneering work was done in the in the fifties and early sixties.[7-10] In the following years, the complex methanol oxidation reaction was researched intensively (for a review see ref. [11]). However, after nearly two decades of research it was concluded that commercialization of the direct methanol fuel cell would be extremely difficult and it was considered unlikely that oil prices would show any substantial rise over the foreseeable future.[12] Consequently, in the 80ies and 90ies fuel cell research was redirected to development of high temperature systems such as the molten carbonate fuel cell (MCFC) and solid oxide fuel cell (SOFC), which show an overall higher efficiency and can be operated utilizing different fuel types. Although several versions working of these fuel cells exist, life expectancy of a high temperature fuel cell is still a problem, which needs to be solved.

One of the last major changes in the direction of fuel cell research occurred in the 1990s and was inspired by developments in the field of membrane technologies. The new membranes greatly improved the properties of so called polymer electrolyte fuel cells (PEFC), which were deemed to be unreliable for space applications back in the sixties. Although high power densities and longer life expectancies were obtained, commercialization was still hindered by the high costs of this type of fuel cell.

2

Introduction

1.1.2. Current Research

It seems that in the last couple of years another shift in research efforts can be detected. Table 1.1. presents a list of the fuel cell types mentioned in the previous section, together with the temperature range in which they operate, electrolyte composition and the major areas of application. (Table 1.1 is based on Table 4-1. in ref. [5]) Although research on most of these fuel cell types still continues, membrane-based fuel cell systems seem to appear as most attractive systems for study.

Fuel cell type Temp

range / ºC Fuel type Electrolyte Application

area (potential) Alkaline Fuel Cell (AFC)

60-90 H2-O2 35-50% KOH Space

Transportation Phosphoric Acid Fuel Cell (PAFC)

160-220 H2-O2 Concentrated

phosphoric acid Stationary

Solid Oxide Fuel Cell (SOFC) 800-1000 H2-O2

Yttrum-stabilized Zirkondioxide (ZrO2/Y2O3)

Stationary

Molten Carbonate Fuel Cell (MCFC)

620-660 H2-O2 Molten carbonate (Li2CO3/Na2CO3)

Stationary

Polymer Electrolyte Fuel Cell (PEFC) 50-80 H2-O2

Polymer membrane

(Nafion/Dow)

Space Transportation

Direct Alcohol Fuel Cell (DAFC)

60-90 Alcohols

Acidic media Alkaline media

Polymer membranes

Portable Transportation

Stationary

Table 1.1. Different fuel cell types listed together with their respective temperature range, fuels,electrolyte composition, and application area.

Although emissions from internal combustion engines (ITC) have become considerably “cleaner” in the past decades due (mostly) to the introduction of the three-way automobile catalyst, further optimalization of the engine, and adjustments and purification to fossil fuels, there is a rising concern about environmental pollution and the effects of global warming. This environmental awareness together with rapidly increasing oil prices and depleting fossil fuel reserves greatly increases the pressure to replace the ITC with a more efficient, reliable fuel cell based power source.

As was already mentioned briefly in the previous section, the use of H2-O2 (air) systems as replacement for combustion engines has an intrinsic disadvantage, namely the transportation of hydrogen gas. In the gas phase or stored in the absorbed form in metal alloys, a severe weight penalty is suffered. Onboard production of hydrogen from solid or liquid C-H fuels is a technically more complex process and the overall efficiency

0.50 V

3

Chapter 1

decreases due to the fact that water and oxygen are required to ensure that the process is exothermic. Moreover, the produced hydrogen gas inevitably contains CO, which poisons the catalyst, thus requiring the use of less active, more CO-tolerant anodes.[12] For these reasons, theoretically the most promising candidate for replacement of the internal combustion engine seems to be the Direct Alcohol Fuel Cell.

Unfortunately, the power output of existing (prototype) DAFCs is (still) insufficient to be used as replacement for the combustion engine. Therefore, a more immediate field of application is as power source of small portable appliances, such as laptops, mobile phones, PDAs, etc.

1.2. Low Temperature Direct Methanol Fuel Cells

Direct oxidation of methanol at the anode of a DAFC has distinct advantages over other alcohols.[11-18] Firstly, methanol is a liquid at room temperature and can, therefore, easily be introduced in the already existing fuel distribution system. Secondly, it can be produced in large quantities and it has an acceptable toxicity. Moreover, using a liquid fuel eliminates the need for complex fuel vaporizes or reformers and the associated heat sources and controls.[5] However, from the fuel cell technology point of view a more important consideration for choosing methanol is the fact that it can be catalytically oxidized on platinum electrodes in aqueous environment yielding CO2 and six electrons per methanol molecule:

−+ ++→+ eHCOOHOHCH 66223 (1.1)

This reaction is usually complimented by the reduction of oxygen from the air at the cathode. Reaction 1.1 has a very promising thermodynamic potential of 0.029 V vs. the saturated hydrogen electrode (SHE) [18] and may, theoretically, allow for a power nearly as high as that of a hydrogen-based fuel cell.[19] Unfortunately, the disadvantages of using methanol as fuel are considerable: - The decomposition reaction of methanol on platinum produces surface poisoning

species, which leads to a low catalytic activity and presents a severe inhibition for the development of a low temperature fuel cell.

- As carbonate formation is a serious problem in alkaline media, acid electrolytes must be used, which result in corrosion problems. More importantly, acid electrolytes are responsible for the slow electrode kinetics of the reduction of oxygen at the air cathode.[5]

- The anode reaction is sluggish near the thermodynamic potential, which results in a considerable loss in over-potential and, therefore also in the efficiency of the system.[11]

- High noble metal catalyst loadings are necessary in order to obtain a sufficiently high power output, which makes the fuel cell costly.

- Crossover of methanol through the membrane of PEMFCs from anode to cathode results in a considerable decrease in the efficiency.

4

Introduction

Despite these difficulties the tantalizing benefits of a working low temperature methanol-based fuel cell over the internal combustion engine has ensured continued interest over the past decades. In order to solve the problems mentioned above, knowledge of the mechanism and kinetics of the reactions are required. Improvements in the following areas would be highly advantageous: - The anode activity must be further improved - Current loadings of noble metals on the anode need to be reduced - Membrane properties for PEMFCs need to be improved

In the following sections we will summarize concepts and results published in the literature on the adsorption and electrooxidation of methanol and detected intermediate species, such as carbon monoxide, formaldehyde and formic acid, on platinum and bi-metallic surfaces.

1.3. Electrooxidation of Methanol

1.3.1. Electrooxidation of Methanol on Bulk Platinum Electrodes

As platinum is a good dehydrogenation catalyst, it is the catalyst of choice when investigating the electrooxidation reaction of methanol. Already in early studies researchers found that in addition to the complete oxidation product CO2, methanol reacting on platinum surfaces also yields the partial oxidation product carbon monoxide, which acts as a surface poison.[11, 20-31] The self-poisoning of the methanol oxidation reaction (MOR) by strongly adsorbed CO is one of the main problems preventing the development of a low temperature direct methanol fuel cell. Additionally, the detection of CO raised a question: Is CO formed in a parallel reaction mechanism, or is it a necessary intermediate in a serial reaction? The answer to this question has a profound influence on the course of DMFC related research. If a serial pathway proves true, future research will be focused intensively towards developing more CO-tolerant catalysts, while in case a parallel pathway mechanism applies, research can also be directed towards designing catalysts, which selectively oxidize methanol in a pathway not involving the formation of adsorbed CO.

Initially, the fact that CO was found as partial oxidation product and because a Differential Electrochemical Mass Spectrometry (DEMS) analysis on polycrystalline Pt showed no CO2 formation at low potentials, a serial pathway was suggested.[31-35] However, evidence in favor of a parallel pathway involving non-CO species in the methanol oxidation scheme is considerable.[11, 36-47] The existence of a parallel mechanism was already suggested by Breiter as early as 1967.[48] Although disputed by Vielstich et al.,[33] based on an analysis of transients obtained on Pt(111), Pt(110) and Pt(100) in 0.2 M MeOH and both perchloric and sulfuric acid, Herrero et al. concluded that an alternative pathway leading directly to CO2 must be active.[36, 37] A similar conclusion was drawn from other detailed coulometric analyses.[39, 41-43, 49] From time-dependent methanol reaction data and complementary CO oxidation experiments at

5

Chapter 1

H 2CO sol + HCOOH sol

diff

CH 3 OH sol CH 2 OH ads

CO ads

CO 2

H 2CO ads + HCOOH ads

(1)

(3)

(2)

(5)

(4)

(6)

Figure 1.1. Schematic representation of the parallel pathway for methanol oxidation on platinumelectrodes.

0.6 V vs. RHE, both Lu et al. and Sriramulu et al. concluded that a serial reaction path involving adsorbed residue is inadequate to explain the observed rate of CO2 production.[41-43] Moreover, the detection of appreciable amounts of formic acid (which can react with methanol to form methylformate) and formaldehyde by various techniques, like DEMS and fluorometric techniques, is also in favor of a parallel pathway mechanism rather than a serial one.[38, 40, 44-46, 50-52]

In an effort to elucidate the nature of the methanol oxidation intermediates and to explain the apparent contradiction between observations made by the groups of Wieckowski and Vielstich, Baltruschat et al. proposed a parallel pathway mechanism in which CO2 could be formed through oxidation of adsorbed CO and/or the oxidation of dissolvable intermediate species like formic acid and formaldehyde.[44, 45] The resulting parallel pathway mechanism is depicted schematically in Fig. 1.1. The mechanism explains the observation of a higher oxidation current than necessary for formation of COads, the appearance of significant amounts of CO2 only at potentials higher than the CO oxidation potential and it incorporates the possibility and identity of intermediate species detected during the incomplete oxidation reaction. Species like methylformate and 1,1-dimethoxymethane, which were found in small amounts by several groups,[16, 44] are not included in this scheme. In Fig. 1.1 the pathway forming CO2 through oxidation of adsorbed CO (reac. (1) and (2)) is commonly referred to as the indirect oxidation pathway, while oxidation through soluble intermediates such as formic acid and formaldehyde is called the direct oxidation pathway (reac. (3) and (5)).

By comparing voltammograms obtained on the three basal planes of platinum, Clavilier and co-workers [53, 54] demonstrated that the MOR is strongly structure sensitive. These experiments were later augmented by chronoamperometric, spectroscopic and kinetic isotope studies, which all indicated a high structure sensitivity and point to crystalline defects as most active centers for the reaction.[36, 38, 55, 56] Of the three platinum basal planes, Pt(111) was found to be the least reactive towards the decomposition of methanol, while (110) was reported to be the most active.[36] As a result, the surface poisoning process is slowest on (111) and fastest on (110). However,

6

Introduction

the Pt(111) electrode was found to give only a small initial current that decays to an even lower level with time, while the Pt(110) surface deactivates much faster, but remains the most efficient in oxidizing methanol.[36] The first dehydrogenation step of methanol to CH2OHads, involving cleavage of the C-H bond, was found to be the rate determining step on Pt(111) and Pt(110),[36, 57] while the Tafel slope recorded for Pt(100) suggests the second dehydrogenation step from CH2OHads to CHOHads as rate determining.[36]

Interestingly, under Ultra High Vacuum (UHV) conditions dissociative adsorption of methanol through O-H bond scission is known to occur readily.[13, 58-61] Franaszczuk et al. explained the differences between the electrochemical and UHV data by assuming that the OH group is solvated in aqueous media and, thus, hindered from approaching the surface.[57] Furthermore, Davis and Barteau noted that breaking of the C-H bond, being 393 kJ⋅mol-1 in energy, should be favored over breaking of the O-H bond, which has an energy of 435 kJ⋅mol-1.[58] Together, these observations can be used to explain the formation of large amounts of soluble intermediates detected during the electrooxidation of methanol on many different platinum surfaces.[40, 52]

In a study on the effect of specific anion adsorption on the selectivity of the direct oxidation pathway measured for Pt(111), Batista et al. concluded that a lack of multiple coordination sites for methanol adsorption due to strong adsorption of (bi)sulfate leads to the formation of soluble intermediates rather than adsorbed CO.[62, 63] They proposed that when multiple sites are available for the adsorption of methanol, i.e. in the absence of a strongly adsorbing anion, C-H bond scission is preferred. This principle is commonly known as the “ensemble site effect” and has been reported previously in the literature.[64, 65] In case only single adsorption sites or insufficiently large ensemble sites are available, as would occur more frequently in the presence of a strongly adsorbing anion, dissociative adsorption of methanol is assumed to occur preferably through O-H bond scission. Batista et al. suggested that C-H bond scission leads to COads formation, while O-H bond scission leads to the formation of soluble intermediates.

The nature of the intermediates between the CH2OHads species and COads (1) has been extensively addressed by both in-situ and ex-situ methodologies.[11, 32, 35, 66-70] Based on a DEMS study where measurements of the charge passed during adsorption of methanol and the charge for the subsequent oxidation of the adsorbate were compared, Iwasita et al. suggested an intermediate with H:C:O stoichiometry. Interestingly, they also reported that the adlayer formed upon dissociative adsorption of methanol contained only a minority of CO. Thus, it was proposed that this H:C:O species may also act as surfaces poison (see ref. [11] and references therein) The presence of hydrogen in the “non-CO” intermediate was demonstrated by Willsau et al.[32] Studies on elucidating the structure of the H:C:O species proved inconclusive and both HCO [32, 35] and COH [66, 67] were found as possible candidates. Other in-situ techniques, like infrared spectroscopy,[71] and ex-situ techniques, like ECTDMS (ElectroChemical Thermal Desorption Mass Spectrometry),[68-70] likewise failed to reveal conclusive evidence as to which intermediate is formed. However, ultimately COads is formed, which may be oxidized at sufficiently high potentials to form CO2.

7

Chapter 1

1.3.2. Electrooxidation of Methanol on Platinum Nanoparticles

From a technological point of view, the oxidation of methanol on platinum nano-scale particles is far more important than on bulk platinum. Proper use of nano-structured electrodes can greatly reduce the catalyst loading needed to obtain a desired power output. Interestingly, compared to bulk electrode materials, nano-scale particles display different, often unexpected, catalytic properties.

For the MOR a pronounced “particle size effect” was reported on Pt particles ranging between 1.4 and 5 nm in diameter, meaning that below a particular particle size the reactivity of the particles decreases considerably. Much controversy exists over the origin and characteristics of this “size effect”. McNicol and co-workers suggested a maximum size effect for Pt clusters of ca. 3 nm in diameter,[72, 73] while Kennedy et al. and Park et al. found an optimal cluster diameter of 2 and 4 nm, respectively [65, 74] Three possible explanations for the decrease in activity with the decreasing particle size are: 1) the coverage of OHads increases as the particles get smaller, thus, blocking empty surface sites, 2) smaller particles have fewer preferential adsorption sites (i.e. “ensemble sites”) for dissociative adsorption of methanol, and 3) the self-poisoning of the MOR on small particles is faster due to slower oxidation of COads (i.e. stronger CO bond).[75] However, at present it is unclear which of these explanations is correct or which effect plays a dominant role in the “particle size effect”. The discrepancy in the optimal particle size was tentatively explained by assuming that the particle morphology is more important than the actual size.[12, 74, 76, 77] Therefore, it may be more correct to speak of “particle shape” rather than “particle size” effect.

1.3.3. Electrooxidation of Methanol on Bi- and Tri-Metallic Surfaces

Platinum itself is not sufficiently active to be useful in commercial fuel cells and considerable efforts have been undertaken to find more active materials. This search has been directed mostly by the results of the mechanistic analyses given in the previous section. It was quickly realized that a catalyst must be both capable of chemisorbing methanol but also oxidizing the resultant chemisorbed fragments. This may be achieved using the so called “bifunctional mechanism”, originally proposed by Watanabe and Motoo.[78] The bi-functional mechanism is based on the idea that sites on the more oxophilic metal act as adsorption centers for oxygen-containing species (generally accepted to be OHads), which can react with CO adsorbed on platinum to form CO2. The reactivity of the surface may also be influenced by affecting the electronic structure. The presence of a second metal can induce a change in the CO (and OH) binding strength on Pt, thus facilitating surface poison oxidation. This effect is often referred to as the “electronic” or “ligand” effect. A third way for improving catalyst performance is based on an “ensemble effect”. Here, adding a catalytically inert material to the active compound is assumed to change the distribution of active sites, thereby opening different

8

Introduction

reaction pathways.[15] Various ways, based on one or more of the explanations given above, have been proposed to enhance the activity of platinum electrodes:[12] 1. The activity and oxophilicity of platinum can be enhanced by generating more

reticulate (rough) surfaces on which Pt-O species form more readily. Incorporating metals like Sn and Cr in a platinum alloy and then dissolving these metals, leads to reticulate surfaces. However, this type of activity enhancement does not seem useful in a fuel cell, as long-term stability of the particles is low.

2. Deposition of ad-atoms on a platinum surface can enhance the activity of the substrate by blocking hydrogen adsorption, altering the electronic properties of the surface (“electronic effect”), acting as redox centers, blocking the adsorption of surface poisons and inducing the formation of oxygen-containing species.[78-85] Again, the stability of the surfaces generated is a major issue when considering their usefulness to fuel cell applications.

3. By alloying platinum with metals, which form surface oxides in the potential range for methanol oxidation, the oxidation of surface poisoning species adsorbed on the platinum is facilitated by the oxide formation on the second metal. Although Pt-Sn, Pt-Ir, and Pt-Os alloys show a catalytic improvement with respect to pure platinum,[86-88] Pt-Ru alloys show by far the best catalytic properties.[45, 88-92]

4. The combination of platinum with a base metal was also reported to promote the MOR, with Nb, Zr and Ta as most active promoters.[93]

5. The opposite of the previously stated methods is also possible. By incorporating platinum directly in an oxide surface, highly active platinum particles can be obtained. Surfaces consisting of SrRu0.5Pt0.5O3 and DyxPt3O4 were found to produce reasonably high current densities.[94, 95] At present, the stability of these surfaces in a working fuel cell with an acidic electrolyte is not well documented.

6. Simply replacing platinum with another noble metal can also increase the activity of the anode. However, despite the fact that some materials, such as Ir, WC and NiZr,[96-99] are capable of oxidizing methanol, the activity remains far below that of platinum.

Despite some promising results, these electrodes still do not have the desired catalytic effect, or still require relatively high Pt loadings. Moreover, electrode stability is a serious problem in many cases. Therefore ongoing fuel cell research is directed towards finding more active bi- or even tri-metallic surfaces, which are capable of generating a high current density and which are stable at the operating conditions of a fuel cell. So far, the Pt-Ru electrode is the only electrode that offers a real catalytic advantage. As the enhancement effect is based primarily on facilitating the CO electrooxidation reaction, we will discuss this mechanism in more detail in Section 1.5.3.

1.4. Electrooxidation of Formic acid and Formaldehyde

Although formic acid and formaldehyde play an important part in the methanol oxidation scheme their oxidation mechanisms on platinum have received less attention

9

Chapter 1

than the oxidation of CO and methanol.[15, 25, 27, 29-31, 54, 65, 90, 100-132] Both reactions were found to depend strongly on the structure of the electrode surface.[25, 29, 110, 131, 132] The oxidation rate of formic acid on the basal planes of platinum was reported to be highest for Pt(111) and lowest for Pt(100).[110] Introducing steps (defects) in a (111) plane results in a decrease in the activity,[15] which distinguishes the formic acid oxidation from the electrooxidation reactions of methanol and CO. The reaction is generally also believed to follow a parallel pathway mechanism.[35, 112-114, 126, 127, 129, 130] Under electrochemical conditions, the C-H bond scission upon formic acid adsorption was found to be the rate determining step on both Pt(111) and Pt(100).[131] Although recent results published by Miki et al. indicate that formate is formed during the electrooxidation of formic acid on Pt nano-particles in perchloric acid media, suggesting O-H bond cleavage.[133] Analogous to methanol, in vacuum O-H cleavage was found as the first step to form adsorbed formate (HCOOads).[134]

Infrared experiments point to CO as the main poisoning species formed during the reaction.[25, 133] However, a non-CO intermediate was also suggested, which again may be in the form of an H:C:O species,[70] or as was suggested by Lamy and Leger, in the form of COOH or HCOO.[132] The Surface Enhanced Infra Red Adsorption (SEIRA) data reported by Miki et al. indicating the presence of formate adsorbed on the surface corroborate this suggestion.[107, 133, 135] It is assumed that the HCO or COH species are responsible for the formation of COads, while formate can be oxidized directly to CO2 in a parallel reaction, or can desorb again as formic acid. Interestingly, dissociation of formic acid leads to a higher coverage of poisoning species compared to dissociative adsorption of methanol,[136] which suggests that formic acid require smaller “ensemble sites” to adsorb. In fact, dissociative adsorption of formic acid demonstrates a higher activity towards CO formation than for methanol.[70, 136]

In contrast to methanol, the oxidation rate of formic acid on small particles was found to be considerably higher than for larger particles.[65] Furthermore, the onset of the reaction was found to be shifted to lower potentials for smaller particles. The difference between the electrooxidation pathway of formic acid as compared to formaldehyde (discussed below) and methanol may be ascribed to the fact that for formic acid oxidation no addition of oxygen is required to produce CO2.[128] Therefore, the reaction rate is not inhibited by stronger OHads bonding on the smaller particles.[137, 138] Moreover, sequential C-dehydrogenation for dissociative adsorption of methanol and formaldehyde requires multiple contiguous surface sites, while this requirement is not necessary for formic acid.

In aqueous solutions formaldehyde is easily hydrolyzed to methylene glycol (H2C(OH)2),[139, 140] which can dissociate on the platinum surface to COads,[105, 141] but can also react directly to CO2.[106] Although linearly bonded CO was reported as primary poisoning species, other intermediates such as C-bonded formyl (CHO) and formate (HCOOads) were also reported.[107, 108] Compared to methanol and formic acid, formaldehyde reacts somewhat differently, as the Pt(100) surface was reported to be more active than Pt(110) and Pt(111).[106, 108] Additionally, the rate of formaldehyde

10

Introduction

oxidation is relatively insensitivity to the size of Pt nano-particles, even though an ensemble site is required for dissociative adsorption to COads and the number of these sites decreases with the particle size.[65]

1.5. Electrooxidation of Carbon Monoxide

1.5.1. Electrooxidation of CO on Bulk Platinum Electrodes

The oxidation of carbon monoxide on noble metal surfaces is probably one of the most extensively studied reactions in heterogeneous catalysis.[15, 142-147] Several reasons have ensured its continued interest over the past decades. As was mentioned earlier, it is a common catalyst poison and, accordingly, hinders the development of low temperature fuel cells.[12, 15, 17, 148, 149] The previous sections demonstrated that CO is formed in nearly all platinum catalyzed oxidation reactions of C1 organic molecules. Moreover, the reaction is of fundamental and practical interest, as CO can be used as a neutral probe to surface activity and structure sensitivity [149-152] and it is a toxic product produced by incomplete combustion of fossil fuels in internal combustion engines.[153, 154] Due to the general interest in this reaction, numerous reviews have been published.[11, 15, 132, 144, 155] In the following text we will briefly outline the mechanism and kinetics of the CO electrooxidation reaction on noble metal surfaces, specifically bulk platinum and platinum-based bi-metallic surfaces.

The CO electrooxidation mechanism in acidic media is assumed to be a Langmuir-Hinshelwood type reaction between adsorbed CO and a surface bound oxygen-containing species.[142] The oxygen-containing species is assumed to be in the form of OHads, generated by the oxidation of water at the electrode surface. The overall reaction mechanism is:

H2O + * OHads + H+ + e- (1.2)

COads + OHads CO2 + H+ + 2e- + 2* (1.3) with * denoting a free adsorption site.

Despite this deceptively simple reaction scheme, oxidative stripping of a CO adlayer on polycrystalline and single crystal platinum electrodes using linear sweep voltammetry exhibits multiple current peaks, attesting to the complex nature of the reaction.[20] The onset potential and rate of the CO adlayer oxidation depend critically on the CO overage, pH of the electrolyte, the nature of the anions in the electrolyte, and the structure of the surfaces. Increasing the CO coverage results in an increase in the onset potential. On Pt(111) this increase can become quite substantial, where the onset potential of the reaction in perchloric acid increases from ca. 0.6 V at low coverages to 0.9 V vs. RHE at coverages close to saturation.[144] As can be expected of a Langmuir-Hinshelwood type reaction where CO has a higher affinity for the free surface sites than OH, continuous readsorption of CO results in a positive shift of the oxidation potential. Like most surface confined reactions, the electrooxidation of CO was shown to be a structure sensitive process.[15, 144] On low-index Pt single crystal surfaces the activity

11

Chapter 1

increases in the order of Pt(111)<Pt(100)<Pt(110).[15, 144] Moreover, steps and crystalline defects in general were found to catalyze the oxidation reaction on Pt [15, 156-159] and Rh [152].

The presence of strongly adsorbing anions has a negative influence on the rate of the reaction, since they inhibit the formation of OHads.[15] Alkaline media, on the other hand, greatly facilitate the CO oxidation reaction, as OH adsorption on defects occurs more readily than in acidic media.[15] However, as was mentioned in Section 1.3., carbonate formation is a serious problem for the oxidation of methanol in alkaline electrolytes. For this reason most research in the field of electrooxidation of “methanolic” CO is carried out in acidic media.

Recently, Lebedeva et al. preformed a systematic study of the role of crystalline defects in the electrocatalytic oxidation of CO on platinum surfaces of [n(111)×(111)] orientation.[160-162] The obtained results indicate that CO preferably adsorbs on the steps, blocking the electrochemical hydrogen adsorption on these sites, which is consistent with results obtained from other electrochemical experiments,[163-166] UHV experiments,[167-171] and quantum chemical calculations.[150] Moreover, as the over-potential for oxidative stripping of saturated CO adlayers as well as submonolayers increases in the order Pt(553)<Pt(554)<Pt(111), they concluded that the actual oxidation reaction occurs preferably on the steps. The higher activity of steps and defects towards the oxidation of CO was explained based on a higher concentration of oxygen-containing species at a given potential on the stepped surfaces and a lower CO packing density.[160] The effect of the OHads coverage on the onset and rate of the reaction was considered more important than the reduced CO packing density due to increasing step density.

A detailed chronoamperometric analysis provided a resolution to an ongoing discussion about which analytical expression of the overall reaction can best describe the kinetics of the CO electrooxidation reaction, the “nucleation and growth” model or the “mean-field approximation”.[162, 172] The mean-field approximation assumes that the reactants are perfectly mixed on the surface, i.e. the mobility of the adsorbed species is very high and the reaction rate is proportional to the average coverage of the reactants. If the nucleation and growth model applies, reactants are assumed to be immobile on the surface. The reaction nucleates by adsorption of OH at ”special” sites, usually defects or steps, and proceeds only at the interface between two reacting phases, causing the formation and growth of islands.[173] Some authors found that the nucleation and growth model provides the best fit of their experimental data,[143, 156] while others favor the mean-field approximation.[159] Based on the fact that the shape of the current-time transients is not affected by the step density, Lebedeva et al. concluded that the mobility of CO on platinum must be high. They also showed that the reaction kinetics of the main CO oxidation peak on clean well-defined platinum surfaces are best described by the mean-field approximation.[162, 172] Additionally, it was pointed out that cleanliness of the system is of critical importance to the accurate interpretation of the results, as the presence of contamination can alter the shape of the current-time transients.

12

Introduction

Finally, based on extensive IRRAS results, Lebedeva et al. proposed a model for the electrooxidation of CO on stepped platinum single crystal electrodes with (110) and (100) oriented steps.[161] CO adsorbed on the (111) terraces was found to be more reactive compared to CO adsorbed on (110) and (100) steps. As water decomposition was proposed to prefer the trough of the step,[160] these sites are concluded to be the active centers for the oxidation of COads.

1.5.2. Electrooxidation of CO on Platinum Nanoparticles

The effect of particle size on the CO electrooxidation reaction has been studied for a number of different systems.[77, 174-181] For gold supported Pt nano-particles, obtained from a colloidal Pt solution, Friedrich et al. showed that the oxidation overpotential increases from 100 to 500 mV on particles of ca. 3 nm in diameter with respect to polycrystalline platinum surfaces.[77, 177] Agglomerates of these particles with a diameter of ca. 10-16 nm also exhibit a positive potential shift, although markedly smaller. Maillard et al. and Cherstiouk et al.[178-180] made similar observations for the stripping of CO adlayers on platinum nano-particles supported on glassy carbon. They found that as the particle size is decreased below 3 nm, the overpotential for CO oxidation shifts to considerably higher potentials. Restricted mobility of COads on smaller particles was suggested to be responsible for this phenomenon.

The origin of this particle size effect was suggested to lay in the geometrical structure of the particles and to a lesser extent the electronic properties. Arenz et al. attributed the difference in CO oxidation potential to the number of defects present on the surface of a larger particle.[76] As defects are able to dissociate water to form OHads, more defects on the larger particles resulting in a shift of the CO oxidation peak to lower potentials. NMR studies revealed that decreasing the size of a well-defined Pt particle can enhance the Fermi level density of states and accordingly alter the CO electrooxidation properties.[182, 183] Infrared experiments showed a pronounced redshift in the stretching frequencies of atop-bound CO adlayers for particles between 2 and 4 nm,[184, 185] which was explained by stronger adsorption of CO on the relatively more numerous edge sites on the well-defined nanoparticles as compared to bulk electrodes and larger particles.

1.5.3. Electrooxidation of CO on Bi- and Tri-metallic Surfaces

Enhancing the catalytic activity of a fuel cell anode in essence depends on the ability of the electrode to oxidize surface poisoning species, while maintaining appreciably high rates for the oxidation of the used fuel. This knowledge has lead to a large body of literature concerning the CO electrooxidation reaction on bi- and even tri-metallic surfaces. The main difference between this reaction and the MOR on bi-metallic surfaces lies in the fact that for the oxidation of methanol both the direct as well as the indirect pathway may be influenced by the structure of the surface. Therefore, it is

13

Chapter 1

important to also know the individual effects of the surface structure on the oxidation of CO as well as methanol oxidation reaction.

Most fundamental studies involve bi-metallic surfaces where Pt is either alloyed or modified with Ru, Re, Sn, Pd, Mo, and Rh,[15, 90, 137, 147, 155, 186-195] and are based on the bi-functional effect. Information on the “electronic and ensemble effects” is far less common. Detailed Density Functional Theory (DFT) calculations show that mixing of Pt with Ru, Rh, Ir, Re (typically those metals located to the upper left of Pt in the Periodic Table) results in a considerable decrease in the Pt-CO binding strength due to a shift in the d-band (the so-called “d-band model”),[196-199] signifying that the electronic effect also plays an important role in increasing catalyst activity. Proof of the ensemble effect can be found in the high activity of well-defined Pt-Ru electrodes towards the oxidation of methanol as well as CO.[147, 200] Although Ru itself is inactive for the methanol oxidation at room temperature, low Ru coverages on Pt lead to facile chemisorption of methanol, while the electrooxidation of CO remains rate limiting, due to slow migration of CO to the Ru clusters. The activity maximizes at a 50-50 Pt-Ru ratio. Interestingly, Gasteiger et al. demonstrated the oxidation activity of methanol on Pt-Ru electrodes requires a higher amount of Platinum, attesting to the “site ensemble effect” reported for this reaction.[200]

DFT results indicate that, in terms of the electronic effect, Pt-overlayer systems may have the best CO-tolerant properties (i.e. the lowest CO binding energy).[15, 137, 191, 194, 198] These strong electronic effects are caused by the fact that the CO binding energy in the overlayer system is determined primarily by the Pt-Pt distance in the overlayer, which is dictated by the underlying substrate. Contraction of the Pt overlayer with respect to pure Pt results in a lower CO bond strength (i.e. PtML on Ru, Rh, Ir, Re), while expansion results in higher binding energies (i.e. PtML on Au). Pt submonolayers on Ru(0001) and Ru(101 0), as well as on Ru nano-particles, were investigated recently by Adžić et al., and indeed showed good CO tolerance for hydrogen oxidation.[201-204]

1.6. Outline of the Thesis

The research presented in this thesis is focused towards elucidating the mechanisms and kinetics of the electrooxidation of small organic molecules (i.e. carbon monoxide, methanol, and formic acid). A qualitative correlation between the mechanism, the reaction kinetics and surface mobility of adsorbed species and the structure of the catalytically active noble metal surfaces was sought. To this end, the reactions were investigated on well-defined platinum and rhodium single crystal electrodes, platinum nano-particles and bi-metallic platinum-based surfaces. Using electrochemical techniques like cyclic voltammetry and chronoamperometry, single crystal electrodes (basal planes and [n(111)×(111)] type electrodes) provide clean high quality surfaces of known orientation, which can act as model surfaces of practical catalysts, while nano-scale particles more closely resemble industrial and commercially available catalysts. Bi-metallic surfaces are used to investigate the effect of metal adlayers on the

14

Introduction

electrooxidative properties of the substrate, with the goal of designing catalysts with better performance and tolerance to poison. Next to characterization by voltammetric techniques, the surfaces were also investigated using Scanning Tunneling Microscopy (STM) and Atomic Force Microscopy (AFM). Surface Enhanced Raman Spectroscopy (SERS) and On-Line Electrochemical Mass Spectrometry (OLEMS) provided insight into the nature of the surface adsorbed species and the production of soluble intermediates.

In Chapter 2 and 3 we report on the electrocatalytic properties of CO adsorbed on stepped rhodium electrodes of [n(111)×(111)] orientation. Voltammetric and chronoamperometric data show that the electrooxidation reaction of CO on these surfaces is strongly structure sensitive. A low CO surface mobility is suggested and addressed extensively.

The effect of anion adsorption on the CO-rhodium system is discussed in Chapter 4. It is suggested that reducing the anion adsorption strength increases the surface mobility of CO and, thus, results in a considerable change in the kinetics and dynamics of the reaction.

From the results resented in Chapters 2-4 a model for the electrooxidation of CO on stepped Rh surfaces emerged, which we tested in Chapter 5 by dynamic Monte Carlo simulations. By altering the CO surface diffusion rate, the MC model can predict the voltammetric and chronoamperometric profiles obtained in sulfuric (Chapter 2 and 3) and perchloric acid (Chapter 4) and simulate results previously obtained in our lab on stepped platinum electrodes.

In Chapter 6 the electrooxidation properties of CO, methanol and formic acid on a platinum modified Rh(111) electrode are investigated using cyclic voltammetry and chronoamperometry. The electrocatalytic properties of the Pt-Rh(111) surfaces are correlated to the surface structure and extensively discussed with respect to single crystal rhodium surfaces.

The methanol oxidation reaction on platinum single crystal electrodes is studied in more detail in Chapter 7 and 8. Results of a model applied to the chronoamperometric transients obtained on Pt[n(111)×(111)] stepped surfaces indicate that the decomposition reaction of methanol preferably takes place on the steps and crystalline defects. Additionally, the formation of soluble intermediates like formic acid and formaldehyde is studied using On-Line Electrochemical Mass Spectrometry measurements, the results of which also indicate the unique role of steps in the formation of these intermediates. Comparison of the methanol oxidation properties of Pt basal planes with stepped surfaces provides new information on the nature of the “ensemble site” required for the decomposition of methanol to carbon monoxide and the nature of the adsorption site leading to the formation of soluble intermediates.

Finally, a study on the reactivity of electrodeposited platinum nanoparticles on gold towards the electrooxidation of small organic molecules is presented in Chapter 9. A pronounced particle size effect was observed for the CO oxidation reaction and explained in terms of the morphology of the nano-particles, as was the decreasing

15

Chapter 1

activity of the methanol oxidation reaction and increasing formic acid oxidation rate for decreasing particle size.

References

[1] G. Sandstede, E. J. Cairns, V. S. Bagotzky, and K. Wiesener, Handbook of Fuel Cells, Chapter 12, John Wiley & Sons Ltd., West Sussex, 2003.

[2] W. R. Grove, Philos. Mag. Ser. 3 (1839) 127. [3] H. Davy, J. Nat. Phil. 144 (1802) [4] L. Mond and C. Langer, Proc. Roy Soc. 46 (1889) [5] K. Kordesch and C. Simander, Fuel Cells and Their Applications, VCH Verlagsgescellschaft

mbH, Weinheim, New York, Basel, Cambridge, Tokye, 1996. [6] Editors:, W. Vielstich, A. Lamm, and H. A. Gasteiger, Handbook of Fuel Cells, John Wily &

Sons, West Sussex, 2003. [7] K. R. Williams, M. R. Andrew, and F. Jones, Hydrocarbon Fuel Cell Technol., Symp., Am.

Chem. Soc., Atlantic City (1965) 143. [8] K. Kordesch and A. Marko, Oesterr. Chemicker. Ztg. 52 (1961) 125. [9] W. Vielstich, Hydrocarbon Fuel Cell Technol., Symp., Am. Chem. Soc., Atlantic City (1965) 79. [10] C. W. Foust and W. J. Sweeney, Chem. Technol. (Will Mitchell, Jr., editor, Academic) 1 (1963)

371. [11] R. Parsons and T. VanderNoot, J. Electroanal. Chem. 257 (1988) 9. [12] A. Hamnett, Comp. Chem. Kin. (1999) 635. [13] N. Kizhakevariam and E. M. Stuve, Surf. Sci. 286 (1993) 246. [14] G. Hopranyi and A. Wieckowski, Proceedings - Electrochemical Society 92 (1992) 70. [15] N. M. Markovic and P. N. Ross, Surf. Sci. Rep. 45 (2002) 117. [16] T. Iwasita, Electrochim. Acta. 47 (2002) 3663. [17] S. Wasmus and A. Kuver, J. Electroanal. Chem. 461 (1999) 14. [18] R. Metkemeijer and P. Achard, International Journal of Hydrogen Energy 19 (1994) 535. [19] Y. Zhu, H. Uchida, T. Yajima, and M. Watanabe, Langmuir 17 (2001) 146. [20] B. Beden, A. Bewick, and C. Lamy, J. Electroanal. Chem. Inter. Electrochem. 148 (1983) 147. [21] B. Beden, S. Juanto, J. M. Leger, and C. Lamy, J. Electroanal. Chem. Inter. Electrochem. 238

(1987) 323. [22] B. Beden, F. Hahn, S. Juanto, C. Lamy, and J. M. Leger, J. Electroanal. Chem. Inter. Electrochem.

225 (1987) 215. [23] B. Beden, F. Hahn, J. M. Leger, C. Lamy, and M. I. d. S. Lopes, J. Electroanal. Chem. Inter.

Electrochem. 258 (1989) 463. [24] B. Beden, F. Hahn, C. Lamy, J. M. Leger, N. R. De Tacconi, R. O. Lezna, and A. J. Arvia, J.

Electroanal. Chem. Inter. Electrochem. 261 (1989) 401. [25] S. C. Chang, L. W. H. Leung, and M. J. Weaver, J. Phys. Chem. 94 (1990) 6013. [26] S. C. Chang, Y. Ho, and M. J. Weaver, Surf. Sci. 265 (1992) 81. [27] K. Kunimatsu, J. Electroanal. Chem. Inter. Electrochem. 213 (1986) 149. [28] S. Motoo and N. Furuya, J. Electroanal. Chem. Inter. Electrochem. 197 (1986) 209. [29] J. Clavilier and S. G. Sun, J. Electroanal. Chem. Inter. Electrochem. 199 (1986) 471. [30] R. R. Adzic, A. V. Tripkovic, and N. M. Markovic, J. Electroanal. Chem. Inter. Electrochem. 150

(1983) 79. [31] T. Iwasita, F. C. Nart, B. Lopez, and W. Vielstich, Electrochim. Acta. 37 (1992) 2361. [32] J. Willsau and J. Heitbaum, J. Electroanal. Chem. Inter. Electrochem. 185 (1985) 181. [33] W. Vielstich and X. H. Xia, J. Phys. Chem. 99 (1995) 10421. [34] M. Krausa and W. Vielstich, J. Electroanal. Chem. 379 (1994) 307. [35] J. Willsau and J. Heitbaum, Electrochim. Acta. 31 (1986) 943. [36] E. Herrero, K. Franaszczuk, and A. Wieckowski, J. Phys. Chem. B 98 (1994) 5074. [37] E. Herrero, W. Chrzanowski, and A. Wieckowski, J. Phys. Chem. 99 (1995) 10423. [38] X. H. Xia, T. Iwasita, F. Ge, and W. Vielstich, Electrochim. Acta 41 (1996) 711. [39] T. D. Jarvi, S. Sriramulu, and E. M. Stuve, J. Phys. Chem. B 101 (1997) 3649. [40] C. Korzeniewski and C. L. Childers, J. Phys. Chem. B 102 (1998) 489. [41] S. Sriramulu, T. D. Jarvi, and E. M. Stuve, Electrochim. Acta. 44 (1998) 1127. [42] S. Sriramulu, T. D. Jarvi, and E. M. Stuve, J. Electroanal. Chem. 467 (1999) 132. [43] G. Q. Lu, W. Chrzanowski, and A. Wieckowski, J. Phys. Chem. B 104 (2000) 5566.

16

Introduction

[44] H. Wang, T. Loffler, and H. Baltruschat, J. Appl. Electrochem. 31 (2001) 759. [45] H. Wang, C. Wingender, H. Baltruschat, M. Lopez, and M. T. Reetz, J. Electroanal. Chem. 509

(2001) 163. [46] Z. Jusys and R. J. Behm, J. Phys. Chem. B 105 (2001) 10874. [47] Z. Jusys, J. Kaiser, and R. J. Behm, Langmuir 19 (2003) 6759. [48] M. W. Breiter, Electrochim. Acta. 12 (1967) 1213. [49] T. D. Jarvi, S. Sriramulu, and E. M. Stuve, Coll Surf. A 134 (1998) 145. [50] K.-I. Ota, Y. Nakagawa, and M. Takahashi, J. Electroanal. Chem. 179 (1984) 179. [51] T. Iwasita and W. Vielstich, J. Electroanal. Chem. 201 (1986) 403. [52] C. L. Childers, H. Huang, and C. Korzeniewski, Langmuir 15 (1999) 786. [53] J. Clavilier, R. Durand, G. Guinet, and R. Faure, J. Electroanal. Chem. 127 (1981) 281. [54] C. Lamy, J. M. Leger, J. Clavilier, and R. Parsons, J. Electroanal. Chem. 150 (1983) 71. [55] P. A. Christensen, A. Hamnett, J. Munk, and G. L. Troughton, J. Electroanal. Chem. 370 (1994)

251. [56] B. I. Podlovchenko and T. D. Gladysheva, Russian Journal of Electrochemistry (Translation of

Elektrokhimiya) 38 (2002) 560. [57] K. Franaszczuk, E. Herrero, P. Zelenay, A. Wieckowski, J. Wang, and R. I. Masel, J. Phys. Chem.

B 96 (1992) 8509. [58] J. L. Davis and M. A. Barteau, Surf. Sci. 187 (1987) 387. [59] J. Wang and R. I. Masel, Surf. Sci. 243 (1991) 199. [60] J. Wang and R. I. Masel, J. Vac. Sci. Technol. A 9 (1991) 1879. [61] B. A. Sexton, Surf. Sci. 102 (1981) 271. [62] E. A. Batista, G. R. P. Malpass, A. J. Motheo, and T. Iwasita, J. Electroanal. Chem. 571 (2004)

273. [63] E. A. Batista, G. R. P. Malpass, A. J. Motheo, and T. Iwasita, Electrochem. Comm. 5 (2003) 843. [64] V. S. Bagostki, Y. B. Vassilev, and O. A. Khazova, J. Electroanal. Chem. Inter. Electrochem. 81

(1977) 229. [65] S. Park, Y. Xie, and M. J. Weaver, Langmuir 18 (2002) 5792. [66] T. Iwasita, W. Vielstich, and E. Santos, J. Electroanal. Chem. 229 (1987) 367. [67] J. Willsau, O. Wolter, and J. Heitbaum, J. Electroanal. Chem. Inter. Electrochem. 185 (1985) 163. [68] S. Wilhelm, T. Iwasita, and W. Vielstich, J. Electroanal. Chem. Inter. Electrochem. 238 (1987)

383. [69] W. Vielstich, P. A. Christensen, S. A. Weeks, and A. Hamnett, J. Electroanal. Chem. Inter.

Electrochem. 242 (1988) 327. [70] S. Wilhelm, W. Vielstich, H. W. Buschmann, and T. Iwasita, J. Electroanal. Chem. Inter.

Electrochem. 229 (1987) 377. [71] M. I. S. Lopes, B. Beden, F. Hahn, J. M. Leger, and C. Lamy, J. Electroanal. Chem. Inter.

Electrochem. 313 (1991) 323. [72] P. A. Attwood, B. D. McNicol, R. T. Short, and J. A. Van Amstel, J. Chem. Soc. Farad. Trans. I

76 (1980) 2310. [73] B. D. McNicol, P. Attwood, and R. T. Short, J. Chem. Soc. Farad. Trans. I 77 (1981) 2017. [74] B. J. Kennedy and A. Hamnett, J. Electroanal. Chem. Inter. Electrochem. 283 (1990) 271. [75] T. Frelink, W. Visscher, and J. A. R. van Veen, J.Electroanal.Chem. 382 (1995) 65. [76] M. Arenz, K. J. J. Mayrhofer, V. Stamenkovic, B. B. Blizanac, T. Tomoyuki, P. N. Ross, and N.

M. Markovic, J. Am. Chem. Soc. (2005) submitted. [77] K. A. Friedrich, F. Henglein, U. Stimming, and W. Unkauf, Electrochim. Acta. 45 (2000) 3283. [78] M. Watanabe and S. Motoo, J. Electroanal. Chem. Inter. Electrochem. 60 (1975) 275. [79] G. Kokkinidis, J. Electroanal. Chem. Inter. Electrochem. 201 (1986) 217. [80] M. Watanabe and S. Motoo, J. Electroanal. Chem. Inter. Electrochem. 60 (1975) 259. [81] M. Watanabe and S. Motoo, J. Electroanal. Chem. Inter. Electrochem. 60 (1975) 267. [82] M. Watanabe, Y. Furuuchi, and S. Motoo, J. Electroanal. Chem. Inter. Electrochem. 191 (1985)

367. [83] S. A. Campbell and R. Parsons, J. Chem. Soc. Farad. Trans. I 88 (1992) 833. [84] E. Herrero, A. Fernandez-Vega, J. M. Feliu, and A. Aldaz, J. Electroanal. Chem. 350 (1993) 73. [85] E. Herrero, K. Franaszczuk, and A. Wieckowski, J. Electroanal. Chem. 361 (1993) 269. [86] M. M. P. Janssen and J. Moolhuysen, Electrochim. Acta. 21 (1976) 869. [87] B. Beden, F. Kadirgan, C. Lamy, and J. M. Leger, J. Electroanal. Chem. Inter. Electrochem. 127

(1981) 75. [88] A. Hamnett and B. J. Kennedy, Electrochim. Acta. 33 (1988) 1613.

17

Chapter 1

[89] A. V. Tripkovic, K. D. Popovic, B. N. Grgur, B. Blizanac, P. N. Ross, and N. M. Markovic, Electrochim. Acta. 47 (2002) 3707.

[90] N. M. Markovic, H. A. Gasteiger, P. N. Ross, Jr., X. Jiang, I. Villegas, and M. J. Weaver, Electrochim. Acta. 40 (1995) 91.

[91] H. Hoster, T. Iwasita, H. Baumgartner, and W. Vielstich, Phys. Chem. Chem. Phys. 3 (2001) 337. [92] T. Iwasita, H. Hoster, A. John-Anacker, W. F. Lin, and W. Vielstich, Langmuir 16 (2000) 522. [93] A. Hamnett, B. J. Kennedy, and S. A. Weeks, J. Electroanal. Chem. Inter. Electrochem. 240

(1988) 349. [94] J. H. White and A. F. Sammells, J. Electrochem. Soc. 140 (1993) 2167. [95] K. Machida, M. Enyo, G. Adachi, and J. Shiokawa, Bull. Chem. Soc. Jap. 60 (1987) 411. [96] R. Y. Shaidullin, A. D. Semenova, G. D. Vovchenko, and Y. B. Vasil'ev, Zh. Fiz. Khim. 57

(1983) 1019. [97] H. Okamoto, G. Kawamura, A. Ishikawa, and T. Kudo, J. Electrochem. Soc. 134 (1987) 1645. [98] S. Y. Vasina, S. A. Stuken, O. A. Petrii, I. L. Gogichadze, and V. A. Mukhin, Elektrokhimiya 23

(1987) 1127. [99] C. C. Hays, R. Manoharan, and J. B. Goodenough, J. Power Sources 45 (1993) 291. [100] S. Motoo and N. Furuya, Ber. Bunsen-Ges. Phys. Chem. 91 (1987) 457. [101] K. Yahikozawa, Y. Yamashita, Y. Matsuda, and Y. Takasu, Chem. Express 7 (1992) 597. [102] J. Stelmach, R. Holze, and M. Beltowska-Brzezinska, J. Electroanal. Chem. 377 (1994) 241. [103] P. Olivi, L. O. S. Bulhoes, J. M. Leger, F. Hahn, B. Beden, and C. Lamy, J. Electroanal. Chem.

370 (1994) 241. [104] J. Wang, M. A. DeAngelis, D. Zaikos, M. Setiadi, and R. I. Masel, Surf. Sci. 318 (1994) 307. [105] S.-G. Sun, G.-Q. Lu, and Z.-W. Tian, J. Electroanal. Chem. 393 (1995) 97. [106] P. Olivi, L. O. S. Bulhoes, J. M. Leger, F. Hahn, B. Beden, and C. Lamy, Electrochim. Acta. 41

(1996) 927. [107] A. Miki, S. Ye, T. Senzaki, and M. Osawa, J. Electroanal. Chem. 563 (2004) 23. [108] C.-F. Mai, C.-H. Shue, Y.-C. Yang, L.-Y. O. Yang, S.-L. Yau, and K. Itaya, Langmuir 21 (2005)

4964. [109] M. Baldauf and D. M. Kolb, J. Phys. Chem. 100 (1996) 11375. [110] R. R. Adzic, W. E. O'Grady, and S. Srinivasan, Surf. Sci. 94 (1980) L191. [111] P. Cao, Q. Zhong, Y. Sun, and R. Gu, Chem. Phys. Lett. 376 (2003) 806. [112] A. Capon and R. Parsons, J. Electroanal. Chem. Inter. Electrochem. 44 (1973) 239. [113] A. Capon and R. Parsons, J. Electroanal. Chem. Inter. Electrochem. 65 (1975) 285. [114] A. Capon and R. Parsons, J. Electroanal. Chem. Inter. Electrochem. 45 (1973) 205. [115] E. Herrero, M. J. Llorca, J. M. Feliu, and A. Aldaz, J. Electroanal. Chem. 394 (1995) 161. [116] H. Kita and H.-W. Lei, J. Electroanal. Chem. 388 (1995) 167. [117] M. J. Llorca, J. M. Feliu, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 376 (1994) 151. [118] G.-Q. Lu, A. Crown, and A. Wieckowski, J. Phys. Chem. B 103 (1999) 9700. [119] M. D. Macia, E. Herrero, and J. M. Feliu, Electrochim. Acta. 47 (2002) 3653. [120] S. Motoo and N. Furuya, J. Electroanal. Chem. Inter. Electrochem. 184 (1985) 303. [121] C. Rice, S. Ha, R. I. Masel, P. Waszczuk, A. Wieckowski, and T. Barnard, J. Power Sources 111

(2002) 83. [122] C. Rice, S. Ha, R. I. Masel, and A. Wieckowski, J. Power Sources 115 (2003) 229. [123] S. P. E. Smith, E. Casado-Rivera, and H. D. Abruna, J. Solid State Electrochem. 7 (2003) 582. [124] G. Stalnionis, L. Tamasauskaite-Tamasiunaite, V. Pautieniene, and Z. Jusys, J. Solid State

Electrochem. 8 (2004) 900. [125] S.-G. Sun, Y. Lin, N.-H. Li, and J.-Q. Mu, J. Electroanal. Chem. 370 (1994) 273. [126] A. Wieckowski and J. Sobkowski, J. Electroanal. Chem. Inter. Electrochem. 63 (1975) 365. [127] O. Wolter, J. Willsau, and J. Heitbaum, J. Electrochem. Soc. 132 (1985) 1635. [128] T. D. Jarvi and E. M. Stuve, Electrocatalysis: Frontiers in Electrochemistry, Chapter 3, Wiley,

New York, 1998. [129] X. H. Xia and T. Iwasita, J. Electrochem. Soc. 140 (1993) 2559. [130] S. G. Sun, J. Clavilier, and A. Bewick, J. Electroanal. Chem. Inter. Electrochem. 240 (1988) 147. [131] A. Tripkovic, K. Popovic, and R. R. Adzic, J. Chim. Phys. Phys. Chim. Bio. 88 (1991) 1635. [132] C. Lamy and J. M. Leger, J. Chim. Phys. Phys. Chim. Bio. 88 (1991) 1649. [133] A. Miki, S. Ye, and M. Osawa, Chem. Commun. (2002) 1500. [134] M. R. Columbia and P. A. Thiel, J. Electroanal. Chem. 369 (1994) 1. [135] Y. X. Chen, A. Miki, S. Ye, H. Sakai, and M. Osawa, J. Am. Chem. Soc. 125 (2003) 3680. [136] S. G. Sun and J. Clavilier, J. Electroanal. Chem. Inter. Electrochem. 236 (1987) 95.

18

Introduction

[137] A. Wieckowski, Interfacial Electrochemistry: Theory, Experiment, and Applications, Marcel Dekker, New York, 2000.

[138] S. Mukerjee and J. McBreen, J. Electroanal. Chem. 448 (1998) 163. [139] J. P. Guthrie, Can. J. Chem. 53 (1975) 898. [140] J. G. M. Winkelman, O. K. Voorwinde, M. Ottens, A. A. C. M. Beenackers, and L. P. B. M.

Janssen, Chem. Eng. Sci. 57 (2002) 4067. [141] K. Nishimura, R. Ohnishi, K. Kunimatsu, and M. Enyo, J. Electroanal. Chem. Inter. Electrochem.

258 (1989) 219. [142] S. Gilman, J. Phys. Chem. 68 (1964) 70. [143] C. McCallum and D. Pletcher, J. Electroanal. Chem. 70 (1976) 277. [144] B. Beden, C. Lamy, N. R. De Tacconi, and A. J. Arvia, Electrochim. Acta. 35 (1990) 691. [145] J. M. Feliu, J. M. Orts, A. Fernandez-Vega, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 296

(1990) 191. [146] M. J. Weaver, S. C. Chang, L. W. H. Leung, X. Jiang, M. Rubel, M. Szklarczyk, D. Zurawski, and

A. Wieckowski, J. Electroanal. Chem. 327 (1992) 247. [147] H. A. Gasteiger, N. Markovic, P. N. Ross, Jr., and E. J. Cairns, J. Phys. Chem. 98 (1994) 617. [148] L. Carrette, K. A. Friedrich, and U. Stimming, Fuel Cells 1 (2001) 5. [149] A. Hamnett, Catal. Today. 38 (1997) 445. [150] B. Hammer, O. H. Nielsen, and J. K. Noerskov, Catal. Lett. 46 (1997) 31. [151] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, Electrochem.Commun. 2

(2000) 487. [152] R. Gómez, J. M. Orts, J. M. Feliu, J. Clavilier, and L. H. Klein, J. Electroanal. Chem. 432 (1997)

1. [153] E. L. Holt, in Brit., (Exxon Research and Engineering Co., USA). Gb, 1975, p. 9 pp. [154] J. T. Kummer, J. Phys. Chem. 90 (1986) 4747. [155] N. M. Markovic and P. N. Ross, Electrochim. Acta. 45 (2000) 4101. [156] B. Love and J. Lipkowski, ACS Symposium Series 378 (1988) 484. [157] E. Santos, E. P. M. Leiva, and W. Vielstich, Electrochim. Acta. 36 (1991) 555. [158] A. V. Petukhov, W. Akemann, K. A. Friedrich, and U. Stimming, Surf. Sci. 402-404 (1998) 182. [159] M. Bergelin, E. Herrero, J. M. Feliu, and M. Wasberg, J. Electroanal. Chem. 467 (1999) 74. [160] N. P. Lebedeva, M. T. M. Koper, E. Herrero, J. M. Feliu, and R. A. van Santen, J. Electroanal.

Chem. 487 (2000) 37. [161] N. P. Lebedeva, A. Rodes, J. M. Feliu, M. T. M. Koper, and R. A. van Santen, J. Phys. Chem. B

106 (2002) 9863. [162] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Phys. Chem. 106 (2002)

12938. [163] C. S. Kim and C. Korzeniewski, Anal. Chem. 69 (1997) 2349. [164] C. S. Kim, W. J. Tornquist, and C. Korzeniewski, J. Phys. Chem. 97 (1993) 6484. [165] C. S. Kim, W. J. Tornquist, and C. Korzeniewski, J. Chem. Phys. 101 (1994) 9113. [166] H. Wang, R. G. Tobin, and D. K. Lambert, J. Chem. Phys. 101 (1994) 4277. [167] J. Xu and J. T. Yates, Jr., J. Chem. Phys. 99 (1993) 725. [168] E. Hahn, A. Fricke, H. Roeder, and K. Kern, Surf. Sci. 297 (1993) 19. [169] B. E. Hayden, K. Kretzschmar, A. M. Bradshaw, and R. G. Greenler, Surf. Sci. 149 (1985) 394. [170] J. S. Luo, R. G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio, Surf. Sci. 274 (1992)

53. [171] M. A. Henderson, A. Szabo, and J. T. Yates, Jr., J. Chem. Phys. 91 (1989) 7255. [172] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Electroanal. Chem. 524-

525 (2002) 242. [173] W. Schmickler, Interfacial Electrochemistry, Oxford University press, Oxford, 1996. [174] Y. Takasu, N. Ohashi, X. G. Zhang, Y. Murakami, H. Minagawa, S. Sato, and K. Yahikozawa,

Electrochim. Acta. 41 (1996) 2595. [175] M. L. Sattler and P. N. Ross, Ultramicroscopy 20 (1986) 21. [176] L. J. Bregoli, Electrochim. Acta. 23 (1978) 489. [177] K. A. Friedrich, F. Henglein, U. Stimming, and W. Unkauf, Colloids and Surfaces, A:

Physicochemical and Engineering Aspects 134 (1998) 193. [178] F. Maillard, M. Eikerling, O. V. Cherstiouk, S. Schreier, E. Savinova, and U. Stimming, Faraday

Discuss. 125 (2003) 357. [179] O. V. Cherstiouk, P. A. Simonov, V. I. Zaikovskii, and E. R. Savinova, J. Electroanal. Chem. 554-

555 (2003) 241. [180] O. V. Cherstiouk, P. A. Simonov, and E. R. Savinova, Electrochim. Acta. 48 (2003) 3851.

19

Chapter 1

20

[181] A. Wieckowski, E. R. Savinova, and C. G. Vayenas, Catalysis and Electrocatalysis at Nanoparticle Surfaces, Marcel Dekker inc., New York, 2003.

[182] Y. Tong, C. Rice, A. Wieckowski, and E. Oldfield, J. Am. Chem. Soc. 122 (2000) 11921. [183] Y. Tong, C. Rice, A. Wieckowski, and E. Oldfield, J. Am. Chem. Soc. 122 (2000) 1123. [184] S. Park, S. A. Wasileski, and M. J. Weaver, J. Phys. Chem. B 105 (2001) 9719. [185] S. Park, Y. Tong, A. Wieckowski, and M. J. Weaver, Electrochem. Comm. 3 (2001) 509. [186] H. A. Gasteiger, N. M. Markovic, and P. N. Ross, Jr., J. Phys. Chem. 99 (1995) 8290. [187] H. A. Gasteiger, N. M. Markovic, and P. N. Ross, Jr., J. Phys. Chem. 99 (1995) 16757. [188] B. N. Grgur, N. M. Markovic, and P. N. Ross, Jr., J. Phys. Chem. B 102 (1998) 2494. [189] B. N. Grgur, N. M. Markovic, and P. N. Ross, Electrochim. Acta. 43 (1998) 3631. [190] R. Ianniello, V. M. Schmidt, U. Stimming, J. Stumper, and A. Wallu, Electrochim. Acta. 39

(1994) 1863. [191] W. Chrzanowski, H. Kim, and A. Wieckowski, Catal. Lett. 50 (1998) 69. [192] J. C. Davies, B. E. Hayden, and D. J. Pegg, Electrochim. Acta 44 (1998) 1181. [193] A. Kabbabi, R. Faure, R. Durand, B. Beden, F. Hahn, J. M. Leger, and C. Lamy, J. Electroanal.

Chem. 444 (1998) 41. [194] W. F. Lin, M. S. Zei, M. Eiswirth, G. Ertl, T. Iwasita, and W. Vielstich, J. Phys. Chem. B 103

(1999) 6968. [195] R. Liu, H. Iddir, Q. Fan, G. Hou, A. Bo, K. L. Ley, E. S. Smotkin, Y. E. Sung, H. Kim, S.

Thomas, and A. Wieckowski, J. Phys. Chem. B 104 (2000) 3518. [196] E. Christoffersen, P. Liu, A. Ruban, H. L. Skriver, and J. K. Norskov, J. Catal. 199 (2001) 123. [197] M. T. M. Koper, T. E. Shubina, and R. A. van Santen, J. Phys. Chem. B 106 (2002) 686. [198] T. E. Shubina and M. T. M. Koper, Electrochim. Acta. 47 (2002) 3621. [199] P. Liu, A. Logadottir, and J. K. Norskov, Electrochim. Acta. 48 (2003) 3731. [200] H. A. Gasteiger, N. Markovic, P. N. Ross, Jr., and E. J. Cairns, J. Phys. Chem. 97 (1993) 12020. [201] S. R. Brankovic, J. X. Wang, and R. R. Adzic, Electrochem. Solid State Lett. 4 (2001) A217. [202] S. R. Brankovic, N. S. Marinkovic, J. X. Wang, and R. R. Adzic, J. Electroanal. Chem. 532

(2002) 57. [203] K. Sasaki, J. X. Wang, M. Balasubramanian, J. McBreen, F. Uribe, and R. R. Adzic, Electrochim.

Acta. 49 (2004) 3873. [204] K. Sasaki, Y. Mo, J. X. Wang, M. Balasubramanian, F. Uribe, J. McBreen, and R. R. Adzic,

Electrochim. Acta. 48 (2003) 3841.

CO oxidation on stepped Rh[n(111)×(111)] single crystal electrodes: a voltammetric study

Abstract

The CO electrooxidation reaction was studied on Rh[n(111)×(111)]-type electrodes in 0.5 M H2SO4 using cyclic voltammetry. Examination of the blank cyclic voltammograms of the electrodes under investigation [Rh(111), Rh(554), Rh(553) and Rh(331)] shows that a distinction can be made between the electrodes on the basis of their voltammetric profile. An increase in reversibility of the peaks in the hydrogen/(bi)sulfate region and an increase in the potential at which surface oxidation occurs, has been observed for decreasing step density. Higher reversibility of the peaks in the low-potential region therefore indicates more step and/or defects on the surface. Cycling the potential to within the oxidation region leads to disordering of the surface, which, unlike reported previously, can be identified in the blank cyclic voltammogram. Calculations on the hydrogen/(bi)sulfate adsorption/desorption charge indicate that (bi)sulfate preferably adsorbs on terrace sites rather than the (110) steps. This idea is supported by the fact that disordering of the surface leads to lower charges in the hydrogen/(bi)sulfate region. Both CO stripping as well as bulk CO oxidation experiments show a pronounced effect of the surface structure on the reaction rate. In general, the activity of the electrodes increases for increasing step density. Also disordering of the surface increases the activity, due to an increasing number of defects. For CO stripping experiments on well-ordered electrodes a pre-shoulder is observed prior to the main oxidation peak, which can be ascribed to CO oxidation next to or at the steps. The presence of the pre-shoulder as well as pronounced tailing of the main oxidation peak suggests that diffusion of CO on rhodium may be slow.

This chapter is published as T.H.M. Housmans, J.M. Feliu, M.T.M. Koper, J. Electroanal.Chem. 572 (2004) 79

Chapter 2

2.1. Introduction

The oxidation of carbon monoxide on noble metal surfaces is probably one of the most intensively studied reactions in heterogeneous catalysis.[1-7] This reaction is both of fundamental as well as practical interest because CO can be used as a neutral probe to surface activity and structure sensitivity, it is a toxic product produced by internal combustion engines, and it is a common poison in the catalytic electrooxidation of organic fuels.[8, 9]

Rhodium is one of the noble metals that has enjoyed continuous interest with respect to CO oxidation. Since the development of the three-way automobile catalyst, researchers have turned their attention primarily to elucidating the mechanism of the CO oxidation by oxygen (molecular or atomic) and NO under UHV and gas-phase conditions [10-22] and to the binding properties of CO on low-index Rh surfaces.[23-27] However, electrooxidation of CO on rhodium has received considerably less attention than on platinum or palladium.

In UHV experiments the oxidation reaction of CO with O2 on rhodium was found to be a structure sensitive [17, 19, 22, 28] Langmuir-Hinshelwood type reaction.[11, 16, 22, 29] Molecular oxygen was found to adsorb dissociatively yielding atomic oxygen, which is mostly immobile on the surface, while the mobility of adsorbed CO remains relatively high.[14] In aqueous media a Langmuir-Hinshelwood type reaction is also assumed and, as is the case on platinum, the overall reaction mechanism may be represented by the following two steps:

H2O + * OHads + H+ + e- (2.1) COads + OHads CO2 + H+ + e- + 2* (2.2) Some of the first results on the electrooxidation of CO on Rh were obtained by

the groups of Weaver [6, 25, 30-33] and Wieckowski.[34-36] Weaver et al. were mainly interested in characterization of carbon monoxide chemisorbed on low-index rhodium single crystal electrodes with Scanning Tunneling Microscopy (STM), Infrared Reflection-Adsorption Spectroscopy (IRAS) and voltammetric techniques. On the basis of combined results of an IRAS and voltammetric analysis Chang et al. concluded that islands are formed during the electrooxidative removal of a CO adlayer on Rh(100) and the actual oxidation reaction proceeds at the boundary of the islands.[32] Moreover, the discrepancies between the CO coverage estimated from the voltammetric profiles and IRAS techniques were addressed extensively.[6, 33]

Wieckowski's group has used CO electrooxidation primarily as part of their surface preparation technique (iodine-CO displacement technique). The CO oxidation peak on Rh(111) was found to be around 0.69 V vs. RHE (in agreement with Weaver et al.[6, 25, 30-33]) and was reported to overlap with surface oxidation processes. Similar results were obtained for Rh(100) and polycrystalline Rh electrodes.[36] In a later study on the structure of electrochemical adsorbates on Rh(111), Sung et al. found two structures, which are distinguishable by cyclic voltammetry.[37] The first structure has a coverage of 0.75 ML and is formed at 0.05 V vs. RHE at long CO absorption times,

22

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a voltammetric study

while the second is formed at short absorption times and has a coverage of 0.65 ML. These findings are in agreement with the STM results reported by Yau et al.,[25] who found two distinctly different potential dependent adlayer structures for saturated CO coverage on ordered Rh(111) in aqueous solutions.

Gómez et al.[38] studied the effect of surface crystalline heterogeneities on the electrooxidation reaction of CO adsorbed on Rh(111). In an effort to explain the discrepancies in electrooxidation rates on electrodes prepared by the iodine-CO displacement [34] and the flame annealing technique (as reported in ref. [39]), they found that the kinetics of the CO stripping process is very sensitive to the long-range order of the (111) domains. Additionally, flame-annealed electrodes have fewer defects than their I/CO prepared counterparts. Since a more ordered surface leads to lower CO oxidation rates, the CO electrooxidation reaction provides a tool for the assessment of the surface quality. This finding is especially important considering that the blank voltammogram of a stepped rhodium electrode, like Rh(554), closely resembles that of Rh(111).

Even if the work by Gómez et al. provided evidence for the strong structure sensitivity of the CO oxidation reaction on Rh, no systematic analysis of the mechanism and structure sensitivity has as yet been performed. Therefore, the goal of this research is to study in more detail the electrooxidation of carbon monoxide on stepped rhodium single crystal electrodes of [n(111)×(111)] orientation. The well-defined surface structure of these electrodes allows us to systematically investigate the effects of crystalline heterogeneities on the kinetics and mechanism of the CO electrooxidation reaction, using cyclic voltammetry, CO adlayer stripping, and bulk CO oxidation experiments.

Results presented in this chapter show that the oxidation of CO on rhodium is a structure sensitive process and that, as for platinum, steps are the active sites for the reaction. However, the kinetics of this process is different from that on platinum, where the onset and main peak potential of the CO oxidation reaction was found to decrease with increasing step density. On rhodium this effect is not clearly observed, which may be explained by a combined effect of stronger anion adsorption and easier surface oxidation on rhodium electrodes compared to platinum surfaces.

2.2. Experimental Setup

The working electrodes used in this study were rhodium bead-type single crystal electrodes of Rh[n(111)×(111)] (identical to Rh[(n-1)(111)×(110)]) orientation (Rh(331), Rh(553), Rh(554), and Rh(111) with n=3, n=5, n=10, and n=200-500 respectively). The electrodes were prepared as described in ref. [39], and oriented, cut and polished according to the Clavilier method.[40] Prior to each measurement the single crystal electrode was flame annealed and cooled down to room temperature in an argon (Hoekloos, N50)-hydrogen atmosphere (ratio 3:1), after which it was transferred to the electrochemical cell under protection of a droplet of deoxygenated water.

A special electrochemical cell, described in ref. [5], contained a small movable

0.55 V

23

Chapter 2

spoon over an electrolyte reservoir, which allowed dosing CO at open circuit potential (ocp) from a saturated CO solution without dissolving CO in the blank electrolyte. The cell was cleaned by boiling in a 1:1 mixture of concentrated sulfuric and nitric acid, followed by repeated boiling (four times) with ultra-pure water (Millipore MilliQ gradient A10 system, 18.2 MΩcm, 2 ppb total organic carbon). The blank electrolyte, 0.5 M H2SO4, was prepared with concentrated sulfuric acid (Merck, "Suprapur") and ultra-pure water. During measurements the blank electrolyte was deoxygenated with argon (N50) and the electrolyte in the container above the spoon was saturated with CO gas (Hoekloos, N47).

For continuous CO oxidation experiments a conventional three-electrode-cell was used. The blank electrolyte was saturated by bubbling CO through the solution for 8-10 minutes while the electrode remained immersed in the electrolyte at 0.1 V vs. RHE. During the oxidation experiments CO was passed over the solution. A coiled platinum wire was used as a counter electrode and the reference electrode was a mercury-mercury sulfate electrode (MMSE: Hg|Hg2SO4|K2SO4 (sat)) connected via a Luggin capillary. However, all potentials in this article were converted to the reversible hydrogen electrode (RHE) scale. Measurements were performed at room temperature (22 ºC) with a computer-controlled Autolab PGSTAT20 potentiostat.

2.3. Results and Discussion

2.3.1. System Cleanliness and Surface Quality

Due to a pronounced effect of contamination on experiments with single crystal electrodes, it is necessary to determine the cleanliness of the electrochemical system. For platinum electrodes this information is readily obtained from the blank cyclic voltammograms (BCV). However, the BCV of a rhodium single crystal electrode is not as sensitive to contamination as that of platinum. Therefore, the system cleanliness was checked by recording the CV of Pt(111) in 0.5 M H2SO4.

The quality of the flame annealed rhodium electrodes was assessed by the blank cyclic voltammograms and CO adlayer stripping, as discussed in the introduction. For a well-ordered Rh(111) surface at a scan rate of 20 mV⋅s-1 up to a potential of 1 V vs. RHE more than thirty cycles were required to completely recover the blank CV (in accordance with ref. [38]). Rh(554), Rh(553), and Rh(331) require five, three, and two cycles, respectively.

2.3.2. Cyclic Voltammetry in Absence of CO

Fig. 2.1a-d show the blank cyclic voltammograms of the flame annealed Rh(111), Rh(554), Rh(553), and Rh(331) single crystal electrodes, respectively, recorded in 0.5 M H2SO4. The CVs of Rh(111) and Rh(554) correspond well with those reported in the literature [35-38] and clearly show the expected overlap of the hydrogen

24

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a voltammetric study

0.0 0.2 0.4 0.6 0.8 1.0

-600

-500

-400

-300

-200

-100

0

100

200

0.0 0.2 0.4 0.6 0.8 1.0

-600

-500

-400

-300

-200

-100

0

100

200

0.0 0.2 0.4 0.6 0.8 1.0

-600

-500

-400

-300

-200

-100

0

100

200

0.0 0.2 0.4 0.6 0.8 1.0

-600

-500

-400

-300

-200

-100

0

100

200

j / µ

A cm

-2

E / V vs. RHE

j / µ

A cm

-2

E / V vs. RHE

j / µ

A cm

-2

E / V vs. RHE

j / µ

A cm

-2

E / V vs. RHE

(d)

(b)

(c)

(a)

Figure 2.1. Blank cyclic voltammograms of (a) Rh(111), (b) Rh(554) , (c) Rh(553), and (d) Rh(331) in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

adsorption/desorption region with the (bi)sulfate adsorption/desorption region.[41, 42] The same characteristic feature is also observed for Rh(553) and Rh(331).

In earlier work on the role of surface defects in the electrooxidation of CO on rhodium electrodes, Gómez and co-workers [38] pointed out that in the negative potential region (0.05V-0.2V) the CVs of Rh(554) and Rh(111) closely resemble each other, despite the fact that on Rh(554) approximately 10% of the surface sites are defects (steps), whereas Rh(111) is a virtually “defect-free” surface. Our results indicate that the same is true for the BCV recorded on Rh(553) (see Fig. 2.1). The anodic and cathodic peaks for these three surfaces are quite similar in shape and height. This contrasts strongly with the BCV characteristics of platinum electrodes of the same orientation, where the presence of defects is clearly visible.[43]

However, even though the rhodium BCVs in Fig 2.1 appear similar, close examination of the low potential region reveals clearly distinguishable features by which electrodes may be identified. Fig. 2.2a shows the hydrogen/(bi)sulfate region of the four electrodes plotted in one graph and demonstrates that the degree of reversibility of the hydrogen/(bi)sulfate peaks increases with the number of steps. Although the change in the degree of reversibility is too insensitive to the number of defects to determine the quality of the surface after the flame annealing procedure, it is nevertheless accurate enough to allow individual electrodes to be distinguished. Examination of the influence

25

Chapter 2

0.05 0.08 0.10 0.13 0.15 0.17 0.20

-400

-200

0

200

Rh(111) Rh(554) Rh(553) Rh(331)

j / µ

A cm

-2

E / V vs. RHE

0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2

0.04

0.06

0.08

0.10

0.12

0.14

E p / V

vs.

RH

E

Log(ν / mV s-1)

(b)

Cathodic peaks

Anodic peaks

(a)

Figure 2.2b. Effect of scan rate on the position of the anodic and cathodic hydrogen/(bi)sulfate peak position for Rh(111) (, solid thin line), Rh(554) (, dashed line), Rh(553) (, dotted line) and Rh(331) ( , solid thick line).

Figure 2.2a. Hydrogen/(bi)sulfate region of the cyclic voltammograms of Rh(111) (solid thin line), Rh(554) (dashed line), Rh(553) (dotted line), and Rh(331) (solid thick line) in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

of the scan rate on the degree of reversibility shows that for each surface the position of both the anodic and cathodic peaks approaches 0.1 V vs. RHE as the scan rate nears 0 mV⋅s-1 (see Fig. 2.2b). Moreover, the peak position of the reduction process appears slightly more sensitive to the scan rate than that of the anodic process, indicating that the removal of adsorbed (bi)sulfate is a more sluggish process than its adsorption.[44]

Another interesting difference between the blank cyclic voltammograms of the electrodes is the charge under the low potential cathodic and anodic peaks, which decreases with increasing step density. Integrating the anodic peaks yields charge densities of 284 ± 14, 268 ± 13, 241 ± 14, and 176 ± 6 µC⋅cm-2 for the (111), (554), (553), and (331) surface, respectively (integration of the cathodic peak yields similar charge densities). From CVs of Rh(111) recorded in 0.01 M HF, which should be virtually unaffected by anion adsorption, Yau et al. estimated that the adsorption/desorption of hydrogen results in a charge of ca. 220 µC⋅cm-2.[45] A similar value was also used by Podlovchenko et al.[46] Subtraction of this value from our anodic charge on the (111) surface (neglecting the double layer charge and other effects and assuming that at 0.05V only H is adsorbed on the surface) leads to a surface ion concentration of approximately 4·1014 ions⋅cm-2, which is roughly equal to 0.25 ML of (bi)sulfate. Of course, because stepped surfaces have a lower surface atom density (due to the nature of the (110) steps) and thus are more open than the (111) surface, the hydrogen adsorption/desorption charge is lower on these electrodes. An estimated value for this charge on Rh(554), Rh(553) and Rh(331) can be found by estimation of the surface atom density of these electrodes and recalculating the hydrogen charge accordingly (as was done for platinum electrodes. [47, 48]). Here, 220 µC⋅cm-2 was taken as maximum coverage on the (111) plane. This calculation yields sulfate coverages of 0.22 ML for Rh(554), 0.17 ML for Rh(553) and 0.006 ML for Rh(331), which seems to

26

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a voltammetric study

suggest that adsorption and desorption of (bi)sulfate takes place primarily on the terraces and not on the steps.

Interestingly, Wan et al.[49] reported a value of 0.2 ML for (bi)sulfate on Rh(111) corresponding to a (√3 x √7) adlayer structure, which falls within our experimental error of 14 µC⋅cm-2 (± 14 µC⋅cm-2: 0.25 ± 0.054 ML). Results of lattice-gas modeling techniques provided by Rikvold et al. also suggest a (√3 x √7) (bi)sulfate adlayer structure.[42] In contrast, UHV experiments performed by Sung et al. indicate a (√3 x √3)R30º adlayer with a maximum coverage of 0.33 ML.[37] In addition, the authors obtained the same maximum coverage from an integration method analogous to ours. However, integration of their Rh(111) cathodic peak yields a charge approximately 100 µC⋅cm-2 (380 µC⋅cm-2) larger than the one we obtained. Moreover, from experiments in perchloric acid electrolyte, a hydrogen adsorption/desorption charge of 262 µC⋅cm-2 was estimated,[37] which is close to the theoretical value assuming one hydrogen atom per rhodium surface atom (1.6⋅1015 atoms⋅cm-2).

In order to verify our previously made observation concerning (bi)sulfate coverages on stepped surfaces, CO charge displacement experiments were performed.(see ref. [50]) The experiment was performed by measuring the charge resulting from displacement of a (bi)sulfate adlayer by CO at a potential of 0.4 V vs. RHE. The recorded displacement charges for Rh(111), Rh(554), Rh(553) and Rh(331) were 155, 124, 116, and 106 µC⋅cm-2, respectively. Using an estimate for the surface atom density of the stepped surfaces, these charges, when taken literally, would amount to a (bi)sulfate coverage of 0.60, 0.48, 0.45 and 0.41, or a sulfate coverage 0.30, 0.24, 0.23 and 0.21 on Rh(111), Rh(554), Rh(553) and Rh(331), respectively. Although we cannot specify the nature of the displaced species (sulfate or (bi)sulfate), the results indicate a decreasing coverage for increasing step density, as was also concluded above. The qualitative agreement between the displaced charges and the calculation from cyclic voltammetry suggests that the terrace-step model is fulfilled for surfaces with terraces larger than 4 atomic rows. In addition, the calculation of the hydrogen adsorption on the Rh(331) surface may perhaps introduce a large error, thus resulting in a unrealistically low value for the (bi)sulfate coverage. The advantage of the charge displacement technique over determination of the (bi)sulfate coverage from CV data is that it does not require an estimation of the hydrogen adsorption charge and, therefore, it is probably more accurate.

At present the reason for the difference in the charge between results reported by Sung et al. and ours is unclear. However, it is well known that traces of molecular oxygen can cause large differences in the shape of rhodium voltammograms and may lead to an overestimation of the charge in the hydrogen/(bi)sulfate region. In addition, which of the two reported hydrogen charges is correct, is another issue that needs to be resolved. Although the correctness of the values obtained from the calculation may be arguable, we believe that the trend shown by our findings is valid and suggests that (bi)sulfate indeed preferably adsorbs on the terraces and not on the steps (and perhaps

27

Chapter 2

crystalline defects in general). Both Mostany et al.[51] and Gómez et al.[52] showed that on stepped platinum electrodes (bi)sulfate indeed seems to adsorb only on the terraces. (Although Marković et al.[53, 54] suggested that (bi)sulfate may adsorb more strongly at or on the (110) steps than on the (111) terraces.) Moreover, this trend provides an explanation for the increasing reversibility of the anodic and cathodic peaks with increasing step density. Because of the lower (bi)sulfate concentration on “defective” (stepped) surfaces, hydrogen absorption/desorption on the terraces and steps is less hindered and thus more reversible.

In the upper potential range (See Fig. 2.3), where adsorption and desorption of oxygenated species takes place, the differences between the four electrodes become more significant. As one would expect, the potential at which these species can be observed is lowest for the most open surface and highest for the nearly defect-free (111) surface, which indicates preferential adsorption of oxygenated species on the steps. The onsets of surface oxidation on Rh(331), Rh(553), Rh(554) and Rh(111) lie at 0.55; 0.60; 0.65; and 0.80 V vs. RHE, respectively. Even though it is well known that cycling up to high potentials leads to surface disordering,[55, 56] Gómez et al. mentioned that, when cycling up to 1 V vs. RHE, adsorption and desorption of oxygenated species on rhodium takes place without modification of the hydrogen/(bi)sulfate region of the voltammetric profiles.[38] However, during our experiments we have observed that exceeding the potential at which oxidation of the electrode starts leads to a detectable change in the voltammetric profile corresponding to the hydrogen-(bi)sulfate region. Fig. 2.4 shows a potentiodynamic curve of Rh(553) scanned several times up to 0.85 V compared to one where the upper potential limit was set below the oxidation potential of 0.60 V. From the broadening and the decrease in height of the hydrogen/(bi)sulfate peaks we can conclude that disordering of the surface occurs upon cycling to potentials well within the oxidation region. In fact, when comparing disordering of Rh(111) with, for instance, Rh(553) and Rh(331) it seems that surfaces with higher step densities disorder faster and more readily

28

0.5 0.6 0.7 0.8 0.9-30

-20

-10

0

10

20

30

40

Figure 2.3. Oxygen-formation region of the cyclic voltammograms of Rh(111) (solid thin line), Rh(554) (dashed line), Rh(553) (dotted line), and Rh(331) (solid thick line) in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

j / µ

A cm

-2

E / V vs. RHE

Rh(111) Rh(554) Rh(553) Rh(331)

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a voltammetric study

0.0 0.2 0.4 0.6 0.8

-600

-400

-200

0

200

0.05 0.10 0.15

-600

-550

0

j / µ

A cm

-2

E / V vs. RHE

Figure 2.4. Cyclic voltammograms of Rh(553) recorded up to 0.55 V (solid line) and 0.85 V (dotted line) in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

than surfaces with large terraces. This can be rationalized by assuming that the presence of steps (and crystalline defects in general), which oxidize more easily than terrace sites, promote disordering by facilitating a place exchange process.

In light of the results shown above, the decreasing charge under the low potential peaks in the voltammograms for the disordered electrodes can be ascribed to a lower (bi)sulfate adsorption, which also indicates an increased number of defects on the surface. Here it should be noted that repetitive cycling (10 to 15 scans) of a rhodium electrode up to high potentials eventually leads to a steady-state voltammogram. We assume that beyond this point cycling no longer affects the macroscopic structure of the surface (The position of defects may change due to surface oxidation and reduction, but the number of defects is constant.)

2.3.3. Cyclic Voltammetry saturated CO-adlayer Oxidation

To obtain the profiles of the ordered electrodes (henceforth denoted as electrodes of Type I), the potential of the blank CV, recorded prior to CO adsorption and subsequent stripping, was kept below the surface oxidation potential (see previous section), while the profiles of the disordered surfaces (henceforth denoted as electrodes of Type II) were obtained by cycling the blank CV up to 0.85 V until a steady-state voltammogram was reached. Shown in Fig. 2.5a and b are voltammetric profiles (first scans) of the oxidation of a saturated CO adlayer on Rh(111), Rh(554), Rh(553), and Rh(331) on electrodes of Type I and II, respectively.

In the CO stripping voltammograms obtained on Type I electrodes, depicted in Fig. 2.5a, each of the stepped electrodes exhibits a small shoulder at low potentials followed by a single asymmetrical oxidation peak, with a maximum occurring at approximately 0.72-0.73 V. Moreover, as was mentioned in Section 2.3.1., stripping of a CO adlayer on the Type I Rh(111), Rh(554), Rh(553) and Rh(331) requires

29

Chapter 2

0.60 0.65 0.70 0.75 0.80 0.85 0.90

0

50

100

150

200

250

300

j / µ

A c

m-2

E / V vs. RHE

Rh(111) Rh(554) Rh(553) Rh(331) 0.5 0.6 0.7 0.8 0.9

-3-2-1012345

j / µ

A c

m-2

E / V vs. RHE

0.60 0.65 0.70 0.75 0.80 0.85 0.90-50

0

50

100

150

200

250

300

350

(a)

j / µ

A c

m-2

E / V vs. RHE

Rh(111) Rh(554) Rh(553) Rh(331)

(b)

Figure 2.5a and b. CO adlayer stripping on ordered “Type I” (a) and “Type II” (b) Rh(111) (solid thin line), Rh(554) (dashed line), Rh(553) (dotted line), and Rh(331) (solid thick line) in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1. The inset in (a) shows a zoom of the CO adlayer stripping on Rh(111).

approximately 30, 5, 3 and 2 scans, respectively. This incomplete oxidative stripping of the CO adlayer, in combination with influences of (bi)sulfate re-adsorption and surface oxidation makes estimation of the stripping charge difficult. If the CO stripping charge of the second scan on Rh(331) is neglected, integration of the oxidation peak yields a charge of ca. 420 µC⋅cm-2. This is higher than the charge of 380 µC⋅cm-2 reported on Rh(111) by Sung et al.[37] The discrepancy is most probably caused by a difference in surface oxidation between the (111) and the more open (331) surface.

Although the CVs in Fig. 2.5a suggest an increasing peak potential for increasing step density, the differences are statistically insignificant (determined by measuring peak potentials of 4-6 curves). In agreement with previous observations, the peaks overlap with the region where oxidation of the surface occurs. Rh(111) remains virtually inactive even at potentials higher than 0.8 V (see the insert in Fig. 2.5a), while the highest activity is achieved for the most open surface, Rh(331). These features underline the strong effect of the surface structure on the kinetics of CO electrooxidation on rhodium and lead us to conclude that the oxidation reaction takes place preferentially on the defects and steps.

When examining the shoulder on the negative potential side of the main oxidation peak, the charge under this part of the curve is found to depend linearly on the step density (see Fig. 2.6). The presence of a shoulder can be explained by assuming that either CO reacts first at the steps at low potential, or by oxidation of a densely packed CO adlayer to form a less dense (and more stable) adlayer structure. As, in general, defects and steps decrease the maximum packing density of a CO adlayer, the latter explanation would imply that the charge recorded should decrease in the following the order: Rh(554)>Rh(553)>Rh(331). Because the opposite order is observed, we assume that the former explanation is more plausible. However, CO oxidation on steps can only lead to an increasing linear relationship between the step density and charge when the mobility of CO on the surface is low. If the mobility of CO on rhodium were high, fast diffusion towards the reactive sites would lead to a situation analogous to platinum

30

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a voltammetric study

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35

0

20

40

60

80

100

q / µ

C c

m-2

θ steps = 1/n

Figure 2.6. Relationship between the charge under the low potential shoulder recorded on the “Type I” electrodes and the step density. Charges were recorded from a potential of 0.6 V vs. RHE to the onset of the main oxidation peak.

electrodes, where for saturated CO adlayers a single oxidation peak is observed upon scanning to high potentials.[43] This suggests that the mobility of CO on rhodium single crystal electrodes may be much lower than on corresponding platinum electrodes. The assumption of a low surface mobility of CO would also agree well with the formation of densely packed CO islands during electrooxidation observed by Chang et al.[32]

Another indication for slow surface diffusion of CO may be found in the presence of an asymmetrical tail at the positive potential side of the main oxidation peak for the stepped electrodes (see Fig. 2.5a). When the CO coverage decreases due to oxidative removal at the steps, diffusion of reactive species over the surface toward the steps starts playing a greater role. This may lead to a surface diffusion limitation of the reaction rate and can result in a tailing in the i-E curve. The presence of strongly adsorbing (bi)sulfate ions may further inhibit free diffusion of CO over the surface. Based on this explanation, one would expect (i) the most pronounced tailing effect on the electrode with the largest terraces and (ii) the most symmetrical peak for the surface with the highest number of steps. Indeed the largest tailing effect can be observed for Rh(554), while Rh(331) displays a nearly symmetrical oxidation peak (Fig. 2.5a). A lower adsorption affinity of (bi)sulfate for more open surfaces may also lead to more symmetrical oxidation peaks for these surfaces. Since, unlike platinum, the surface mobility of adsorbed species on the rhodium surface seems low, it may well be possible that the overall reaction kinetics on this material can best be described by the nucleation and growth model (reactant species are assumed immobile on the surface), instead of by a mean-field approximation (reactant species are assumed to be perfectly mixed, i.e. highly mobile, on the surface). It should also be noted that, as rhodium is a metal that oxidizes easily, it is conceivable that surface oxides grow from the steps over the terraces. Chronoamperometry or potential step experiments may provide more detailed information on the diffusion parameters and kinetics of this reaction. The results of this analysis will be reported in the next chapter.

31

Chapter 2

Interestingly, the position of the main oxidation peak in Fig. 2.5a does not decrease with increasing step density, as is the case for platinum electrodes.[43] On platinum, the decrease in overpotential was attributed primarily to preferential adsorption of oxygen containing species at the steps in conjunction with fast CO diffusion over the surface. Although we assume that the same process also occurs on rhodium, it apparently does not have the same effect. Possible explanations for the difference between these two metals may be found in a slow CO diffusion over the surface, stronger adsorption of anions on rhodium, or in the higher oxidizability of rhodium surfaces.

Depicted in Fig. 2.5b are the CVs of the stripping experiments on the disordered “Type II” electrodes. Compared to the Type I CVs they show markedly different features. The maximum currents recorded on the disordered electrodes are higher and the number of scans required to completely strip the CO adlayer are lower (approx. 10 cycles for Rh(111), 2-3 for Rh(554), 2-1 for Rh(554) and 1 for Rh(331)) than on their well-ordered counterparts. This indicates a large increase in the number of defects on the surface compared to the ordered surfaces. Furthermore, the disappearance of a clearly distinguishable shoulder prior to the main oxidation peak, less pronounced tailing, and a shift of the peaks to more negative potentials, also underline this conclusion. More specifically, the disappearance of the distinguishable features in the CVs of the Type II electrodes indicates that the formed defects are randomly distributed over the surface. (When integrating the CO stripping charge for the Type II Rh(331) electrode, a charge of 520 µC⋅cm-2 is obtained. A possible explanation of this higher charge can be found in a more complete CO adlayer stripping and an increased effect of surface oxidation due to the presence of randomly distributed defects.)

Remarkably, the order in which the CO oxidation overpotential decreases for increasing step density is rather unexpected. By analogy with platinum electrodes, one would expect the main peak potential to decrease from Rh(111) to Rh(554), Rh(553) and Rh(331). However, the main peak potential of Rh(553) occurs at a higher potential than on Rh(554). A similar observation can also be made from the onset potential of the CO oxidation reaction. On both the disordered Type II as well as the ordered Type I electrodes, Rh(331) starts reacting first, followed by Rh(554), Rh(553) and Rh(111) (Fig. 2.5a and b). We suspect that the cause of this unexpected trend may be found in a complex interplay of surface oxidation, oxidation of CO at the steps, and diffusion of reacting species over the terraces.

In order to reduce the effects of low surface mobility on the response of the system, CO stripping experiments on the Type I electrodes were also performed at low scan rate, i.e. 2 mV⋅s-1 (see Fig. 2.7). As can be expected, at low scan rates the tailing at the positive side of the main oxidation peak almost completely disappeared, while the characteristic shoulder at the low potential side of the main oxidation peak is still discernable. Interestingly, at this scan rate, the order in which the CO oxidation overpotential decreases for increasing step density is Rh(111)>Rh(554)>Rh(553)>Rh(331), which coincides with the order recorded on analogous platinum electrodes. Because CVs recorded at higher scan rates show a

32

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a voltammetric study

different order, we conclude that slow mobility of CO must be have a pronounced influence on the reaction rate.

2.3.4. Cyclic Voltammetry of Bulk CO Oxidation

Potentiodynamic curves of continuous CO electrooxidation recorded on the electrodes under investigation in a 0.5 M H2SO4 electrolyte, are presented in Fig. 2.8a and b. The curves shown in Fig. 2.8a were obtained by cycling Type I electrodes up to 0.85 V in the CO-saturated electrolyte. Apart from the (331) surface, steady-state voltammograms were reached after the first scan for each surface. When cycling up to 1 V (Fig 2.8b), approximately 10-15 scans were required to obtain steady-state voltammograms.

0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90-5.0

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

j / µ

A cm

-2

E / V vs. RHE

0.6 0.7 0.8 0.9 1.0

-2.0

-1.0

0.0

1.0

2.0

3.0

j / µ

A cm

-2

E / V vs. RHE

Figure 2.7. CO adlayer stripping on ordered “Type I” Rh(111) (solid thin line), Rh(554) (dashed line), Rh(553) (dotted line), and Rh(331) (solid thick line) in 0.5 M H2SO4 at a scan rate of 2 mV⋅s-1. The inset shows a zoom of the CO adlayer stripping on Rh(111).

0.78 0.80 0.82 0.84 0.86

0

10

20

j / µ

A cm

-2

E / V vs. RHE

Rh(111) Rh(554) Rh(553) Rh(331)

0.70 0.75 0.80 0.85 0.90-5

0

5

10

15

20

j / µ

A cm

-2

E / V vs. RHE

0.4 0.6 0.8 1.0

0

100

200

300

400

500

600

700

j / µ

A cm

-2

E / V vs. RHE

Rh(111) Rh(554) Rh(553) Rh(331)

(a) (b)

Figure 2.8a and b. Bulk CO oxidation on “Type I” (a) and “Type II” (b) Rh(111) (solid thin line), Rh(554) (dashed line), Rh(553) (dotted line), and Rh(331) (solid thick line) in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

33

Chapter 2

Continuous readsoption of CO leads to a considerable positive shift in the onset of the electrooxidation reaction when compared to CO stripping experiments.[57] Similar observations were made on the corresponding platinum electrodes.[43] Such a positive shift is expected for Langmuir-Hinshelwood type reactions where CO has a much higher affinity for the adsorption sites than OH, which is somewhat surprising as rhodium surfaces oxidize easily and the binding strength of OH with the surface is expected to be considerably higher than for platinum. However, a CO stripping experiment performed on a rhodium electrode pre-covered by OH (by cycling to within the oxidation region, breaking contact with the electrolyte, and performing CO adsorption and stripping), showed normal stripping characteristics. Therefore, we conclude that CO can displace the (ir)reversibly adsorbed surface oxides and, thus, indeed adsorbs more strongly than OH. A similar observation was also made by Oh et al., who found that under UHV conditions CO adsorbs more strongly on rhodium than oxygen.[14]

The differences in the onset of the oxidation reaction, shown in Fig. 2.8a and b, on Type I and Type II electrodes illustrate the structure sensitivity of the process. The positive shift in the onset of the CO oxidation potential is largest for the ordered surfaces, being around 0.16 V, whereas for the disordered surfaces the average positive shift is approximately 0.13 V, with disordered Rh(111) having the highest shift of 0.18 V.

On well-ordered electrodes the reactivity increases in the order of Rh(111)<Rh(554)<Rh(553)<Rh(331). This order has also been observed on corresponding platinum electrodes by Lebedeva et al. and was explained by preferential adsorption of oxygen-containing species on the steps.[43] Introducing more randomly distributed defects, by cycling up to 1 V (Fig. 2.8b), shifts the onset of bulk CO oxidation to slightly more negative potentials and changes the order of reactivity to Rh(111)<Rh(553)<Rh(554)<Rh(331). The decrease in the onset of the oxidation potential induced by disordering also indicates that the steps and generally crystalline defects are the active sites for the formation of adsorbed OH. Apparently, disordering of the (554) surface leads to a higher CO oxidation activity than disordering of the (553) surface. This is remarkable because a similar effect was also observed in the CO stripping voltammetry performed on the disordered electrodes.

2.4. Conclusion

Blank voltammetric profiles, oxidative stripping of CO adlayers, and bulk CO oxidation were studied on Rh[n(111)×(111)] single crystal electrodes in 0.5 M H2SO4. The blank CVs recorded on Rh(111), Rh(554), Rh(553) and Rh(331), show a distinct increase in the reversibility of the hydrogen/(bi)sulfate region as the number of steps on the surface increases. Moreover, the charge under the hydrogen/(bi)sulfate peaks was found to decrease with the step density. These observations may be explained by assuming that (bi)sulfate adsorption does not occur on the steps and defects, but rather on the terraces. Increasing the step density leads to lower (bi)sulfate adsorption and, thus, to

34

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a voltammetric study

a higher reversibility of the hydrogen/(bi)sulfate peaks. Results from CO displacement experiments also show a decreasing coverage of (bi)sulfate for increasing step density.

At higher potentials, increasing the step density of the rhodium electrodes leads to decreasing surface oxidation potentials, analogous to the platinum electrodes. However, contrary to previous observations, cycling the potential of rhodium electrodes into the surface oxidation region does not result in the formation of reversibly adsorbed surface oxides, but rather in a disordering of the electrode surface due to formation of irreversibly adsorbed species. This disordering is reflected in a decrease in height of the hydrogen/(bi)sulfate peaks in the low potential region, which may be explained by a decreased adsorption of sulfate on the more defective disordered surfaces.

CO stripping experiments performed on both well-ordered and disordered electrodes show the strong structure sensitivity of the reaction. On all electrodes the rate of the CO oxidation reaction was found to increase with the step density. The higher activity of the disordered electrodes was ascribed to an increased number of defects.

Unlike for platinum, stripping of a CO adlayer on rhodium does not result in a single symmetrical oxidation peak, but rather in an oxidation peak with a pre-shoulder and a distinct tailing in the high potential region. Charge calculations show that the charge under the shoulder increases linearly with the step density, which leads to the conclusion that the shoulder is associated with CO reacting directly on or next to the steps and that surface diffusion of the reacting species must be slow. Since the shoulder is recorded at lower potentials than the main oxidation peak, we conclude that steps (and crystalline defects in general) are the active sites for the CO oxidation reaction. The presence of a distinct tailing of the main oxidation peak and the fact that this tailing disappears when the scan rate, and thus the influence of mass transport processes, is decreased, also suggests slow surface diffusion of CO. A more symmetrical shape of the oxidation peaks recorded on the disordered electrodes was ascribed to an increased number of randomly distributed surface defects.

Bulk CO oxidation experiments again show strong structure sensitivity and the positive shift in the oxidation potential for continuous CO oxidation indicates that CO and OH adsorb competitively.

Finally, we conclude that steps (and defects) on rhodium have the same catalytic effect on the CO oxidation reaction as on platinum electrodes and that they act as nucleation sites for the formation of OHads. Therefore, at a given overpotential, the concentration of OHads is higher on more open surfaces, resulting in an enhanced CO oxidation reaction rate. However, even though the mechanism of the reaction may be similar to that on platinum, the dynamics of the reaction on rhodium differs significantly. It seems that the surface mobility of CO is lower on Rh than on Pt. Therefore, it is conceivable that the best model description for the overall reaction rate should be based on dynamic Monte Carlo simulations (reported in Chapter 5) rather than on the “mean-field approximation”. A more suitable technique for analyzing reaction kinetics of CO adlayer electrooxidation is chronoamperometry or potential-step experiments. Moreover, as (bi)sulfate anions play a dominant role in both the catalysis of CO on Pt(hkl) as well as

35

Chapter 2

on Rh(hkl) electrodes, a deeper understanding of the adsorption properties of anions and their influence on the reaction and diffusion rate of CO can be obtained by repeating the experiments published in this chapter in perchloric or fluoric acid electrolytes. The results of these studies can be found Chapters 3 and 4.

Acknowledgement: This research was supported by the Netherlands Foundation for Scientific Research (NWO) and MCYT (Spain) through project BQU2003-04029.

References

[1] N. M. Markovic and P. N. Ross, Surf. Sci. Rep. 45 (2002) 117. [2] S. Gilman, J. Phys. Chem. 68 (1964) 70. [3] C. McCallum and D. Pletcher, J. Electroanal. Chem. 70 (1976) 277. [4] B. Beden, C. Lamy, N. R. De Tacconi, and A. J. Arvia, Electrochim. Acta. 35 (1990) 691. [5] J. M. Feliu, J. M. Orts, A. Fernandez-Vega, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 296

(1990) 191. [6] M. J. Weaver, S. C. Chang, L. W. H. Leung, X. Jiang, M. Rubel, M. Szklarczyk, D. Zurawski, and

A. Wieckowski, J. Electroanal. Chem. 327 (1992) 247. [7] H. A. Gasteiger, N. Markovic, P. N. Ross, Jr., and E. J. Cairns, J. Phys. Chem. 98 (1994) 617. [8] A. Hamnett, Catal. Today. 38 (1997) 445. [9] A. Hamnett, Comp. Chem. Kin. (1999) 635. [10] J. T. Kummer, J. Phys. Chem. 90 (1986) 4747. [11] S. B. Schwartz, L. D. Schmidt, and G. B. Fisher, J. Phys. Chem. 90 (1986) 6194. [12] D. W. Goodman and C. H. F. Peden, J. Phys. Chem. B 90 (1986) 4839. [13] C. H. F. Peden, D. W. Goodman, D. S. Blair, P. J. Berlowitz, G. B. Fisher, and S. H. Oh, J. Phys.

Chem. 92 (1988) 1563. [14] S. H. Oh, G. B. Fisher, J. E. Carpenter, and D. W. Goodman, J. Catal. 100 (1986) 360. [15] J. Siera, F. Rutten, and B. E. Nieuwenhuys, Catal. Today. 10 (1991) 353. [16] M. Bowker, Q. Guo, and R. W. Joyner, Surf. Sci. 280 (1993) 50. [17] A. Baraldi, L. Gregoratti, G. Comelli, V. R. Dhanak, M. Kiskinova, and R. Rosei, App. Surf. Sci.

99 (1996) 1. [18] B. A. Gurney, L. J. Richter, J. S. Villarrubia, and W. Ho, Journal of Chemical Physics 87 (1987)

6710. [19] A. Baraldi, S. Lizzit, D. Cocco, G. Comelli, G. Paolucci, R. Rosei, and M. Kiskinova, Surf. Sci.

385 (1997) 376. [20] C. Castellarin Cudia, S. W. Hla, G. Comelli, Z. Sljivancanin, B. Hammer, A. Baraldi, K. C.

Prince, and R. Rosei, Phys. Rev. Lett. 87 (2001) 196104/1. [21] M. J. P. Hopstaken, W. J. H. Van Gennip, and J. W. Niemantsverdriet, Surf. Sci. 433-435 (1999)

69. [22] M. J. P. Hopstaken and J. W. Niemantsverdriet, J. Chem. Phys. 113 (2000) 5457. [23] V. R. Dhanak, A. Baraldi, G. Comelli, G. Paolucci, M. Kiskinova, and R. Rosei, Surf. Sci. 295

(1993) 287. [24] L. W. H. Leung, J. W. He, and D. W. Goodman, J. Chem. Phys. 93 (1990) 8328. [25] S. L. Yau, X. Gao, S. C. Chang, B. C. Schardt, and M. J. Weaver, J. Am. Chem. Soc. 113 (1991)

6049. [26] P. Cernota, K. Rider, H. A. Yoon, M. Salmeron, and G. Somorjai, Surf. Sci. 445 (2000) 249. [27] A. Beutler, E. Lundgren, R. Nyholm, J. N. Andersen, B. Setlik, and D. Heskett, Surf. Sci. 371

(1997) 381. [28] V. Nehasil, I. Stara, and V. Matolin, Surf. Sci. 352-354 (1996) 305. [29] L. S. Brown and S. J. Sibener, J. Chem. Phys. 89 (1988) 1163. [30] L. W. H. Leung and M. J. Weaver, J. Phys. Chem. 93 (1989) 7218. [31] L. W. H. Leung, S. C. Chang, and M. J. Weaver, J. Chem. Phys. 90 (1989) 7426. [32] S. C. Chang and M. J. Weaver, J. Electroanal. Chem. 285 (1990) 263. [33] S. C. Chang, Y. Ho, and M. J. Weaver, J. Electrochem. Soc. 139 (1992) 147. [34] M. Hourani and A. Wieckowski, J. Electroanal. Chem. 227 (1987) 259. [35] M. Hourani, M. Wasberg, C. Rhee, and A. Wieckowski, Croatica Chemica Acta 63 (1990) 373.

36

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a voltammetric study

[36] C. K. Rhee, M. Wasberg, P. Zelenay, and A. Wieckowski, Catal. Lett. 10 (1991) 149. [37] Y. E. Sung, S. Thomas, and A. Wieckowski, J. Phys. Chem. 99 (1995) 13513. [38] R. Gomez, J. M. Orts, J. M. Feliu, J. Clavilier, and L. H. Klein, J. Electroanal. Chem. 432 (1997)

1. [39] J. M. Feliu, J. M. Orts, R. Gomez, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 372 (1994)

265. [40] J. Clavilier, D. Armand, S. G. Sun, and M. Petit, J. Electroanal. Chem. 205 (1986) 267. [41] J. Clavilier, A. Rodes, K. El Achi, and M. A. Zamakhchari, J. Chim. Phys. Phys. Chim. Bio. 88

(1991) 1291. [42] P. A. Rikvold, M. Gamboa-Aldeco, J. Zhang, M. Han, Q. Wang, H. L. Richards, and A.

Wieckowski, Surf. Sci. 335 (1995) 389. [43] N. P. Lebedeva, M. T. M. Koper, E. Herrero, J. M. Feliu, and R. A. van Santen, J. Electroanal.

Chem. 487 (2000) 37. [44] R. Albalat, J. Claret, A. Rodes, and J. M. Feliu, J. Electroanal. Chem. 550-551 (2003) 53. [45] S.-L. Yau, Y.-G. Kim, and K. Itaya, J. Am. Chem. Soc. 118 (1996) 7795. [46] B. I. Podlovchenko and T. D. Gladysheva, Russian Journal of Electrochemistry (Translation of

Elektrokhimiya) 38 (2002) 349. [47] J. Clavilier, K. El Achi, and A. Rodes, Chem. Phys. 141 (1990) 1. [48] J. Clavilier, K. El Achi, and A. Rodes, J. Electroanal. Chem. 272 (1989) 253. [49] L. J. Wan, S. L. Yau, and K. Itaya, J. Phys. Chem. B 99 (1995) 9507. [50] J. M. Orts, J. M. Feliu, A. Aldaz, and J. Clavilier, Electrochim. Acta. 39 (1994) 1519. [51] J. Mostany, E. Herrero, J. M. Feliu, and J. Lipkowski, J. Phys. Chem. B 106 (2002) 12787. [52] R. Gomez, V. Climent, J. M. Feliu, and M. J. Weaver, J. Phys. Chem. B 104 (2000) 597. [53] N. M. Markovic, N. S. Marinkovic, and R. R. Adzic, J. Electroanal. Chem. Interfacial

Electrochem. 241 (1988) 309. [54] N. M. Markovic, N. S. Marinkovic, and R. R. Adzic, J. Electroanal. Chem. 314 (1991) 289. [55] L.-J. Wan, S.-L. Yau, G. M. Swain, and K. Itaya, J. Electroanal. Chem. 381 (1995) 105. [56] K. Sashikata, N. Furuya, and K. Itaya, J. Vac. Sci. & Tech. B 9 (1991) 457. [57] H. A. Gasteiger, N. M. Markovic, and P. N. Ross, Jr., J. Phys. Chem. 99 (1995) 8290.

37

Chapter 2

38

Chapter 2

38

CO oxidation on stepped Rh[n(111)×(111)] single crystal electrodes: a chronoamperometric study

Abstract

The CO electrooxidation reaction was studied on Rh[n(111)×(111)]-type electrodes in 0.5 M H2SO4 using chronoamperometry. The transients recorded on Rh(111), Rh(554), Rh(553) and Rh(331) are characterized by a current plateau, visible directly after charging of the double layer, followed by a main oxidation feature, which consists of two peaks, a pre-peak and a main-peak. The current density in the plateau region is nearly constant over time and, thus, is of (quasi) zero-th order in CO coverage. A plot of log(jplateau) vs. Efinal gives a linear relationship with a slope of ca. 45 mV⋅dec-1, suggesting a second electron transfer as rate determining step. Analogous to platinum, the current plateau was ascribed to a Langmuir-Hinshelwood type reaction between COads and OHads with no effective freeing of sites for OH adsorption due to relaxation of the CO adlayer. The presence of two peaks rather than one in the main oxidation feature can be explained by assuming a low surface mobility of CO and high oxidizability of the rhodium surfaces. Indeed, dual step chronoamperometry shows that the mobility of CO on rhodium surfaces in aqueous media is very low. As rhodium is known to oxidize readily, the pre-peak and main-peak can be ascribed to CO reacting with OH, which adsorbs fast at the steps and more slowly at terrace sites. Because the geometry of the steps is nearly the same for each surface, the pre-peak appears structure insensitive, while the main-peak shifts considerably with the step density. Introducing randomly distributed crystalline defects by cycling the electrodes repeatedly up to the surface oxidation region prior to each potential step experiment, results in a negative shift of the main-peak, while the position of the pre-peak remains fixed. From the data presented we conclude that the reaction nucleates at the steps and grows in the direction of the terraces.

This chapter is published as T.H.M. Housmans, M.T.M. Koper, J. Electroanal. Chem. 575(2005) 39

Chapter 3

3.1. Introduction

The electrooxidation of CO on noble metal surfaces is one of the most interesting and most intensively investigated reactions in electrocatalysis. Not only does CO act as catalyst poison and, thus, hinders the development of a low temperature fuel cell,[1-4] this relatively simple model reaction is also of fundamental interest because CO acts as a neutral probe to surface activity and structure sensitivity.

Rhodium is one of these noble metals that has enjoyed continued interest with respect to the oxidation of CO. Its ability, when combined with platinum, to catalytically convert nitric oxide, carbon monoxide, and unburned hydrocarbons to N2, CO2 and H2O have made it a vital component of the three-way automobile catalyst.[5-17] However, the electrooxidation of CO on rhodium has received much less attention, certainly in comparison to platinum. Even if rhodium may not be a particularly attractive metal for application in low-temperature fuel cells (being considerably less active than platinum), a detailed study of its CO electrooxidation kinetics, similarly to studies carried out recently for platinum,[18-20] may give considerable insight, especially in relation to the still much debated issue of CO surface mobility.[21]

Analogous to platinum electrodes, the electrochemical oxidation of CO on rhodium was found to be a strongly structure sensitive reaction,(see Chapter 2 and refs. [22, 23]) and is assumed to be of the Langmuir-Hinshelwood type with the overall mechanism represented by the following two steps [24] (* denotes an empty surface site):

H2O + * OHads + H+ + e- (3.1) COads + OHads CO2 + H+ + 2e- + 2* (3.2) In a recent voltammetric study of this reaction on stepped rhodium single crystal

electrodes of [n(111)×(111)] orientation in sulfuric acid,(Chapter 2, ref. [23]) we concluded that steps (and crystalline defects in general) are the nucleation sites for the reaction, but also that the mobility of CO on the surfaces is very low. Evidence for these conclusions was found in a shoulder-like feature prior to the main oxidation peak, the charge of which varied linearly with the step density, and a pronounced tailing of the main oxidation peaks recorded for Rh(111) and Rh(554), surfaces with relatively wide terraces. As results of UHV studies indicate a very high surface mobility of CO,[25-27] either or both a strong anion adsorption and ease of oxidizability of rhodium surfaces were suggested as possible causes for such a low diffusion rate.

Since the total reaction rate is dependent on the spatial distribution of the reactants, it is influenced by their surface mobility. Analytical expressions for the total reaction rate can be obtained in two limiting cases, namely the mean-field approximation and the nucleation and growth model. The mean-field approximation assumes that reactants are perfectly mixed on the surface and that the reaction rate is proportional to the average coverages of the reacting species.[28] This condition is satisfied when surface diffusion is faster than the reaction itself. In the nucleation and growth model reacting species are assumed to be immobile on the surface. The reaction nucleates by adsorption of OH at ”special” sites, usually defects or steps, and proceeds only at the

40

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

interface between two reacting phases, causing the formation and growth of islands.[29] If diffusion and reaction rates are comparable, a full numerical solution of the kinetic equations is required, which may be obtained by dynamic Monte Carlo simulations or a differential-equation approach involving more advanced approximations than the mean-field approximation or nucleation and growth model.[30]

Chronoamperometry (or potential step experiments) is arguably the most appropriate electrochemical technique for qualitative and quantitative studies of the kinetics of the CO electrooxidation reaction. For platinum numerous chronoamperometric studies show current-time curves that are characterized by a current plateau, followed by a single main oxidation peak.[19, 20, 24, 31-33] The current plateau was initially explained by an Eley-Rideal type reaction between adsorbed CO and non-adsorbed water.[32] However, since the Tafel slope of the plateau is close to that of the main oxidation peak (80mV⋅dec-1) and because the current in the plateau region increases with increasing step density, Lebedeva et al.[19, 20] considered an Eley-Rideal mechanism unlikely and suggested a Langmuir-Hinshelwood type reaction with no effective freeing of sites for OH adsorption, due to relaxation of the CO adlayer. Moreover, they concluded that the reaction kinetics in the main peak region could be modeled well using the mean-field approximation for the Langmuir-Hinshelwood mechanism.

Given that chronoamperometric CO oxidation transients on well-defined rhodium surfaces are absent from the literature, we have performed chronoamperometry on rhodium single crystal electrodes of [n(111)×(111)] orientation. Our prime aim is to elucidate the differences and similarities between the CO oxidation transients on Rh and Pt and determine which model best describes the reaction kinetics on rhodium surfaces. In agreement with our earlier voltammetric results, the chronoamperometric transients indicate that the mobility of CO on rhodium surfaces is indeed very low and that on rhodium the CO electrooxidation reaction can best be described by a nucleation and growth Langmuir-Hinshelwood type reaction, rather than by a mean-field approximation.

3.2. Experimental Setup

The working electrodes used in this study were rhodium bead-type single crystal electrodes of Rh[n(111)×(111)] (identical to Rh[(n-1)(111)×(110)]) orientation (Rh(331), Rh(553), Rh(554), and Rh(111) with n=3, n=5, n=10, and n=200-500 respectively). The electrodes were prepared as described in ref. [34], and oriented, cut and polished according to the Clavilier method.[35] Prior to each measurement the single crystal electrode was flame annealed and cooled down to room temperature in an argon (Hoekloos, N50)-hydrogen atmosphere (ratio 3:1), after which it was transferred to the electrochemical cell under protection of a droplet of deoxygenated water.

A special electrochemical cell, described in ref. [36], contained a small movable spoon over an electrolyte reservoir, which allowed dosing of CO at open circuit potential (ocp) from a saturated CO solution without dissolving CO in the blank electrolyte. The

41

Chapter 3

CO coverage was checked by cycling in the hydrogen/(bi)sulfate region. The electrochemical cell was cleaned by boiling in a 1:1 mixture of concentrated sulfuric and nitric acid, followed by repeated boiling (four times) with ultra-pure water (Millipore MilliQ gradient A10 system, 18.2 MΩcm, 2 ppb total organic carbon). The blank electrolyte, 0.5 M H2SO4, was prepared with concentrated sulfuric acid (Merck, "Suprapur") and ultra-pure water. During measurements the blank electrolyte was deoxygenated with argon (N50) and the electrolyte in the container above the spoon was saturated with CO gas (Hoekloos, N47).

A coiled platinum wire was used as a counter electrode and the reference electrode was a mercury-mercury sulfate (MMSE: Hg|Hg2SO4|K2SO4 (sat)) electrode connected via a Luggin capillary. However, all potentials in this article were converted to the RHE scale. Measurements were performed at room temperature (22 ºC) with a computer-controlled Autolab PGSTAT20 potentiostat.

3.3. Results and Discussion

3.3.1. System Cleanliness and Surface Quality

As is customary for single crystal electrode experiments, the system cleanliness was verified prior to each measurement. As the blank cyclic voltammograms (BCVs) of rhodium single crystal electrodes are not as sensitive to contamination as those of platinum electrodes, the cleanliness of the cell was checked by recording the BCV of Pt(111) in 0.5 M H2SO4 electrolyte.

The quality of the flame annealed rhodium electrodes was checked by cyclic voltammetry and CO adlayer stripping. Stripping of a CO adlayer on a well-ordered Rh(111) surface should require more than 30 cycles up to 1 V vs. RHE at a scan rate of 20mV⋅s-1 to completely recover the blank CV.[22] Under the same conditions Rh(554), Rh(553) and Rh(331) require five, three, and two cycles, respectively. However, as we pointed out in the previous chapter (also see ref. [23]) repeated cycling to potentials higher than the surface oxidation potential leads to disordering of the rhodium surfaces. Keeping the upper potential limit of the cyclic voltammogram below the potential at which surface oxidation commences can prevent this disordering (for details, see ref. [22]). For more details on the cyclic voltammetry on Rh single crystal surfaces we would like to refer to Chapter 2 and ref. [37-47].

Here, we would like to introduce a method for estimating the surface quality of well-ordered rhodium electrodes (with blank CVs cycled up to the surface oxidation potential). Previously, we found that the degree of reversibility of the hydrogen/(bi)sulfate peaks, present at low potential in the blank CVs of rhodium single crystal electrodes in 0.5 M H2SO4 (Fig. 3.1), increases with an increasing step density.[23] Therefore, the separation between the anodic and cathodic peaks in this region can be taken as a rough measure for the surface order of the electrodes. For a well-

42

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

0.0 0.2 0.4 0.6 0.8 1.0

-600

-500

-400

-300

-200

-100

0

100

200

0.0 0.2 0.4 0.6 0.8 1.0

-600

-500

-400

-300

-200

-100

0

100

200

0.0 0.2 0.4 0.6 0.8 1.0

-600

-500

-400

-300

-200

-100

0

100

200

0.0 0.2 0.4 0.6 0.8 1.0

-600

-500

-400

-300

-200

-100

0

100

200

j / µ

A cm

-2

E / V vs. RHE

j / µ

A cm

-2

E / V vs. RHE

j / µ

A cm

-2

E / V vs. RHE

j / µ

A cm

-2

E / V vs. RHE

(d)

(b)

(c)

(a)

Figure 3.1. Blank cyclic voltammograms of Rh(111) (a), Rh(554) (b), Rh(553) (c), and Rh(331) (d) in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

ordered Rh(111) this separation should be 60 mV or larger, while for Rh(554), Rh(553) and Rh(331) a minimum separation of 50, 30 and 25 mV is required, respectively.

3.3.2. CO Adlayer Oxidation on Rhodium

Investigating the electrooxidation of CO adlayers on well-defined rhodium single crystal electrodes is in many ways more complicated than on platinum. Not only is establishing the system cleanliness and quality of the surfaces more difficult compared to platinum, dealing with surface disordering during experiments is also a major issue. Since the CO electrooxidation potential and surface oxidation potential overlap (see Chapter 2, ref. [23]), the process of surface disordering occurs simultaneously with the oxidation of CO. As was pointed out by Gómez et al.,[22] oxidation of CO is a strongly structure sensitive reaction, which, in combination with surface disordering, makes interpreting the recorded current-time transients more problematic. For this reason the chronoamperometric experiments were performed on rhodium electrodes that were already disordered by repeated cycling up to 0.85 V vs. RHE until a steady voltammogram was reached (in Chapter 2 called “Type II” electrodes). Although this electrode type has more (randomly distributed) defects on the surface, it is considered more stable over the potential range under investigation. In order to obtain an indication

43

Chapter 3

of the influence of surface disordering on the reaction rate, current-time transients of well-ordered electrodes (in Chapter 2 called “Type I” electrodes) at two final potentials (0.635 and 0.68 V) were also recorded.

Another problem that arises when using rhodium single crystal electrodes is the estimation of the CO coverages prior to, and after measurements. Slow oxidative stripping of CO combined with overlapping hydrogen and (bi)sulfate adsorption/desorption regions makes an accurate determination of the CO coverage difficult. However, a scan of the hydrogen/(bi)sulfate region before and after an experiment compared to the blank CV can give an indication of the initial and final CO coverage.

3.3.2.1. Oxidation of saturated CO Adlayers on “Type II” Electrodes

The current-time transients recorded at a number of final potentials on Type II Rh(111), Rh(554), Rh(553) and Rh(331) are shown in Figure 3.2a-d. All transients are characterized by a current plateau, visible directly after charging of the double layer,

0 10 20 30 40 50-0.05

0.00

0.05

0.10

0.15

0.20

j / µ

A cm

-2

t / s

0 50 100 150 200 250 300

0.00

0.50

1.00

1.50

2.000 10 20 30 40 50 60

0.0

2.0

4.0

6.0

8.0

10.0

t / s

j / µ

A cm

-2

(5)

(3) 0.635 V (4) 0.620 V (5) 0.590 V

(1) 0.665 V (2) 0.650 V

(5)

(4)

(3)

(2)

(1)

Figure 3.2a. Current transients of the oxidation of saturated CO adlayers on “Type II” Rh(111). Step potentials are listed in the figure.

44

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

followed by a main oxidation peak. The shape of the peak depends greatly on the structure of the surface. In general, for surfaces with low step density, i.e. Rh(111) and Rh(554), the main oxidation peak is asymmetrical in shape and it is characterized by a fast current increase followed by a slow current decay. On Rh(553) the increase of the reaction rate with time is slower than on Rh(111) and Rh(554) and it decreases more rapidly after reaching its maximum. Further increase of the step density by going to Rh(331) results in nearly symmetrical peaks.

On platinum electrodes several authors have observed current plateaus prior to either a symmetrical or asymmetrical main oxidation peak.[19, 20, 24, 31-33] However, as, to the best of our knowledge, no plateau regions have been observed previously on rhodium, we shall discuss this process first before addressing the main oxidation peak.

45

0 10 20 300.0

25.0

50.0

75.0

0 2 4 60.0

50.0

100.0

150.0

200.0

250.0

0 50 100 150 200 250 300

0.0

2.0

4.0

6.0

8.0

10.0

j / µ

A c

m-2

t / s

0.60 V

Figure 3.2b. Current transients of the oxidation of saturated CO adlayers on “Type II” Rh(554). Step potentials are listed in the figure.

(7) 0.605 V(8) 0.590 V(9) 0.575 V

(9)

(8)

(6)

(3)

(4) 0.650 V(5) 0.635 V(6) 0.620 V

(1) 0.695 V(2) 0.680 V(3) 0.665 V

(7)

(5)

(4)

(2)

(1)

Chapter 3

3.3.2.1.1. Current Plateau

The inset in Fig. 3.2a shows a zoom of the current plateau recorded on Rh(111) at 0.59 V vs. RHE. It is clear from the graph that the current density is nearly constant over a time interval of approximately 20 seconds, which indicates that the apparent order of this plateau is either zero, as modeled by the following equation,

kdt

dj CO −=∝

θ (Eq. 3.1)

or “quasi zero” with respect to the CO coverage, as represented by the rate law in equation 2. [20]

COCO k

dtd

j θθ

−=∝ with θCO ≈ constant (Eq. 3.2)

While the main oxidation peak indicates a second order rate law with respect to the CO coverage, a different reaction mechanism must apply to the plateau region. A plot of the logarithm of the plateau current, jplateau, versus the final potential (see Fig. 3.3) of

0 50 100 150 200

0.0

10.0

20.0

30.00 10 20 30 40

0.0

50.0

100.0

150.0

200.0

t / s

j / µ

A cm

-2

Figure 3.2c. Current transients of the oxidation of saturated CO adlayers on “Type II” Rh(553). Step potentials are listed in the figure.

(5) 0.635 V (6) 0.620 V (7) 0.605 V (8) 0.590 V

(1) 0.695 V (2) 0.680 V (3) 0.665 V (4) 0.650 V

(8)

(7)

(6)

(5)

(4)

(3)

(2)

(1)

46

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

the chronoamperometric experiments shows the effect of the step density on the reaction rate in this region. Similar to platinum, the introduction of a small number of defects, as illustrated by going from Rh(111) to Rh(554), results in a relatively large increase in the plateau current, while increasing the step density further by going from Rh(554) to Rh(553) has virtually no effect on the current density.[19, 20]

The slope of log(jplateau) versus Efinal gives a linear relationship with a value of 44, 42 and 48 mV⋅dec-1 for Rh(111), Rh(554), Rh(553), respectively, in contrast to Pt(111), Pt(554) and Pt(553) for which Lebedeva et al. found Tafel slopes of 81, 82 and 111 mV⋅dec-1, respectively.[19, 20] From the Tafel slope of ca. 60-80 mV⋅dec-1, several authors [48, 49] have suggested that the following chemical step may be rate determining:

COads + OHads COOHads + * (3.3)

0.58 0.60 0.62 0.64 0.66

-1.5

-1.0

-0.5

0.0

0.5

log(

j plat

eau /

µA c

m-2)

E / V vs. RHE

0 10 20 30 400.0

50.0

100.0

150.0

200.0

250.0

300.0

Figure 3.2d. Current transients of the oxidation of saturated CO adlayers on “Type II” Rh(331). Step potentials are listed in the figure.

j / µ

A cm

-2

t / s

(1) 0.680 V(2) 0.635 V

(2)

(1)

Figure 3.3. Logarithm of the current in the plateau region, jplateau, versus the step potential on Rh(111) ( , solid line), Rh(554) ( , dashed line) and Rh(553) ( , dotted line).

47

Chapter 3

(The change in the slope from ca. 80 to 111 mV⋅dec-1 was interpreted as a change in the rate-determining step from the chemical step on electrodes with large (111) terraces to the first electron transfer process on electrodes with a high number of (110) steps.) On rhodium on the other hand, the Tafel slope of ca. 40 mV⋅dec-1 suggests that a second electron transfer may be rate determining. Reaction (3.2) may now be represented by:

COads + OHads COOHads + * (3.4)

COOHads CO2 + H+ + e- + * (rds) (3.5)

As was mentioned in Section 3.1., on platinum, current plateaus were observed in numerous studies.[19, 20, 24, 31-33] Initially Bergelin et al. explained the plateau region by assuming an Eley-Rideal type reaction between adsorbed CO and a non-adsorbed water molecule.[32] However, from a transient analysis on Pt(111) in 0.1 M HClO4, Akemann et al. concluded that the charge under the current plateau resulted from oxidation of CO in the direct vicinity of defects and steps on the surface.[33] Following this theory, the discrepancies between results reported by different groups can be explained by differences in the quality of the single crystal electrodes used. Based on the presence of a low quantity of free sites at high CO coverages [20] and a dependence of the plateau current on the step density,[19] Lebedeva et al. considered an Eley-Rideal mechanism unlikely and instead proposed a Langmuir-Hinshelwood type mechanism with no effective freeing of sites for OH adsorption as a possible cause of the current plateau. They assume that the free sites generated by the oxidation of the first few CO molecules are reoccupied upon relaxation of the compact CO adlayer, thus keeping the number of adsorption sites available for OH nearly constant.

A scan of the hydrogen/(bi)sulfate region of our CO covered surfaces indicates that, even at saturated CO coverages, there are still some free adsorption sites present on the surface. Moreover, as the formation of different types of CO adlayers on rhodium is well known,[43, 46, 50-53] a mechanism similar to the one described for platinum electrodes, where the current plateau is associated with a relaxing compact CO adlayer, and where the second electron transfer is the rate-determining step, may apply to our system.

3.3.2.1.2. Main Oxidation Peak

The strong asymmetry of the main oxidation peaks recorded on the Rh[n(111)×(111)] electrodes indicates that the CO electrooxidation process probably does not proceed through a mean-field Langmuir-Hinshelwood type reaction, as is the case for platinum.[19, 20] As a matter of fact, it is possible that the oxidation transients, shown in Fig. 3.2a-d, consist of more than a single peak, which would indicate more than one process. When considering the transients as single peaks, several observations can be made. Firstly, the maximum current density of the oxidation peaks increases with the step density and final potential, which points to higher activities for more open surfaces (more

48

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

open = higher step density). Secondly, the shape of the peak depends greatly on the step density. The fast initial current increase followed by slow current decay recorded on surfaces with low step density [Rh(111) and Rh(554)] suggests slow diffusion of CO towards the active sites.[19] As the terrace width is decreased to 5 atoms per terrace for Rh(553), the current increases slower towards its maximum, after which it decays exponentially. For even smaller terraces, like on Rh(331), nearly symmetrical peaks are obtained. Considering the width of the terraces, the symmetry of the Rh(331) peaks and exponential decay of the Rh(553) peak are not surprising. As the terrace width is decreased the influence of diffusion of CO over the terrace towards the active sites diminishes, which results in a change of the shape of the peak and current decay. For narrow terraces the contribution of diffusion becomes negligible so that nearly symmetrical peaks are obtained.

Analysis of the hydrogen/(bi)sulfate regions of the different electrodes after a potential step experiment shows that there is still CO present on the electrode surfaces. The amount of remaining CO is found to decrease for increasing final potentials and increasing step density. This observation is somewhat surprising, since for long times the current in the transients reaches zero, signifying that there is no longer any reaction occurring. It seems that, for a given final potential, some CO adsorbed on the surface remains unreactive. As the potential is increased more of this “less reactive” CO is oxidized. Increasing the step density of the surface also results in a higher reactivity, e.g. on Rh(111) approximately 10% of the initial CO adlayer was oxidized at 0.68V after 400 seconds, whereas under the same conditions on Rh(331) more than 90% of the initial CO adlayer is removed. It is possible that the (lack of) surface mobility of CO on large terraces plays an important role in rationalizing this observation.

Fig. 3.4 illustrates how the position of the oxidation peak shifts to lower times as the final potential of the experiment is increased. Comparing the time of the oxidation peaks recorded on rhodium with those recorded on platinum [19, 20] reveals noticeably shorter times for the former, even at lower final potentials. This indicates that, at least

0.60 0.65 0.702.5

2.0

1.5

1.0

0.5

0.0

-0.5

log(t m

ax /

s)

E / V vs. RHE

Figure 3.4. Logarithm of time of the main peak versus the step potential on Rh(111) ( , solid line), Rh(554) ( , dashed line) and Rh(553) ( , dotted line).

49

Chapter 3

initially, the CO electrooxidation reaction on rhodium proceeds faster and occurs at lower potentials than on platinum. Although a fast oxidation reaction seems in contradiction with incomplete CO adlayer stripping even for long experimental times at high final potentials, it can be rationalized by either the formation of a surface deactivating species during the experiment, or the effect of anions. Since it is well known that rhodium oxidizes more readily than platinum (compare Fig. 2.3 in Chapter 2 with Fig 2d. in ref. [18]) and since the CO electrooxidation potential overlaps with the surface oxidation potential, the formation of an inactive, irreversibly adsorbed surface oxide seems a likely candidate for the deactivating species. Moreover, the potential at which surface oxidation occurs is dependent on the structure of the electrode. Thus, the shape and position of the main oxidation peak may be explained by a difference in surface oxidation rate for the stepped electrodes under investigation.

The reaction may also be blocked by the strong adsorption of (bi)sulfate. As was indicated in Chapter 2, the adsorption of (bi)sulfate is dependent on the structure of the surface and, therefore, it is possible that the shape of the main oxidation peak is likewise dependent on the adsorption on anions. We shall discuss this theory in more detail in Section 3.3.3.

A common way to obtain information on the mechanism and kinetics of the CO electrooxidation reaction is by analyzing the slope of the log(tmax) vs. Efinal plot. Lebedeva et al. pointed out that this method may be ambiguous, because the time of the current maximum depends on the kinetics of the reaction initiation as well as the reaction in the main peak area.[20] On rhodium electrodes this matter is further complicated by the possibility that not one, but two peaks are present in the transients. Despite these limitations, the analysis has been used in a number of previous studies.[31-33, 48] For our system, we obtained slopes of 55, 49, 61 and 53 mV⋅dec-1 for Rh(111), Rh(554), Rh(553) and Rh(331) respectively, all being relatively close to 60 mV⋅dec-1. For platinum, both Lebedeva et al.[20] and Santos et al.[48] reported a linear relationship between log(tmax) and Efinal with a slope of ca. 70 mV⋅dec-1. On the other hand, Love et al.[31] and Bergelin et al.[32] found non-linear relationships with slopes changing from 67 mV⋅dec-1 at low potential to 190 mV⋅dec-1 at high potential, and from 100 mV⋅dec-1 to ca.260 mV⋅dec-1, respectively. In all cases, a Tafel slope of ca. 60 mV⋅dec-1 was ascribed to a slow chemical reaction between COads and OHads in the Langmuir-Hinshelwood reaction scheme (3.3). Apparently, despite many dissimilarities with the transients recorded on platinum, the rate-determining step on rhodium surfaces could be the same slow chemical reaction.

3.3.2.1.3. Surface Mobility of CO

In contrast to numerous UHV studies claiming high mobility of CO, both the chronoamperometric results presented here and the previously reported voltammetric results (Chapter 2) indicate a low diffusion rate of CO across rhodium surfaces.[25, 26, 54] As the diffusion rate has a pronounced effect on the total reaction rate and,

50

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

accordingly, on the shape of the current-time transients, we attempted to clarify this process by dual-step chronoamperometry. In the dual-step experiments the effect of surface diffusion is investigated by allowing the system time to recover after part of the CO adlayer was oxidized. The measurements were performed on Rh(553) according to the following procedure: part of the CO adlayer was oxidized by stepping the potential from 0.1 V vs. RHE to 0.665 V for 4 seconds, after which a resting period of 5, 10, or 60 seconds at the initial potential of 0.1 V was introduced. Next, the potential is again stepped up to the final potential and held constant until the current drops to zero. In principle, the shape of the curve after the resting period is dependent on the extent to which the CO adlayer configuration has changed.

Fig. 3.5 depicts the results of the dual-step experiments together with an uninterrupted current-time curve of Rh(553) recorded at 0.665 V vs. RHE. Integration of the dual-step transients shows that the total charge recorded is the same as for the original uninterrupted current-time curve, which signifies that the same amount of CO has been oxidized in both experiments. The shape of the current-time curves in the initial 4 seconds at 0.665 V is virtually the same for each experiment, indicating identical initial reaction conditions and surface morphology. The fact that the shape of the transients recorded after the resting period closely resembles the shape of the original curve irrespective of the duration of the resting period, shows that the mobility of CO across the surface is indeed very low. Even after 60 seconds at 0.1 V the characteristic features of the main oxidation peak can still be observed. As the resting time is increased a small increase in the current density directly after the resting period (current spike visible in Fig 3.5. starting at 9, 14, and 64 seconds), as well as the current density of the main oxidation peak (with the exception of the transient recorded after a resting period of 60 seconds) can be detected.

The increasing current density directly after the resting period can be attributed to

51

0 10 20 60 70 80

0

20

40

60

80

100

1 2 3

j / µ

A cm

-2

t / s

Figure 3.5. Dual-step chronoamperometry on CO covered Rh(553). Initial potential of 0.1 V is stepped to 0.665 V for 4 seconds (arrow 1), followed by a resting period of 5, 10, and 60 seconds at 0.1 V (arrow 2). The rest of the transient is then recorded at 0.665 V vs. RHE (arrow 3).

Chapter 3

oxidation of CO, which has diffused to the reactive sites that were emptied in the first 4 seconds. As the resting time is increased, more CO diffuses to these empty sites, which results in a larger oxidation current directly after the potential is stepped up to 0.665 V. If the mobility of CO were high, this initial current density should be very high even for very short resting times and the shape of the transients following the resting period should differ greatly from the normal current-time curve. The increase in the maximum current density of the main oxidation peak for increasing resting times can be ascribed to a decrease in the coverage of the remaining CO adlayer. As the resting time is increased more CO diffuses towards the sites emptied in the initial 4 seconds, resulting in a decrease in the total coverage of the remaining CO adlayer. As it is well known that stripping of sub-saturated adlayers is easier (more adsorption sites for OH), the time to reach the maximum decreases and, thus, the maximum current density recorded increases. Consequently, the time needed to remove the remaining CO decreases, as can be seen from the decreasing width of the current-time curves recorded after the resting period. For the transient recorded after a resting period of 60 seconds diffusion of CO towards the empty sites resulted in a “depletion” of the remaining CO adlayer. Even though oxidation of this less dense adlayer is faster, the reduced number of remaining reactants leads to a lower total charge under the peak and, thus, to a lower maximum oxidation current.

From the data shown in Fig. 3.5 we can conclude that the mobility of CO on rhodium in 0.5 M H2SO4 is indeed very low. This is surprising, since the mobility of CO on platinum surfaces under similar conditions was reported to be very high.[19, 55, 56] Moreover, under UHV conditions the diffusion rate of CO on rhodium surfaces was also found to be high,[25, 26, 54] in some cases even higher than on platinum. Possible causes for the discrepancy between results from UHV and electrochemical experiments may be found in strong anion adsorption and the ease of oxidizability of rhodium surfaces. Performing similar experiments in electrolytes containing different anions like perchloric or fluoric acid may provide more insight into the diffusion properties of CO on rhodium surfaces (see Chapter 4).

3.3.2.2. Oxidation of saturated CO Adlayers on “Type I” Electrodes

The transients of the Type I electrodes, shown in figure 3.6a-d, were recorded at final potentials of 0.68 and 0.635 V vs. RHE. After a small current plateau, the transients of well-defined electrodes, with the exception of Rh(111), exhibit two peaks, one at short times (henceforth called pre-peak) and one at longer times (henceforth called main-peak). The transient recorded on Type I Rh(111) only shows a single peak at short times. Unlike for platinum electrodes, where only a single peak is observed for saturated CO adlayers,[19, 20] the existence of two peaks on rhodium electrodes hints at the occurrence of two different processes.

52

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

3.3.2.2.1. Pre-peak

With the exception of the Rh(331) transient recorded at 0.68 V, pre-peaks are visible for all surfaces at the two final potentials under investigation. In the transient of Rh(331) recorded at 0.68V the main-peak overlaps with the pre-peak, which results in a single, nearly symmetrical peak. Interestingly, for any final potential, the position of the pre-peak is relatively structure insensitive and the charge corresponding to the peak increases with the step density. Increasing the final potential of the experiment shifts the position of the peak to shorter times, which shows the effect of potential on the reaction rate.

0 50 100 150 200

0.0

0.5

1.0

1.5

2.0

80 sec

10 sec

j / µ

A cm

-2

t / s

0.0 0.1 0.2 0.3 0.40.0

0.5

1.0

1.5

2.0

80 sec

10 sec

j / µ

A c

m-2

t-0.5 / s-0.5

0 20 40 60 800.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

Figure 3.6b. Current transients of the oxidation of saturated CO adlayers on “Type I” Rh(554). Step potentials are listed in the figure. The insert shows the Cottrellian current decay of the transient recorded on Rh(554) at 0.68 V as is shown by the linear relationship between the current density and t-0.5.

80 sec

55 secj / µ

A c

m-2

t / s

0.11 0.12 0.13 0.14 0.150

1

2

3

4

5

6

7

80 sec

55 sec

j / µ

A cm

-2

t-0.5 / s-0.5

(1) 0.680 V (2) 0.635 V

(2)

(1)

Figure 3.6a. Current transients of the oxidation of saturated CO adlayers on “Type I” Rh(111). Step potentials are listed in the figure. The insert shows the Cottrellian current decay of the transient recorded on Rh(111) at 0.68 V as is shown by the linear relationship between the current and t-0.5.

(1) 0.680 V(2) 0.635 V

(2)

(1)

53

Chapter 3

0 20 40 60 80 100 120

0.0

10.0

20.0

30.0

40.0

50.0

j / µ

A cm

-2

t / s

(1) 0.680 V(2) 0.635 V

(2)

(1)

Figure 3.6c. Current transients of the oxidation of saturated CO adlayers on “Type I” Rh(553). Step potentials are listed in the figure.

The fact that the charge associated with the pre-peak increases with the step density, in combination with the structure insensitivity of the position of the peak, suggests that the process associated with the pre-peak is restricted to (the neighborhood of) the steps. This implies that adsorption of oxygenated species at a step and subsequent reaction with a neighboring CO molecule occurs at nearly the same rate on the surfaces investigated, irrespective of the step density. In fact, as CO is known to adsorb more strongly on step sites than on terrace sites,[27] it is likely that the CO molecules adsorbed at the terraces are more reactive than those adsorbed on the steps. If the step sites were the only active sites on the surface, low mobility of CO across the surface would lead to a single peak with a pronounced tailing.[19] The observation of the second main-peak can therefore not be explained solely by slow diffusion of CO from the terraces to the active step site.

As was mentioned earlier, the shape of the Type II transients already hinted at the

0 25 50 75 100 1250.0

50.0

200.0

225.0

j / µ

A cm

-2

t / s

Figure 3.6d. Current transients of the oxidation of saturated CO adlayers on “Type I” Rh(331). Step potentials are listed in the figure.

(1) 0.680 V(2) 0.635 V

(2)

(1)

54

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

presence of two overlapping peaks, which we assume to be identical to the pre- and main-peak recorded on the Type I electrodes. Because the position of the first current maximum visible in the Type II transients coincides with the position of the pre-peaks in the Type I transients for the same final potentials, we assume that the overlap is largely caused by the effect of (randomly distributed) surface defects on the main-peak. This assumption shall be discussed in more detail in the next section and Section 3.3.

3.3.2.2.2. Main-peak

Contrary to the pre-peak, the position and shape of the main peak recorded on Rh(554), Rh(553) and Rh(331) is strongly structure sensitive. At the final potentials applied, no main-peak can be detected on the well-ordered Rh(111) surface. With the exception of Rh(331), all peaks are asymmetrical in shape and show a slow current increase, followed by a current decay. On the (554) surface the current maximum is characterized by a distinct tailing, which is Cottrellian in nature (as can be seen from a plot of the current density versus 1/√t in the inset in Fig. 3.6b).[57] The same is true for the tail following the pre-peak on Rh(111) (see inset in Fig. 3.6a) .The transients recorded on Rh(553) and Rh(331) clearly do not display Cottrellian characteristics.

A Cottrellian relationship between the current density and time is expected when the reaction rate is limited by diffusion of the reacting species across the surface towards the reactive site, which seems to be the case for CO on the surfaces with low step densities, Rh(111) and Rh(554). As the terrace width is decreased from 10 to 5 and 3 atoms, i.e. from (554) to (553) and (331), the influence of diffusion on the reaction rate becomes smaller and, consequently, the shape of the current decay changes.

Similar to Type II electrodes, on Type I electrodes some CO also remains on the surface after a transient was recorded. In general, as a result of the absence of crystalline defects, the amount of CO remaining is higher for the well-ordered than for the disordered surfaces.

The question that now remains is: if the pre-peak is caused by fast OH adsorption at the steps and subsequent oxidation of neighboring CO molecules, what causes the appearance of a slower, structure sensitive main-peak, which is characterized by a shape that is apparently dependent on the diffusion rate of CO adsorbed on the surface, and on which “reactive” and “unreactive” CO can be present? The answer to this question may be found in the ease of oxidizability of rhodium surfaces. In Chapter 2, we observed that the potential at which the electrooxidation of CO occurs on rhodium coincides with the surface oxidation potential. Therefore, it is likely that during our potential-step experiments, OH does not only adsorb at the steps, but on terrace sites as well, albeit more weakly on the latter. Terrace oxidation, in combination with slow CO surface diffusion, could explain the appearance of a second peak at longer times in our current-time curves, as we will discuss in some detail in the next section.

55

Chapter 3

3.3.3. Mechanism of CO Electrooxidation on Rh[n(111)×(111)]

In this section, we will draw a picture of the mechanism of the CO electrooxidation reaction on Rh[n(111)×(111)] that emerges from our experimental data. For ease of comparison, we shall first discuss the oxidation mechanism on stepped platinum electrodes.

As CO chemisorbed on (111) terraces is believed to be more reactive than CO adsorbed on step sites [18, 27, 58, 59] and, since oxygen containing species preferably adsorb at the steps,[18] the CO electrooxidation reaction on platinum is assumed to be initiated at the steps (Fig. 3.7a), or more specifically in the troughs of the steps. After initialization, rapid diffusion of CO across the terraces towards the generated empty sites next to the steps, replenishes the reaction (see Fig. 3.7b). The remaining “less-reactive” CO adsorbed on the steps oxidizes slowly and mainly through conversion to “terrace” CO followed by rapid diffusion towards the reactive site.[21, 60] Due to the high mobility of CO on platinum, only a single, mostly symmetrical oxidation peak is observed in the current-time transient, which is best described by a mean-field Langmuir-Hinshelwood type mechanism.

On rhodium, the low CO surface mobility related to the ease of oxidizability of the surface and probably also strong anion adsorption, leads to a very different situation. After the current plateau, the reaction starts at the steps in much the same way as on platinum (see Fig. 3.7a). However, instead of CO diffusing towards the generated empty sites near the steps, our results suggest that OH is formed on the free terrace sites, which can subsequently react with neighboring CO molecules adsorbed on the terraces (see Fig. 3.7c). The reaction now proceeds at the interface between a COads and OHads phase. It is important to point out that this is the simplest explanation by which the appearance of two peaks in the current-time transients can be rationalized. Since it is generally accepted

(c)

(a) (b)

Figure 3.7. Schematic representation of the initialisation (a) and progression of the CO electrooxidation reaction on platinum (b) and rhodium (c) electrode surfaces.

56

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

that the dissociation of water to produce adsorbed OH is faster at the steps than on the terraces, the pre- and main-peak can be associated with CO reacting rapidly with OH on the steps (pre-peak) and more slowly with OH on the terraces (main peak). In other words, the reaction nucleates at the steps and then grows onto the terraces.

As the (local) geometry of the steps is nearly identical on all surfaces used, it is not surprising that the position of the pre-peak is relatively structure insensitive. The position of the main-peak, on the other hand, is strongly structure sensitive because the more open surfaces provide more nucleation sites and, most likely, also because the oxidizability of the terrace depends on its width (see oxidation region in Fig. 3.1a-d and Fig. 2.3 in Chapter 2). This assumption would explain the absence of a main-peak on Rh(111) as well. At the final potentials under investigation, dissociation of water (and possibly subsequent surface oxidation) on large (111) terraces is unfavorable and, therefore, only one peak is observed with a pronounced tailing due to slow diffusion of CO across the terraces. On Rh(331) the rates of OH adsorption on the steps and terraces are so close together that the reaction on the terraces will occur immediately after free sites are generated by the reaction on the steps, which leads to the nearly symmetrical peaks observed in the current-time transients.

The mechanism described above provides an attractive explanation for the appearance of the pre- and main-peak as it explains the difference in position and structure sensitivity of the peaks. However, it does not account for the fact that there is still some CO left on the surface of the electrodes, even after the current recorded in the transients has dropped to zero. It also does not describe the change in current decay from Cottrellian to exponential decay as the terrace width is decreased from 10 to 5 (Rh(554) to Rh(553)) atoms/terrace. Without affecting the validity of the mechanism described above, we can propose several tentative explanations for these observations. The explanation for the presence of CO after an experiment may be found in the strong anion adsorption on rhodium surfaces. As the reaction progresses, (bi)sulfate can adsorb in the generated empty sites and hinder the mobility of CO or even block sites for OH adsorption. In Chapter 2 we pointed out that increasing the amount of steps and defect sites on a rhodium surface leads to a decrease in the maximum (bi)sulfate coverage, which may account for the apparent higher mobility of CO on the more open surfaces.

Another explanation for the presence of un-reacted CO can be found in the formation of surface oxides. Perhaps during the slow oxidation of CO on surfaces with wide terraces, surface oxides can be formed before all CO has been stripped. These inactive surface oxides can act as an effective barrier for either the adsorption of OH, or the electrooxidation reaction. On highly stepped surfaces the rates of both surface oxidation and CO electrooxidation are higher, which may result in a nearly complete stripping of the original adlayer.

57

Chapter 3

3.4. Conclusion

The CO electrooxidation reaction on single crystal Rh[n(111)×(111)] type electrodes in 0.5 M H2SO4 was investigated by means of chronoamperometric experiments. All the current-time transients recorded on Rh(111), Rh(554), Rh(553) and Rh(331) are characterized by an apparent zero-order current plateau, followed by the main oxidation region, which consists of two peaks, a pre-peak at short times and a main-peak at longer times.

Similar to stepped platinum electrodes, the origin of the current plateau on rhodium is ascribed to a Langmuir-Hinshelwood type reaction between COads and OHads, with no effective freeing of sites for OH adsorption. The value for the Tafel slope of the plateau region of ca. 45 mV⋅dec-1 indicates a second electron transfer process as rate determining, rather than a slow chemical step, as was suggested for platinum.

The presence of a structure insensitive pre-peak at short times, and a strongly structure sensitive main-peak at longer times in the transients recorded implies the existence of more than one process. In addition, for the peaks to be clearly distinguishable, the mobility of the reacting species on the surface has to be low and the quality of the surface high. Results obtained from dual step experiments indeed point at a low surface mobility of CO. This low diffusion rate of CO on rhodium electrodes may be ascribed to the strong influence of anions or the formation of inactive surface oxides. From the data presented here and results reported in Chapter 2 we conclude that the presence of the pre- and main-peaks can both be ascribed to CO reacting with OH, with the latter adsorbing fast at the steps and more slowly on the terraces. The overall mechanism of the CO electrooxidation reaction on Rh[n(111)×(111)] single crystal electrodes can now be visualized by a nucleation and growth Langmuir-Hinshelwood type mechanism (contrary to the mean-field Langmuir-Hinshelwood type mechanism reported for platinum). The reaction nucleates at the steps by OH adsorption in the trough of the steps followed by reaction with a neighboring terrace-adsorbed CO molecule, giving rise to the pre-peak. Next, the reaction grows on the terraces by adsorption of OH at the generated empty terraces sites and subsequent reaction with the remaining CO, resulting in a main-peak. For surfaces with wide terraces, OH adsorption at the terrace sites is energetically less favorable than for surfaces with narrow terraces, which leads to incomplete oxidation of the CO adlayer and to main-peaks with a Cottrellian type current decay. As the step density is increased, the main-peak shifts to lower times and overlaps completely with the pre-peak, leading to nearly symmetrical peaks.

As the adsorption of (bi)sulfate may play an important role in the catalysis of CO oxidation on rhodium surfaces, experiments on the CO electrooxidation reaction on Rh in the absence of specifically adsorbing anions are desirable and will be reported in Chapter 4.(ref. [61]) Furthermore, the validity of the mechanism described above will be verified by kinetic Monte Carlo simulations, the results of which are presented in Chapter 5.(ref. [62])

58

CO oxidation on stepped Rh[n(111)×(111)] electrodes: a chronoamperometric study

Acknowledgement: This research was supported by the Netherlands Foundation for Scientific Research (NWO). Special thanks go to Professor Juan Feliu of the Departamento de Química-Física at University of Alicante in Spain for supplying the high quality stepped single crystal electrodes.

References

[1] L. Carrette, K. A. Friedrich, and U. Stimming, Fuel Cells 1 (2001) 5. [2] A. Hamnett, Catal. Today. 38 (1997) 445. [3] N. M. Markovic and P. N. Ross, Surf. Sci. Rep. 45 (2002) 117. [4] S. Wasmus and A. Kuver, J. Electroanal. Chem. 461 (1999) 14. [5] J. T. Kummer, J. Phys. Chem. 90 (1986) 4747. [6] S. B. Schwartz, L. D. Schmidt, and G. B. Fisher, J. Phys. Chem. 90 (1986) 6194. [7] S. H. Oh, G. B. Fisher, J. E. Carpenter, and D. W. Goodman, J. Catal. 100 (1986) 360. [8] C. H. F. Peden, D. W. Goodman, D. S. Blair, P. J. Berlowitz, G. B. Fisher, and S. H. Oh, J. Phys.

Chem. 92 (1988) 1563. [9] D. W. Goodman and C. H. F. Peden, J. Phys. Chem. B 90 (1986) 4839. [10] J. Siera, F. Rutten, and B. E. Nieuwenhuys, Catal. Today. 10 (1991) 353. [11] M. Bowker, Q. Guo, and R. W. Joyner, Surf. Sci. 280 (1993) 50. [12] A. Baraldi, S. Lizzit, D. Cocco, G. Comelli, G. Paolucci, R. Rosei, and M. Kiskinova, Surf. Sci.

385 (1997) 376. [13] A. Baraldi, L. Gregoratti, G. Comelli, V. R. Dhanak, M. Kiskinova, and R. Rosei, App. Surf. Sci.

99 (1996) 1. [14] B. A. Gurney, L. J. Richter, J. S. Villarrubia, and W. Ho, Journal of Chemical Physics 87 (1987)

6710. [15] C. Castellarin Cudia, S. W. Hla, G. Comelli, Z. Sljivancanin, B. Hammer, A. Baraldi, K. C.

Prince, and R. Rosei, Phys. Rev. Lett. 87 (2001) 196104/1. [16] M. J. P. Hopstaken, W. J. H. Van Gennip, and J. W. Niemantsverdriet, Surf. Sci. 433-435 (1999)

69. [17] M. J. P. Hopstaken and J. W. Niemantsverdriet, J. Chem. Phys. 113 (2000) 5457. [18] N. P. Lebedeva, M. T. M. Koper, E. Herrero, J. M. Feliu, and R. A. van Santen, J. Electroanal.

Chem. 487 (2000) 37. [19] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Phys. Chem. 106 (2002)

12938. [20] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Electroanal. Chem. 524-

525 (2002) 242. [21] M. T. M. Koper, N. P. Lebedeva, and C. G. M. Hermse, Faraday Discuss. 121 (2002) 301. [22] R. Gomez, J. M. Orts, J. M. Feliu, J. Clavilier, and L. H. Klein, J. Electroanal. Chem. 432 (1997)

1. [23] T. H. M. Housmans, J. M. Feliu, and M. T. M. Koper, J. Electroanal. Chem. 572 (2004) 79. [24] S. Gilman, J. Phys. Chem. 68 (1964) 70. [25] J. V. Barth, Surf. Sci. Rep. 40 (2000) 75. [26] E. G. Seebauer, A. C. F. Kong, and L. D. Schmidt, J. Chem. Phys. 88 (1988) 6597. [27] J. T. Yates, Jr., J. Vac. Sci Technol. A 13 (1995) 1359. [28] R. I. Masel, Principles of adsorption and reaction on solid surfaces, John Wiley & Sons Inc, New

York, 1996. [29] W. Schmickler, Interfacial Electrochemistry, Oxford University press, Oxford, 1996. [30] M. T. M. Koper, A. P. J. Jansen, and J. J. Lukkien, Electrochim. Acta. 45 (1999) 645. [31] B. Love and J. Lipkowski, ACS Symposium Series 378 (1988) 484. [32] M. Bergelin, E. Herrero, J. M. Feliu, and M. Wasberg, J. Electroanal. Chem. 467 (1999) 74. [33] W. Akemann, K. A. Friedrich, and U. Stimming, J. Chem. Phys. 113 (2000) 6864. [34] J. M. Feliu, J. M. Orts, R. Gomez, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 372 (1994)

265. [35] J. Clavilier, D. Armand, S. G. Sun, and M. Petit, J. Electroanal. Chem. 205 (1986) 267. [36] J. M. Feliu, J. M. Orts, A. Fernandez-Vega, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 296

(1990) 191. [37] M. Hourani and A. Wieckowski, J. Electroanal. Chem. 227 (1987) 259.

59

Chapter 3

60

[38] M. Hourani and A. Wieckowski, J. Electroanal. Chem. 244 (1988) 147. [39] M. Hourani, M. Wasberg, C. Rhee, and A. Wieckowski, Croatica Chemica Acta 63 (1990) 373. [40] M. Wasberg, M. Hourani, and A. Wieckowski, J. Electroanal. Chem. 278 (1990) 425. [41] P. Zelenay, G. Horanyi, C. K. Rhee, and A. Wieckowski, J. Electroanal. Chem. 300 (1991) 499. [42] M. J. Weaver, S. C. Chang, L. W. H. Leung, X. Jiang, M. Rubel, M. Szklarczyk, D. Zurawski, and

A. Wieckowski, J. Electroanal. Chem. 327 (1992) 247. [43] Y. E. Sung, S. Thomas, and A. Wieckowski, J. Phys. Chem. 99 (1995) 13513. [44] L. W. H. Leung and M. J. Weaver, J. Phys. Chem. 93 (1989) 7218. [45] L. W. H. Leung, S. C. Chang, and M. J. Weaver, J. Chem. Phys. 90 (1989) 7426. [46] S. L. Yau, X. Gao, S. C. Chang, B. C. Schardt, and M. J. Weaver, J. Am. Chem. Soc. 113 (1991)

6049. [47] S. C. Chang and M. J. Weaver, J. Electroanal. Chem. 285 (1990) 263. [48] E. Santos, E. P. M. Leiva, and W. Vielstich, Electrochim. Acta. 36 (1991) 555. [49] E. Herrero, J. M. Feliu, S. Blais, Z. Radovic-Hrapovic, and G. Jerkiewicz, Langmuir 16 (2000)

4779. [50] P. Cernota, K. Rider, H. A. Yoon, M. Salmeron, and G. Somorjai, Surf. Sci. 445 (2000) 249. [51] V. R. Dhanak, A. Baraldi, G. Comelli, G. Paolucci, M. Kiskinova, and R. Rosei, Surf. Sci. 295

(1993) 287. [52] A. Beutler, E. Lundgren, R. Nyholm, J. N. Andersen, B. Setlik, and D. Heskett, Surf. Sci. 371

(1997) 381. [53] H. Over, S. Schwegmann, G. Ertl, D. Cvetko, V. De Renzi, L. Floreano, R. Gotter, A. Morgante,

M. Peloi, F. Tommasini, and S. Zennaro, Surf. Sci. 376 (1997) 177. [54] T. W. Root, L. D. Schmidt, and G. B. Fisher, Surf. Sci. 150 (1985) 173. [55] K. A. Friedrich, K. P. Geyzers, U. Stimming, and R. Vogel, Z. Phys. Chem. 208 (1999) 137. [56] H. A. Gasteiger, N. Markovic, P. N. Ross, Jr., and E. J. Cairns, J. Phys. Chem. 98 (1994) 617. [57] A. J. Bard and L. R. Faulkner, Electrochemical Methods: Fundamentals and Applications, John

Wiley & Sons, New York, 2001. [58] J. Xu and J. T. Yates, Jr., J. Chem. Phys. 99 (1993) 725. [59] J. Xu, P. Henriksen, and J. T. Yates, Jr., J. Chem. Phys. 97 (1992) 5250. [60] N. P. Lebedeva, A. Rodes, J. M. Feliu, M. T. M. Koper, and R. A. van Santen, J. Phys. Chem. B

106 (2002) 9863. [61] T. H. M. Housmans and M. T. M. Koper, Electrochem. Comm. 7 (2005) 581. [62] T. H. M. Housmans and M. T. M. Koper, to be submitted (2005)

CO oxidation on stepped Rh[n(111)×(111)] single crystal electrodes: anion effects on CO surface mobility

Abstract

The influence of anion adsorption on the CO electrooxidation reaction on stepped Rh[n(111)×(111)] electrodes was investigated by comparing voltammetric and chronoamperometric data obtained in 0.5 M HClO4 with previously published results obtained in 0.5 M H2SO4. Compared to sulfuric acid media, complete stripping of the CO adlayer requires fewer cycles in perchloric acid and the resulting stripping voltammetry peaks are shifted to considerably lower potentials, attesting to the reduced influence of more weakly adsorbed anions. The absence of a shoulder prior to the main oxidation peak and the higher symmetry of the main-peaks implies that the electrooxidation reaction in perchloric acid is not only faster than in sulfuric acid but probably also not diffusion limited, which suggests the mean-field approximation as the best mathematical model for the reaction kinetics rather than nucleation and growth. The chronoamperometric transients recorded at various potentials show only a single oxidation peak with a slight tailing at longer times. Only the Rh(111) transients display tailing, which is Cottrellian in nature. The surface diffusion coefficient of CO deduced from the Cottrell plots is more than 4 orders of magnitude larger in perchloric than in sulfuric acid (HClO4: 1⋅10-12 <D< 2⋅10-11 cm2⋅s-1 vs. H2SO4: 1⋅10-16 <D< 8⋅10-16 cm2⋅s-1), which also suggests a mean-field model for HClO4 rather than nucleation and growth. Apparently, specific anion adsorption on rhodium surfaces not only affects the rate, but also the dynamics of the CO electrooxidation reaction. Thus, by varying the adsorption strength of the anion, we can, in principle, influence the diffusion rate of adsorbates on the surface and, therefore, the reaction dynamics and the overall reaction rate.

This chapter is published as T.H.M. Housmans, M.T.M. Koper, Electrochem. Comm. 7 (2005)581

Chapter 4

4.1. Introduction

Understanding the overall reactivity of surface-catalyzed reactions requires insight into the mobility of adsorbed species. In a Langmuir-Hinshelwood type mechanism the overall rate of the reaction depends critically on the extent to which reactants are mixed on the surface and, accordingly, on their respective rate of diffusion.[1] In electrocatalysis the issue of surface mobility has recently received considerably attention in relation to the mechanism of CO electrooxidation.[2-12]

The oxidation reaction of CO on noble metal surfaces is generally accepted to be of Langmuir-Hinshelwood type with the overall mechanism represented by the following steps:[13]

H2O + * OHads + H+ + e- (4.1) COads + OHads COOHads (4.2) COOHads CO2 + H+ + e- + 2* (4.3)

Where “*” denotes a free surface site, reaction 4.1 and 4.3 are assumed to be fast, and reaction 2 is rate determining (at relatively high overpotentials).

Because the rate of reaction 4.2 depends on the spatial distribution of the reactants, the surface mobility of the adsorbates is of great importance. Analytical expressions for the total reaction rate can be obtained in two limiting cases, namely the mean-field approximation and the nucleation and growth model. The mean-field approximation assumes perfectly mixed reactants and a reaction rate that is proportional to the average coverages of the reacting species,[1] implying a surface diffusion rate higher than the reaction rate. In the nucleation and growth model reacting species are assumed to be immobile on the surface. The reaction nucleates by adsorption of OH at ”special” sites, usually defects or steps, and proceeds only at the interface between two reacting phases, causing the formation and growth of islands.[14]

A noble metal to have recently gained renewed interest with respect to CO electrooxidation is rhodium.[15-17] Even though rhodium may not be a particularly attractive metal for application in low temperature fuel cells, it displays CO oxidation characteristics, which deviate considerably from those of platinum and which, therefore, make it an interesting surface to study, especially in relation to the still much debated issue of CO surface mobility.[18] Recently we studied this reaction on stepped Rh[n(111)×(111)] electrodes in 0.5 M H2SO4 and concluded that, contrary to what was reported for stepped Pt electrodes,[2] the kinetics of the reaction can best be described by the nucleation and growth model rather than the mean-field approximation due to a low CO surface mobility.(Chapters 2 and 3, ref. [15, 16]) Moreover, it was suggested that the low mobility of CO on rhodium might be caused by the influence of the strongly adsorbed (bi)sulfate anion.

In this chapter we investigate in more detail the influence of specifically adsorbing anions on the catalysis of the CO electrooxidation reaction on stepped rhodium electrodes of [n(111)×(111)] orientation. CO stripping and chronoamperometric experiments were performed in 0.5 M HClO4 and the results will be compared to those

62

CO oxidation on stepped Rh[n(111)×(111)] electrodes: anion effects

obtained previously in 0.5 M H2SO4.(Chapters 2 and 3) Results presented herein show that, like the electrooxidation of CO on Pt surfaces, the rate of the oxidation reaction is higher in perchloric acid than in sulfuric acid.[19, 20] However, contrary to sulfuric acid, current-time transients recorded in perchloric acid display only a single oxidation peak followed by a tail at longer times. For Rh(111) this tail is Cottrellian in nature (similar to sulfuric acid media), which in principle could be taken as evidence for a surface diffusion limited reaction. We will interpret these results in terms of an anion-dependent CO surface diffusion coefficient.

4.2. Experimental Setup

The working electrodes used in this study were rhodium bead-type single crystal electrodes of Rh[n(111)×(111)] (identical to Rh[(n-1)(111)×(111)]) orientation (Rh(331), Rh(553), Rh(554), and Rh(111) with n=3, n=5, n=10, and n=200-500 respectively). The electrodes were prepared as described in ref. [21], and oriented, cut and polished according to the Clavilier method.[22] Prior to each measurement the single crystal electrode was flame annealed and cooled down to room temperature in an argon (Hoekloos, N50)-hydrogen atmosphere (ratio 3:1), after which it was transferred to the electrochemical cell under protection of a droplet of deoxygenated water.

A special electrochemical cell, described in ref. [3], contained a small movable spoon over an electrolyte reservoir, which allowed dosing of CO at open circuit potential (ocp) from a saturated CO solution without dissolving CO in the blank electrolyte. The CO coverage was checked by cycling in the hydrogen adsorption/desorption region. The electrochemical cell was cleaned by boiling in a 1:1 mixture of concentrated sulfuric and nitric acid, followed by repeated boiling (four times) with ultra-pure water (Millipore MilliQ gradient A10 system, 18.2 MΩcm, 2 ppb total organic carbon). The blank electrolyte, 0.5 M HClO4, was prepared with concentrated perchloric acid (Merck, "Suprapur") and ultra-pure water. During measurements the blank electrolyte was deoxygenated with argon (N50) and the electrolyte in the container above the spoon was saturated with CO gas (Hoekloos, N47).

A coiled platinum wire was used as a counter electrode and the reference electrode was a reversible hydrogen electrode (RHE) connected via a Luggin capillary. Measurements were performed at room temperature (22 ºC) with a computer-controlled Autolab PGSTAT20 potentiostat (Ecochemie).

4.3. Results and Discussion

4.3.1. Cyclic Voltammetry in Absence of CO

In sulfuric acid media it is difficult to establish the system cleanliness and surface quality of a Rh[n(111)×(111)] electrode from the blank cyclic voltammogram (BCV) due to the absence of well-defined, reversible anion and hydrogen adsorption/desorption

63

Chapter 4

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

-150

-100

-50

0

50

100

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8-100

-75

-50

-25

0

25

50

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8-100

-80

-60

-40

-20

0

20

40

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8-80

-60

-40

-20

0

20

40j / µ

A cm

-2

E / V vs. RHE

(d) (c)

(b) (a)

Figure 4.1. Cyclic voltammograms of Rh(111) (a), Rh(554) (b), Rh(553) (c), and Rh(331) (d) in 0.5 M HClO4 at a scan rate of 20 mV⋅s-1.

peaks.(see Chapters 2 and 3 and ref. [15, 16]) However, in perchloric acid the cyclic voltammogram of Rh(111) is characterized not only by the reduction of perchloric acid at low potentials but also by a “butterfly” at approximately 0.64 V vs. RHE.[23-28] This butterfly is ascribed to the adsorption and desorption of perchloric acid and oxygen containing species (OHads) in the same manner as (bi)sulfate on Pt(111). Similar to platinum, the stability and height of the peaks are indicative for the cleanliness of the system and quality of the surface.[29]

Figure 4.1a-d shows the BCVs of Rh(111), Rh(554), Rh(553) and Rh(331) recorded in 0.5 M HClO4. The Rh(111) BCV displays the characteristic perchloric acid reduction and a well-developed, stable butterfly at 0.64 V vs. RHE indicating good surface quality and system cleanliness. As the step density is increased from Rh(111) to Rh(331), the perchlorate reduction reaction is enhanced while the butterfly feature decreases. The increase in reduction current density is not surprising as Rhee et al. already observed that more “open” surfaces are more active towards the reduction of perchlorate than Rh(111).[30] Analogous to (bi)sulfate adsorption on platinum electrodes, the disappearance of the reversible peaks at 0.64 V when going from Rh(111) to Rh(331) can be ascribed to disruption of the anion disorder/order transition by the narrowing of the terraces.[31] The figures also clearly demonstrate that the surface oxidation potential decreases considerably for increasing step density.

64

CO oxidation on stepped Rh[n(111)×(111)] electrodes: anion effects

0.5 0.6 0.7 0.8 0.9

0

100

200

300

400

500

j / µ

A cm

-2

E / V vs. RHE

Rh(111) Rh(554) Rh(553) Rh(331)

Figure 4.2. CO stripping voltammetry of saturated CO adlayers on Rh(111) (solid thin line), Rh(554) (dashed line), Rh(553) (dotted line), and Rh(331) (solid thick line) in 0.5 M HClO4 at a scan rate of 20 mV⋅s-1.

4.3.2. Cyclic voltammetry: saturated CO Adlayer Oxidation

The CO stripping voltammograms of Rh(111), Rh(554), Rh(553) and Rh(331) in 0.5 M HClO4 can be found in Fig 4.2. In order to avoid poisoning of the electrode surface by chloride formed in the reduction of perchlorate, the electrodes were flame annealed, cooled down in argon-hydrogen and subsequently submerged in a CO saturated solution prior to the stripping voltammetry experiments. Analogous to CO stripping on Pt[n(111)×(111)] electrodes,[32] the onset of the oxidation peaks lies at lower potentials in perchloric than in sulfuric acid (compare Fig. 4.2 to Fig. 2.5a in Chapter 2), which demonstrates the strong influence of specific anion adsorption on the oxidation rate. The reduced influence of perchloric acid on the CO electrooxidation reaction compared to sulfuric acid is also evident from the number of scans required to completely strip a saturated CO adlayer from the electrode surfaces. Complete stripping of the CO adlayer requires 2-3 cycles for Rh(111), 1-2 cycles for Rh(554) and only 1 scan for Rh(553) and Rh(331), compared to 30-40, 5, 3 and 2 scans in sulfuric acid, respectively.

However, compared to sulfuric acid media, most noticeable are the absence of a shoulder prior to the main oxidation peak and the higher symmetry of the main-peak. Also, no tailing can be detected at the high potential side of the oxidation peaks. These observations suggest that CO stripping in HClO4 is not only faster than in H2SO4 but possibly also not limited by surface diffusion. In perchloric acid media the overpotential necessary for oxidation of a saturated CO adlayer increases in the order Rh(553)<Rh(554)<Rh(111) as deduced from both onset potential as well as main oxidation peak. This trend was also observed for analogous platinum electrodes in sulfuric acid,[2] but was not observed for rhodium electrodes in sulfuric acid electrolyte.(Chapter 2, [16]) The positive shift in the potential of the CO oxidation peak recorded on the (331) surface may be explained by competition between the formation of

65

Chapter 4

surface oxygen containing species and the electrooxidation of CO. We shall discuss this observation in more detail in the next section.

4.3.3. Chronoamperometry: saturated CO Adlayer Oxidation

The kinetics of the CO electrooxidation reaction on the well-ordered rhodium electrodes (referred to as “Type I” electrodes in Chapters 2 and 3) in perchloric acid was also investigated by chronoamperometry. The final potentials examined were 0.600; 0.635 and 0.680 V vs. RHE and the corresponding current-time transients are shown in Fig. 4.3a-d.

With the exception of Rh(553), all transients recorded in perchloric acid display a single oxidation peak and slight tailing at higher times. Only for the (111) surface the tailing seems Cottrellian in nature at all step potentials (see insets in Fig. 4.3a-d). As on Rh(553) at 0.636 V vs. RHE and Rh(331), surface oxidation plays a considerable role, we believe that the straight line in the Cottrell plot should not be ascribed to diffusion of CO, but rather to some other process, most likely surface oxidation. Compared to the transients recorded on the same electrodes in 0.5 M H2SO4 the oxidation peaks are shifted towards shorter times and the maximum current densities are noticeably higher (see Fig. 3.6a-d in Chapter 3), which signifies a reduced influence of the perchlorate anion on the reaction rate. Moreover, the Rh[n(111)×(111)] transients recorded in sulfuric acid media almost exclusively exhibit multiple peaks, while here only single

0 20 40 60 80 100

0

25

175

200

225

j / µ

A cm

-2

t / s

0.05 0.10 0.15 0.20 0.25 0.30

2

4

6

8

100.600 V

120 s25 sec

j / µ

A c

m-2

1/t0.5

0.10 0.15 0.20 0.252.0

4.0

6.0

8.0

10.0

12.00.635 V

120 s

20 secj /

µA

cm

-2

1/t0.5

Figure 4.3a. Current transients of the oxidation of saturated CO adlayers on a “Type I” Rh(111) electrode. Step potentials are listed in the figure. The insets show the Cottrell plots of the recorded transients at 0.600 and 0.635 V vs. RHE.

(1) 0.680 V (2) 0.635 V (3) 0.600 V

(3)

(2)

(1)

66

CO oxidation on stepped Rh[n(111)×(111)] electrodes: anion effects

peaks were recorded.(Chapter 3, ref. [15]) These observations imply that on rhodium specific anion adsorption not only affects the rate (as is the case for platinum), but also the surface dynamics of the CO electrooxidation reaction.

A single oxidation peak indicates a mechanism similar to that on stepped platinum electrodes, where reaction 4.2 (see introduction) occurs exclusively at the steps and the CO adsorbed on the terraces diffuses towards the reactive site.[2, 33-35] The Cottrellian nature of the small tail located at longer times in the Rh(111) transient suggests that, like in sulfuric acid media, at low coverages (after part of the CO adlayer has been oxidized) the diffusion from the terraces towards the steps becomes rate limiting.

The Cottrell equation as it can be found in textbooks (Eq. 4.1) [36] normally

0 20 40 60 80

0

50

300

350

400

j / µ

A cm

-2

t / s

(1) 0.680 V(2) 0.635 V(3) 0.600 V

(2)

(3)

0.1 0.2 0.3 0.4-5

0

5

10

15

20

25

0.635 V

0.600 V

130 sec

12 sec

j / µ

A cm

-2

1/t0.5

0 10 20 300

50

100

150

200

250

500

550

600

j / µ

A cm

-2

t / s

(1) 0.680 V(2) 0.635 V(3) 0.600 V

(2)

(3) 0.1 0.2 0.3 0.4

0

10

20

30

40

50

0.600 V

0.635 V130 sec

12 sec

j / µ

A c

m-2

1/t0.5

(1)

Figure 4.3b. Current transients of the oxidation of saturated CO adlayers on a “Type I” Rh(554) electrode. Step potentials are listed in the figure. The inset shows the Cottrell plots of the recorded transients at 0.600 and 0.635 V vs. RHE.

(1)

Figure 4.3c. Current transients of the oxidation of saturated CO adlayers on a “Type I” Rh(553) electrode. Step potentials are listed in the figure. The inset shows the Cottrell plots of the recorded transients at 0.600 and 0.635 V vs. RHE. 0.65 V

67

Chapter 4

0 10 20 30 400

100

200

300

400

500

600

700

800

j / µ

A cm

-2

t / s

0.1 0.2 0.3 0.4-10

0

10

20

30

40

50

60

0.600 V

0.635 V130 sec

12 sec

j / µ

A c

m-2

1/t0.5

(1) 0.680 V(2) 0.635 V(3) 0.600 V

(3)

(2)

(1)

Figure 4.3d. Current transients of the oxidation of saturated CO adlayers on a “Type I” Rh(331) electrode. Step potentials are listed in the figure. The inset shows the Cottrell plots of the recorded transients at 0.600 and 0.635 V vs. RHE.

applies to systems where a reactant diffuses from the bulk electrolyte towards the catalytically active surface.

tDnFACtiπ0)( = (Eq. 4.1)

In this “3D” model the current, i(t), is generated by the diffusion (diffusion coefficient D / cm2⋅s-1) of a reactant (bulk concentration, C0 / mol⋅cm-3) through the geometric surface area A of the flat catalyst surface. Applied to a surface adsorbed species reacting at a one-dimensional step, this equation can be rewritten as:

tD

LnFC

jt

b

π= (Eq. 4.2)

The current density, j, is generated by diffusion of the adsorbed reactant, with concentration Cb (in mol⋅cm-2) at large distances from the reactive site, over the terraces towards the step (distance between steps, Lt). This equation is valid only for surfaces where Lt→∞, i.e. in our case the Rh(111) surface. (The derivation of Eq. 4.2 has been relegated to the Appendix) If the length of the terraces and the concentration of reactant remaining on the surface are known, a value for the diffusion coefficient of CO on the Rh(111) surface in both sulfuric as well as perchloric acid can be obtained from the Cottrell plots. For the calculation of Lt and Cb we refer to the Appendix, while the calculated diffusion coefficient together with the values assumed for Lt and Cb can be found in Table 4.1.

Depending on the values use for Cb and Lt, the diffusion coefficient of CO on Rh(111) in perchloric acid lies in the interval 1⋅10-12 - 2⋅10-11 cm2⋅s-1, while in sulfuric acid it is nearly four orders of magnitude lower (1⋅10-16 - 8⋅10-16 cm2⋅s-1). These values are considerably lower than the CO diffusion coefficient of 10-9 cm2⋅s-1 reported for CO on Pt(111) in vacuum (at high coverages and 300 K) [37] and lower (as it should be) than the minimum diffusion coefficient of 10-11 cm2⋅s-1 reported for Pt(111) in H2SO4.[2, 18]

68

CO oxidation on stepped Rh[n(111)×(111)] electrodes: anion effects

Slope of Cottrell plot / A⋅cm-2⋅s-½ Cb /mol⋅cm2 Lt / cm D / cm2⋅s-1

4.6⋅10-6 9.8⋅10-13 1.99⋅10-9 1.2⋅10-5 6.1⋅10-12 4.6⋅10-6 2.8⋅10-12

0.600 V 4.67⋅10-5

1.19⋅10-9 1.2⋅10-5 1.8⋅10-11 4.6⋅10-6 7.7⋅10-13 1.99⋅10-9 1.2⋅10-5 4.8⋅10-12 4.6⋅10-6 1.8⋅10-12

HClO4

0.635 V 4.12⋅10-5 1.31 ⋅10-9

1.2⋅10-5 1.1⋅10-11 4.6⋅10-6 1.3⋅10-16

1.99⋅10-9 1.2⋅10-5 8.0⋅10-16

4.6⋅10-6 1.3⋅10-16 H2SO4[15] 0.635 V 5.31⋅10-7

1.96⋅10-9 1.2⋅10-5 8.2⋅10-16

Table 4.1. Calculation of the CO surface diffusion coefficient. From the above results, we conclude that the adsorption strength of anions on

rhodium surfaces has a significant influence on the mobility of the CO adsorbate. It is conceivable that a similar situation applies to platinum surfaces as well. In this respect, it would be interesting to perform potential step experiments on platinum electrodes in the presence of anions, which adsorb more strongly than (bi)sulfate, in order to see if indeed a change in the dynamics of the reaction can be observed. On the other hand, it must be avoided that the ion adsorbs so strongly that it replaces CO from the surface.

As mentioned earlier, the overpotential needed for the electrooxidation of CO on rhodium surfaces increases in the order Rh(553)<Rh(554)<Rh(111). However, for Rh(331) the onset of the CO oxidation lies between that of Rh(554) and Rh(111). It was suggested that, as the surface oxidation potential decreases for increasing step density, this trend may be ascribed to surface oxidation. This assumption was verified by a potential step experiment in which the current-time transient was terminated at the maximum of the oxidation peak, and immediately afterwards a scan was recorded from 0.70 V in the negative going direction. The resulting voltammogram indeed shows a surface oxide reduction peak characteristic for Rh(331). If a complete transient is recorded immediately followed by a scan in the negative going direction, the oxide reduction peak is considerably larger, indicating that on Rh(331) surface oxidation indeed occurs during the potential-step experiments and, thus, may hinder the oxidation of CO. The simultaneous occurrence of CO oxidation and surface oxidation is probably also the cause of the multiple peaks observed in the (553) transients.

Unfortunately, the symmetry and position of the peaks in Fig. 4.3a-d do not allow for an accurate determination of which model, mean-field approximation or nucleation and growth, describes the reaction kinetics best. However, the fact that the graphs shown here resemble more closely transients recorded on analogous platinum electrodes [2] than the transients recorded on rhodium electrodes in 0.5 M H2SO4 (Chapter 3, ref. [15])

69

Chapter 4

strongly suggests that the best model for the reaction kinetics lies closer to mean-field approximation than to the nucleation and growth model. Moreover, the large differences in diffusion coefficients for perchloric acid and sulfuric acid media corroborate this conclusion. Since the diffusion rate of CO on Rh(111) in perchloric acid is not much smaller than the minimum value reported for CO on Pt(111) in sulfuric acid,[2, 18] the mean-field approximation indeed seems the most plausible mechanism, whereas the much lower diffusion coefficient recorded in sulfuric acid media strongly suggests nucleation and growth as the best model. It is possible that if a less strongly adsorbing anion than perchlorate were used, the characteristics of the CO electrooxidation on rhodium surfaces may approach or equal those of platinum surfaces.

As a last comment, we would like to point out that in a true nucleation and growth mechanism no Cottrellian decay should be observed in the potential step transients. Based on the results reported in the Chapters 2 and 3 and the data presented here, we would like to give a tentative explanation for the presence of the Cottrellian current decay in the transients of Rh(111) (in both HClO4 as well as H2SO4) and Rh(554) (in H2SO4). The appearance of two peaks in both the cyclic voltammetry and current-time transients was explained by “immobile” CO reacting with OH, which adsorbs fast at the steps and more slowly at the terrace sites. On Rh(111), if we assume that below 0.85 V no appreciable amounts of OH are formed on the terraces and only step adsorbed OHads is formed, even very slow diffusion of CO from the terrace to the step sites would be visible as a Cottrellian decay. However, on Rh(554) the terrace sites do in fact oxidize at these potentials and, thus, this assumption cannot adequately explain the Cottrellian decay observed in the transients recorded. This discrepancy may be explained if we assume partial oxidation of the terraces on Rh(554). It is possible that the steps induce oxidation (formation of OHads but also surface oxides) of neighboring terrace atoms. As this surface oxidation progresses from the (110) steps to the (111) terrace sites (resulting in a nucleation and growth mechanism), the more stable (111) sites become increasingly difficult to oxidize, at which point the nucleation-and-growth-like response of the current-time transient changes to a Cottrellian decay. On the other surfaces, the small terraces can completely oxidize and therefore no Cottrellian decay is observed.

4.4. Conclusion

The effect of anion adsorption on the CO electrooxidation reaction on single crystal rhodium electrodes of [n(111)×(111)] orientation was investigated by performing cyclic voltammetry and chronoamperometry in 0.5 M HClO4. The obtained results were compared to results of previously preformed studies in 0.5 M H2SO4.(Chapters 2 and 3, ref. [15, 16])

Compared to sulfuric acid media, the stripping voltammetry peaks in perchloric acid are shifted to lower potentials, attesting to the reduced influence of the weaker adsorbed perchlorate anions compared to more strongly bound (bi)sulfate anions. In addition, in HClO4 complete stripping of the CO adlayer requires considerably less

70

CO oxidation on stepped Rh[n(111)×(111)] electrodes: anion effects

cycles than in sulfuric acid. The absence of a shoulder prior to the main oxidation peak (as was recorded in sulfuric acid) and the higher symmetry of the main-peaks suggests that the electrooxidation reaction in perchloric acid is not only faster than in sulfuric acid but probably also not diffusion limited. Thus, the nucleation and growth mechanism, as suggested for sulfuric acid media, would not apply to the perchloric acid electrolyte.

The chronoamperometric transients recorded at various potentials show only a single oxidation peak with some slight tailing at longer times. Contrary to H2SO4, only the tailing recorded on the Rh(111) surface is Cottrellian in nature. From an analysis of the slope of the Cottrell plot we conclude that the surface diffusion coefficient of CO on Rh(111) in perchloric acid is more than 4 orders of magnitude larger than in sulfuric acid (HClO4: 1⋅10-12 <D< 2⋅10-11 cm2⋅s-1; H2SO4: 1⋅10-16 <D< 8⋅10-16 cm2⋅s-1). The large difference in the surface diffusion coefficient of CO on Rh(111) also strongly suggests that previous assumptions regarding the proper kinetic model are correct.

Summarizing, specific anion adsorption on rhodium surfaces not only affects the rate (as is the case for platinum), but also the dynamics of the CO electrooxidation reaction. In media with strongly adsorbing anions the diffusion of CO on the surface is greatly reduced and the nucleation and growth model describes the kinetic mechanism best. As the anion adsorption strength decreases, the surface diffusion of CO increases and the mechanism shifts from nucleation and growth to mean-field approximation. This implies that, by varying the adsorption strength of the anion, we could in principle influence the diffusion rate of adsorbates on the surface and, therefore, the reaction dynamics and the overall reaction rate.

Acknowledgement: This research was supported by the Netherlands Foundation for Scientific Research (NWO).

Appendix

In this Appendix, we briefly outline the derivation of the Cottrell equation as it applies to diffusion of an adsorbed species on a two-dimensional surface reacting at a one-dimensional step. If the surface of a single crystal electrode can be represented as a series of terraces with of length Lt and width Ls (see Fig. 4.4) the flux of reactants per unit step length, pertaining to one step:

x = 0 x = Lt

Ls

Figure 4.4. Schematic representation of the surface of a stepped electrode.

71

Chapter 4

0=∂∂

−=xx

CDJ (A4.1)

where the surface concentration, C, is in mol⋅cm-2 and the flux, J, is in mol⋅s-1⋅cm-1. The total flux per step (i.e. number of molecules reacting per unit of time) is then:

ssteptot LJJ ⋅= (A4.2)

In case of Cottrell behavior the flux is the same as for 3D systems,[36] but with a different dimension as Cb (saturation concentration far from the step) is in mol⋅cm-2:

tDCJ b π

−= (A4.3)

Therefore, the total flux per step equals:

tDLCJ sb

steptot π

−= (A4.4)

As the total flux on the electrode is simply the flux per step times the number of steps (Jtot

electrode = Jtotstep * Nstep) the total current density can be written as:

AnFJ

jelectrodetot−= (A4.5)

where the surface area A is given by Ls*Lt*Nstep (this is exact for a rectangular electrode; the summation requires indexing for step and step length for any other shape). Substituting equation A4.3 and A4.4 in equation A4.5 yields:

tD

LnFC

jt

b

π= (A4.6)

This “2D” Cottrell equation is valid for surfaces where Lt approaches infinity (i.e. the limit in which A4.3 may be obtained).

Assuming that the average size of the terraces lies between 200 and 500 atoms per terrace, a value for the length of the terraces can be estimated (Lt (200 atoms)= 2.3⋅10-6 cm; Lt (500 atoms)= 5.8⋅10-6 cm). However, estimation of the CO surface concentration Cb at the moment the transients start displaying Cottrellian behavior is more difficult. Before a Cottrellian decay is observed, part of the adsorbed CO has already been oxidized. Depending on the diffusion rate the remaining CO may spread evenly over the surface, resulting in a surface concentration lower than the saturation concentration, or, if the mobility is sufficiently low, the concentration of the remaining CO may still be close to the saturation concentration. These two limiting cases have both been used in determining the diffusion coefficient. If the surface diffusion is negligible, the value for Cb can be estimated from the maximum surface coverage (0.75 for CO adsorbed on Rh(111)), the surface atom density (Rh(111) = 1.6⋅1015 atoms⋅cm-2) and Avogadro’s number, resulting in 2.0⋅10-9 mol⋅cm-2. If surface diffusion is not negligible, the CO surface concentration can be estimated using the ratio of the charge under the non-Cottrellian part and the Cottrellian part of the current-time transient. For instance, at 0.600 V vs. RHE on Rh(111) in perchloric acid approximately 40% of the CO is oxidized before the transient starts displaying a Cottrellian decay. Hence, we can assume that the

72

CO oxidation on stepped Rh[n(111)×(111)] electrodes: anion effects

surface concentration is 40% lower than if no diffusion occurs, thus Cb = 1.2⋅10-9 mol⋅cm-2. At 0.635 V a Cb of 1.3⋅10-9 mol⋅cm-2 can be calculated and in H2SO4 at 0.635 V a Cb of 1.96⋅10-9 (2.0⋅10-9) mol⋅cm-2 was found.

References

[1] R. I. Masel, Principles of adsorption and reaction on solid surfaces, John Wiley & Sons Inc, New York, 1996.

[2] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Phys. Chem. 106 (2002) 12938.

[3] J. M. Feliu, J. M. Orts, A. Fernandez-Vega, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 296 (1990) 191.

[4] A. V. Petukhov, Chem. Phys. Lett. 277 (1997) 539. [5] A. V. Petukhov, W. Akemann, K. A. Friedrich, and U. Stimming, Surf. Sci. 402-404 (1998) 182. [6] M. T. M. Koper, A. P. J. Jansen, R. A. van Santen, J. J. Lukkien, and P. A. J. Hilbers, J. Chem.

Phys. 109 (1998) 6051. [7] K. A. Friedrich, K. P. Geyzers, U. Stimming, and R. Vogel, Z. Phys. Chem. 208 (1999) 137. [8] M. T. M. Koper, J. J. Lukkien, A. P. J. Jansen, and R. A. van Santen, J. Phys. Chem. B 103 (1999)

5522. [9] C. Korzeniewski and D. Kardash, J. Phys. Chem. B 105 (2001) 8663. [10] H. Massong, H. Wang, G. Samjeske, and H. Baltruschat, Electrochim. Acta. 46 (2000) 701. [11] Y. Tong, E. Oldfield, and A. Wieckowski, Anal. chem. 70 (1998) 518A. [12] Y. Tong, H. S. Kim, P. K. Babu, P. Waszczuk, A. Wieckowski, and E. Oldfield, J. Am. Chem.

Soc. 124 (2002) 468. [13] S. Gilman, J. Phys. Chem. 68 (1964) 70. [14] W. Schmickler, Interfacial Electrochemistry, Oxford University press, Oxford, 1996. [15] T. H. M. Housmans and M. T. M. Koper, J. Electroanal. Chem. 575 (2005) 39. [16] T. H. M. Housmans, J. M. Feliu, and M. T. M. Koper, J. Electroanal. Chem. 572 (2004) 79. [17] R. Gomez, F. Javier Gutierrez de Dios, and J. M. Feliu, Electrochim. Acta. 49 (2004) 1195. [18] M. T. M. Koper, N. P. Lebedeva, and C. G. M. Hermse, Faraday Discuss. 121 (2002) 301. [19] N. M. Markovic and P. N. Ross, Surf. Sci. Rep. 45 (2002) 117. [20] N. M. Markovic, C. A. Lucas, A. Rodes, V. Stamenkovic, and P. N. Ross, Surf. Sci. 499 (2002)

L149. [21] J. M. Feliu, J. M. Orts, R. Gomez, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 372 (1994)

265. [22] J. Clavilier, D. Armand, S. G. Sun, and M. Petit, J. Electroanal. Chem. 205 (1986) 267. [23] M. Hourani and A. Wieckowski, J. Electroanal. Chem. 244 (1988) 147. [24] C. K. Rhee, M. Wasberg, P. Zelenay, and A. Wieckowski, Catal. Lett. 10 (1991) 149. [25] M. Hourani, M. Wasberg, C. Rhee, and A. Wieckowski, Croatica Chemica Acta 63 (1990) 373. [26] Y. E. Sung, S. Thomas, and A. Wieckowski, J. Phys. Chem. 99 (1995) 13513. [27] J. Clavilier, M. Wasberg, M. Petit, and L. H. Klein, J. Electroanal. Chem. 374 (1994) 123. [28] C. K. Rhee, M. Wasberg, G. Horanyi, and A. Wieckowski, J. Electroanal. Chem. 291 (1990) 281. [29] L.-J. Wan, S.-L. Yau, G. M. Swain, and K. Itaya, J. Electroanal. Chem. 381 (1995) 105. [30] C. K. Rhee, M. Wasberg, G. Horanyi, and A. Wieckowski, J. Electroanal. Chem. 291 (1990) 281. [31] M. T. M. Koper, J. J. Lukkien, N. P. Lebedeva, J. M. Feliu, and R. A. van Santen, Surf. Sci. 478

(2001) L339. [32] E. Herrero, B. Alvarez, J. M. Feliu, S. Blais, Z. Radovic-Hrapovic, and G. Jerkiewicz, J.

Electroanal. Chem. 567 (2004) 139. [33] N. P. Lebedeva, A. Rodes, J. M. Feliu, M. T. M. Koper, and R. A. van Santen, J. Phys. Chem. B

106 (2002) 9863. [34] N. P. Lebedeva, M. T. M. Koper, E. Herrero, J. M. Feliu, and R. A. van Santen, J. Electroanal.

Chem. 487 (2000) 37. [35] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Electroanal. Chem. 524-

525 (2002) 242. [36] A. J. Bard and L. R. Faulkner, Electrochemical Methods: Fundamentals and Applications, John

Wiley & Sons, New York, 2001. [37] E. G. Seebauer and C. E. Allen, Prog. Surf. Sci. 49 (1995) 265.

73

Chapter 4

74

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

Abstract

In this chapter we investigate the validity of the previously proposed model for the mechanism of the electrooxidation of CO on stepped Rh[n(111)×(111)] electrodes by dynamic Monte Carlo simulations. The mechanism described in Chapters 2, 3 and 4 assumes initiation of the reaction at step sites by fast adsorption of OH on the steps and reaction with a neighboring CO molecule, resulting in a pre-peak, followed by growth onto the terraces by slow OH adsorption on the terraces which reacts with neighboring terrace adsorbed CO, resulting in a main-peak. The presence of these two peaks in both the cyclic voltammetry as well as chronoamperometry can only be explained by assuming a low surface mobility of the reactants. In the absence of CO surface diffusion Monte Carlo simulations indeed produce voltammetric and chronoamperometric profiles that resemble the experimentally obtained curves. The shape and position of the pre-peak are determined largely by a reaction in one-dimension at the steps. Initiation of the reaction at empty sites results in the formation of two reaction fronts proceeding left and right of the initiation point, leading to a strong initial current growth followed by a gradual decay as multiple “shrinking rows” extinguish one another. This characteristic feature can also be identified in the experimental data. After stripping of the step and the first row of CO atoms adsorbed on the terraces, the reaction grows over the terraces by slower adsorption of OH. The presented Core parameter set also predicts a Cottrellian decay in absence of surface oxidation. Increasing the surface mobility of CO results in cyclic voltammograms (CVs) and transients similar to experiments performed on a rhodium surface in perchloric acid and to those of analogous Pt electrodes. Therefore, we conclude that this model, albeit with different parameters, also holds for platinum noble metal surfaces.

This chapter is published as T.H.M. Housmans, C.G.M. Hermse, M.T.M. Koper, to be submitted

Chapter 5

5.1. Introduction

Knowledge of the mobility of adsorbates in surface-catalyzed reactions is of pivotal importance in understanding the overall reactivity of the catalyst. In a Langmuir-Hinshelwood type reaction the overall reactivity is largely determined by the rate of diffusion of the reactants on the surface.[1] A low diffusion rate of reactants may lead to a low number of reactive configurations and, thus, to a low catalytic activity. The issue of surface mobility has recently received considerable attention for one of the most extensively studied reactions in electrocatalysis, namely the electrooxidation of CO on noble metal and bi-metallic surfaces (especially platinum and platinum-ruthenium surfaces).[2-13]

This reaction is generally accepted to be of Langmuir-Hinshelwood type with the overall mechanism represented by the following steps [14]:

H2O + * OHads + H+ + e- (5.1) COads + OHads COOHads (rds) (5.2) COOHads CO2 + H+ + e- + 2* (5.3)

where “*” denotes a free surface site, reaction 5.1 and 5.3 are assumed to be fast, and reaction 5.2 is rate determining (at relatively high overpotentials).

Since the spatial distribution of the reactants influences the rate of the slowest step (reaction 5.2), the diffusion rate of the reactants plays an important role in the overall reactivity of the surface. Two limiting cases for the analytical expression of the reaction rate can be given, namely the mean-field approximation and the nucleation and growth model. The mean-field approximation assumes that reactants are perfectly mixed on the surface (i.e. high surface mobility) and the reaction rate is proportional to the average coverages of the reacting species i.e. θCO and θOH.[1] If the reactants are assumed immobile on the surface, the nucleation and growth model applies. In this case, adsorption of OH at “special” sites, usually defects or steps, nucleates the reaction after which it proceeds only at the interface between the reacting species, causing the formation and growth of islands.[15]

Recent results on the electrooxidation reaction of CO on stepped Pt and Rh electrodes of [n(111)×(111)] orientation, showed markedly different oxidation characteristics.(see Chapters 2-4 and refs. [16-18]) On both metals steps (and crystalline defects in general) act as nucleation sites for the reaction.[2, 16-24] However, on stepped platinum surfaces Lebedeva et al. showed that the CO oxidation kinetics can best be modeled by the mean-field approximation,[2] while the presence of a shoulder prior to the main-peak and a distinct tailing in the CO stripping voltammetry on stepped rhodium surfaces indicate a nucleation and growth type mechanism. (Chapter 2, ref. [16]) As the charge corresponding to the pre-peak was found to increase linearly with the step density, it was concluded that the shoulder is associated with CO reacting directly next to the steps and, consequently, that the surface mobility of CO must be low. A low surface diffusion rate is consistent with the tailing visible in the cyclic voltammograms recorded on Rh(111) and Rh(554). Moreover, the presence of multiple peaks and a Cottrellian type

76

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

current decay in the current-time transients recorded on the rhodium electrodes also indicates a much lower surface diffusion rate of the reactants compared to analogous platinum electrodes. Assuming that CO is virtually immobile on the surface, the structure insensitive pre-peak at short times and strongly structure sensitive peak at longer times can be ascribed to CO reacting with OH, which adsorbs first at the steps and subsequently more slowly at empty terrace sites.(Chapter 3, ref. [17]) Strong anion adsorption and the ease of oxidizability of rhodium surfaces were given as possible explanations for the low surface mobility of CO.

Recently the influence of anion adsorption on the mobility of CO on Rh stepped electrodes was investigated by performing CO stripping voltammetry and potential-step experiments in perchloric acid and comparing the results to those reported for sulfuric acid media.(Chapter 4, ref. [18]) Contrary to sulfuric acid media the CO stripping peaks appear at lower potentials and no pre-peak or pronounced tailing could be detected, which indicates a reduced effect of the less strongly adsorbing perchlorate and hints at a higher surface mobility of CO. As the current-time transients recorded in 0.5 M HClO4 display only a single oxidation peak, the position of which shifts to shorter times with the step density, it was concluded that specific anion adsorption on rhodium surfaces not only affects the rate (as is the case for platinum), but also the dynamics of the CO electrooxidation reaction. In media with strongly adsorbing anions the diffusion of CO on the surface is greatly reduced and the nucleation and growth model gives a better description of the kinetic mechanism. As the anion adsorption strength decreases, the surface mobility of CO increases and the mechanism shifts from nucleation and growth to mean-field. From an analysis of the Cottrell plot it was concluded that the surface diffusion coefficient of CO on Rh(111) in perchloric acid is 4 orders of magnitude larger than in sulfuric acid (HClO4: D in the order of 1-20⋅10-12 cm2⋅s-1; H2SO4: D in the order of 1-8⋅10-16 cm2⋅s-1), which is in agreement with the previously made suggestions.

In this chapter we will investigate the role of CO diffusion on the electrooxidation characteristics of stepped surfaces using a general model for CO electrooxidation on stepped surfaces as studied by dynamic Monte Carlo simulations. We will demonstrate that by varying the CO surface mobility, the voltammetric and chronoamperometric profiles of CO oxidation on both platinum as well as rhodium can be generated using the same model.

5.2. Model

The general model considered here is a modification of a previously suggested model,[6] by distinguishing two different sites on the surface, i.e. terrace “T” and step “S” sites. Water activation leading to OHads formation can take place at the T and S sites:

H2O + *T ↔ OHads,T + H+ + e- (5.4) H2O + *S ↔ OHads,S + H+ + e- (5.5)

which leads to four different possible sites for the CO oxidation: COads,T + OHads,T → CO2 + 2 *T + e- + H+ (5.6)

77

Chapter 5

COads,S + OHads,S → CO2 + 2 *S + e- + H+ (5.7) COads,T + OHads,S → CO2 + *T + *S + e- + H+ (5.8) COads,S + OHads,T → CO2 + *T + *S + e- + H+ (5.9) In this model the reaction between CO and OH is only allowed in the trough of

the step (as indicated by arrow 3 in Fig 5.1 and reaction 5.7) and in the direction of the lower terrace (as indicated by arrow 2 in Fig. 5.1 and reaction 5.8 and 5.9). It is assumed that OH cannot react with CO adsorbed on the top of the steps (arrow 5 and reaction 5.8). The complete model, as described above, will henceforth be referred to as the “master model”. For simplicity both OH and CO are assumed to compete for the same on-top sites on a hexagonal lattice. This is not entirely accurate as at coverages close to saturation (θ = 0.75) the CO adlayer on Rh(111) has a (2×2)-3CO structure comprising both atop and multifold coordination.[25-27] Similarly, on Pt(111) CO adsorbs predominantly linearly at lower coverages, but may also adsorb in bridged configuration.[28, 29] Calculations performed by Vassilev et al. indicate that OH, when co-adsorbed with water, prefers the atop coordination.[30]

In our model simulations the rates of the reactions 5.4-5.9 have to be specified, as well as the rates of diffusion of the adsorbates. The model has two different adsorption sites, which leads multiple diffusion pathways:

COads,T + *T → *T + COads,T (5.10) COads,T + *S → *T + COads,S (5.11) COads,S + *T → *S + COads,T (5.12) COads,S + *S → *S + COads,S (5.13) These processes occur with a rate Di (s-1), implying a diffusion coefficient of

a2⋅D, with a the lattice constant (Doverall indicates a constant value for D5.10=D5.11=D5.12=D5.13). Moreover, diffusion reactions 5.11 and 5.12 are restricted to only one direction, i.e. CO adsorbed in the step cannot jump up to a higher terrace, nor

(b)

(a)

COads

OHads

4 5

3 4

1

2

Figure 5.1. Schematic representation of the different reactions in a) top view, and b) side view on a stepped surface used in the Core parameter set. In the model reactions (1)-(4) are allowed, while reaction (5) is not.

78

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

can CO adsorbed on the step jump down to a lower terrace site. All charge transfer reaction rates are assumed to obey the Butler-Volmer law and are given by:

)/exp( 01011, TkEekkk B

TadsOH α== (5.14)

)/)1(exp( 01011, TkEekkk B

TdesOH α−−== −− (5.15)

)/exp( 02022, TkEekkk B

SadsOH α== (5.16)

)/)1(exp( 02022, TkEekkk B

SdesOH α−−== −− (5.17)

)/exp( 03033

,,2

TkEekkk BOHCOdesCO

TT α== (5.18)

)/exp( 04044

,,2

TkEekkk BOHCOdesCO

SS α== (5.19)

)/exp( 05055

,,2

TkEekkk BOHCOdesCO

ST α== (5.20)

)/exp( 06066

,,2

TkEekkk BOHCOdesCO

TS α== (5.21)

where the rate constants, ki, are in s-1. All transfer coefficients, αi, are 0.5, E is the electrode potential, and e0, kB, and T have their usual meaning. The rate of the ith reaction, vi (in cm-2⋅s-1), can be calculated by the dynamic Monte Carlo method (DMC) described in refs. [6] and [31] and are obtained from:

tLaN

v ii ∆= 22 (5.21)

with Ni as the number of times reaction i has occurred in the time ∆t, a is the lattice constant and L2 is the size of the rectangular lattice in the simulations (256×300 in all calculations). The total current density, j, follows Faraday’s law and can be calculated by:

)( 654322110 vvvvvvvvej ++++−+−= −− (5.22)

The rate constants of the OH adsorption and desorption reaction were chosen such that at 20 mV⋅s-1 the OH adsorption appears reversible. Analogous to the experiments in Chapters 2-4 and refs. [2, 16-18, 23], the stepped surfaces (554), (553) and (331) were used in the modeling. Because an ideal (111) surface does not have step sites on which the reaction can nucleate, a (151514) surface, with terraces of 30 atoms wide, was used to approximate the electrochemical characteristics of the (111) plane. The temperature in the calculations was fixed at 300 K. The Monte Carlo simulations were carried out using the general-purpose code CARLOS. [6, 31] The CO saturated surfaces were generated by running a MC simulation with a CO adsorption and desorption rate set to 0.99 and 0.01, respectively on both step and terrace sites. All presented results are obtained by averaging the results of 10 simulations.

5.3. Results and Discussion

With the “master model” presented in the previous section we have modeled the experimental voltammetric and chronoamperometric data as obtained for CO electrooxidation on stepped platinum and rhodium single crystal electrodes of

79

Chapter 5

[n(111)×(111)] orientation.(Chapters 2 and 3, ref. [2, 16, 17, 22, 23]) The main attention was focused on deriving a set of parameters for stepped rhodium surfaces, for which a pre- and main-peak in both the linear sweep voltammetry as well as the potential-step experiments have been observed. The tailing of the main-peak in the voltammetry and chronoamperometric transients was investigated by modifying the diffusion rate of the CO adsorbate. We suggested recently that Rh[n(111)×(111)] electrodes start displaying characteristics similar to those of the analogous platinum surfaces if the mobility of CO is increased by reducing the anion adsorption strength.(Chapter 4) Thus, by simply altering the surface diffusion rate of CO on the stepped surfaces we should be able to model both the characteristics of stepped rhodium electrodes in sulfuric acid and perchloric acid solutions as well as stepped platinum electrodes.

In search for a “Core parameter set” we assumed that the rate constants of the CO + OH reactions are slower than the OH adsorption reaction. This allowed us to find a simple parameter set that produces both the cyclic voltammetry data as well as the potential-step transients. We believe that this assumption is justified as ab-initio quantum-chemical calculations as well as experimental results show that OH binds more strongly on rhodium than on platinum [16, 32-34] and Koper et al. showed that the most suitable model for the CO stripping voltammetry on Rh(100) is obtained when the CO + OH reaction rate is slower than the OH adsorption rate.[6] Moreover, in contrast to stepped platinum electrodes, the potential at which the CO electrooxidation reaction on Rh[n(111)×(111)] type electrodes occurs overlaps with the surface oxidation potential, which, in model terms, translates as the CO-OH reaction being slower than the OH adsorption.

5.3.1. OH Ad- and Desorption

The rate constants for the adsorption and desorption of OH on step and terrace sites were chosen such that at 20 and 2 mV⋅s-1 the reaction is in equilibrium.

Reaction Rate constant

Core parameter set / s-1

OH adsorption T (5.14) k1 5·10-5

OH desorption T (5.15) k-1 2·1010

OH adsorption S (5.16) k2 5·10-4

OH desorption S (5.17) k-2 4·108

CO oxidation COT – OHT (5.18) k3 5·10-4

CO oxidation COS – OHS (5.19) k4 5·10-4

CO oxidation COT – OHS (5.20) k5 5·10-4

CO oxidation COS – OHT (5.21) k6 5·10-4

CO diffusion on terrace sites (5.10) D5.10 0 CO diffusion from step to terrace site (5.11) D5.11 0 CO diffusion from terrace to step site (5.12) D5.12 0 CO diffusion from step to step site (5.13) D5.13 0 Table 5.1. Kinetic parameters for the Monte Carlo model

80

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

0.5 0.6 0.7 0.8 0.9 1.0

0

10

20

30

40

50

j / µ

A cm

-2

E / V vs. RHE

(151514) (554) (553) (331)

Figure 5.2. Cyclic voltammogram of the OH formation obtained from the Core parameter set for (151514) (solid thin line), (554) (dashed line), (553) (dotted line), and (331) (solid thick line). Scan rate is 20 mV⋅s-1.

Additionally, the adsorption of OH on terrace sites is assumed to take place at more positive potentials than on step sites.[16, 23] The intrinsic rate constants used in our simulations can be found in Table 5.1 and the resulting CVs for the OH adsorption on (151514), (554), (553) and (331) at 20 mV⋅s-1 are shown in Fig 5.2. The first peak located at ca. 0.7 V represents the formation of OH at step sites (in the trough of the steps), while the second peak at 0.87 V is caused by the formation of OH on the terrace sites. We recognize that exactly the same curves could have been easily generated using mean-field equations, but we present only MC-generated curves here.

The choice of the intrinsic rate constants of the OH ad- and desorption is based on the experimentally measured surface oxidation region of stepped Rh[n(111)×(111)] type electrodes (See Fig. 2.3 in Chapter 2). Here, the surface oxidation region is characterized by two features at high potential, which shift to lower potential as the step density is increased. We ascribed these features to OH adsorption at the steps at low potential and at the terraces at higher potentials. In fact, the features visible at high potentials on single crystal rhodium electrodes are most likely also due to the formation of irreversibly adsorbed surface oxides. However, we assume that prior to the formation of surface oxides first OHads is generated on the surface. Thus, by using the experimentally determined surface oxidation potential as a benchmark for the modeling of the OH formation, we most likely overestimate the potential at which OH is formed. This systematic error should be kept in mind when dealing with the reaction parameters obtained from the simulation. Moreover, contrary to the experimentally obtained blank cyclic voltammograms (Chapter 2, ref. [16]) the OH formation reaction modeled in Fig. 5.2 is not structure sensitive i.e. the peaks are located at a fixed potential for each stepped surface. In principle, this reaction should be modeled to each surface individually in order to incorporate the structure sensitivity of the surface oxidation reaction. However,

81

Chapter 5

for the sake of simplicity we decided to use only a single set of reaction parameters for all surfaces.

5.3.2. Core Parameter Set: Nucleation and Growth Type CO Oxidation

As was mentioned in the previous section, we will first focus on modeling the experimental results obtained for the CO electrooxidation reaction on stepped rhodium surfaces. These results are characterized by two clearly distinguishable peaks in both the voltammetry and chronoamperometry. The first peak was ascribed to CO reacting with OH formed at relatively low potentials near or at the steps (designated as “pre-peak”), while the second peak was ascribed to CO being oxidized more slowly at the terraces (designated as “main-peak”).(Chapters 2 and 3,[16, 17]) The peaks are distinguishable because the mobility of CO on the surface is assumed to be very low.

The simplest model that is able to reproduce both pre- and main-peak assumes a single CO oxidation rate constant regardless of the adsorption site in the absence of diffusion (see Table 5.1). The total reaction rate is now determined by the potential of OH adsorption. The pre-peak feature is generated by oxidation of CO adsorbed in the trough of the steps and on the first row of terrace atoms with step-adsorbed OH (reac. 5.7 and 5.8), while the main-peak results from COads,T reacting with OHads,T adsorbing at higher potentials (reac. 5.6). The influence of reaction 5.9 is negligible as most of the step adsorbed CO reacts away in the pre-peak feature.

5.3.2.1. Cyclic Voltammetry in Absence of Diffusion

The cyclic voltammograms and chronoamperometric transients obtained with the core parameter set under nucleation and growth conditions for the (151514), (554), (553)

0.5 0.6 0.7 0.8 0.9 1.0

0

50

100

150

200

250

300

350

j / µ

A cm

-2

E / V vs. RHE

(151514) (554) (553) (331)

0.60 0.65 0.70 0.75

0

50

100

150

j / µ

A cm

-2

E / V vs. RHE

Total current COT+OHT COS+OHS COT+OHS COS+OHT

Figure 5.3. Cyclic voltammograms obtained from the Core parameter set for (151514) (solid thin line), (554) (dashed line), (553) (dotted line), and (331) (solid thick line). Scan rate is 20 mV⋅s-1. The inset shows the individual contributions of reac. 5.6-5.9 to the overall calculated current density on (554).

82

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

and (331) surfaces are shown in Fig. 5.3 and 5.4, respectively. The CO stripping voltammograms in Fig. 5.3 show many similarities with the experimentally obtained curves for CO oxidation on stepped rhodium electrodes: they display a pre-peak at low potentials and a main-peak at higher potentials.(Chapter 2, ref. [16]) The peak visible between 0.8 and 0.95 V vs. RHE is due to the adsorption of OH on the emptied terrace sites. The simulated current densities are close to those measured experimentally. In comparison to the experimental results, the simulated pre-peak is less pronounced as it overlaps partially with the main-peak (see the inset in Fig. 5.3).

Because the total charge corresponding to the pre-peak is generated by OHads,S reacting with COads,S and COads,T adsorbed on the terrace row right next to the step, it is equal to twice the total number of step sites. Consequently, the total charge of the main-peak decreases as the terrace width decreases. This shift in pre- and main-peak charge with the step density does not agree with the experimental data, where the total charge corresponding to the main-peak was found to increase with the step density as a result of increasingly more complete oxidative stripping of the CO adlayer (compare Fig. 5.3 with Fig. 2.5a in Chapter 2). Moreover, for the (331) surface the model predicts that the charge in the pre-peak is nearly as large as the charge in the main-peak, which disagrees with the experimental results. The reason for this discrepancy is probably related to the extensive surface oxidation of the Rh(331) surface, which increases the total charge corresponding to the main-peak.

It was reported that the oxidation potential of CO on stepped rhodium electrodes overlaps with the surface oxidation potential. However, comparing the blank CVs with the CO stripping voltammograms (Fig. 2.3 and 2.5a in Chapter 2, respectively) shows that the oxidation of CO only overlaps with the onset of the surface oxidation region. The feature visible in the stripping CVs between 0.75 and 0.85 V vs. RHE is due to the oxidation of the surface, which in our model was simulated by OH adsorption on the terraces (the feature located at ca. 0.8-0.9 V in Fig. 5.3). The core parameter set was

0 10 20 30

0

50

100

150

200

250

j / µ

A cm

-2

t / s

(151514) (554) (553) (331)

Figure 5.4. Chronoamperometric transients obtained from the Core parameter set for (151514) (solid thin line), (554) (dashed line), (553) (dotted line), and (331) (solid thick line). Step potential is 0.68 V vs. RHE.

83

Chapter 5

chosen such that the position of the CO oxidation peaks is located at approximately the same position as was observed experimentally. To validate the value of the CO oxidation rate constant, we also performed a simulation with a lower value of the CO oxidation rate constant. Fig. 5.5a shows that when the overall CO oxidation rate constants (reac. 5.18-5.21 in Table 5.1) are lowered to 1⋅10-5 s-1, the pre- and main-peaks in the voltammetry shift to higher potentials and the CO stripping potential overlaps considerably with the formation of OH on the terraces. In analogy to the experimentally obtained results, this would mean a considerable overlap of the CO oxidation peak with the surface oxidation region. Because this is in disagreement with the CVs presented in Chapter 2, we believe that the oxidation rates chosen for the core parameter set are justified, at least qualitatively. Moreover, the absence of clearly distinguishable pre- and main-peaks in the current time transients obtained at 0.68 V vs. RHE for an overall reaction rate constant of 1⋅10-5 s-1 would indicate that the value of 1⋅10-5 s-1 for the rate constant is unrealistically low.

Fig. 5.6 shows some snapshots of the (554) surface as the reaction is proceeding. An ani

0.5 0.6 0.7 0.8 0.9 1.0

0

50

100

150

200

250

300

350

j / µ

A cm

-2

E / V vs. RHE

(151514) (554) (553) (331)

0 10 20 30 40 50

0

2

4

6

j / µ

A cm

-2

t / s

(151514) (554) (553) (331)

(a) (b)

Figure 5.5. Cyclic voltammograms (a) and chronoamperometric transients (b) obtained from the Core parameter set (no diffusion) using 1⋅10-5 as reaction rate constant for reac. 5.6 – 5.9 for (151514) (solid thin line), (554) (dashed line), (553) (dotted line), and (331) (solid thick line). Scan rate is 20 mV⋅s-1.

mation generated by the Monte Carlo simulations of the stripping reaction on the (554) surface can been seen in the right-hand lower corner of the page by quickly flipping the pages from front to back of the thesis. White areas and light gray lines represent CO adsorbed on terrace and step sites, respectively. Darker gray areas and black dots represent empty sites and adsorbed OH, respectively. The figures and the animation clearly show that, in the absence of diffusion, the reaction mechanism is best described by nucleation and growth. The reaction nucleation occurs by removal of CO at the steps and neighboring terrace sites, followed by a slower growth onto the terraces until the remaining terrace adsorbed CO is oxidized.

84

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

0.69 VRHE 0.66 VRHE

0.73 VRHE 0.71 VRHE

Figure 5.6. Snapshots of the (554) surface during the MC simulations of a cyclic voltammogram using the Core parameter set. The potentials at which the snapshots are taken are 0.66; 0.69; 0.71 and 0.73 V vs. RHE. (light gray) empty step site (*S), (dark gray) Empty terrace site (*T), (light blue) step adsorbed CO (COads,S ), (dark blue) terrace adsorbed CO (COads,T), (dark red) step adsorbed OH (OHads,S), and (light red) terrace adsorbed OH (OHads,S).

5.3.2.2. Chronoamperometry in Absence of Diffusion

The current-time transients obtained from the Monte Carlo simulations shown in Fig. 5.4 also demonstrate the two expected features: a pre-peak at short times that is generated by the CO reacting in or near the step (reac. 5.7 and 5.8) and a main-peak at longer times that is generated by oxidation of COads,T (reac. 5.6). The shape of the main-peaks in the simulation resembles the experimentally obtained data presented in Chapter 3 and the position of the peaks shifts to shorter times as the step density is increased.[17]

However, the shape of the pre-peak differs markedly from what one would expect for a Langmuir-Hinshelwood type reaction. Immediately after the start of the simulation the pre-peak reaches its maximum after which it slowly decays to zero. This is surprising as the reaction which gives rise to the pre-peak feature assumes competitive adsorption of OHads,S and COads,S and should, therefore, be peak shaped in nature. Interestingly, in the

85

Chapter 5

experimentally obtained current-time transients the pre-peaks recorded on Rh(111) and Rh(554) are also characterized by a fast current increase followed by a steady decay (see Chapter 3). The decay of the (pre-)peak on Rh(111) was found to be Cottrellian in nature and, accordingly, ascribed to diffusion of CO from the terraces to the reactive step. For Rh(554) no comments were made previously about shape of the pre-peak, however, it seems that apart from the instantaneous current increase the basic shape is similar to that predicted by the Monte Carlo simulations. We will discuss the mechanism and kinetics of the step reaction in more detail in Section 5.3.4.

5.3.3. Core Parameter Set: Influence of Diffusion

The influence of the overall diffusion rate (Doverall) on the voltammetric and chronoamperometric curves for the (151514), (554), (553), and (331) surfaces are shown in Fig. 5.7a-d and 5.8a-d, respectively.

The cyclic voltammograms in Fig. 5.7a-d show that increasing the surface diffusion rate of CO results in a shift of the pre- and main-peak to lower potentials. Moreover, the total charge of the main-peak decreases in favor of the charge in the pre-

0.5 0.6 0.7 0.8 0.9 1.0-50

0

50

100

150

200

250

300

350

j / µ

A cm

-2

E / V vs. RHE

Doverall=0 Doverall=10 Doverall=100

0.5 0.6 0.7 0.8 0.9 1.0

0

100

200

300

400

500

j / µ

A cm

-2

E / V vs. RHE

Doverall=0 Doverall=10 Doverall=100

0.5 0.6 0.7 0.8 0.9 1.0

0

100

200

300

400

500

600

j / µ

A c

m-2

E / V vs. RHE

Doverall=0 Doverall=10 Doverall=100

0.5 0.6 0.7 0.8 0.9 1.0

0

100

200

300

400

500

600

j / µ

A c

m-2

E / V vs. RHE

Doverall=0 Doverall=10 Doverall=100

0.6 0.7 0.8

0

50

100

150

200

250

300

j / µ

A cm

-2

E / V vs. RHE

Total current density Pre-peak

(d) (331) (c) (553)

(b) (554) (a) (151514)

Figure 5.7. Influence of diffusion on the cyclic voltammograms obtained on a) (151514) (solid thin line), b) (554) (dashed line), c) (553) (dotted line), and d) (331) (solid thick line) using the Core parameter set and Doverall = 0, 10 and 100 s-1. The scan rate is 20 mV⋅s-1. The inset shows the shape of the pre-peak in the CV simulated for (151514) and Doverall = 10 s-1.

86

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

0 5 10 15 20 25

0

100

200

300

400

500

j / µ

A c

m-2

t / s

Doverall=0 Doverall=10 Doverall=100

0 5 10 15 20 25

0

250

500

750

1000

1250

j / µ

A cm

-2

t / s

Doverall=0 Doverall=10 Doverall=100

0 5 10 15

0

250

500

750

1000

1250

1500

1750

2000

j / µ

A c

m-2

t / s

Doverall=0 Doverall=10 Doverall=100

0 5 10

0

250

500

750

1000

1250

1500

1750

2000

j / µ

A cm

-2

t / s

Doverall=0 Doverall=10 Doverall=100

0.0 2.5 5.0 7.5 10.0-20

0

20

40

60

80

100

120

140

j / µ

A cm

-2

t / s

Total current density pre-peak

(d) (331) (c) (553)

(a) (151514) (b) (554)

Figure 5.8. Influence of diffusion on the chronoamperometric transients obtained on a) (151514) (solid thin line), b) (554) (dashed line), c) (553) (dotted line), and d) (331) (solid thick line) using the Core parameter set and Doverall = 0, 10 and 100 s-1. Step potential is 0.68 V vs. RHE. The inset shows the shape of the pre-peak in the transient.

peak. As OHads,S is formed at lower potentials than OHads,T, the reaction still nucleates at the steps by reaction between OHads,S and a neighboring CO molecule. Next, diffusion of terrace adsorbed CO supplies the reaction at the step (reac. 5.7 and 5.8). When the surface mobility is low, the contribution of CO diffusing from the terraces to the steps is small. At very high diffusion rates the reaction proceeds almost solely at the steps (see Fig. 5.7 for D = 100) and CO electrooxidation characteristics similar to single crystal Pt electrodes are obtained.[2, 6, 13, 23, 35-37] In the current-time transients, where the driving force of the reaction is constant, the charge of the pre-peak increases at the expense of the main oxidation peak as Doverall increases. At high diffusion rates a single symmetrical oxidation peak is obtained. Interestingly, the shape of the pre-peak in the presence of CO diffusion (see the insets in Fig. 5.7a and 5.8a), is markedly different from the pre-peaks shown in the CVs and transients in Fig. 5.3 and 5.4, respectively. We will discuss this observation in more detail in the next section.

Compared to analogous stepped platinum surfaces, the transients obtained from the MC model at high diffusion rates are shifted towards shorter times.[2, 6, 13, 23, 35-37] This, of course, is caused by the OH ad- and desorption rate constants used in the Core parameter set, which were obtained by modeling the surface oxidation of rhodium

87

Chapter 5

single crystal electrodes. Notably, the obtained MC results closely resemble current-time transients for CO oxidation obtained on stepped rhodium electrodes in perchloric acid media.(Chapter 4, ref. [18]) Here, it was also suggested that the lower binding strength of the perchlorate anion compared to (bi)sulfate results in an considerable increase in the mobility of CO adsorbed on the surface, thereby leading to a reaction best described by the mean-field approximation rather than nucleation and growth.

Simulations in which the diffusion rates in reaction 5.10-5.13 are varied individually reveal that of the four different diffusion coefficients, the D5.10 and D5.13 play a predominant role in the shape of the CVs and transients. The influence of CO diffusing from the step sites to the terraces and from the terraces to step sites (D5.11 and D5.12., respectively) is limited, as OH formed in the trough of the step reacts rapidly with any neighboring CO. The effect of D5.11 and D5.12 only becomes more pronounced at high diffusion rates when the reaction rate becomes slower than the supply of COads,T to the steps.

In a true nucleation and growth type mechanism no tailing of the current-time transients is expected. The fact that for stepped rhodium surface nucleation and growth was suggested, while a Cottrellian decay was measured experimentally was tentatively explained in Chapter 4 by partial oxidation of the terraces. It was suggested that the presence of steps induces surface oxidation of the neighboring terrace atoms. If the terraces are wide, only a small portion of the terrace will be oxidized, and the remaining COads,T diffuses to the reactive sites. When the terraces get narrower, surface oxidation is more complete and the Cottrellian decay disappears. Of course, our model does not incorporate partial surface oxidation and can, therefore, not predict a pre-peak and a main-peak with tailing at longer times. However, in the absence of terrace oxidation, the model should be able to predict the current-time transients as they were recorded experimentally for Rh(111). The current-time transient under these conditions was simulated on the (151514) surface and the results are shown in Fig. 5.9. Indeed, in the

0 10 20 30 40 50-10

0

10

20

30

40

50

60

70

j / µ

A cm

-2

t / s

0.0 0.5 1.0 1.5-10

0

10

20

30

40

50

60

70

45 sec

3 sec

j / µ

A cm

-2

1/t0.5

Figure 5.9. Chronoamperometric transients obtained on (151514) in the absence of terrace oxidation (i.e. k1 and k-1 are 0). The inset shows the Cottrellian current decay of the transient at 0.68 V vs. RHE.

88

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

absence of reaction 5.4 and with an overall diffusion rate, Doverall, of 10 s-1 for reaction 5.10–5.13 the model predicts a single peak at short times similar to the transient recorded on Rh(111),[17] followed by tailing due to slow diffusion of CO on the terraces to the steps. The Cottrell plot of this transient shows a straight line after ca. 3 seconds (see inset in Fig. 5.9). Based on these results we conclude that a combination of a nucleation and growth model and partial surface oxidation with slow diffusion of CO on the terraces can indeed generate a pre-peak and a main peak together with a Cottrellian like decay.

5.3.4. Modeling the CO Electrooxidation on Step Sites

In the transients obtained in the absence of CO diffusion the recorded pre-peaks show a sharp increase in the current density followed by a slow decay, while in the presence of surface diffusion the initial current increase becomes more gradual and the

0.5 0.6 0.7 0.8 0.9 1.0

0.0

1.0x10-4

2.0x10-4

3.0x10-4

4.0x10-4

5.0x10-4

j / µ

A m

E / V vs. RHE

no diffusion high diffusion

0.0 0.5 1.0 1.5 2.0

0.0

1.0x10-3

2.0x10-3

3.0x10-3

4.0x10-3 no diffusion high diffusion

j / µ

A m

t / s

(a) (b)

(c) 0.50 V

0.60 V

0.63 V

0.70 V

1.00 V

D =100D = 0

0.57 V

1.00 V

0.63 V

0.58 V

0.55 V

0.50 V

Figure 5.10. Core parameter set applied to a single step. a) Influence of diffusion on the cyclic voltammogram obtained on a single step using the Core parameter set and D5.13 = 0 (solid thick line) and 100 s-1 (dashed line). b) Influence of diffusion on the chronoamperometric transients obtained on a single step using the Core parameter set and D5.13 = 0 (solid thick line) and 100 s-1 (dashed line). c) Snapshots of the reaction on the modeled step for different potentials at D5.13 = 0 (left hand picture) and 100 s-1 (right hand picture). (light gray) Empty step site (*S), (light blue) step adsorbed CO (COads,S ), and (dark red) step adsorbed OH (OHads,S).

0.70 V

89

Chapter 5

pre-peak is more peak-like in shape. In order to investigate in more detail the characteristics of the CO electrooxidation reaction at the step itself we have applied our Core parameter set (Table 5.1) to a surface consisting solely of step sites. Furthermore, the two limiting cases for the analytical expression of the total reaction rate of CO at the steps, the nucleation and growth model and the mean-field approximation, will be derived. As was mentioned in the introduction, in the nucleation and growth model species are assumed to be immobile on the surface (in this case D5.13 = 0).[15] For the mean-field approximation reactants are assumed to be perfectly mixed on the surface and the reaction rate is proportional to the average coverages of the reacting species.[1] This condition is satisfied when surface diffusion is faster than the reaction itself. The resulting voltammetric and chronoamperometric profiles, together with a schematic representation of the adsorbates at the surface for both zero as well as high diffusion rates, can be found in Fig. 5.10a-c, respectively.

Fig. 5.10a demonstrates the influence of the diffusion rate on the voltammetric profiles of CO being oxidized at the steps. In the absence of diffusion the oxidation peak is symmetrical and shows a typical nucleation and growth peak shape. Moreover, it overlaps considerably with the OHads,S adsorption peak located between 0.65 and 0.8 V vs. RHE. The nucleation of the reaction occurs by adsorption of OH in an empty step site followed by reaction with a neighboring CO molecule (see Fig. 5.10c), thus generating two empty sites. The reaction now grows in two directions along the step, generating “shrinking rows”. When the potential is increased, the OH adsorption becomes faster and the reaction rate increases (i.e. the production rate of OH increases), which eventually strips the CO from the step and results in a fast drop in the current density.

As can be expected, at high diffusion rates the oxidation peak is shifted to more negative potentials due to better mixing of the reactants (see Fig. 5.10a and c). The CVs obtained from the MC simulations for this one-dimensional system with high and low diffusion rates are in agreement with the expected results for a (two-dimensional) system in the mean-field approximation and nucleation and growth limits, respectively.[6]

For the chronoamperometric transients the differences between the one-dimensional step system and a normal two-dimensional surface become more pronounced. Fig. 5.10b shows that, in the absence of diffusion, the MC simulations predict a rapidly increasing current density, followed by a slow decay to zero, while for high diffusion rates much less, but still significant tailing is obtained. This is remarkable as in two dimensions, for both the nucleation and growth model and the mean-field approximation, much more symmetrical peaks are predicted.[2, 6, 13] The fast current increase is caused by the fact that two reaction fronts are formed from a single empty site immediately after the potential step. Since the OH adsorption rate and, accordingly, the total reaction rate is constant for a constant potential these fronts grow along the length of the step at a constant rate resulting in a constant current. The reaction proceeds along the step and the “shrinking rows” of CO are gradually depleted, which leads to a stepwise decrease in the total current (See Fig. 5.10c “no diffusion”). In principle, if the empty sites are spaced evenly along the step, the current should increase rapidly initially, remain

90

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

constant while the rows of CO are consumed, and then rapidly drop to zero as the reaction fronts extinguish each other, as indeed is predicted by our model under these circumstances (See Fig. 5.11b). A similar situation holds for the cyclic voltammetry. The OH adsorption rate increases for increasing potentials, which results in an increasing reaction rate. Therefore, the current increases until the “shrinking rows” extinguish one another simultaneously, leading to a sudden current drop, as can be seen in Fig. 5.11a. The peak at 0.7 V vs. RHE in the cyclic voltammetry and the current between 1.5 and 2.5 seconds in the chronoamperometry are due to oxidation of the last rows of CO near the border of the model surface.

Because the step-model is a fairly simple system an analytical expression for the one-dimensional nucleation and growth is easily obtained. The derivation of the equations for the instantaneous and progressive nucleation and growth can be found in the appendix. A plot of the analytical expressions for the instantaneous and progressive

0 2 4 6 8 10

0.0

0.2

0.4

0.6

0.8

1.0

j/jm

ax

t/tmax

Instantaneous N&G Progressive N&G

0.5 0.6 0.7 0.8 0.9 1.0

0.0

0.5

1.0

1.5

Figure 5.11. Core parameter set applied to a single step with defects in the CO adlayer evenly spaced along the step. a) Cyclic voltammetry with a scan rate of 20 mV⋅s-1, b) chronoamperometry with a step potential of 0.68 V vs. RHE

j / N

tot s

tep-1

E / V vs. RHE0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0.0

0.5

1.0

1.5

j / N

tot s

tep-1

t / s

(b)(a)

Figure 5.12. Normalized current transients for the analytical expression of the instantaneous and progressive nucleation and growth equations.

91

Chapter 5

nucleation and growth in dimensionless form, as described in the appendix (equations A5.12 and A5.13, respectively), indeed shows a current decay in the absence of diffusion and a current maximum in case of high diffusion rates (see Fig. 5.12). In the MC simulations, setting the CO diffusion rate to zero results in a transient, which resembles the analytical expression for the instantaneous nucleation and growth, while the transient for high CO diffusion rates is similar to the progressive nucleation and growth transient.

In addition to these results we would like to note that including a row of terrace atoms next to the step sites (thus allowing reaction 5.8), while maintaining the values given in Table 5.1, only increases the total current density recorded, but does not alter the shape of the pre-peak. Therefore, we can conclude that the position and shape of the pre-peak are determined almost exclusively by the reaction occurring directly at the step.

Interestingly, the basic shape of the pre-peak recorded on Rh(554), and to a lesser extend Rh(553), corresponds to that of the pre-peak obtained in the Monte Carlo simulations (see Fig. 3.6b and c in Chapter 3, ref. [17]). The fact that, after a small current plateau,[2, 35-38] the experimentally obtained voltammograms also display a rapidly increasing current, followed by a slower decrease, suggests that the one-dimensional oxidation mechanism proposed for the steps is applicable. As surface oxidation starts playing an increasingly more important role for higher step densities, this characteristic shape is not as well defined on Rh(553) as it is on Rh(554). On Rh(331) it is completely absent.

5.3.5. Parameter Set for CO Electrooxidation on Rh[n(111)×(111)]

Although the core parameter set is able to predict the appearance of a pre- and main-peak in the both the voltammetric as well as the chronoamperometric transients, and it is able to explain the Cottrellian decay of the main-peak observed in the transients

0.5 0.6 0.7 0.8 0.9 1.0

0

50

100

150

200

250

300

350

j / µ

A cm

-2

E / V vs. RHE

(151514) (554) (553) (331)

0 10 20 30

0

50

100

150

200

250

j / µ

A c

m-2

t / s

(151514) (554) (553) (331)

(a) (b)

Figure 5.13. a). Cyclic voltammograms obtained from the Improved model for (151514) (solid thin line), (554) (dashed line), (553) (dotted line), and (331) (solid thick line). The scan rate is 20 mV⋅s-1. b) Chronoamperometric transients obtained from the Final model (151514) (solid thin line), (554) (dashed line), (553) (dotted line), and (331) (solid thick line). Step potential is 0.68 V vs. RHE.

92

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

obtained on low step density surfaces, there are still discrepancies with the experimental results that may be of importance. More specifically, the MC calculations simulated a pre-peak with considerable overlap with the main oxidation peak. The shape of this feature is quite flat and broad due to the absence of CO diffusion along the trough of the step. In a previous paper we concluded that (bi)sulfate preferably adsorbs and desorbs on the terraces but not so on the steps (Chapter 2, ref. [16]). Because strong anion adsorption is the primary reason for low CO surface mobility, we believe that the diffusion of CO in the trough of the step is higher than on the terraces. By setting D5.13 to 10 s-1 and keeping all other diffusion coefficients zero the voltammetric and chronoamperometric profiles resemble more closely those recorded experimentally (see Fig. 5.13a and b, respectively). For this reason we believe that the best model for the electrooxidation of CO on stepped Rh[n(111)×(111)] electrodes consists of mobile COads,S and immobile COads,T which reacts fast with OHads,S, followed by COads,T reacting more slowly with OHads,T.

5.4. Conclusions

In this chapter we used dynamic Monte Carlo simulations to test the validity of a model proposed for the electrooxidation of CO on stepped Rh[n(111)×(111)] electrodes suggested in Chapter 3. In this mechanism the reaction is assumed to nucleate at step sites by fast adsorption of OH and subsequent reaction with a neighboring CO molecule resulting in a pre-peak, followed by growth onto the terraces by OH, which adsorbs more slowly at the terraces, with terrace adsorbed CO, resulting in the main-peak. In a study on the effect of anion adsorption on the reaction kinetics presented in Chapter 4 the low surface diffusion of CO on rhodium surfaces was attributed to strongly adsorbing anions. As the anion adsorption strength, is decreased the surface mobility of CO is increased and, similar to analogous platinum surfaces, a mean-field type reaction rate is observed. Therefore, the model should also be able to produce results similar to those published for analogous Pt surfaces by simply increasing the surface mobility of CO.

In the absence of CO surface diffusion dynamic Monte Carlo simulations performed with the model presented above indeed produce voltammetric and chronoamperometric profiles that resemble the experimentally obtained curves on Rh electrodes in H2SO4. Increasing the surface mobility of CO results in CVs and transients similar to those of Rh electrodes in HClO4 and Pt electrodes. This leads us to conclude that the model proposed, albeit with different parameters, also holds for platinum surfaces.

Moreover, we found that the shape and position of the pre-peak are determined mainly by a one-dimensional reaction along the steps. Initiation of the reaction at empty step sites results in the formation of two reaction fronts proceeding to the left and right of the initiation point. In the case of instantaneous nucleation and growth this results in a strong initial current growth followed by a gradual decay as the reaction fronts extinguish one another. If the initiation sites are evenly spaced along the step a constant current is

93

Chapter 5

obtained in the chronoamperometry. Remarkably, experimentally, pre-peaks of a similar shape are obtained. In the absence of surface oxidation the Core parameter set can also predict a Cottrellian decay in the current-time transients. If partial oxidation of the terraces is assumed a pre-peak and a main-peak followed by a Cottrellian decay can be expected.

Based on an analysis of the experimental results presented in Chapters 2-4 and the results presented in this chapter we also conclude that the mobility of CO along the trough of the step presumably is higher than on the terraces due to a reduced adsorption of (bi)sulfate at the steps. For this reason, the shape of the pre-peak is sharper and more pronounced. However, in the Core parameter set the pre-peak consists of two reactions, namely, the reaction between OHads,S and both COads,T and COads,S, which probably results in an overestimation of the charge under the pre-peak.

Acknowledgement: This research was supported by the Netherlands Foundation for Scientific Research (NWO). Special thanks go to Peter Vassilev for his assistance with the simulation procedures and data processing.

Appendix

An analytical expression for a nucleation and growth mechanism in one dimension can be drawn in much the same way as for the two dimensional system.[15] Analogous to the 2D-system, the number of active growing nuclei, M(t), is defined by:

]1[)( )(0

tkNeMtM −−= , (A5.1)

where M0 is the number of active sites at t = 0 and kN is the rate constant of the formation of a nucleus. Two limiting cases are of particular interest:

1. Instantaneous nucleation and growth for kN⋅t >> 1 and M(t) = M0

2. Progressive nucleation and growth for kN⋅t << 1 and M(t) = kN⋅M0⋅t The approximate laws for the stripping of an one-dimensional CO adlayer can be

obtained by assuming a model where a single empty site in the step grows upon reaction of OH with a neighbouring CO molecule and that all expanding empty rows formed are linear and do not touch each other. The change in the number of empty sites added to the row is now:

42 kdtdN

⋅= , (A5.2)

where k4 is the rate constant of empty site generation and (i.e. CO oxidation). The change in the length of the empty row, L, is:

42 kadtdL

⋅⋅= , which results in L(t) = 2a⋅k4⋅t (A5.3)

The extent of overlap between growing nuclei is estimated by Avrami’s Theorem, which, in the 1D case, relates the length that is actually covered by coalescing expanding rows, L(t), to the extended area, Lex(t), that would be covered if the rows did not overlap:

94

CO oxidation on stepped single crystal electrodes: a Monte Carlo study

)(1)( exLav etL −−= (A5.4)

For the instantaneous nucleation and growth Lex(t) and thus L(t) are: tMkatLex ⋅⋅⋅⋅= 042)( (A5.5)

)2( 041)( tMkaetL ⋅⋅⋅⋅−−= (A5.6) The concomitant current density is obtained from the derivative of N(t)=L(t)/a;

)2(0400

042)( tMkaeMkezdtdNeztj ⋅⋅⋅⋅−⋅⋅⋅⋅⋅=⋅⋅= , (A5.7)

where z is the number of electrons transferred during the oxidation of CO and e0 has its usual meaning. For progressive nucleation and growth, new growing rows are formed at a constant rate of kN⋅M0 . Thus, Lex(t) and L(t) are:

)'(2)( 4 ttaktLsex −= (A5.8)

2044

000 )'(2)()( tMkkadtttakMktLMktL N

t

NsexNex ⋅⋅⋅⋅=−== ∫ (A5.9)

)( 2041)( tMkka NetL ⋅⋅⋅⋅−−= , (A5.10)

which leads to the following expression for the current density: )(

040

2042)( tMkka

NNetMkkeztj ⋅⋅⋅⋅−⋅⋅⋅⋅⋅⋅⋅= (A5.11)

The equations A5.7 and A5.11 can easily be converted to a dimensionless form by introducing the maximum current, jmax, at the time, tmax (or, as there is no maximum, trelax for instantaneous nucleation and growth) at which it is attained. The dimensionless expressions for the instantaneous and progressive nucleation and growth are respectively,

)(

max

)(relax

relax

ttt

ej

tj−−

= , and (A5.12)

)2

(

maxmax

2max

2max

2

)( t

tt

et

tj

tj−−

= . (A5.13)

References

[1] R. I. Masel, Principles of adsorption and reaction on solid surfaces, John Wiley & Sons Inc, New York, 1996.

[2] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Phys. Chem. 106 (2002) 12938.

[3] J. M. Feliu, J. M. Orts, A. Fernandez-Vega, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 296 (1990) 191.

[4] A. V. Petukhov, Chem. Phys. Lett. 277 (1997) 539. [5] A. V. Petukhov, W. Akemann, K. A. Friedrich, and U. Stimming, Surf. Sci. 402-404 (1998) 182. [6] M. T. M. Koper, A. P. J. Jansen, R. A. van Santen, J. J. Lukkien, and P. A. J. Hilbers, J. Chem.

Phys. 109 (1998) 6051. [7] K. A. Friedrich, K. P. Geyzers, U. Stimming, and R. Vogel, Z. Phys. Chem. 208 (1999) 137. [8] M. T. M. Koper, J. J. Lukkien, A. P. J. Jansen, and R. A. van Santen, J. Phys. Chem. B 103 (1999)

5522. [9] C. Korzeniewski and D. Kardash, J. Phys. Chem. B 105 (2001) 8663. [10] H. Massong, H. Wang, G. Samjeske, and H. Baltruschat, Electrochim. Acta. 46 (2000) 701. [11] Y. Tong, E. Oldfield, and A. Wieckowski, Anal. chem. 70 (1998) 518A.

95

Chapter 5

96

[12] Y. Tong, H. S. Kim, P. K. Babu, P. Waszczuk, A. Wieckowski, and E. Oldfield, J. Am. Chem. Soc. 124 (2002) 468.

[13] M. T. M. Koper, N. P. Lebedeva, and C. G. M. Hermse, Faraday Discuss. 121 (2002) 301. [14] S. Gilman, J. Phys. Chem. 68 (1964) 70. [15] W. Schmickler, Interfacial Electrochemistry, Oxford University press, Oxford, 1996. [16] T. H. M. Housmans, J. M. Feliu, and M. T. M. Koper, J. Electroanal. Chem. 572 (2004) 79. [17] T. H. M. Housmans and M. T. M. Koper, J. Electroanal. Chem. 575 (2005) 39. [18] T. H. M. Housmans and M. T. M. Koper, Electrochem. Comm. 7 (2005) 581. [19] V. Climent, R. Gomez, and J. M. Feliu, Electrochim. Acta. 45 (1999) 629. [20] A. Rodes, R. Gomez, J. M. Feliu, and M. J. Weaver, Langmuir 16 (2000) 811. [21] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, Electrochem.Commun. 2

(2000) 487. [22] N. P. Lebedeva, A. Rodes, J. M. Feliu, M. T. M. Koper, and R. A. van Santen, J. Phys. Chem. B

106 (2002) 9863. [23] N. P. Lebedeva, M. T. M. Koper, E. Herrero, J. M. Feliu, and R. A. van Santen, J. Electroanal.

Chem. 487 (2000) 37. [24] R. Gómez, J. M. Orts, J. M. Feliu, J. Clavilier, and L. H. Klein, J. Electroanal. Chem. 432 (1997)

1. [25] Y. E. Sung, S. Thomas, and A. Wieckowski, J. Phys. Chem. 99 (1995) 13513. [26] S. L. Yau, X. Gao, S. C. Chang, B. C. Schardt, and M. J. Weaver, J. Am. Chem. Soc. 113 (1991)

6049. [27] M. Hourani, M. Wasberg, C. Rhee, and A. Wieckowski, Croatica Chemica Acta 63 (1990) 373. [28] B. Beden, C. Lamy, A. Bewick, and K. Kunimatsu, J. Electroanal. Chem. Inter. Electrochem. 121

(1981) 343. [29] Y. X. Chen, A. Miki, S. Ye, H. Sakai, and M. Osawa, J. Am. Chem. Soc. 125 (2003) 3680. [30] P. Vassilev, M. T. M. Koper, and R. A. van Santen, Chem. Phys. Lett. 359 (2002) 337. [31] J. J. Lukkien, J. P. L. Segers, P. A. J. Hilbers, R. J. Gelten, and A. P. J. Jansen, Phys. Rev. E 58

(1998) 2598. [32] M. Chen, S. P. Bates, R. A. van Santen, and C. M. Friend, J. Phys. Chem. B 101 (1997) 10051. [33] M. Mavrikakis, J. Rempel, J. Greeley, L. B. Hansen, and J. K. Norskov, J. Chem. Phys. 117

(2002) 6737. [34] E. M. Patrito, P. P. Olivera, and H. Sellers, Surf. Sci. 306 (1994) 447. [35] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Electroanal. Chem. 524-

525 (2002) 242. [36] M. Bergelin, E. Herrero, J. M. Feliu, and M. Wasberg, J. Electroanal. Chem. 467 (1999) 74. [37] W. Akemann, K. A. Friedrich, and U. Stimming, J. Chem. Phys. 113 (2000) 6864. [38] B. Love and J. Lipkowski, ACS Symposium Series 378 (1988) 484.

CO oxidation on Pt modified Rh(111) electrodes Abstract

The electrooxidation reactions of CO, methanol and formic acid were studied on platinum modified Rh(111) electrodes in 0.5 M H2SO4 using cyclic voltammetry and chronoamperometry. The Pt-Rh(111) electrodes were generated during voltammetric cycles at 50 mV⋅s-1 in a 30 µM H2PtCl6 and 0.5 M H2SO4 solution. Surfaces generated by n deposition cycles (Ptn-Rh(111) with n=2, 4, 6, 8, 10, and 16) were characterized by voltammetry and Scanning Tunneling Microscopy. The blank cyclic voltammograms show a pronounced sharpening of the hydrogen/(bi)sulfate adsorption/desorption peaks typical for Rh(111), and the appearance of contributions between 0.1 and 0.4 V, ascribed to hydrogen/(bi)sulfate adsorption/desorption on the deposited platinum. The STM characterization shows Pt islands of monoatomic height for n=2, which grow three-dimensionally for increasing number of deposition cycles. Two peaks characterize the CO stripping voltammetry on the Ptn-Rh(111) surfaces. At low potential a structure insensitive peak is ascribed to CO reacting at the platinum monolayer islands. The onset of the peak is shifted 150, 250 and 100 mV negatively with respect to pure Rh(111), Pt(111) and polycrystalline Pt, respectively, indicating the enhanced CO electrooxidation properties of the Pt overlayer system. A structure sensitive peak at higher potentials was ascribed to CO reacting on the islands of multi-atomic height. Current-time transients recorded on monolayer islands also indicate enhanced CO oxidation kinetics. Comparison of these current-time transients to those of pure Rh(111) and Pt(111) shows reduced reaction times and pronounced tailing, indication surface diffusion limitation. For the oxidation of methanol and formic acid, on the other hand, no pronounced influence of the platinum overlayers can be observed. The results presented in this chapter show that, as indicated by DFT calculations, the CO adlayer oxidation for platinum monolayer systems is enhanced compared to both pure Rh and Pt.

This chapter is published as T.H.M. Housmans, J.M. Feliu, R. Gómez, M.T.M. Koper, Chem.Phys. Chem. 6 (2005) 1522

Chapter 6

6.1. Introduction

A major impediment in the development of low temperature fuel cells is the deactivation of the platinum-based anode by even traces (e.g. 10-100 ppm) of carbon monoxide. Metals other than platinum often show inferior fuel cell efficiency and power output, and are, therefore, unattractive. Hence, the development of a highly active, CO-tolerant platinum-based anode catalyst is of great importance.[1, 2] To this end research in this field has focused extensively on bi-metallic (or even tri-metallic) platinum-based catalysts as a solution for the problem of CO poisoning.

The usage of well-defined single crystal electrodes provides an excellent tool for elucidating the factors involved in the adsorption and oxidation of CO on these bi-metallic systems. Most fundamental studies involve bi-metallic surfaces where Pt is either alloyed or modified with Ru, Re, Sn, Pd, Mo, and Rh.[3-17] The aim of such studies is to develop an anode electrocatalyst that either adsorbs very small amounts of CO while retaining the ability to oxidize hydrogen at an acceptable rate, or that is able to oxidize CO at considerably lower overpotential than pure Pt. The enhanced CO oxidation properties of bi-metallic surfaces are usually explained by the so-called bifunctional mechanism, originally proposed by Watanabe and Motoo,[18] which is based on the idea that sites on the more oxophilic metal act as adsorption centers for oxygen-containing species (generally accepted to be OHads), which can react with CO adsorbed on platinum to form CO2.(see Chapter 1)

However, it has been repeatedly pointed out that the bifunctional mechanism does not take into account a possible change in the CO binding strength on Pt induced by the presence of the second metal, an effect often referred as the “electronic” or “ligand” effect. Indeed, detailed Density Functional Theory (DFT) calculations show that mixing of Pt with Ru, Rh, Ir, Re (typically those metals located to the upper left of Pt in the Periodic Table) results in a considerable decrease in the CO binding strength to platinum due to a shift in the d-band (the so-called “d-band model”).[19-22] DFT results also indicated that, in terms of the electronic effect, Pt-overlayer systems may have the best CO-tolerance properties (i.e. the lowest CO binding energy).[3, 12, 13, 16, 21] These strong electronic effects are caused by the fact that the CO binding energy in the overlayer system is determined primarily by the Pt-Pt distance in the overlayer, which is dictated by the underlying substrate. Contraction of the Pt overlayer with respect to pure Pt results in a lower CO bond strength (i.e. PtML on Ru, Rh, Ir, Re), while expansion results in higher binding energies (i.e. PtML on Au). Pt submonolayers on Ru(0001) and Ru(101 0), as well as on Ru nano-particles, were investigated recently by Adžić et al., and indeed showed good CO tolerance for hydrogen oxidation.[23-26]

Even if the DFT calculations predict a weaker CO binding to these Pt monolayer systems their possible role in the bifunctional mechanism for CO oxidation is not clear, as the OH binding to the Pt monolayers should also be weakened.[19-22] Brankovic et al.[24] reported CO oxidation on Pt-modified Ru(101 0) to be faster than on bare Pt, but slower compared to Ru(101 0). These experiments were carried out in a CO-saturated

98

CO oxidation on Pt modified Rh(111) electrodes

solution however, which has the added complication of re-adsorption of CO and the associated self-inhibition. Furthermore, electrochemical investigations on Ru single-crystal electrodes can be problematic due to more difficult preparation methods for clean and well-defined surfaces, certainly in comparison to the more widely studied Pt and Rh surfaces. Since Pt-modified Rh is predicted to have electronic properties very similar to Pt-modified Ru,[19-22] and stripping of CO adlayers has been extensively studied on these metals in our laboratories,(see Chapters 2-5, refs. [27-35]) PtML-Rh(111) may be a good model system for studying the kinetics of the CO adlayer oxidation on a Pt-monolayer system.

Using the data on the CO electrooxidation reaction on Rh[n(111)×(111)] type electrodes reported in Chapters 2-5 we present in this chapter results of an investigation of the CO electrooxidation reaction on Pt-modified Rh(111) surfaces of varying Pt coverage. The surfaces, prepared by cycling well-ordered Rh(111) in a platinum containing solution, were characterized by cyclic voltammetry and Scanning Tunneling Microscopy (STM). The CO stripping voltammetry and chronoamperometry reveal a relationship between the potential at which CO oxidizes and the surface structure and composition. The results will be discussed and explained in terms of the electronic effect mentioned above at low platinum coverages, where islands of mono- and di-atomic height are present on the surface. At higher Pt coverages we will discuss the possibility of a platinum particle size effect on CO oxidation properties. The electrocatalytic properties of the Pt modified Rh electrodes were also investigated for methanol and formic acid.

6.2. Experimental Setup

The working electrode used in this study was a rhodium (111) bead-type single crystal electrode. The electrode was oriented, cut and prepared according to the Clavilier method as described in ref. [36]. Prior to each measurement the single crystal electrode was flame annealed and cooled down to room temperature in an argon (Hoekloos, N50)-hydrogen atmosphere (ratio 3:1), after which it was transferred to the electrochemical cell under protection of a droplet of deoxygenated water.

A special electrochemical cell, described in ref. [37], contained a small movable spoon over an electrolyte reservoir, which allowed dosing of CO at open circuit potential (ocp) from a saturated CO solution without dissolving CO in the blank electrolyte. The saturation of the CO adlayer (reached after 2 minutes of emersion in the CO-containing solution) was checked by cycling in the hydrogen/(bi)sulfate region. The electrochemical cell was cleaned by boiling in a 1:1 mixture of concentrated sulfuric and nitric acid, followed by repeated boiling (four times) with ultra-pure water (Millipore MilliQ gradient A10 system, 18.2 MΩcm, 2 ppb total organic carbon). The blank electrolyte, 0.5 M H2SO4, was prepared with concentrated sulfuric acid (Merck, "Suprapur") and ultra-pure water. A second standard three-electrode cell was used for platinum deposition. The platinum was deposited on the electrode surface by cycling between 0.0 and 0.80 V vs.

99

Chapter 6

RHE at 50 mV/s in an electrolyte containing 0.5 M H2SO4 and 30 µM H2PtCl6 (Drijfhout). The methanol and formic acid containing electrolytes consisted of the blank electrolyte and 0.1 M methanol (Merck, pro analysi) or 0.1 M formic acid (Merck, extra pure).

During measurements and platinum deposition the electrolyte was deoxygenated with argon (N50) and the electrolyte in the container above the spoon was saturated with CO (Hoekloos, N47). A coiled platinum wire was used as a counter electrode and the reference electrode was a mercury-mercury sulfate electrode (MMSE: Hg|Hg2SO4|K2SO4

(sat)) connected via a Luggin capillary. However, all potentials in this chapter were converted to the RHE scale. Measurements were performed at room temperature (22 ºC) with a computer-controlled Autolab PGSTAT20 potentiostat.

Scanning Tunnelling Microscopy (STM) experiments were performed in air using a Nanoscope III system (Digital Instruments). The tips were prepared from 0.25-mm diameter Pt-Ir wires (80:20 alloy) by an electrochemical AC melt etching procedure. Thermal drift was found to be the main difficulty for obtaining correct measurements of interatomic spacing and non-distorted images. Thus, we used scan rates from 15 to 60 Hz to minimize the influence of drift.

After each experiment the deposited platinum was removed by repeated heating and subsequent quenching of the electrode in concentrated nitric acid. This procedure, although successful in removing the platinum (as is apparent from the blank cyclic voltammetry, which is very sensitive to the presence of platinum), damages the surface of the electrode. With time the defects introduced in this way may start to affect the measurements. In this respect alloying could be an issue of some importance and it is probably behind the final damage of the electrodes. During the nitric acid quenching and subsequent flame annealing Pt adatoms are probably introduced in the bulk of the Rh crystal. In oxidation atmospheres the surface will be enriched in Rh, as its avidity for adsorbates (in particular oxygen) is higher. For marcocrystals introducing a few ML of adatoms in the bulk is negligible, but in the long term alters the behavior, leading to an irreversible damage of the Rh single crystal.

6.3. Results and Discussion

6.3.1. System Cleanliness and Surface Quality

As is customary for single crystal electrode experiments, the system cleanliness was verified prior to each measurement. Because the CVs of rhodium single crystal electrodes are not as sensitive to contamination as those of platinum the cleanliness of the cell was checked by recording the CV of Pt(111) in a 0.5 M H2SO4 working electrolyte.

The quality of the flame-annealed rhodium (111) electrode was checked by cyclic voltammetry and CO adlayer stripping. At a scan rate of 20 mV⋅s-1 stripping of a saturated CO adlayer on well-ordered Rh(111) should take more than 30 cycles up to 1 V

100

CO oxidation on Pt modified Rh(111) electrodes

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9-1200

-1000

-800

-600

-400

-200

0

200

400

600

800 Pt-Rh(111) Rh(111)

j / µ

A cm

-2

E / V vs. RHE

Figure 6.1. Cyclic voltammograms of Rh(111) (solid line) and Pt4-Rh(111) (dashed line: 4 cycles of platinum deposition from 30 µM H2PtCl6) in 0.5 M H2SO4 at 20 mV⋅s-1.

vs. RHE.[31] Furthermore, as was pointed out in Chapter 3, the separation between the anodic and cathodic hydrogen/(bi)sulfate peaks should be 60mV or larger.[28] Cycling to potentials higher than the surface oxidation potential should be avoided in order to prevent surface disordering.(Chapter 2, refs. [27, 31])

6.3.2. Platinum Deposition and voltammetric Characterization

After a blank cyclic voltammogram (BCV) was recorded the Rh(111) electrode was transferred to the cell with the deposition solution and platinum was deposited by repeated cycling between 0.0 and 0.80 V vs. RHE. Ptn-Rh(111) surfaces with n=2, 4, 6, 8, 10 and 16 deposition cycles were made. Spontaneous (similar to a procedure used by the group of Adžić and coworkers [23, 24, 38, 39]) and galvanostatic deposition were also attempted but proved to be unreliable with respect to the amount of platinum deposited and, thus, also with respect to the reproducibility of the results.

Performing the deposition experiment without prior recording of a BCV results in a fast deposition of large quantities of platinum. As the deposition rate is difficult to control under these conditions, the reproducibility of the results is low. The absence of a strongly adsorbed (bi)sulfate adlayer is the most likely cause of this phenomenon.

In Fig. 6.1 a BCV of Rh(111) is plotted together with a voltammogram of a platinum modified Rh(111) surface (Pt4-Rh(111)). The influence of platinum on the blank cyclic voltammogram can be clearly seen. The anodic and cathodic hydrogen/(bi)sulfate peaks of the Pt4-Rh(111) electrode are sharper than for well-ordered Rh(111) and the separation between the peaks is considerably smaller. The increase in sharpness and reversibility of the hydrogen/(bi)sulfate peaks already occurs at low platinum concentrations (<5% of the surface covered, as determined by STM imaging) and remains as the concentration is increased further. Why platinum adsorption leads to a sharpening of the hydrogen/(bi)sulfate peaks is currently not understood and would need

101

Chapter 6

further investigation. Similar observations were also made by Gómez et al. for Rh-modified Pt(111) electrodes.[40] The small peak at 0.075 V in the negative-going scan (see also Fig. 6.2a) corresponds to the cathodic hydrogen/(bi)sulfate peak on Rh(111) and is, therefore, probably due to hydrogen adsorption and (bi)sulfate desorption on bare Rh(111) which is unaffected by the platinum regions. The corresponding oxidation feature presumably overlaps with the Pt-related peak at 0.12 V. Since the height of the sharpened hydrogen/(bi)sulfate peaks decreases for increasing platinum coverage, we believe these peaks are due to hydrogen/(bi)sulfate adsorbing/desorbing on the rhodium regions affected by nearby platinum particles or islands.

Increasing the platinum coverage on Rh(111) by increasing the number of deposition cycles (from 2 to 16 cycles, see Fig. 6.2a and b) results in a decrease in the height of the sharp hydrogen/(bi)sulfate peaks due to masking of the rhodium surface (not shown in Fig. 6.2a). Fig. 6.2a demonstrates that the increase in the platinum surface concentration also results in the appearance of distinct features between 0.1 and 0.4 V. The position and shape of the reversible peaks visible at ca. 0.27 V vs. RHE for the Pt16-Rh(111) surface correspond to those recorded for polycrystalline platinum in 0.5 M H2SO4. Therefore, we attribute the features between 0.1 and 0.4 V to hydrogen/(bi)sulfate adsorption/desorption on poly-oriented platinum.

It is interesting to note that the position and shape of the small peak at 0.075 V in the negative-going scan seems independent of the number platinum deposition cycles. Apparently, addition of platinum does not greatly affect the quantity of hydrogen adsorbing and (bi)sulfate desorbing on the remaining bare rhodium, which suggests three-dimensional growth of the platinum islands rather than epitaxially growth. This assumption is also supported by the fact that the total charge in the hydrogen/(bi)sulfate region increases as the Pt coverage is increased when going from 2 to 16 deposition cycles (suggesting an increased surface area), whereas the amount of “free” rhodium does not change significantly in this range. Only at very high platinum concentrations

102

Figure 6.2. Zoom of the low (a) and high potential regions (b) of the cyclic voltammograms of Ptn-modified Rh(111) (n=2, 4, 6, 8, 10, and 16 cycles of platinum deposition from 30 µM H2PtCl6) electrodes in 0.5 M H2SO4 at 20 mV⋅s-1. The graphs were corrected for the influence of oxygen traces.

0.5 0.6 0.7 0.8 0.9-10

-5

0

5

10

j / µ

A c

m-2

E / V vs. RHE

2 scans 4 scans 6 scans 8 scans 10 scans 16 scans

0.1 0.2 0.3 0.4

-40

-30

-20

-10

0

10

20

30

40

2 scans 4 scans 6 scans 8 scans 10 scans 16 scans

j / µ

A cm

-2

E / V vs. RHE

Rh(111)

(b) (a)

CO oxidation on Pt modified Rh(111) electrodes

does this peak disappear, meaning full coverage of the electrode surface. In the high potential region (Fig. 6.2b) the differences between Rh(111) and Pt-

Rh(111) are quite pronounced. In the previous chapters we reported that a well-ordered Rh(111) surface starts to oxidize at ca. 0.8 V vs. RHE,(Chapters 2 and 3, ref. [27, 28]) while Lebedeva et al. pointed out that water oxidation on well-ordered Pt(111) takes place at potentials well above 1.0 V vs. RHE.[34] Polycrystalline platinum surfaces start to oxidize at approximately 0.75 V vs. RHE. Yet, as is apparent from Fig. 6.2b incipient oxidation of the Pt-Rh(111) electrode can occur at potentials as low as 0.55 V. Remarkably, increasing the amount of platinum on the surface results in a decrease of the oxidation charge in this potential region. It seems that addition of platinum to a Rh(111) surface results in an enhancement of the surface oxidation process and, thus, a decrease in the surface oxidation potential. Because the oxidation charge decreases for increasing platinum surface concentration (i.e. decreasing “platinum free” rhodium surface area) and the onset potential for surface oxidation is considerably lower than for polycrystalline Pt, we assume that the oxidation potential of the substrate is affected by the platinum and not visa versa. When comparing Fig. 6.2b to Fig. 3.3 in Chapter 3 it can be seen that the surface oxidation potential of a Rh(111) electrode can be lowered by the deposition of platinum to a value close to that of the stepped Rh(331) surface.

Interestingly, it may be noted that such an effect is not unexpected from a theoretical point of view. While mixing Pt with Ru or Rh weakens the bond of CO and OH to Pt, it strengthens their bond to Ru or Rh.[20] A stronger OH bond to Rh would imply a lower surface oxidation potential.

6.3.3. STM Characterization

The surface structure of the Pt-Rh(111) electrodes with varying platinum coverages was investigated by Scanning Tunneling Microscopy (STM). Fig. 6.3a-c show STM images of Rh(111) with low, medium and high platinum surface concentrations, equivalent to 2, 8, and 16 deposition cycles, respectively.

In the STM images, rhodium can be identified by the dark atomic structure of hexagonal orientation visible in Fig. 6.3a and b and the darker patches in Fig. 6.3c. As the STM images were taken in air the apparent ordered array is probably due to the spontaneous formation of an oxygen adlayer on the Rh(111) substrate with a (2×2) structure on the basis of the symmetry and distance between adjacent maxima.[36] The lighter colored patches are the platinum islands. The STM images show that platinum does not grow epitaxially, but rather forms islands (Volmer-Weber island growth [41, 42]) on the rhodium substrate. For bi-metallic systems (in thermodynamic equilibrium) a Volmer-Weber growth mode is expected when the sum of the relative free surface energies (in this case ∆γ = γPt + γRh-Pt - γRh) of the pure metal (γPt), the pure substrate (γRh) and the interface (γRh-Pt) is smaller than 0,[41-43] as is apparently the case for our platinum-rhodium system. This makes it difficult to produce platinum islands of mono-atomic height on a rhodium substrate using electrodeposition. However, Fig. 6.3a shows

103

Chapter 6

(a) (b)

(c)

Figure 6.3. STM image and cross section analysis of a Pt modified Rh(111) surface with (a) low (i.e.equivalent to 2 deposition cycles), (b) medium (i.e. equivalent to 8 deposition cycles) and (c) highplatinum coverages (i.e. equivalent to 16 deposition cycles).

that, at low platinum surface concentrations, small islands of mono-atomic height (as indicated by the section analysis) and approximately 2-4 nm in diameter are formed. Increasing the amount of deposited platinum leads to an increase in the size and height of

104

CO oxidation on Pt modified Rh(111) electrodes

the islands. At medium platinum concentrations the average height of the islands is 1 to 2 monolayers and the average diameter of the platinum particles has grown from 2 to 5-10 nm (see Fig. 6.3b). Most of the platinum islands are of monoatomic height with a few di-atomically high particles. Fig.6.3b also shows in the upper part a monoatomic step of the Rh(111) substrate, where platinum deposition seems to be more favorable compared to the terrace sites. When 16 cycles of platinum are deposited (Fig. 6.3c) the surface is covered with relatively homogeneously shaped particles of ca. 10-15 nm in diameter and approximately 5-6 layers high. The amount of exposed rhodium has decreased slightly with respect to the other surfaces and no more “monolayer” islands can be found.

Using the hydrogen adsorption/desorption charge obtained from the cyclic voltammograms and the results from the STM analysis, the maximum platinum coverage before multilayer growth occurs was estimated to be 0.16 (calculated for the Pt8-Rh(111) surface).

6.3.4 Cyclic voltammetry saturated CO Adlayer Oxidation

Presented in Fig. 6.4 are the voltammetric profiles (first scans) for the oxidation of a saturated CO adlayer on the Pt-modified Rh(111) surfaces. The raw stripping charges in each CV are similar and lie within 497 ± 35 µC cm-2 (obtained by integrating from 0.5 to 0.9 V vs. RHE and dividing by the geometrical surface area of the rhodium electrode), which indicates that the initial CO coverage is similar on all surfaces. It should be noted that this value is ca. 80 µC cm-2 higher than what we previously reported for CO stripping at Rh[n(111)×(111)] electrodes (Chapter 2) and 110 µC cm-2 higher than the stripping charge reported for pure Rh(111) by Sung et al. and Lam-wing et al.[44, 45] The discrepancy may be due to the enhanced surface oxidation and increased surface area of the Ptn-Rh(111) electrodes resulting from the three-dimensional growth

0.65 0.70 0.75 0.80 0.85 0.90

0

50

100

150

200

250

300

2

4

6

8

10

16

0.5 0.6 0.7 0.8 0.9-3-2-1012345

j / µ

A c

m-2

E / V vs. RHE

j / µ

A cm

-2

E / V vs. RHE

Figure 6.4. CO adlayer stripping on Pt modified Rh(111) in 0.5 M H2SO4.at 20 mV⋅s-1. The numeric labels indicate the number of deposition cycles in 30 µM H2PtCl6. The inset shows the first scan of CO adlayer stripping on well-ordered Rh(111).

105

Chapter 6

of the platinum islands. In fact, because the surface area increases as the platinum islands start to grow three-dimensionally, one would expect an increase in the integrated stripping charge density. However, this increase cannot be detected within the error of our experiments. It is possible that the expected increase in CO stripping charge is (at least partially) offset by the decreasing charge of the “enhanced” surface oxidation as the surface is covered more and more by platinum islands. Or perhaps adsorption of CO on the three dimensional islands is hindered, leading to a lower saturation CO coverage than on a platinum monolayer of identical surface area.

Contrary to well-defined Rh(111) the Pt-modified Rh(111) electrodes require only a single scan at 20 mV⋅s-1 to completely strip a CO adlayer and produce a blank cyclic voltammogram. The oxidation wave for surfaces with a low Pt coverage is shifted approximately 150 mV negative with respect to the main oxidation peak on well-ordered Rh(111) (see insert in Fig. 3.4 and Chapter 3), more than 250 mV more negative than reported for Pt(111) [34] and ca. 100 mV more negative than for polycrystalline Pt. This demonstrates the increased activity of the Pt-Rh(111) surface towards CO oxidation with respect to the pure substrate.[24]

At low and medium platinum coverages (less than 10 deposition cycles) the shape of the curves is characterized by two regions: a feature at low potentials starting at approx. 0.63 V vs. RHE (maximum at ca. 0.675 V) and a peak at higher potentials, which shifts positively for increasing platinum surface concentrations. Even though the position of the low potential peak is insensitive to the structure of the surface its corresponding charge initially increases when going from 2 to 4 deposition cycles, after which it decreases to zero for surfaces with more than 10 cycles of platinum deposition. The position of the second peak depends markedly on the surface structure and shifts from 0.75 V for 2 and 4 deposition cycles to approximately 0.82 V vs. RHE for 10 cycles. Further addition of platinum results in a symmetrical peak, which is shifted negatively to 0.77 V vs. RHE, and which resembles the CO stripping voltammetry recorded on polycrystalline platinum.

The changing shape of the stripping CVs can be related to changes in the surface structure as deduced from our STM results and the blank cyclic voltammetry. At low and medium platinum coverage, where STM images indicate the presence of Pt islands of mono- and diatomic height, the CO overpotential is considerably lower than for the pure substrate or pure ad-metal. [27, 34] Since, on well-defined Rh(111), CO oxidation does not occur appreciably below ca. 0.8 V vs. RHE the low potential peak in the stripping voltammetry may be attributed to either CO oxidizing on the Pt monolayer, or close to Pt-Rh(111) boundaries. The second, high potential peak may be attributed to CO oxidizing on the platinum islands of di- or multi-atomic height. As the islands grow progressively larger and higher with each scan, the monolayer islands disappear, which results in the disappearance of the low potential CO stripping peak, and a shift of the second peak towards higher potentials.

It is interesting to compare the CO oxidation profiles presented here with those corresponding to CO saturated layers on Rh-Pt(111) electrodes.[46] In the submonolayer

106

CO oxidation on Pt modified Rh(111) electrodes

range these surfaces display a behavior similar to that presented here for low Pt coverages. In fact, the introduction of Rh adatoms on the Pt(111) surface also provoked a negative shift in the oxidation peak for the saturated CO adlayer, which showed consistently two clearly discernible contributions. Therefore, in the absence of nanoparticle formation (three-dimensional growth), the appearance of bi-metallic sites at the edges of the islands seems to be the main factor determining the electrocatalytic properties of the bi-metallic surfaces. The fact that the catalytic promotion is more important for Pt-Rh(111) than for Rh-Pt(111) is probably due to the increased adsorption capability of the Rh surface atoms in the former case, which leads to an increase in the underpotential shift for the adsorption of OH. Seemingly, for CO electrooxidation the formation of adsorbed hydroxyls is more important than the possible effect of the different bi-metallic surfaces on the CO adsorption energy.

Thus, apart from the expected weakening of the Pt-CO bond, the decrease in the surface oxidation potential of rhodium induced by the presence of platinum (shown in Fig. 6.2b) may contribute considerably to the enhancement of the CO oxidation reaction. Still, we believe that this effect cannot fully account for the lowering of the CO oxidation overpotential. The low potential peak, visible in the CO stripping voltammetry, disappears for higher platinum concentrations (>10 deposition cycles), while the blank CV still shows bare Rh(111) sites and surface oxidation enhancement even for high platinum concentrations (up to Pt16-Rh(111)).

A third explanation for the lower CO oxidation potential may be found in a reduced anion adsorption at the Pt-Rh(111) boundaries. If (bi)sulfate cannot adsorb strongly at the boundary between the Pt islands and Rh(111) surface, while OH can, the CO oxidation potential is expected to be considerably lower. Although the extent of this effect cannot be estimated from the obtained results, it may play an important role in lowering the oxidation potential of CO and should not be ignored.

The positive shift of the CO oxidation peaks with increasing Pt coverages to potentials higher than reported for polycrystalline platinum may be explained by the particle size effect. The effect of particle size on the CO electrooxidation reaction has been studied for a number of different systems.[47-55] For gold supported Pt nano-particles, obtained from a colloidal Pt solution, Friedrich et al. showed that the electrooxidation overpotential grows from 100 to 500 mV on particles of ca. 3 nm in diameter with respect to polycrystalline platinum surfaces.[50, 51] Larger particles of 10-16 nm in diameter also exhibit a positive shift, although markedly smaller. Similar observations were also made by Maillard et al. and Cherstiouk et al.[52-54] for CO adlayer stripping on platinum nano-particles supported on glassy carbon (generated by reduction of dried droplet of a H2PtCl4 solution in H2 gas). They found that as the particle size is decreased below 3 nm the overpotential for CO oxidation shifts to considerably higher potentials. Restricted mobility of COads on smaller particles was suggested to be responsible for this phenomenon.

This particle size effect may also apply to our Pt-Rh(111) system and can account for the positive shift of the second oxidation peak. Since the 2-4 nm large islands, visible

107

Chapter 6

in Fig. 6.3a, are mono-atomic in height they behave as Pt monolayer systems. When the islands grow to diatomic and multi-atomic height the predicted overlayer behavior diminishes and is gradually replaced by the particle size effect. That the positive shift on the Pt-Rh(111) particles is less pronounced than reported in the literature may be due to the fact that the geometry of the particles generated by voltammetric deposition is different. The particles are flatter and larger in diameter than particles generated from colloidal solutions [50, 51] or reduction of a droplet of platinum containing solution [52-54] and, accordingly, can be expected to react differently. As the particles reach a size of about 10nm in diameter, the properties of polycrystalline platinum start to dominate. Consequently, the CO oxidation peak shifts to lower potentials.

Finally, we would like to mention that multiple peaks in the CO stripping voltammetry have also been observed on other bi-metallic systems (for a recent review, see ref. [56] and references therein). The electrooxidation of CO on Ru-Pt(111) electrodes displays two well-resolved peaks, one at low and one at higher potentials. As in our case, the low potential peak was ascribed to CO reaction on or next to the Ru islands, while the peak located at high potentials was ascribed to CO chemisorbed on “pure” platinum regions.

6.3.5 Chronoamperometry for the saturated CO-adlayer Oxidation

The kinetics of the CO electrooxidation reaction on Pt modified Rh(111) electrodes was also investigated by chronoamperometry. Fig. 6.5 shows three current-time transients recorded on a Pt4-Rh(111) surface. According to the stripping voltammetry this surface has the highest area of “monolayer” Pt islands. The final potentials under investigation lie just before the onset of the oxidation wave (0.635 V vs. RHE) and at the maximum of the first (0.675 V vs. RHE) and the second oxidation peaks (0.750 V vs. RHE). When comparing the obtained current-time curves to transients recorded on well-ordered Rh(111) (see Chapter 3) and stepped Pt electrodes (see refs. [32, 33]) the enhanced catalytic activity of the Pt-Rh(111) system for the CO electrooxidation reaction becomes apparent. For identical final potentials the oxidation wave on Pt-Rh(111) lies at noticeably shorter times than for Rh(111), Pt(111) and stepped Pt electrodes (compare Fig 6.5 to Fig 3.6a in Chapter 3, Fig 3 in ref. [32] and Fig 4 in ref. [33]). Moreover, contrary to well-ordered Rh(111), complete stripping of a CO adlayer from the Pt-Rh(111) electrode is already possible at potentials as low as 0.675 V. Even when the potential is stepped to the onset of the CO oxidation current (0.635 V vs. RHE) a large part (ca. 80%) of the CO adlayer can be oxidized within 250 seconds.

The current-time transients recorded are characterized by a sharp peak at short times, followed by a second peak at longer times. Based on the STM and cyclic voltammetry results we conclude that the peak visible at short times is most likely due to CO oxidizing on the mono-atomically high islands, whereas the second peak at longer times may be ascribed to CO, which is adsorbed on the Rh(111) sites. The fact that the current decay in the transients recorded at 0.635 and 0.675 V is Cottrellian in nature

108

CO oxidation on Pt modified Rh(111) electrodes

0 20 40 60

0

25

50

75

100

125

150

175

j / µ

A cm

-2

t / s

0.10 0.12 0.14 0.16 0.18 0.20

01234567

j / µ

A c

m-2

t-0.5 / s-0.5

0.635 V

0.12 0.16 0.20 0.24 0.28

0

2

4

6

8

10

120.675 V

j / µ

A c

m-2

t-0.5 / s-0.5

(3)

(2)

(1) 0.750 V(2) 0.675 V(3) 0.635 V

(1)

Figure 6.5. Current transients for the oxidation of saturated CO adlayers on Pt4-Rh(111). Step potentials are listed in the figure. The insets show the Cottrellian current decay of the transient recorded on Pt4-Rh(111) at 0.635 and 0.675 V as is shown by the linear relationship between the current density and t-0.5.

agrees well with this assumption (see the insets in Fig. 6.5), as such transients have been observed for Rh(111) and Rh(554).(Chapter 3, ref. [28]) At higher final potentials the tail of the second peak disappears due to enhanced surface oxidation.

For pure Cottrellian behavior the current should drop to 0 when the time goes to infinity. The facts that the Cottrell plots in Fig. 6.5 do not pass the origin is probably due to effects of a second faradaic process, most likely surface oxidation.

6.3.6 Cyclic Voltammetry in Methanol and Formic Acid

The effects of the Pt modified Rh(111) electrodes on the electrooxidation of methanol and formic acid were also investigated by cyclic voltammetry. The CVs for the methanol-containing electrolyte are shown in Fig. 6.6, while the CVs recorded in formic acid are shown in Fig. 6.7.

As is apparent from Fig. 6.6, the activity of Rh(111) towards the oxidation of methanol is low and addition of a small amount of platinum to the Rh(111) surface results in a pronounced increase in the activity (compare the CV for Rh(111) to the Pt2-Rh(111) and Pt4-Rh(111) CVs in Fig. 6.6). However, when the platinum coverage is increased beyond Pt4-Rh(111) the current density increases only slowly for each additional Pt deposition cycles (n = 6 and 8). This slow current increase is probably related to the three-dimensional growth, which becomes more pronounced for more than deposition cycles. The onset of the methanol oxidation lies at ca. 0.5 V vs. RHE and is similar to the onset potential recorded on stepped Pt surfaces and Pt nanoparticles (see Chapters 7-9, refs. [57, 58]).

109

Chapter 6

0.5 0.6 0.7 0.8 0.9 1.0

0

25

50

75

100

125

150

j / µ

A cm

-2

E / V vs. RHE

Rh(111) Pt2-Rh(111) Pt4-Rh(111) Pt6-Rh(111) Pt8-Rh(111) Pt10-Rh(111)

0.2 0.4 0.6 0.8 1.00

100

200

300

400

j / µ

A cm

-2

E / V vs. RHE

Rh(111) Pt2-Rh(111) Pt4-Rh(111) Pt6-Rh(111) Pt8-Rh(111) Pt10-Rh(111)

Pt10-Rh(111)

Rh(111)

Figure 6.6. MeOH oxidation on Pt modified Rh(111) in 0.1 M MeOH, 0.5 M H2SO4 at 20 mV⋅s-1. The numeric labels indicate the number of deposition cycles in 30 µM H2PtCl6.

Figure 6.7. HCOOH oxidation on Pt modified Rh(111) in 0.1 M HCOOH, 0.5 M H2SO4 at 20 mV⋅s-1. The numeric labels indicate the number of deposition cycles in 30 µM H2PtCl6.

For formic acid (Fig. 6.7), increasing the platinum surface concentration simply leads to an increase in the maximum current density. This behavior is typical for formic acid electrooxidation on Pt nanoparticles. Therefore, we can conclude that although the Ptn-Rh(111) electrodes displays an enhanced CO tolerance, the methanol and formic acid oxidation characteristics are as one would be expected of Pt nanoparticles.

6.4. Conclusion

The CO electrooxidation reaction on platinum modified Rh(111) was studied by means of cyclic voltammetry, CO stripping voltammetry, chronoamperometry and STM imaging. Platinum was deposited on the surface by cycling the electrode in a H2PtCl6 containing solution. The deposition of a small amount of platinum causes the hydrogen/(bi)sulfate peaks of the rhodium substrate to become sharper and shift closer together. The presence of surface platinum also leads to a pronounced enhancement of the surface oxidation process, which is visible in the CVs from ca. 0.55 V vs. RHE. Because the (enhanced) surface oxidation charge decreases when the substrate becomes covered with more platinum, we conclude that platinum enhances the oxidation of the rhodium surface, and not vice versa.

STM analysis of the generated surfaces shows that, although energetically unfavorable, it is possible to produce mono-layer platinum islands on the rhodium substrate by keeping the number of deposition cycles low (<4 cycles). When the surface platinum concentration is increased, the number of islands of mono-atomic height decreases due to three-dimensional growth, and finally, multi-atom high particles are formed.

The CO stripping voltammetry on Pt-Rh(111) is characterized by two peaks: a peak at low potential, which we ascribe to CO reacting on the Pt monolayer, and a peak at higher potential, which was ascribed to CO oxidizing from Pt islands of multi-layer

110

CO oxidation on Pt modified Rh(111) electrodes

height. The signature of the enhanced catalytic properties of the PtML-Rh(111) system with respect to CO oxidation is the onset of the low potential peak, which lies approximately 150 mV lower than reported for pure Rh(111), (see Chapter 3) over 250 mV lower than for pure Pt(111) [34] and ca. 100 mV lower than for polycrystalline Pt. This oxidation onset is also lower by 60 mV than that corresponding to the RhML-Pt(111) bimetallic surface.[59] The unique catalytic properties of these Pt overlayer islands may result from a weakened CO bond to PtML on Rh(111) with increased OH bonding to the free Rh(111) sites, in correspondence with predictions based on DFT calculations. Comparison with the parent Rh-Pt(111) surface suggests that the second effect is more important. As the islands grow in size and height, the catalytic effect of the monolayer diminishes and is gradually replaced by a particle size effect, which causes the CO oxidation potential to shift to potentials well over that reported for polycrystalline platinum. Further increase of the platinum concentration leads to larger particles and to polycrystalline CO stripping characteristics.

Current-time transients recorded for the surface with the highest amount of monolayer islands also indicate enhanced CO oxidation kinetics. Comparison of the Pt4-Rh(111) current time transients to those of pure Rh(111) and Pt(111) shows greatly reduced reaction times. A Cottrellian decay at long times indicates surface diffusion limited CO oxidation on the bare Rh(111) surface, while the peak visible at short times is indicative of CO reacting at the monolayer platinum islands.

For the electrooxidation of methanol and formic acid no anomalous behavior of the Pt modified Rh(111) electrodes was found. The electrocatalytic properties correspond to those of small Pt nanoparticles.

The results presented in this chapter show that it is possible to generate small monolayer islands of platinum on a Rh(111) surface. Moreover, as predicted by DFT calculations, CO adlayer oxidation on this system is enhanced compared to both pure Rh and Pt. Therefore, these surfaces do not only have a positive electronic effect on CO tolerance, but are also beneficial to the bifunctional mechanism of CO electrooxidation.

Acknowledgements: The research was supported by the Netherlands Foundation for Scientific Research (NWO) and MCYT (Spain) through project BQU2003-04029.

References

[1] A. Hamnett, Catal. Today. 38 (1997) 445. [2] G. Hoogers and D. Thompsett, Cattech 3 (2000) 106. [3] N. M. Markovic and P. N. Ross, Surf. Sci. Rep. 45 (2002) 117. [4] H. A. Gasteiger, N. Markovic, P. N. Ross, Jr., and E. J. Cairns, J. Phys. Chem. 98 (1994) 617. [5] H. A. Gasteiger, N. M. Markovic, and P. N. Ross, Jr., J. Phys. Chem. 99 (1995) 8290. [6] H. A. Gasteiger, N. M. Markovic, and P. N. Ross, Jr., J. Phys. Chem. 99 (1995) 16757. [7] N. M. Markovic, H. A. Gasteiger, P. N. Ross, Jr., X. Jiang, I. Villegas, and M. J. Weaver,

Electrochim. Acta. 40 (1995) 91. [8] B. N. Grgur, N. M. Markovic, and P. N. Ross, Jr., J. Phys. Chem. B 102 (1998) 2494. [9] B. N. Grgur, N. M. Markovic, and P. N. Ross, Electrochim. Acta. 43 (1998) 3631. [10] N. M. Markovic and P. N. Ross, Electrochim. Acta. 45 (2000) 4101.

0.75 V

111

Chapter 6

[11] R. Ianniello, V. M. Schmidt, U. Stimming, J. Stumper, and A. Wallu, Electrochim. Acta. 39 (1994) 1863.

[12] W. Chrzanowski, H. Kim, and A. Wieckowski, Catal. Lett. 50 (1998) 69. [13] A. Wieckowski, Interfacial Electrochemistry: Theory, Experiment, and Applications, Marcel

Dekker, New York, 2000. [14] J. C. Davies, B. E. Hayden, and D. J. Pegg, Electrochim. Acta 44 (1998) 1181. [15] A. Kabbabi, R. Faure, R. Durand, B. Beden, F. Hahn, J. M. Leger, and C. Lamy, J. Electroanal.

Chem. 444 (1998) 41. [16] W. F. Lin, M. S. Zei, M. Eiswirth, G. Ertl, T. Iwasita, and W. Vielstich, J. Phys. Chem. B 103

(1999) 6968. [17] R. Liu, H. Iddir, Q. Fan, G. Hou, A. Bo, K. L. Ley, E. S. Smotkin, Y. E. Sung, H. Kim, S.

Thomas, and A. Wieckowski, J. Phys. Chem. B 104 (2000) 3518. [18] M. Watanabe and S. Motoo, J. Electroanal. Chem. Inter. Electrochem. 60 (1975) 275. [19] E. Christoffersen, P. Liu, A. Ruban, H. L. Skriver, and J. K. Norskov, J. Catal. 199 (2001) 123. [20] M. T. M. Koper, T. E. Shubina, and R. A. van Santen, J. Phys. Chem. B 106 (2002) 686. [21] T. E. Shubina and M. T. M. Koper, Electrochim. Acta. 47 (2002) 3621. [22] P. Liu, A. Logadottir, and J. K. Norskov, Electrochim. Acta. 48 (2003) 3731. [23] S. R. Brankovic, J. X. Wang, and R. R. Adzic, Electrochem. Solid State Lett. 4 (2001) A217. [24] S. R. Brankovic, N. S. Marinkovic, J. X. Wang, and R. R. Adzic, J. Electroanal. Chem. 532 (2002)

57. [25] K. Sasaki, J. X. Wang, M. Balasubramanian, J. McBreen, F. Uribe, and R. R. Adzic, Electrochim.

Acta. 49 (2004) 3873. [26] K. Sasaki, Y. Mo, J. X. Wang, M. Balasubramanian, F. Uribe, J. McBreen, and R. R. Adzic,

Electrochim. Acta. 48 (2003) 3841. [27] T. H. M. Housmans, J. M. Feliu, and M. T. M. Koper, J. Electroanal. Chem. 572 (2004) 79. [28] T. H. M. Housmans and M. T. M. Koper, J. Electroanal. Chem. 575 (2005) 39. [29] T. H. M. Housmans and M. T. M. Koper, Electrochem. Comm. 7 (2005) 581. [30] T. H. M. Housmans and M. T. M. Koper, to be submitted (2005) [31] R. Gomez, J. M. Orts, J. M. Feliu, J. Clavilier, and L. H. Klein, J. Electroanal. Chem. 432 (1997)

1. [32] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Electroanal. Chem. 524-

525 (2002) 242. [33] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Phys. Chem. 106 (2002)

12938. [34] N. P. Lebedeva, M. T. M. Koper, E. Herrero, J. M. Feliu, and R. A. van Santen, J. Electroanal.

Chem. 487 (2000) 37. [35] N. P. Lebedeva, A. Rodes, J. M. Feliu, M. T. M. Koper, and R. A. van Santen, J. Phys. Chem. B

106 (2002) 9863. [36] J. Clavilier, M. Wasberg, M. Petit, and L. H. Klein, J. Electroanal. Chem. 374 (1994) 123. [37] J. M. Feliu, J. M. Orts, A. Fernandez-Vega, A. Aldaz, and J. Clavilier, J. Electroanal. Chem. 296

(1990) 191. [38] S. R. Brankovic, J. McBreen, and R. R. Adzic, J. Electroanal. Chem. 503 (2001) 99. [39] S. R. Brankovic, J. X. Wang, Y. Zhu, R. Sabatini, J. McBreen, and R. R. Adzic, J. Electroanal.

Chem. 524-525 (2002) 231. [40] R. Gómez and J. M. Feliu, Electrochim. Acta. 44 (1998) 1191. [41] E. Bauer, Z. Krist. 110 (1958) 395. [42] E. Bauer and J. H. Van der Merwe, Phys. Rev. B 33 (1986) 3657. [43] A. Zangwill, Physics at Surfaces, Cambridge Univ. Press, Cambridge, 1988. [44] Y. E. Sung, S. Thomas, and A. Wieckowski, J. Phys. Chem. 99 (1995) 13513. [45] S.-C. C. Lam-Wing H. Leung, and Michael J. Weaver, J. Phys. Chem. 90 (1989) 7426. [46] R. Gómez, F. J. Gutierrez de Dios, and J. M. Feliu, Electrochim. Acta. 49 (2004) 1195. [47] Y. Takasu, N. Ohashi, X. G. Zhang, Y. Murakami, H. Minagawa, S. Sato, and K. Yahikozawa,

Electrochim. Acta. 41 (1996) 2595. [48] M. L. Sattler and P. N. Ross, Ultramicroscopy 20 (1986) 21. [49] L. J. Bregoli, Electrochim. Acta. 23 (1978) 489. [50] K. A. Friedrich, F. Henglein, U. Stimming, and W. Unkauf, Electrochim. Acta. 45 (2000) 3283. [51] K. A. Friedrich, F. Henglein, U. Stimming, and W. Unkauf, Colloids and Surfaces, A:

Physicochemical and Engineering Aspects 134 (1998) 193. [52] F. Maillard, M. Eikerling, O. V. Cherstiouk, S. Schreier, E. Savinova, and U. Stimming, Faraday

Discuss. 125 (2003) 357.

112

CO oxidation on Pt modified Rh(111) electrodes

[53] O. V. Cherstiouk, P. A. Simonov, V. I. Zaikovskii, and E. R. Savinova, J. Electroanal. Chem. 554-555 (2003) 241.

[54] O. V. Cherstiouk, P. A. Simonov, and E. R. Savinova, Electrochim. Acta. 48 (2003) 3851. [55] A. Wieckowski, E. R. Savinova, and C. G. Vayenas, Catalysis and Electrocatalysis at

Nanoparticle Surfaces, Marcel Dekker inc., New York, 2003. [56] J. S. Spendelow and A. Wieckowski, Phys. Chem. Chem. Phys. 6 (2004) 5094. [57] T. H. M. Housmans, A. H. Wonders, and M. T. M. Koper, J. Am. Chem. Soc. submitted (2005) [58] T. H. M. Housmans and M. T. M. Koper, J. Phys. Chem. B 107 (2003) 8557. [59] R. Gomez and J. M. Feliu, Electrochim. Acta. 44 (1998) 1191.

113

Chapter 6

114

Chapter 6

114

Methanol oxidation on stepped Pt[n(111)×(111)] single crystal electrodes: a chronoamperometric study

Abstract

The methanol oxidation reaction has been studied on Pt[n(111)×(111)] type electrodes in a 0.5 M sulfuric acid and 0.025 M methanol solution, using cyclic voltammetry and chronoamperometry. The voltammetric profiles of methanol on the three electrodes under investigation (Pt(111), Pt(554) and Pt(553)) shows that the overall oxidation rate increases with increasing step density and that the defects are affected more by the presence of methanol than terraces. The latter implies that either the decomposition products of methanol or methanol itself preferably sits at the steps. Investigation of the chronoamperometric data shows that the steady-state current, recorded at 900 seconds after the start of the experiment, increases with increasing step density. Moreover, surfaces with a higher step density display a faster dropping current, which suggests that the decomposition of methanol into CO poisoning species also preferentially takes place on the steps and defects. Unlike the stepped electrodes, most transients recorded on Pt(111) showed an initial current increase, which may be explained by the CO oxidation being faster than the methanol decomposition. This low decomposition rate is probably the result of a sufficiently low defect density and the low methanol concentration used in our experiments. Fitting the chronoamperometric data with a mathematical model, which includes the methanol decomposition reaction, the CO oxidation reaction and the direct methanol oxidation reaction, suggests that steps and defects catalyze all these reactions. Furthermore, the model indeed predicts that when the CO oxidation rate is faster than the decomposition rate a rising current transient can be expected, as was seen for Pt(111).

This chapter is published as T.H.M. Housmans, M.T.M. Koper, J. Phys. Chem. 107 (2003) 8557

Chapter 7

7.1. Introduction

From a practical and theoretical point of view, one of the most promising fuels for a fuel cell is methanol.[1-7] Firstly, it is a liquid at room temperature and can, therefore, easily be introduced in the already existing fuel distribution system and secondly, it can be produced in large quantities and it is relatively safe in handling. However, from the fuel cell technology point of view a more important consideration for choosing methanol is the fact that it can be catalytically oxidized on a platinum surface in aqueous environment yielding CO2 and 6 electrons per methanol molecule:

−+ ++→+ eHCOOHOHCH 66223 (7.1) This reaction has a very promising thermodynamic potential of 0.04 V versus the

reversible hydrogen electrode (RHE) and may, theoretically, allow for a power as high as that of a hydrogen-based fuel cell. Unfortunately, the fact that the decomposition reaction of methanol on platinum produces surface poisoning species, leads to a low catalytic activity and presents a severe inhibition for the development of the Direct Methanol Fuel Cell (DMFC).

Carbon monoxide has been identified in many studies as the primary poisoning species.[1, 4-6, 8] The electrocatalytic oxidation of CO, both in adsorbed and dissolved form, was found to be a structure-sensitive process.[4, 9, 10] In a study on the role of crystalline defects on the electrocatalytic oxidation of carbon monoxide, Lebedeva et al.[11-13] showed that CO is both preferentially adsorbed and oxidized at steps and defects in the platinum electrode surface. The reaction rate constant was found to depend linearly on the fraction of steps of (110) orientation on surfaces with (111) terrace orientation.[12]

Like the oxidation of CO, the methanol decomposition on platinum has also been reported to have a pronounce structure sensitivity. Under Ultra-High-Vacuum (UHV) conditions, Gibson and Dubois [14] proved that the thermal decomposition of methanol is structure sensitive by showing that a small amount of defects on the surface resulted in CO being formed at 200 K, while on a defect free surface no carbon monoxide could be detected. In a theoretical study of the methanol oxidation reaction, Desai et al.[15] showed, using density functional calculations, that defect-free Pt(111) is unreactive towards the decomposition of methanol to form carbon monoxide. Under electrochemical conditions, Clavilier et al.[16, 17] demonstrated structure sensitivity by comparing voltammograms of the methanol oxidation reaction on the three basal planes of platinum. These experiments were later augmented by chronoamperometric, spectroscopic and kinetic isotope studies, which all indicated a high structure sensitivity of the reaction.[18, 19] Of the three basal planes of platinum, Pt(111) was found to be the least reactive towards methanol decomposition, while (110) was found to be the most active. As a result, the surface poisoning process is slowest on (111) and fastest on (110). However, the Pt(111) electrode was found to give only a small initial current that degrades to an even lower level with time, while the Pt(110) surface deactivates much faster, but remains the most efficient in oxidizing methanol.[19]

116

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

Despite these numerous studies still no clarity exists on how exactly the step density influences the reaction. From an infrared study of the CO formation on step and terrace sites on a platinum electrode surface during methanol oxidation, Shin and Korzeniewski [20] concluded that methanolic CO formation is inhibited on Pt(111) at potentials in the classical hydrogen adsorption region and that the corrugated Pt(335) (a Pt[n(111)×(100) type electrode with n=4) surface plane promotes methanol dissociative chemisorption in this potential region, thus suggesting that defects catalyze methanol decomposition. On the other hand, Tripković et al.[21] reported that the initial surface activity of methanol on stepped electrodes decreases with increasing step density in the sequence: Pt(755)>Pt(211)> Pt(311) (Pt[n(111)×(100)] type electrodes with n=6, 3 and 2 respectively), which seems contradictory to Shin and Korseniewski’s findings. In a recent article on methanol oxidation on carbon-supported platinum nanoparticles, Weaver et al.[22] suggested that in order to explain the observation that the methanol oxidation rate decreases with decreasing particle size, methanol decomposition requires terrace sites rather than defect or step sites.

Given this rather unsatisfactory understanding of the influence of surface defects on methanol oxidation reaction, our goal in this chapter is to investigate the decomposition and oxidation of methanol on a series of stepped platinum single crystal electrodes of Pt[n(111)×(111)] orientation. Conventional techniques like cyclic voltammetry will be used to investigate the overall electrooxidation rates of methanol on the different stepped surfaces. However, the rates of the different reaction steps will be investigated by potential step experiments and a new model, similar to models proposed by the Stuve [8, 23, 24] and Wieckowski groups [25-27], will be introduced to fit the obtained chronoamperometric data and gain insight into the kinetics of the decomposition and oxidation reaction.

7.2. Experimental Setup

A conventional electrochemical cell was cleaned by boiling in a 1:1 mixture of concentrated sulfuric and nitric acid, followed by repeated boiling (4 times) with ultra-pure water. The counter electrode consisted of a coiled platinum wire and the reference consisted of a mercury-mercury sulfate electrode (MMSE: Hg|Hg2SO4|K2SO4 (sat)) electrode connected via Luggin capillary. However, all potentials in this chapter were converted to the RHE scale.

The working electrodes were platinum bead type single crystals of Pt[n(111)×(111)] (which is identical to Pt[(n-1)(111)×(110)]) orientation - Pt(553) having n=5, Pt(554) with n=10 and Pt(111) with n=200-500 - which were prepared according to Clavilier’s method.[28] Prior to each measurement the single crystal electrode was flame annealed and cooled down to room temperature in an argon (N50)-hydrogen atmosphere, after which it was transferred to the electrochemical cell under protection of a droplet of deoxygenated water.[13]

117

Chapter 7

The blank electrolyte consisting of 0.5 M H2SO4 was prepared with concentrated sulfuric acid (Merck, “Suprapur”) and ultra-pure water (Millipore MilliQ gradient A10 system, 18.2 MΩcm, 2 ppb total organic carbon). The working solution consisted of the blank electrolyte with 0.025 M Methanol (Merck, pro analysi, 99.8%). Argon (Hoekloos, N50) was used to deoxygenate all solutions.

All measurements were performed at room temperature (22°C) using a computer-controlled potentiostat (Eco chemie PGSTAT20, autolab).

7.3. Mechanism and Modeling of the Methanol Oxidation Reaction

Although the decomposition and oxidation reaction of methanol on platinum has been studied extensively in the past, the exact mechanism containing all the reactions and kinetic parameters still remains unclear. Early studies assumed that the reaction does not proceed through a series of consecutive reactions,[1] but rather proceeds through a parallel path mechanism (see Fig. 7.1a), where the poisoning species, now generally accepted to be surface-bonded carbon monoxide, is formed in a unwanted parallel reaction (indirect pathway, rate rdec). CO2 may be formed by the direct oxidation of the methanol (direct pathway, rate rd) or by oxidation of the adsorbed CO at increased overpotential (rate rox). This model has been applied to reactions on single crystal surfaces by a number of authors.[8, 19, 23-25, 27]

In a chronoamperometric analysis of methanol oxidation on Pt(111), (110) and (100), Franaszczuk and co-workers [25] modeled the poisoning effect, visible in the current-time transients, using only the direct methanol oxidation and COad formation reaction from Fig. 7.1a. Their expression for the methanol-related oxidation current density recorded in these transients contained four fitting parameters.

rdec , indirect pathway (b)(a)

Figure 7.1. (a) Simplified schematic representation of the parallel pathway for methanol oxidation on platinum electrodes. (b) A more advanced schematic representation of the parallel pathway mechanism for methanol oxidation on platinum electrodes, which incorporates DEMS data on formation on formic acid and formaldehyde.

H2COsol + HCOOHsol

diff

(1)

(3)

(2)

(5)

(4)

(6) CH2OHads

H2COads+HCOOHads

COads COads

rp, direct pathway

rs rCO

CO2 CH3OHsol

CO2 CH3OH

118

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

2

max

3max

max

2max

0 )1

1(4

)1

1()(tkm

eNktk

tkiti

dec

Ptdec

dec

dect θ

θθ

θ+

++

−= = (Eq. 7.1)

Experiments were designed such that at t = 0 the surface could be assumed to be free of poisoning intermediates. In Eq. 7.1, it=0 is the initial current density due to the direct oxidation at the poison-free surface, kdec is the rate constant of the site blocking process (i.e., decomposition of methanol), θmax is the maximum CO coverage on the surface, e is the elementary charge, NPt is the surface atom density of Pt(111) and m is the number of surface sites needed to decompose methanol into a CO molecule. On the basis of least squares curve fitting results, the most suitable value for m was found to be 2. This equation was able to fit the recorded transients quite well.

A plot of the electrode potential vs. the logarithm of the extrapolated initial current density yielded a Tafel slope of 120 mV⋅dec-1 for Pt(110). Combined with the results of an isotope-effect study, the corresponding rate-determining step for methanol oxidation was suggested to be:

−+ ++→ eHOHCHOHCH adsrds

sol 23 (7.2) Later Lu et al.[26] modified this model to fit their experiments for methanol

oxidation on polycrystalline platinum. After addition of an exponential current decay to Eq. 7.1, which served to account for the rapid decaying current component that is observed in their transients at short times (<40ms) and which was ascribed to species formed during the initial oxidation other than adsorbed CO, they reported a noticeable improvement in the fit.

The apparent success of the Wieckowski group model can be explained by its simplicity and the wide potential range over which it can be used. However, an important element missing in the model is the CO to CO2 reaction, which may be especially important at higher potentials. Furthermore, as was pointed out by Vielstich et al.,[29] the usage of it=0 as a fitting parameter is somewhat ambiguous. In their analysis of the chronoamperometric transients Wieckowski et al. found that at potentials where adsorbed COad is not oxidized to CO2, the oxidation current density of methanol on Pt(111), (110) and (100) exceeds the current density needed to form COad. It was assumed that, according to Fig. 7.1a, this excess current was due to the direct formation of CO2. However, this assumption was disputed by Vielstich et al.,[29] who showed, by using differential electrochemical mass spectrometry (DEMS), that CO2 evolution occurs only at potentials higher than the CO oxidation potential, which implies that at t = 0 methanol does not oxidize completely to CO2, thus leading to a less straightforward interpretation of the parameter it=0.

To the best of our knowledge, the only model that does incorporate the CO oxidation reaction along with the direct methanol oxidation and decomposition reaction, was proposed by Stuve et al.[8, 23, 24] Rather than using the chronoamperometric data directly, they chose to model the relationship between the charge obtained from the transient (the reaction charge, qr), the charge obtained after stripping the electrode of adsorbed species in a different cell (stripping charge, qs) and time. The results clearly showed that at short times (< 30 ms) only partial oxidation products are formed and

119

Chapter 7

virtually no CO2. At times longer than 5 seconds CO2 production is greatly favored over residue formation. Stuve and co-workers also concluded that for methanol oxidation on Pt(111) electrodes a serial reaction path, involving adsorbed intermediates, is inadequate to describe the observed rate of CO2 formation. It was proposed that methanol oxidation on CO covered electrodes proceeds through a parallel pathway mechanism, where the pathway not involving strongly adsorbed species accounts for the majority of the CO2 produced. This model not only includes the CO oxidation reaction, but it is also less sensitive to the exact rate equations used to model the system, since it uses charge rather than its derivative. Another advantage of this technique is that the stripping charge, qs, and, thus, the CO coverage on the electrode surface can be determined accurately. However, unaccounted for is the nature of the intermediate species in the direct methanol oxidation pathway.

In an effort to elucidate the nature of these species and to explain the apparent contradiction between Wieckowski’s and Vielstich’s observations, Baltruschat et al.[30, 31] proposed a parallel pathway mechanism in which CO2 could be formed through oxidation of adsorbed CO and/or the oxidation of intermediate dissolvable species like formic acid and formaldehyde. They based their model on DEMS experiments, the results of which clearly showed that H2CO and HCOOH can be detected during methanol oxidation, and on findings reported by Ota et al.[32] and Iwasita et al.[33], who already demonstrated the formation of these species under certain conditions. Korzeniewski and Childers showed that during methanol oxidation the formaldehyde yield could be as high as 30%.[34] The resulting more advanced parallel pathway mechanism is schematically depicted in Fig. 7.1b. The mechanism explains the observation of a higher oxidation current than necessary for COads formation, the appearance of CO2 only at potentials higher than the CO oxidation potential and it incorporates the possibility and identity of intermediate species detected during the incomplete oxidation reaction. Species like methylformate and 1,1-dimethoxymethane, which were found in small amounts by several groups,[5, 31] are not included in this scheme.

In the model we would like to suggest here, the adsorption and subsequent decomposition of methanol and the direct methanol oxidation reaction are treated in the same manner as proposed by Franaszczuk et al.[25] However, instead of assuming that methanol reacts with adsorbed water to form CO2, six protons and six electrons (7.1), we simply assume that a parallel pathway reaction exists, giving rise to a current, id. The main difference between our model and Franaszczuk’s is the incorporation of the CO oxidation reaction. The mechanism is therefore as follows:

−+ ++→ eHCOOHCH adskdec 443 Decomposition current, idec (7.3)

−+ ++↔+ eHOHOH ads*2

*22 +++→+ −+ eHCOOHCO oxkadsads Oxidation current, iox (7.4)

Direct oxidation current, id (7.5)

CH3OH → intermediates → CO2

Diffuse away

120

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

The adsorption of water and the subsequent formation of OHads is assumed to be in equilibrium. The total current is now;

oxdecd iiii ++= (Eq. 7.2) and the rate of surface CO formation and oxidation is given by:

)1()1( 2COCOoxCOdec

CO kkdt

dθθθ

θ−−−= (Eq. 7.3)

The second-order decomposition rate law is the same as that used by Franaszczuk et al., whereas the second-order oxidation rate law, which expresses the Langmuir-Hinshelwood mechanism underlying CO oxidation, was recently found to satisfactorily fit CO adlayer oxidation experiments.[35] This differential equation can be solved to yield the following equation for the CO coverage over time:

tk

tk

CO ox

ox

eXet

)1(11)(+−−

=θ with dec

ox

kk

X = (Eq. 7.4)

Completing equation 7.3 gives: )](1)[(2)](1[4)](1[)( 2 ttkeNtkeNtiti COCOoxPtCOdecPtCOd θθθθ −+−+−= (Eq. 7.5)

Here id is the direct oxidation current from reaction 7.5, kdec is the rate constant for the site blocking process (reac. 4 in Fig. 7.1b), kox is the effective oxidation rate constant (reac. 5 in Fig. 7.1b), NPt is number of platinum atoms per square cm (1.5⋅1015 atoms⋅cm-2), e is the elementary charge (1.6022⋅10-19 C) and 4 and 2 are the number of electrons involved in the respective reactions. In this model it is assumed that the maximum coverage of CO on the surface equals one. Making this assumption allows the differential equation, Eq. 7.3, to be solved analytically and, because it only normalizes the CO coverage on the surface, it should not affect the correctness of our model. It should also be noted that in our model possible reaction and/or adsorption of anions and non-CO adsorbates are not taken into account. Eq. 7.5 can be used to fit the chronoamperometric data directly.

An interesting prediction that follows from Eq. 7.5 is that the shape of the current transient depends sensitively on the value of X and, thus, on the relative numerical values of kox and kdec. If id is set to 0 at t = 0 in Eq. 7.5, we can determine a value for X at which the derivative of the function equals zero, namely for X=4. Fig. 7.2 demonstrates the effect of the value of X on the shape of the transient. As predicted for X ≥ 4, the transient changes from a current density decreasing with time to an increasing current density with time. This means that when the oxidation reaction of COads is faster than the methanol decomposition reaction (X > 4) we would expect the recorded current density to increase due to a slow increase in CO coverage, which is then rapidly oxidized. It is interesting to see that for values a little higher than 4, the curve can go through a maximum as is shown in the inset in Fig 7.2. In this situation the initial CO oxidation reaction is fast but slows down after some time due to increased poisoning of the surface. For X < 4 we obtain the expected decreasing current transient, as also predicted by Eq. 1. Experimentally, we expect the condition of X > 4 to be most easily satisfied by working at low methanol concentrations.

121

Chapter 7

0 50 100 150 200

10

20

30

40

50

60

70

80

j / µ

A cm

-2

t / s

X=10 X=4.55 X=0.1

0 50 100 150 20022.10

22.15

22.20

22.25

22.30

22.35

22.40

j / µ

A cm

-2

t / s

Figure 7.2. Current-time transient predicted with equation 10 for different values of X. In all curves id = 1·10-6 A⋅cm-2. For X = 10 kox and kdec are 0.1 s-1 and 0.01 s-1, respectively. For X = 4.55 these values are 0.1 and 0.022 s-1and for X = 0.1, 0.1 s-

1and = 1 s-1, respectively. The graph for X = 4.55 is shown enlarged in the inset.

Because of the incorporation of the CO oxidation reaction, the potential range for which our model can be applied should be considerably broader than for Franaszczuk’s model. Of course, setting the current produced in the direct pathway equal to id, without making any assumptions regarding the nature of the reaction steps, the amount of electrons produced per methanol molecule, or the formed products, is not an improvement over Franaszczuk’s model. Nevertheless, since Wang et al.[30, 31] showed that the contribution of the direct pathway can be considerable on CO covered electrode surfaces, addition of id is more or less a necessity.

7.4. Results and Discussion

7.4.1. Cyclic Voltammetry

As is customary for single crystal experiments, the cleanliness of the system was checked by recording a cyclic voltammogram of the electrode in contact with the blank electrolyte. The cyclic voltammogram of Pt(111) in 0.5 M H2SO4 is shown in Fig 7.3a. It shows the characteristic well-developed sharp “butterfly” peak at 0.45 V, which is ascribed to the (bi)sulfate adlayer disorder-order phase transition [36, 37] and has only small amounts of (110) and (100) defects, as can be seen from the two small peaks in the hydrogen adsorption/desorption region at 0.125 and 0.27 V.[38]

Both stepped surfaces in Figure 7.3b and c show sharp peaks at 0.125 V due to adsorption and desorption of hydrogen on the (110) steps [39] (although Markoviç et al. claimed that the pseudocapacitance under this peak corresponds not only to hydrogen adsorption/desorption, but also to adsorption/desorption of (bi)sulfate [40, 41]), while the broad features negative of 0.35 V and positive of 0.4 V are due to hydrogen and sulfate adsorption/desorption on the terraces, respectively.[11, 42, 43] All the recorded blank

122

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

0.0 0.2 0.4 0.6 0.8 1.0

-300-200-100

0100200300

0.0 0.2 0.4 0.6 0.8 1.0-200

-100

0

100

2000.0 0.2 0.4 0.6 0.8 1.0

-150

-100

-50

0

50

100

150

E / V vs. RHE

j / µ

A c

m-2

(a) Pt(111)

(b) Pt(554)

(c) Pt(553)

Figure 7.3. Cyclic voltammograms of Pt(111) (a), Pt(554) (b) and Pt(553) (c) in 0.5 M H2SO4. Scan rate 50 mV⋅s-1.

cyclic voltammograms (BCVs) are in good correspondence with those previously reported in the literature and satisfy the criteria of system cleanliness proposed by Lebedeva et al.[11]

Cyclic voltammograms of the three electrodes in methanol containing electrolyte compared to the BCVs are shown in Fig. 7.4. When examining the hydrogen region on the stepped electrodes it can be noticed that the hydrogen adsorption/desorption on the steps is affected most by the presence of methanol and that for Pt(111) the signals corresponding to the (110) and (100) defects sites have disappeared (see Fig 7.4a). This can be rationalized by assuming that either methanol or species formed in the decomposition reaction adsorb preferentially at the defect sites. It is also clear that at this scan rate the (bi)sulfate disorder-order transition peak is largely suppressed in the positive going sweep on Pt(111), indicating that species, again probably adsorbed CO, are present on the terraces. However, from these observations little can be said about where methanol preferably decomposes. In the returning sweep, the disorder-order peak is no longer suppressed, indicating that (bi)sulfate is the most abundant species on the terraces and previously adsorbed species have been removed. The (110) and (100) defects in the hydrogen region remain absent in the negative going sweep, again

123

Chapter 7

0.0 0.2 0.4 0.6 0.8 1.0-150-100-50

050

100

0.0 0.2 0.4 0.6 0.8 1.0-200-100

0100200300

0.0 0.2 0.4 0.6 0.8 1.0

-2000

200400600

j / µ

A c

m-2

E / V vs. RHE

(c)

(b)

(a)

Figure 7.4. Cyclic voltammograms for the Pt(111) (a), Pt(554) (b) and Pt(553) (c) in 0.5 M H2SO4 (dotted line) and in 0.5 M H2SO4 + 0.025 M CH3OH (continuous line). Scan rate 50 mV⋅s-1.

signifying that methanol either preferably adsorbs and/or decomposes on the steps or that the decomposition products diffuse over the surface and get trapped in the defect sites.

The negative going scan on the stepped surfaces shows virtually no difference with the positive going scan (see Fig. 7.4b and c). Apparently, at this scan rate the adsorption/decomposition reaction on stepped surfaces is fast enough to “refill” the steps after oxidation of adsorbed species occurred at higher potentials. However, it is noteworthy that in the voltammograms recorded in MeOH containing electrolyte on the stepped surfaces the hydrogen peak corresponding to the steps is shifted to more positive potentials when compared to the blank cyclic voltammogram, which suggests that adsorption of hydrogen is easier in the presence of adsorbed methanol and/or CO on the steps. A similar effect is also observed when only CO is adsorbed on these surfaces.[11] Also noteworthy is the fact that for Pt(111) the two small peaks at approximately 0.7 V in the positive and negative going sweep, which have been ascribed to OH formation and subsequent reduction,[44, 45] are still visible in the voltammogram. This indicates that at the current methanol concentration, OH species, which are not used in oxidizing adsorbed CO, are formed on the surface.

Judging from the measured methanol oxidation current densities the order in reaction rate is Pt(111)<Pt(554)<Pt(553), thus implying that the oxidation reaction is structure-sensitive and that the rate of the reaction increases with the step density. We also find that the current density at the onset of methanol oxidation increases for increasing step density. These results contradict the findings of Tripković and Popović, who reported that the initial surface activity for methanol on Pt(755), (211) and (311)

124

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

decreases for increasing step density.[21] However, the difference may be due to the fact that these authors used an electrode type different (Pt[n(111)×(100)] type electrodes) from ours (Pt[n(111)×(111)] type electrodes). In a study of methanol oxidation on low index platinum single crystals performed by Herrero et al.[19] methanol was found to react slower on Pt(100) than on (110). For Pt(110) and Pt(100) different values for the Tafel slope were found (120 and 60 mV·dec-1, respectively), indicating a difference in the rate-determining step, which may account for the difference in observations. Our observation compared to the results of Tripković and Popović may mean that methanol decomposes significantly faster at (110) than at (100) steps. Interestingly, in an infrared study on CO oxidation on stepped single crystals, Lebedeva et al.[46] found no significant difference in reactivity of CO oxidation on electrodes with either (110) or (100) steps.

The important question is now, which reaction is observed at a certain potential: the methanol decomposition reaction, the direct methanol oxidation reaction (pathway 3, 4 and 5, Fig 7.1b) or the indirect pathway via CO oxidation (pathway 1 and 2, Fig. 7.1b)? Of these three reactions the decomposition reaction as well as the CO oxidation reaction are known to be structure-sensitive,[11] and the direct methanol oxidation reaction may be as well. We attempt to address this issue with the help of chronoamperometry in the next section.

7.4.2. Chronoamperometry

After addition of 0.025 M MeOH and prior to the chronoamperometric measurement, the single crystal was cleaned of preadsorbed methanol decomposition products by stepping the potential from a potential in the hydrogen region (0.085 V) to 0.855 V for 20 seconds. It is assumed that at this potential all adsorbed species are oxidized and the electrode surface is effectively cleaned. Directly after the cleaning procedure the potential was stepped down to the desired potential and the current-time transient was recorded.

Transients were recorded from 0.305 V to 0.805 V with a potential step increase of 50 mV per experiment. In the range of 0.655 V to 0.755 V four additional data series at 0.675 V, 0.690 V, 0.720 V and 0.735 V vs. RHE were recorded. All measurements were performed at least three times. Some examples of the resulting chronoamperometric curves can be found in Fig. 7.5. The current-time curves shown were recorded at 0.655 V, 0.705 V, 0.755 V and 0.805 V vs. RHE in 0.5 M H2SO4 and 0.025 M methanol. Comparing the current densities measured in the transients with current densities given by the Cottrell equation shows that the reaction is not diffusion limited.[47]

The transients for the methanol decomposition on Pt(554) and (553) show a current decrease in the first 200 seconds, followed by a more or less constant current density at longer times. These transients resemble the transients recorded by Wieckowski and Stuve et al.[8, 23-27] on Pt(111) in 0.5 M H2SO4 and 0.2 M methanol. However, for potentials below ca. 0.7 V, the curve for Pt(111) shows an increase in current density in

125

Chapter 7

Pt(554)

Pt(553)

0.755 V Pt(111)

0 200 400 600 800 10005

10152025

0 200 400 600 800 100050

100

150

200

0 200 400 600 800 1000

200

300

400

500

j / µ

A cm

-2

t / s

0.705 V

0.755 V

0.705 V

0.655 V

0.805 V

0.805 V 0.755 V

0.705 V

0.805 V

0.655 V

0.655 V 0.705 V

Figure 7.5. Current transients of the decomposition of methanol on Pt(111) (continuous line), Pt(554) (dashed line) and Pt(553) (dotted line). The step potentials are 0.655 V, 0.705 V, 0.755 V and 0.805 V vs. RHE.

the first 50–100 seconds before reaching the steady-state current density. At higher potentials the initial shape of the transient becomes flatter, although a change to an initial current density decrease analogous to the stepped crystals cannot be observed unambiguously. This anomalous shape may be rationalized by assuming that the oxidation rate of chemisorbed intermediates is higher than the decomposition rate of methanol because of the low methanol concentration in our experiments. In this case the value for X in our model should be well over 4. Note that at the applied potentials oxygen-containing surface species should react much faster than a few seconds and, therefore, the associated reduction current cannot explain the shape of the recorded transients. Note also that the qualitative shape of the transients on Pt(111) can be explained by our model but not by the models suggested by Franaszczuk et al.[25] and Lu et al.[26]

The data shown in Fig. 7.5 clearly demonstrate that increasing the step density leads to an increase in the overall current density. This is true for the entire potential range under investigation at all times during the transient. Also, in general, the steepness of the initial current drop increases for increasing step density, which means that surfaces with a high amount of (110) steps poison faster than surfaces with little or no defects, again indicating that the reaction is highly structure sensitive in agreement with the literature.[14, 15, 18, 20]

A plot of the current density measured at t = 900 second versus the step potential is shown in Fig. 7.6. The enlarged area shows the behavior at low potentials. It is immediately clear that all three surfaces exhibit nearly the same qualitative trends over the entire potential region. First, the steady-state current density increases slightly with

126

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

0.30 0.35 0.40 0.45 0.50-2

0

2

4

6

8

10

j / µ

A cm

-2

E / V vs. RHE

Pt(111) Pt(554) Pt(553)

0.3 0.4 0.5 0.6 0.7 0.8

0

100

200

300

400

j / µ

A cm

-2

E / V vs. R H E

Figure 7.6. Steady state current density, iss, of the methanol decomposition/oxidation on Pt(111) (), Pt(554) () and Pt(553) () in 0.5 M H2SO4 and 0.025 M methanol. The steady state current density was recorded from the chronoamperometric transients at t=900 seconds. Enlarged is the steady state current-potential relationship at low potentials.

the potential when going from 0.3 to 0.5 V, followed by a fast increase at higher potentials up to approximately 0.7 V. At potentials above ca. 0.7-0.75 V the current density decreases again for all three electrodes.

Taking into consideration our model for methanol decomposition and subsequent oxidation, the observed trend can be rationalized by assuming that at low potentials the CO oxidation is the rate-determining step and that the rate of this reaction increases only weakly with increasing potentials below 0.5 V. When the potential exceeds 0.5 V the CO oxidation reaction occurs faster and the current density increases more rapidly. It can be assumed that the rate-determining step is now the adsorption/decomposition of methanol on the surface. The subsequent decrease in the steady-state current density for potentials higher than 0.7-0.75 V is generally ascribed to a decrease in free adsorption sites for methanol as a result of increasing OH coverage.[1]

7.4.3. Surface Species Coverage

Directly after the steady-state measurements a cyclic voltammogram of the hydrogen region was recorded in order to obtain an indication of the hydrogen coverage and, thus, indirectly of the coverage of adsorbed species on the surface. The relationship between hydrogen and co-adsorbed CO on the investigated single crystal electrodes has been determined empirically by Lebedeva et al.[35] The cyclic voltammograms were recorded in the methanol-containing electrolyte at a scan rate of 750 mV⋅s-1 to ensure that the renewed adsorption of methanolic species on the surface during the scan is negligible. The resulting relationship between the hydrogen coverage and the step potential is shown in Fig. 7.7. Admittedly, this procedure is not the most accurate for

127

Chapter 7

0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.850.4

0.5

0.6

0.7

0.8

0.9

1.0

θ H

E / V vs. RHE

Pt(111) Pt(554) Pt(553)

Figure 7.7. Hydrogen coverage vs. step potential on Pt(111) (), Pt(554) () and Pt(553) (). The coverage was determined from the hydrogen adsorption/desorption charge obtained from a cyclic voltammogram of the hydrogen region. The CV was recorded at 750 mV⋅s-1 in the methanol-containing electrolyte. Each point is the average of 3-4 experiments.

determining the coverage of adsorbed species (presumably mostly CO), but it avoids cumbersome transfer experiments and the qualitative general trends can still be useful.

Clearly at low potentials the number of free sites for hydrogen to adsorb is limited, meaning that the coverage of decomposition products is high. Assuming that CO is the primary decomposition product, this is in agreement with our previous explanation that at these potentials the CO oxidation reaction is slow and rate determining. Also, it can be observed that below 0.6 V the amount of adsorbed species on the electrode surface generally increases with increasing step density.

At potentials above 0.6 V the coverage of adsorbed decomposition species becomes much lower, which implies that the formed carbon monoxide reacts off the surface rapidly and methanol decomposition has become rate limiting. This is consistent with our previous interpretation of the data in Fig. 7.6

7.4.4. Modeling the chronoamperometric Data

Both Eq. 7.1 and our model, Eq. 7.5, were used to fit the chronoamperometric data of the methanol decomposition and oxidation on the single crystal electrodes. Of our model, two versions were used: version 1 excludes the direct oxidation current, id, and version 2 includes id. Only the first 200 seconds of the transients were used in the fitting procedure, because for t > 200 sec the current density still decreases slowly but continuously. Whether this decrease results from the methanol-related reactions occurring at the electrode surface, or from blockage of reactive sites due to adsorption of contaminating species, is still unclear. Examples of the fit produced by the version 1 are shown in Fig. 7.8, while fits produced by version 2 and Franaszczuk’s model can be found in Fig. 7.9. The graphs of the obtained values for the variable parameters versus the step potential found with the various models are shown in Fig. 7.10 and 7.11. Values

128

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

0 50 100 150 200140

160

180

200

0 50 100 150 200

400450500550600

0 50 100 150 200

1516171819

j / µ

A cm

-2

t / s0 50 100 150 200

400

440

480

520

560

600

0 50 100 150 200

15

16

17

18

19

0 50 100 150 200140

150

160

170

180

190

t / s

j / µ

A c

m-2

(c)

(b)

(a)

(c)

(b)

(a)

Figure 7.8. Current transients for methanol decomposition fitted with our model WITHOUT id (continuous line) on (a) Pt(111), (b) Pt(554) and (c) Pt(553). Transients shown were recorded at 0.705 V vs. RHE in 0.5 M H2SO4 and 0.025 M CH3OH.

Figure 7.9. Current transients for methanol decomposition fitted with our model WITH id (continuous line) and Franaszczuk’s model (dotted line) on (a) Pt(111), (b) Pt(554) and (c) Pt(553). Transients shown were recorded at 0.705 V vs. RHE in 0.5 M H2SO4 and 0.025 M CH3OH.

for kdec and kox resulting from version 1 of our model are given in Fig 7.10a, while the values for id and kdec and kox obtained from version 2 are presented in figures 7.10b and c, respectively. Fig. 7.11a, b and c present plots of θmax, log(it=0) and kdec respectively, versus the potential, obtained from Franaszczuk’s model (Eq. 7.1).

Examination of the curve-fitting shown in Fig. 7.8 and 7.9 reveals that, unlike both versions of our model, the model proposed by Franaszczuk et al. is not able to predict the initial current density increase measured on the platinum (111) electrode, due to the absence of the CO oxidation reaction in this model. Interestingly, the introduction of another reaction parameter in our model, namely id, has almost no influence on the curve fitting for the (111) surface. The reason for the lack of an important influence of id on the fit is that the CO coverage under these conditions is very low, meaning that the term id(1-θ) in Eq. 7.5 is essentially constant. As the poorness of the fit is related to the time-dependence of the current density rather than the absolute values, adding id makes no great difference. It seems that for low methanol concentrations the direct pathway current density is very small on essentially defect free Pt(111), as can be deduced from Fig. 7.10b, which shows that for Pt(111) the value for id is very low over the entire potential region under investigation. We note that the fits shown in Fig. 7.8 cannot be

129

Chapter 7

0.4 0.5 0.6 0.7 0.8

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

reac

tion

rate

con

stan

t / s

-1

E / V vs. RHE

kox Pt(111) kdec Pt(111) kox Pt(554) kdec Pt(554) kox Pt(553) kdec Pt(553)

0.55 0.60 0.65 0.70 0.75 0.80

0

100

200

300

400

500

i d / µ

A cm

-2

E / V vs. RHE

Pt(111) Pt(554) Pt(553)

0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85

0.00

0.02

0.04

0.06

0.08

0.10

0.12 kox Pt(554) kdec Pt(554) kox Pt(553) kdec Pt(553)

reac

tion

rate

con

stan

t / s

-1

E / V vs. RHE

0.3 0.4 0.5 0.6 0.7 0.80.1

0.2

0.3

0.4

0.5

0.6

0.7

θ m

ax

E / V vs. RHE

Pt(554) Pt(553)

0.3 0.4 0.5 0.6 0.7 0.8

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

k dec /

s-1

E / V vs. RHE

Pt(111) Pt(554) Pt(553)

-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.00.3

0.4

0.5

0.6

0.7

0.8

E /

V vs

. RH

E

log(it=0/µA)

Pt(554) Pt(553)

(a) (a)

(b) (b)

108 mV/dec

95 mV/dec

(c) (c)

Figure 7.10. (a) CO oxidation rate constant, kox, (straight line) and methanol decomposition rate constant, kdec, (dotted line) vs. potential on Pt(111) (), Pt(554) (), and Pt(553) () obtained from our model without id. (b) Direct methanol oxidation current, id, determined by our model on Pt(111) (), Pt(554) (), and Pt(553) () in 0.5 M H2SO4 and 0.025 M methanol. (c) CO oxidation rate constant, kox, (straight line) and methanol decomposition rate constant, kdec, (dotted line) vs. potential on Pt(554) () and Pt(553) () obtained from version 2 of our model (with id).

Figure 7.11. (a) Maximum CO coverage on Pt(554) () and (553) () obtained by fitting with eq. 2. (b) Electrode potential vs. log(it=0) for methanol decomposition/oxidation on Pt(554) () and (553) () in 0.5 M H2SO4 and 0.025 M CH3OH. (c) Dependence of the adsorption rate constant kdec, obtained by fitting with eq. 2, on the potential.

130

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

improved by changing the total time over which the current density is fitted. The reason for the poor fits obtained with version 1 of our model is more fundamental. Whereas the transients produced with version 1 of our model are approximately exponential (i(t)=A+Be-kt), a log-log analysis of the experimental transients shows that they follow more closely a power law type description (i(t)=A+Btα).

On stepped surfaces, however, incorporating the direct current in the model greatly improves the accuracy of the fit. When comparing Franaszczuk’s model with version 2 of our model, one is inclined to say that the former produces a somewhat better fit. However, as we will argue below, we believe this better fit is mainly the result of the fact that their fit parameter θmax gives more flexibility to the model than our kox. We also tested if the fit of our model (version 2) would be more accurate if the time interval under investigation is shortened to the first 50 seconds of the transients instead of the first 200 seconds. Performing the same transient analysis again on the stepped electrodes for t ≤ 50 seconds, showed that also for this shortened time interval Franaszczuk’s model gives a more accurate fit of the data.

In their model, Franaszczuk et al. used the maximum CO coverage on the surface as a variable, where as theoretically θmax is expected to be virtually potential-independent.[25] Only after addition of an exponential decay factor to Eq. 7.1 did the same group report a potential-independent θmax.[26] When fitting our chronoamperometric data with Eq. 7.1 the resulting θmax is not potential independent as can be seen from Fig. 7.11a. For the stepped surfaces, the calculated maximum CO coverage increases to about 0.7 when increasing the potential from 0.4 to 0.55 V, after which it drops again to an unrealistically low value of approximately 0.2 when the potential is further increased from 0.55 to 0.8 V (Due to the current density increase observed on Pt(111), Eq. 7.1 produced unrealistic values for the fitted parameters. These values were, therefore, omitted from Fig. 7.11a and b). This demonstrates that θmax is an important parameter necessary for an accurate fit of the data, which, we believe, should not be the case.

Following Franaszczuk et al. a Tafel slope for the direct current, id, can be obtained by plotting log(it=0) versus E (Fig. 7.11b). After applying their model to our data we find Tafel slopes for Pt(554) and (553) of 110 and 95 mV·dec-1 respectively in the potential range from 0.405 to 0.655 V. These values are close to those reported by Franaszczuk et al., who found a Tafel slope of 106 and 120 mV·dec-1 for Pt(111) and Pt(110) respectively in 0.1 M H2SO4 and 0.2 M methanol. The slight deviation may be caused by the fact that we chose not to correct our transients with a current-transient recorded in the blank electrolyte

The fit results shown in Fig. 7.10 confirm the qualitative conclusions drawn in the previous sections. The results obtained for the decomposition and oxidation rate constant from both versions 1 and 2 indicate that the steps and defects catalyze methanol decomposition and subsequent CO oxidation, suggesting that both processes take place preferentially at the steps. Note that the rate constants obtained with version 1 of the model are substantially higher than those obtained with version 2, due to the absence of

131

Chapter 7

the direct path in the former. Another observation to be made from Fig. 7.10a and 7.10b is the existence of the transition potential of ca. 0.6-0.65 V.

Below this potential, decomposition is fast compared to oxidation, whereas above this potential, oxidation is faster than decomposition. This is again in good accordance with our previous deductions, and the existence of this transition potential is related to the low methanol concentration used in our experiments. Here, we would like to remark that the values found for the CO oxidation rate constant lie within the range of 10-2 – 102 s-1, which agrees well with the CO oxidation rate constants on these surfaces determined experimentally by Lebedeva et al.[35]

The results shown in Fig. 7.10b would suggest that the direct pathway through soluble intermediates like formic acid and formaldehyde is also strongly structure sensitive and dictated by a process that takes place preferentially at the step sites. The plot of id versus the step potential shown in Fig. 7.10b closely resembles that of the steady-state current density in Fig 7.6, the only difference being that the direct methanol current density is higher than the measured steady state current density. On all surfaces the direct current is virtually zero for potentials below 0.55 V, while it rises fast for potentials over 0.6 V to fall again after a potential of 0.7 V has been reached. This finding is in agreement with recent data reported by Wang et al.,[30, 31, 48] who revealed that the formation of formic acid and formaldehyde starts at a potential of about 0.6 V and reaches a maximum at a potential, which coincides with the maximum of the methanol oxidation peak recorded in the cyclic voltammogram.

The reason for id being higher than the steady state current density, iss, may be found in the fact that iss was recorded at approximately t = 900 seconds while the direct current was determined by modeling the transient in the first 200 seconds. In the remaining 700 seconds the current density still changes, which can account for the discrepancy.

Fig. 7.11 shows the results obtained for the potential dependence of the fit parameters in Franaszczuk’s model Eq. 7.1. The results in Fig. 7.11b and c confirm that both the decomposition rate and the direct current increase with increasing step density. We do believe, however, that for potentials higher than ca. 0.6-0.65 V, the model is less applicable to the actual process as above these potentials CO oxidation (the indirect pathway) can no longer be neglected. We believe that this is the reason for the sudden drop found in kdec (Fig. 7.11c) and especially θmax (Fig. 7.11a) above 0.6-0.65 V. The almost instantaneous drop in θmax from 0.6-0.7 (which seems a reasonable number) to ca. 0.2 is unrealistic and in our view must be due to a process, which takes place at potentials above 0.6-0.65 V and which is not incorporated in the model. Note that this transition potential of 0.6-0.65 V is exactly the potential for which the fits to our model suggest that CO oxidation (and hence the indirect pathway) becomes a fast process. Also, the strong potential dependence of θmax suggests that it is a powerful fit parameter, which may explain why the fits obtained with Eq. 7.1 are always slightly better than those produced with Eq. 7.5. However, since such a potential dependence does not seem

132

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

realistic, it can be questioned whether this more accurate fit is also physically more meaningful.

Finally, we would like to conclude this section with a strong word of caution regarding the fitting of chronoamperometric methanol oxidation transients. We have used the mathematical model presented in Section 7.3. mainly for two purposes. Firstly, the model explains the rising transients obtained on Pt(111) at low methanol concentrations as a consequence of the combination of slow decomposition and rapid subsequent oxidation of the decomposition product. Secondly, the fitting helped in substantiating the qualitative conclusions of the earlier sections, namely that both the methanol decomposition and poison oxidation take place preferentially at the steps and that for the methanol concentration used here there is transition potential of ca. 0.6-0.65 V at which there is a change in rate determining step from decomposition to oxidation. The fit results also suggest that the direct pathway is catalyzed by step sites, which may in fact be interpreted as methanol decomposition and methanol direct oxidation going through the same initial intermediates, as suggested by the scheme in Fig 7.1b. However, the accuracy of the rate parameters obtained with the model, especially those for the direct pathway, remains highly questionable. For instance, it is known from the literature that the amount of formic acid and formaldehyde detected depends on the flow rate and the methanol concentration of the electrolyte solution. As in our system convection is absent and the methanol concentration is rather low, which, according to the data reported in the literature, would favor the direct pathway, it may be argued whether the introduction of id, or the model employed for it, is justified. This problem can only be solved by measuring the direct pathway by another method simultaneously, for instance using on-line mass spectrometry (see Chapter 8). Furthermore, the actual rate laws employed in Eq. 8. for methanol decomposition and subsequent CO oxidation may be incorrect or oversimplified. The “ensemble effect” suggested by the quadratic dependence of the decomposition rate on the number of free sites, is still poorly understood. The Langmuir-Hinshelwood kinetics assumed for CO oxidation is known to give satisfactory results for high CO coverages, but is much less accurate for the oxidation of sub-monolayer coverages of CO as was discussed by Lebedeva et al.[35] Unfortunately, the low CO coverage regime seems the relevant one for experiments presented here. As long as these uncertainties regarding the importance of the direct pathway and the most accurate rate laws for describing the different processes exist, modeling and fitting methanol oxidation transients should only be used as an additional tool to put qualitative conclusions obtained by other methods on a semi-quantitative footing.

7.5. Conclusion

The methanol oxidation reaction was studied on Pt[n(111)×(111)] type electrodes in a 0.5 M H2SO4 solution with 0.025 M CH3OH. By combining voltammetry, chronoamperometry, and fitting the chronoamperometric transients with a new mathematical model incorporating methanol decomposition, poison oxidation, and

0.80 V

133

Chapter 7

methanol direct oxidation, we have shown that the methanol oxidation reactivity is strongly catalyzed by the presence of step and defect sites. The cyclic voltammetry data showed that the overall oxidation rate increases with increasing step density. Moreover, the defect or step sites on all surfaces are affected by the presence of methanol more than the terraces sites. This implies that either the decomposition products of methanol or the methanol itself preferably sits at the steps.

Analysis of the chronoamperometric data shows that the steady-state activity increases with increasing step density over the entire potential range. The drop in transient activity is faster on a surface with a higher step density, suggesting that the methanol decomposition into the CO poisoning species takes place preferentially at or near steps. From the observation that the CO coverage at the end of the transient is low for potentials higher than 0.55 V, it is concluded that the CO oxidation is faster at these potentials, whereas at lower potentials methanol decomposition is faster than CO oxidation. The mathematical model predicts that when CO oxidation is faster than methanol decomposition, a rising current transient should be obtained, which has indeed been observed for Pt(111). The decomposition rate on this surface is low because of the low defect density and because of the low methanol concentration used in our experiments. Fitting the transients to our mathematical model suggests that methanol decomposition, CO oxidation, and methanol direct oxidation are catalyzed by the steps. The fit results also confirm the transition from CO oxidation being rate determining below ca. 0.6 V to methanol decomposition being rate determining above this potential.

The mathematical model suggested in this chapter presents an extension and, we believe, an improvement over an earlier model for fitting chronoamperometric data, suggested by Franaszczuk et al.,[25] as their model does not incorporate the CO oxidation process. However, fitting methanol oxidation transients remains a problematic and potentially deceptive undertaking. First of all, the correct rate laws for both methanol decomposition and CO oxidation at low CO coverage are not well known. It may be mentioned that the coulometry-based method proposed by Stuve et al. partially circumvents this problem, as the time-dependent charge is presumably less sensitive to the exact nature of the rate laws than the time-dependent current density. Secondly, the direct methanol pathway, and especially its intermediates and final products as a function of potential and other variables of the system, need to be much better documented and understood in order to perform meaningful quantitative modeling. In this respect, it would clearly be a substantial improvement if the present experiments could be augmented by transient on-line mass spectroscopy.

Nevertheless, despite these reservations, we believe our results strongly suggest that steps of (110) orientation catalyze the methanol decomposition, the CO oxidation, and presumably also the direct methanol oxidation. The latter process may in fact be related to the initial stages of the methanol decomposition. Our observations seem to be in conflict with the earlier results of Tripković et al.,[21] who found that an increasing density of steps of (100) orientation leads to a lower methanol oxidation activity. This difference may be due to a different methanol concentration or due to the different orientation of the steps, as the Pt(100) surface is indeed known to be much less active for

134

MeOH oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study

methanol oxidation than the Pt(110) surface. Hence, elucidation the role of methanol concentration and the effect of the step orientation is an important goal for future work. Finally, since it is known that the methanol oxidation currents on the three basal planes of platinum are very sensitive to the electrolyte anion, a systematic study comparing the activity of stepped Pt in sulfuric and perchloric acid with different concentrations of methanol would be very valuable. Without inclusion of the specific anion adsorption, modeling the methanol decomposition/oxidation reaction correctly, may indeed be very difficult if not impossible. Acknowledgement: This research was supported by the Netherlands Foundation for Scientific Research (NWO). Special thanks go to Professor Juan Feliu of the Departamento de Química-Física at University of Alicante in Spain for supplying the stepped single crystals.

References

[1] R. Parsons and T. VanderNoot, J. Electroanal. Chem. 257 (1988) 9. [2] N. Kizhakevariam and E. M. Stuve, Surf. Sci. 286 (1993) 246. [3] G. Hopranyi and A. Wieckowski, Proceedings - Electrochemical Society 92 (1992) 70. [4] N. M. Markovic and P. N. Ross, Surf. Sci. Rep. 45 (2002) 117. [5] T. Iwasita, Electrochim. Acta. 47 (2002) 3663. [6] A. Hamnett, Comp. Chem. Kin. (1999) 635. [7] S. Wasmus and A. Kuver, J. Electroanal. Chem. 461 (1999) 14. [8] T. D. Jarvi, S. Sriramulu, and E. M. Stuve, J. Phys. Chem. B 101 (1997) 3649. [9] N. M. Markovic, C. A. Lucas, A. Rodes, V. Stamenkovic, and P. N. Ross, Surf. Sci. 499 (2002)

L149. [10] B. Beden, C. Lamy, N. R. De Tacconi, and A. J. Arvia, Electrochim. Acta. 35 (1990) 691. [11] N. P. Lebedeva, M. T. M. Koper, E. Herrero, J. M. Feliu, and R. A. van Santen, J. Electroanal.

Chem. 487 (2000) 37. [12] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Phys. Chem. B 106 (2002)

12938. [13] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, Electrochem.Commun. 2

(2000) 487. [14] K. D. Gibson and L. H. Dubois, Surf. Sci. 233 (1990) 59. [15] S. K. Desai, M. Neurock, and K. Kourtakis, J. Phys. Chem. B 106 (2002) 2559. [16] J. Clavilier, R. Durand, G. Guinet, and R. Faure, J. Electroanal. Chem. 127 (1981) 281. [17] C. Lamy, J. M. Leger, J. Clavilier, and R. Parsons, J. Electroanal. Chem. 150 (1983) 71. [18] X. H. Xia, T. Iwasita, F. Ge, and W. Vielstich, Electrochim. Acta 41 (1996) 711. [19] E. Herrero, K. Franaszczuk, and A. Wieckowski, J. Phys. Chem. B 98 (1994) 5074. [20] J. Shin and C. Korzeniewski, J. Phys. Chem. B 99 (1995) 3419. [21] A. V. Tripkovic and K. D. Popovic, Electrochim. Acta 41 (1996) 2385. [22] S. Park, Y. Xie, and M. J. Weaver, Langmuir 18 (2002) 5792. [23] S. Sriramulu, T. D. Jarvi, and E. M. Stuve, Electrochim. Acta. 44 (1998) 1127. [24] S. Sriramulu, T. D. Jarvi, and E. M. Stuve, J. Electroanal. Chem. 467 (1999) 132. [25] K. Franaszczuk, E. Herrero, P. Zelenay, A. Wieckowski, J. Wang, and R. I. Masel, J. Phys. Chem.

B 96 (1992) 8509. [26] G. Q. Lu, W. Chrzanowski, and A. Wieckowski, J. Phys. Chem. B 104 (2000) 5566. [27] E. Herrero, W. Chrzanowski, and A. Wieckowski, J. Phys. Chem. 99 (1995) 10423. [28] J. Clavilier, D. Armand, S. G. Sun, and M. Petit, J. Electroanal. Chem. 205 (1986) 267. [29] W. Vielstich and X. H. Xia, J. Phys. Chem. 99 (1995) 10421. [30] H. Wang, C. Wingender, H. Baltruschat, M. Lopez, and M. T. Reetz, J. Electroanal. Chem. 509

(2001) 163. [31] H. Wang, T. Loffler, and H. Baltruschat, J. Appl. Electrochem. 31 (2001) 759. [32] K.-I. Ota, Y. Nakagawa, and M. Takahashi, J. Electroanal. Chem. 179 (1984) 179.

135

Chapter 7

136

[33] T. Iwasita and W. Vielstich, J. Electroanal. Chem. 201 (1986) 403. [34] C. Korzeniewski and C. L. Childers, J. Phys. Chem. B 102 (1998) 489. [35] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Electroanal. Chem. 524-

525 (2002) 242. [36] A. M. Funtikov, U. Stimming, and R. Vogel, J. Electroanal. Chem. 428 (1997) 147. [37] M. T. M. Koper and J. J. Lukkien, J. Electroanal. Chem. 485 (2000) 161. [38] J. Clavilier, K. El Achi, M. Petit, A. Rodes, and M. A. Zamakhchari, J. Electroanal. Chem. 295

(1990) 333. [39] M. T. M. Koper, J. J. Lukkien, N. P. Lebedeva, J. M. Feliu, and R. A. van Santen, Surf. Sci. 478

(2001) L339. [40] N. M. Markovic, N. S. Marinkovic, and R. R. Adzic, J. Electroanal. Chem. 241 (1988) 309. [41] N. M. Markovic, N. S. Marinkovic, and R. R. Adzic, J. Electroanal. Chem. 314 (1991) 289. [42] J. Clavilier, K. El Achi, and A. Rodes, J. Electroanal. Chem. 272 (1989) 253. [43] J. Clavilier and A. Rodes, J. Electroanal. Chem. 348 (1993) 247. [44] H. A. Gasteiger, N. Markovic, P. N. Ross, Jr., and E. J. Cairns, J. Phys. Chem. 98 (1994) 617. [45] C. Saravanan, M. T. M. Koper, N. M. Markovic, M. Head-Gordon, and P. N. Ross, Phys. Chem.

Chem. Phys. 4 (2002) 2660. [46] N. P. Lebedeva, A. Rodes, J. M. Feliu, M. T. M. Koper, and R. A. van Santen, J. Phys. Chem. B

106 (2002) 9863. [47] A. J. Bard and L. R. Faulkner, Electrochemical Methods: Fundamentals and Applications, John

Wiley & Sons, Inc., New York, 2001. [48] H. Wang and H. Baltruschat, Proc. - Electrochem. Soc. 2001-4 (2001) 50.

Structure sensitivity of methanol electrooxidation pathways on platinum: an On-Line Electrochemical Mass Spectrometry study

Abstract

By monitoring the mass fractions of CO2 (m/z 44) and methylformate (m/z 60, formed from CH3OH + HCOOH) with On-Line Electrochemical Mass Spectrometry (OLEMS), the selectivity and structure sensitivity of the methanol oxidation pathways were investigated on the basal planes -Pt(111), Pt(110) and Pt(100)- and stepped Pt electrodes -Pt(554) and Pt(553)- in sulfuric and perchloric acid electrolytes. The maximum reactivity of the MeOH oxidation reaction on Pt(111), Pt(110) and Pt(100) increase in the order Pt(111)<Pt(110)<Pt(100). Mass spectrometry results indicate that the direct oxidation pathway through soluble intermediates plays a pronounced role on Pt(110) and Pt(111), while on Pt(100) the indirect pathway through adsorbed carbon monoxide is predominant. In 0.5 M H2SO4, introducing steps in the (111) plane increases the total reaction rate, whilst the relative importance of the direct pathway decreases considerably. In 0.5 M HClO4, however, introducing steps increases both the total reaction rate and the selectivity towards the direct oxidation pathway. Anion (sulfate) adsorption on (111) leads to a more prominent role of the direct pathway, but on all the other surfaces (bi)sulfate seems to block the formation of soluble intermediates. For both electrolytes, increasing the step density results in more methylformate being formed relative to the amount of CO2 detected, indicating that the [110] steps themselves catalyze the direct oxidation pathway. A detailed reaction scheme for the methanol oxidation mechanism is suggested based on the literature and the results obtained here.

This chapter is published as T.H.M. Housmans, A.H. Wonders, M.T.M. Koper, J. Phys. Chem. Bsubmitted

Chapter 8

8.1. Introduction

Knowledge of the methanol oxidation reaction mechanism on platinum electrodes is of pivotal importance to the development of a low temperature direct methanol fuel cell (DMFC) and has, therefore, enjoyed continued interest over the past decades.[1, 2] In order for these fuel cells to be economically viable, the turnover rate for complete oxidation of methanol must be higher than 0.1 s-1 between 0.2 – 0.4 V vs. RHE.[3] However, at these low potentials the oxidation is incomplete and poisoning species, including carbon monoxide [1, 4-7] and other carbonaceous species (assumed to be HCO or COH),[8-12] are formed during the decomposition of methanol.

One of the main mechanistic issues, which seems to have been resolved recently, is whether adsorbed carbon monoxide (COads) formed during the oxidation of methanol is an intermediate in a serial pathway mechanism or a by-product in a parallel reaction mechanism. Analyzing oxidation currents and CO coverages, the groups of Wieckowski [13-15] and Stuve [4, 16, 17] independently showed that oxidation of methanol can occur at potentials at which adsorbed CO is stable on the surface. These findings were disputed by Vielstich et al., who, on the basis of differential electrochemical mass spectrometry (DEMS), claimed that CO2 can only be formed at potentials corresponding to the CO oxidation potential.[18] However, as was pointed out by Wang et al., these studies did not take into account the possible formation of soluble side products such as formic acid and formaldehyde.[19-21] From calculations of current efficiencies recorded on polycrystalline Pt, which under certain circumstances are considerably less than 100%, they concluded that a parallel pathway involving incomplete methanol oxidation must indeed exist.

Thus, from a fuel cell technology point of view, the study of the individual reaction steps in the overall methanol oxidation scheme has been of great interest. Based on the Tafel slope obtained from an analysis of the instantaneous methanol oxidation current and an isotope effect analysis, Herrero et al. concluded that the rate-determining step of the methanol adsorption on Pt(110) and Pt(111) is:

−+ ++→ eHOHCHOHCH adsrds

sol 23 , (8.1)

whereas on Pt(100) it is:[13, 14] −+ ++→ eHOHCHOHCH adssol 23 (8.2)

−+ ++→ eHCHOHOHCH adsrds

ads2 (8.3)

This mechanism is to be contrasted with the methanol decomposition on platinum in ultra high vacuum, which involves the formation of methoxy (CH3O) rather than C-H bond breaking. In Chapter 7, we reported that the methanol decomposition reaction to form CO, and possibly also soluble intermediates, is structure sensitive and that both the overall rate and the self-poisoning rate increase with the step density.[22] The decomposition of methanol to form COads is often identified with the indirect pathway, while the reaction via soluble intermediates is often referred to as the direct pathway. The platinum-catalyzed electrooxidation of CO to CO2 has been extensively researched by

138

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

numerous groups.[23-26] On stepped single crystal electrodes the oxidation of COads was found to occur primarily at step sites.[26-29] In a recent paper by the Wieckowski and Neurock groups the dual path mechanism for methanol decomposition on well-defined low Miller index platinum single crystal electrodes was investigated at potentials where surface adsorbed CO is stable, using a combination of chronoamperometry, fast scan cyclic voltammetry and theoretical methods.[30] They concluded that at potentials between 0.2 – 0.5 V vs. RHE the dual pathway producing products other than adsorbed CO is active on Pt(111) and Pt(110) above 0.35 V vs. RHE, whereas on Pt(100) in the entire potential region under investigation only complete dehydrogenation of methanol to COads occurs, thus signifying a significant structure sensitivity of the dual path mechanism.

Based on ab initio Density Functional Theory (DFT) calculations, they also suggested that a dual dehydrogenation pathway is initiated already at the first methanol dehydrogenation step. On Pt(111) in aqueous media, the most favorable dehydrogenation path occurs via the adsorbed hydroxymethyl CH2OH intermediate, which is then further dehydrogenated to COads. Alternatively, O-H bond cleavage leads to adsorbed methoxy, CH3Oads, which binds to Pt via the oxygen rather than the carbon atom. Methoxy can be dehydrogenated to H2CO, leading to dissolved formaldehyde. On Pt(111), methoxy has a much lower stability than hydroxymethyl, but on a stepped Pt(211) surface, their stabilities are comparable, indicating that the presence of steps can significantly alter the reaction mechanism and the pathways chosen.[30]

A similar idea, namely that the indirect and direct pathway involve different dehydrogenation precursors, i.e. HC2OHads vs. H3COads respectively, was recently invoked by Iwasita et al. to explain the strong influence of adsorbed anions on the amount of formaldehyde formed during MeOH oxidation on Pt(111) at low overpotential.[31, 32] Adsorbed (bi)sulfate would effectively block “ensemble sites” for H2COHads but not for CH3Oads formation, rationalizing the enhanced influence of the direct pathway in sulfuric acid vs. perchloric acid. This explanation is based on the “ensemble site” theory, which has been quite popular in the methanol oxidation literature,[33, 34] for instance to explain the optimum Pt:Ru ratio (3:1) for PtRu catalysts,[35, 36] and to explain the particle-size dependence of supported nano-particles catalysts. As the methanol oxidation rate was found to decrease with the nanoparticle size, Park et al. concluded that the decomposition of methanol requires an ensemble of terrace sites.[33] Tripković et al. found that when the terrace width of Pt[n(111)×(100)] electrodes is decreased from n=6 to n=3 and 2 atoms per terrace (i.e. Pt(755), Pt(211) and Pt(331), respectively) the total methanol oxidation activity decreases,[34] in agreement with the conclusion of Park et al. However, recent results obtained in our group (see Chapter 7) and results published by Shin and Korzeniewski seem to contradict these findings.[22, 37] In both cases increasing the step density resulted in an increase of the methanol oxidation rate.

In this chapter, we study the formation of soluble intermediates at the basal planes of platinum and stepped Pt surfaces of [n(111)×(111)] orientation both in sulfuric and

139

Chapter 8

perchloric acid, using a tip-based modification of the differential electrochemical mass spectrometry technique (DEMS),[38-44] following an original design of Kita et al.[45] and modifications introduced by Jambunathan et al.[46, 47] The On-Line Electrochemical Mass Spectrometry (OLEMS) technique is described in more detail in ref. [48]. The main advantage of this technique over traditional DEMS is that experiments can be performed on bead-type single crystal electrodes in the hanging meniscus configuration. Moreover, as the gas-inlet is small and tip shaped, no differential pumping is necessary, thereby considerably reducing the detection limit of the system.[48]

With regard to detecting intermediates, however, the interference of species in solution poses a problem, since mass fragments of methanol and formaldehyde coincide. This can be partially circumvented by measuring methylformate, which is formed by reaction of methanol and formic acid and, thus, provides an indirect way of monitoring the activity of the direct oxidation pathway. However, results obtained by the groups of Korzeniewski [49, 50] and Iwasita [31, 32] indicate that formaldehyde is the most abundant soluble intermediate formed during the oxidation of methanol (next to CO2). Measuring methylformate as an indicator for the direct oxidation pathway implies that formic acid is formed in the same pathway as formaldehyde, which is not obvious from the reaction mechanism. Yet, in aqueous solutions formaldehyde is completely hydrolyzed to methylene glycol (H2C(OH)2),[51, 52] which can dissociate to COads,[53] but which can also form formate on the Pt surface (HCOOads).[54-56] These adsorbed formate species can be oxidized to CO2, but also imply the existence of formic acid in solution, thus indicating that the latter species would indeed be formed in the same reaction pathway as formaldehyde. Therefore, monitoring the formation of methylformate seems a viable procedure for measuring the activity of the direct methanol oxidation pathway.

By using the OLEMS setup for the methanol oxidation on stepped Pt electrodes of [n(111)×(111)] orientation and the basal planes in perchloric and sulfuric acid electrolytes, a more detailed insight into the structure sensitivity and the effect of specific anion adsorption on the direct oxidation pathway to soluble intermediates is gleaned. Additionally, the nature of the “ensemble sites” will be discussed. Our findings will be considered within the framework of the current views on platinum-catalyzed methanol electrooxidation and we will propose a new overall oxidation scheme comprising the direct and indirect oxidation pathways.

8.2. Experimental Setup

For a detailed description of the OLEMS setup we refer to ref. [48]. Briefly, the OLEMS setup consists of a small gas inlet tip made of Peek in which a porous Teflon plug is pressed. The tip is connected to the mass spectrometer and can be brought in close proximity (10-20 µm) to the electrode in the hanging meniscus configuration, by means of a micrometer positioning system and a camera.

140

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

The electrochemical cell used in the OLEMS experiments is a standard three-electrode cell made of glass, modified to allow the working electrode to be placed near the side of the cell in order to aid in the positioning of the OLEMS tip close to the electrode surface using the camera. The working electrodes were platinum bead type single crystals, with the basal plane orientation and Pt[n(111)×(111)] (equivalent to Pt[m(111)×(110)] with m = n-1) orientation – Pt(110), Pt(100), Pt(553) n=5, Pt(554) with n=10 and Pt(111) with n=200-500 – which were prepared according to Clavilier’s method.[57] Prior to each measurement the single crystal electrode was flame annealed and cooled down to room temperature in an argon (Hoekloos, N50)-hydrogen atmosphere (3:1 ratio), after which it was transferred to the electrochemical cell under protection of a droplet of deoxygenated water.[58] Glassware, OLEMS tip, and glass tubing are cleaned by boiling in a mixture of concentrated H2SO4/HNO3 followed by repeated boiling with ultra pure water (Millipore MilliQ gradient A10 system, 18.2 MΩcm, 2 ppb total organic carbon). The blank electrolytes, 0.5 M H2SO4 and 0.5 M HClO4, were prepared with concentrated sulfuric and perchloric acid (Merck, "Suprapur") and ultra-pure water. During measurements the blank electrolyte was deoxygenated with argon (Hoekloos, N50). Measurements were performed at room temperature (22 ºC) with a computer-controlled Autolab PGSTAT20 potentiostat (Ecochemie).

8.3. Results and Discussion

8.3.1. System Cleanliness and Surface Structure

As is customary, prior to each experiment the structure of the surface and the cleanliness of the system were checked by measuring a blank cyclic voltammogram (BCV) of the electrode under investigation. The resulting BCVs recorded in the sulfuric and perchloric acid electrolytes are shown in the supporting information section at the end of the chapter (S8.1a-e) and agree well with those previously published in the literature.[13, 16, 28, 30, 59-65]

Equally important when using the OLEMS setup is the influence of the tip on the electrode surface and cleanliness of the system. This influence was checked by measuring and comparing BCVs of Pt(111) in the absence and presence (close proximity) of the tip. The resulting BCVs closely resemble each other, signifying that the tip does not contaminate the system significantly and it has virtually no influence on the quality of the CVs recorded.[48] All obtained blank voltammograms satisfy the criteria of system cleanliness as proposed by Lebedeva et al.[26]

The performance of the OLEMS setup was checked by performing a CO stripping experiment on Pt(111). The shape of the CO oxidation peak agrees well with stripping curves reported in the literature.[26, 48] Similar to the tip-based setups of Kita [66] and Hillier [46], the corresponding signal for CO2 (m/z 44, most intensive ion for CO2) has a

141

Chapter 8

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

1 10-10

ion

curr

ent /

a.u

.

E / V vs. RHE

1 10-9

m/z 440.5 0.6 0.7 0.8 0.9

0

200

400

600

800

1000

1200

1400

j / µ

A cm

-2

E / V vs. RHE

Pt(111) Pt(110) Pt(100)

0.0 0.2 0.4 0.6 0.8 1.0

m/z 44

1 10-11

ion

curr

ent /

a.u

.

E / V vs. RHE

m/z 60

5 10-8

0.0 0.2 0.4 0.6 0.8 1.0

m/z 44

5 10-11

ion

curr

ent /

a.u

.

E / V vs. RHE

m/z 60

5 10-8

0.4 0.5 0.6 0.7 0.8 0.90

20

40

60

80

100

120

140

j / µ

A c

m-2

E / V vs. RHE

(a) (b) Pt(111)

(c) Pt(110) (d) Pt(100)

Figure 8.1. (a) Cyclic voltammograms of Pt(111) (solid line), Pt(110) (dashed line) and Pt(100) (dotted line) in 0.5 M MeOH and 0.5 M H2SO4 at a scan rate of 2 mV⋅s-1. The top mass shown in the associated MSCVs for Pt(111) (b), Pt(110) (c), and Pt(100) (d) displays the methylformate signal (m/z 60) and the bottom one the carbon dioxide signal (m/z 44). The inset shows a zoom of the Pt(111) cyclic voltammogram.

delay of ca. 10-15 seconds between the CO oxidation peak and the peak in the recorded mass, which is characteristic for the OLEMS setup.

8.3.2. Methanol Oxidation on Single Crystal Platinum Surfaces

We wish to stress at the outset that our OLEMS setup does not yet allow for an accurate quantitative calibration. The intensity of the detected mass signals depends on many factors (e.g. the tip-electrode distance, the porosity of the Teflon plug, the quality of the vacuum inside the system, etc.[48]), which vary from one experiment to the other. Consequently, a quantitative comparison of the recorded masses for the different surfaces is not possible at present. Therefore, the strategy we adopted was to compare the ratio between the masses of methylformate (m/z 60) and CO2 (m/z 44) during the oxidation and, thereby, the relationship between the direct and indirect methanol oxidation pathway and the surface structure of the electrode. In order to avoid problems that arise from the delay time, we first compared the integrated charges obtained from the mass spectrometry cyclic voltammograms (MSCVs) for mass signals 60 and 44. Although informative with respect to the influence of the surface structure on the selectivity of the

142

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

0.5 0.6 0.7 0.8 0.9-200

0

200

400

600

800

1000

1200

1400

1600

1800

j / µ

A c

m-2

E / V vs. RHE

Pt(111) Pt(554) Pt(553)

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

1 10-10

ion

curr

ent /

a.u

.

E / V vs. RHE

1 10-9

m/z 44

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

5 10-11

ion

curr

ent /

a.u

.

E / V vs. RHE

5 10-8

m/z 44

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

1 10-10

ion

curr

ent /

a.u

.

E / V vs. RHE

1 10-7

m/z 44

(a) (b) Pt(111)

(c) Pt(554) (d) Pt(553)

Figure 8.2. (a) Cyclic voltammograms of Pt(111) (solid line), Pt(554) (dashed line) and Pt(553) (dotted line) in 0.5 M MeOH and 0.5 M H2SO4 at a scan rate of 2 mV⋅s-1. The top mass shown in the associated MSCVs for Pt(111) (b), Pt(110) (c), and Pt(100) (d) displays the methylformate signal (m/z 60) and the bottom one the carbon dioxide signal (m/z 44).

direct and indirect oxidation pathways, in this analysis any potential-dependent information is lost.

The cyclic voltammograms and MSCVs recorded with the OLEMS on the basal Pt planes, Pt(111), Pt(110) and Pt(100) in 0.5 M MeOH and 0.5 M H2SO4 at a scan rate of 2 mV⋅s-1 are shown in Fig. 8.1a-d, while the results for the stepped surfaces Pt(554) and Pt(553) together with Pt(111) are presented in Fig. 8.2a-d. The effect of specific anion adsorption on the electrooxidation reaction of methanol was investigated by repeating the experiments in 0.5 M HClO4. The results for the basal planes in this electrolyte are presented in Fig 8.3a-d and the stepped surfaces in Fig. 8.4a-d. For the sake of clarity we shall discuss the methanol oxidation on the basal planes and stepped platinum surfaces in separate sections. Fig. 8.5 and 8.6 visualize a tentative relationship between the mass 60/44 ratio and the potential (obtained by dividing the measured m/z 60 signal by the m/z 44 signal) for the basal planes recorded in sulfuric and perchloric acid, while Fig. 8.7 and 8.8 represent similar curves for the stepped surfaces, respectively.

143

Chapter 8

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

1 10-10

ion

curr

ent /

a.u

.

E / V vs. RHE

2 10-8

m/z 44

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

2 10-10

ion

curr

ent /

a.u

.

E / V vs. RHE

1 10-7

m/z 44

0.5 0.6 0.7 0.8 0.9

0

500

1000

1500

2000

2500

j / µ

A cm

-2

E / V vs. RHE

Pt(111) Pt(110) Pt(100)

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

2 10-11

ion

curr

ent /

a.u

.

E / V vs. RHE

1 10-8

m/z 44

(a) (b) Pt(111)

(c) Pt(110) (d) Pt(100)

Figure 8.3. (a) Cyclic voltammograms of Pt(111) (solid line), Pt(110) (dashed line) and Pt(100) (dotted line) in 0.5 M MeOH and 0.5 M HClO4 at a scan rate of 2 mV⋅s-1. The top mass shown in the associated MSCVs for Pt(111) (b), Pt(110) (c), and Pt(100) (d) displays the methylformate signal (m/z 60) and the bottom one the carbon dioxide signal (m/z 44).

8.3.2.1. Methanol Oxidation on basal Pt Planes

The CVs recorded on the basal planes of platinum (Fig. 8.1a and 8.3a, respectively) correspond well with those recorded by Xia et al. in 0.1 M HClO4 at a scan rate of 50 mV⋅s-1,[67] and display an increase in (maximum) activity towards methanol oxidation in the order Pt(111)<Pt(110)<Pt(100). However, our results are in contrast to results published by Herrero et al., who reported an increasing activity in the order of Pt(111)<Pt(100)<Pt(110) in sulfuric as well as perchloric acid.[13] Comparing the Pt(110) blank cyclic voltammogram of Herrero et al. to ours, reveals significant differences in the surface structure and system cleanliness. We believe this suggests that the Pt(110) surface used by Herrero et al. was more disordered than ours and consisted mostly of the (1×2) reconstructed surface and other defects. Marković et al. indeed reported an increased activity for the oxidation of CO on the reconstructed (1×2) Pt(110) surface compared to (1×1) Pt(110).[68] Moreover, the maximum current density of the hydrogen adsorption/desorption peaks suggests that the cleanliness of the system was not

144

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

0.5 0.6 0.7 0.8 0.9

0

500

1000

1500

2000

j / µ

A cm

-2

E / V vs. RHE

Pt(111) Pt(554) Pt(553)

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

2 10-11

ion

curr

ent /

a.u

.

E / V vs. RHE

1 10-8

m/z 44

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

2 10-10

ion

curr

ent /

a.u

.

E / V vs. RHE

1 10-7

m/z 44

0.0 0.2 0.4 0.6 0.8 1.0

m/z 60

2 10-10

ion

curr

ent /

a.u

.

E / V vs. RHE

1 10-7

m/z 44

(a)

(c) Pt(554) (d) Pt(553)

(b) Pt(111)

Figure 8.4. (a) Cyclic voltammograms of Pt(111) (solid line), Pt(554) (dashed line) and Pt(553) (dotted line) in 0.5 M MeOH and 0.5 M HClO4 at a scan rate of 2 mV⋅s-1. The top mass shown in the associated MSCVs for Pt(111) (b), Pt(554) (c), and Pt(553) (d) displays the methylformate signal (m/z 60) and the bottom one the carbon dioxide signal (m/z 44).

optimal. Based on the onset of the reaction, the Pt(110) surface seems to be the most active, followed by Pt(111) and Pt(100).

Turning to product formation, Fig 8.1b shows that on Pt(111) small amounts of methylformate can be detected. In an in-situ FTIR study, Xia et al. also identified weak bands corresponding to methylformate on Pt(111) during the electrooxidation of methanol (0.1 M) in 0.1 M HClO4.[67] As the current density increases from Pt(110) to Pt(100) (Fig. 8.1c and d) the amounts of CO2 and methylformate produced increase. It is noteworthy that on Pt(111) CO2 can be detected in the negative going scan at potentials as low as 0.4 V vs. RHE, which is well below the potential at which surface adsorbed CO is oxidized. As the delay of the system for the detection of CO2 is in the order of seconds, we do not believe that the CO2 signal detected at 0.4 V is entirely due to the delay and tailing of the m/z 44 signal. Moreover, close examination of the m/z 44 signal obtained on Pt(111) in sulfuric acid always reveals a slight increase in signal between 0.43 and 0.35 V vs. RHE in the negative going scan, which cannot be explained by delay or tailing. This observation contradicts earlier findings reported by the group of Vielstich [18] and indicates that a parallel pathway is active at low potentials. Jarvi et al. also

145

Chapter 8

reported non-zero CO2 yields for potentials where surface adsorbed CO is stable on Pt(100) and Pt(111).[4, 69] In addition, our findings agree well with results published by the groups of Wieckowski [13, 14, 30] and Stuve,[16, 17, 69], who based on a coulometric analysis independently concluded that a second pathway to CO2, not involving surface-bonded CO, must be active at low potentials.

As is well known, changing the electrolyte from sulfuric to perchloric acid leads to an increase in the overall current densities recorded due to the lower adsorption strength of the latter anion (see Fig. 8.3a). The intensities of the m/z 60 and 40 signals shown in Fig. 8.3b-d are correspondingly higher.

We mentioned earlier that, by calculating the ratio between the amount of methylformate and CO2 detected during the voltammetric scan, the effect of surface structure and specific anion adsorption on the methanol oxidation pathway can be investigated without the interference of the typical OLEMS delay. The mass 60/44 ratios were calculated by dividing the total amount of methylformate corresponding to the m/z 60 MSCV (obtained by integration of the MSCV from the baseline) with the m/z 44 signal (obtained in the same way). The calculated values, which are an average of 2-3 independent measurements, are given in Table 8.1.

mass 60/44 ratio * 10-3 Surface H2SO4 HClO4 Pt(111) 3.3 1.5 Pt(110) 2.4 6.8 Pt(100) 1.2 1.7 Pt(554) 1.4 1.9 Pt(553) 1.8 3.1

Table 8.1. Ratio of the methylformate (m/z 60) and carbon dioxide (m/z 44) mass signals calculated for Pt(111), Pt(110), Pt(100), Pt(554) and Pt(553) in 0.5 M H2SO4 and 0.5 M HClO4 integrated over one voltammetric cycle at 2 mV⋅s-1. The values give are an average of two-three separate experiments.

As in the positive going scan no soluble products or CO2 are formed at low potentials, the signal detected between 0 and 0.5 V vs. RHE depends only on the background pressure in the OLEMS system. If the background pressure is constant, the detected constant mass signals can be used as baselines for the integration procedure. For background pressure fluctuations, the baseline is corrected accordingly. The tabulated values obtained accordingly have an average deviation of less than 10 - 20%, lending credit to the reliability of the method.

In sulfuric acid, the values for the mass 60/44 ratio clearly demonstrate that although Pt(100) has the highest overall activity, it has the lowest selectivity towards the direct oxidation pathway (through soluble intermediates). Compared to the CO2 formation, both Pt(111) and Pt(110) produce reasonably large quantities of methylformate, suggesting that the direct pathway plays an important role in the

146

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

oxidation of methanol on these surfaces. The results for Pt(111) and Pt(110) agree well with the literature, as both formaldehyde and formic acid have been reported to be among the primary species formed.[7, 31, 32] Using a combination of chronoamperometry, fast cyclic voltammetry and theoretical methods, Cao et al.[30] also found a low selectivity towards the direct methanol oxidation pathway on Pt(100), while both Pt(111) and Pt(110) display a relatively high selectivity for the direct pathway. Interestingly, Herrero et al. observed that Pt(100) reacts differently compared to Pt(111) and Pt(110).[13] On the latter two surfaces they reported a Tafel slope of 120 mV⋅dec-1, while 60 mV⋅dec-1 was found for Pt(100), suggesting a different rate determining step on the (100) surface.

Although the CVs recorded in perchloric acid appear similar to those recorded in sulfuric acid media, a markedly different trend in the methylformate-carbon dioxide ratio is observed. Compared to sulfuric acid the Pt(110) and Pt(100) surfaces are more active towards the direct oxidation pathway, as is apparent from the increased mass 60:44 ratio in Table 8.1, while Pt(111) favors the indirect oxidation pathway.

Batista et al. explained the enhanced role of the direct pathway on Pt(111) in sulfuric acid by a lack of multiple coordination sites for methanol adsorption due to strong adsorption of (bi)sulfate, leading to the formation of soluble intermediates rather than CO.[31, 32] It was suggested that dissociative adsorption of methanol on an ensemble of empty surface sites favors C-H bond scission and, thus, leads to CO formation. If such an “ensemble site” is absent, or if only a single adsorption site is available, methanol would interact with the surface by O-H bond scission and lead to the formation of soluble intermediates. This theory implies a considerable dissociation of methanol on (111) terrace sites, which seems in conflict with the idea that decomposition of methanol preferably takes place at steps.(see Chapter 7, ref. [22]) Perhaps the reaction rate on the terraces is much slower than on the steps, but still plays a prominent role in the formation of intermediates on surfaces with a low step or defect density.

Following the anion theory proposed by Batista et al., reducing the anion adsorption strength by switching from sulfuric acid electrolyte to perchloric acid should lead to an increase in the number of ensemble sites available for methanol decomposition to CO and, accordingly, to a lower mass 60/44 ratio. Indeed, as is apparent from Table 8.1, the mass 60:44 ratio for Pt(111) in sulfuric acid is twice as high as in perchloric acid. On all other surfaces, however, the formation of soluble intermediates is favored in perchloric acid media.

These findings indicate that if no adsorbed CO is present on the surface, competitive adsorption between (bi)sulfate and methanol will occur. As (bi)sulfate adsorbs strongly on the (111) surface and only few crystalline defects are present, the total reaction rate is low and the number of sites where methanol decomposes to soluble intermediates via O-H bond scission will be relatively high compared to the number of ensemble sites on which methanol can decompose to CO via C-H bond scission. Additionally, Olivi et al. suggested that oxidation of formaldehyde on Pt(110) and Pt(100) may occur by two different processes,[70] one involving the formation of COads and the other involving direct oxidation of methylene glycol. As methylene glycol

147

Chapter 8

oxidation to CO2 can occur at lower potentials on Pt(100) but is inhibited on Pt(110), the former surface produces more CO2 and, accordingly, has a lower mass 60/44 ratio than the latter.

The mass 60/44 ratios presented in Table 8.1 illustrate the effect of the structure sensitivity on the methanol oxidation pathway integrated over the forward and back potential scans. To give insight into the potential dependence of the direct and indirect pathway on each surface, Fig. 8.5 and 8.6 show the potential dependent mass 60/44 ratios of the basal planes of platinum recorded in sulfuric and perchloric acid, respectively. In these figures the ratio of the background levels detected by the mass spectrometer was set to zero. Therefore, the figures depict the deviation of the m/z 60:44 ratio (∆ ratio) from the zero line, rather than the actual ratios. We would like to note at this point that even

0.0 0.2 0.4 0.6 0.8 1.0

-6.00

-4.00

-2.00

0.00

2.00

Pt(100)

Pt(110)

Pt(111)

∆ ra

tio m

ass

60:4

4 si

gnal

s *1

0-3

E / V vs. RHE

0.0 0.2 0.4 0.6 0.8 1.0

-12.00

-8.00

-4.00

0.00

4.00

8.00

Pt(100)

Pt(110)

Pt(111)

∆ ra

tio m

ass

60:4

4 si

gnal

*10

-3

E / V vs. RHE

Figure 8.5. Potential dependent plot of the m/z 60:44 ratio measured for Pt(111) (solid line), Pt(110) (dashed line) and Pt(100) (dotted line) recorded in 0.5 M MeOH and 0.5 M H2SO4 at a scan rate of 2 mV⋅s-1.

Figure 8.6. Potential dependent plot of the m/z 60:44 ratio measured for Pt(111) (solid line), Pt(110) (dashed line) and Pt(100) (dotted line) recorded in 0.5 M MeOH and 0.5 M HClO4 at a scan rate of 2 mV⋅s-1.

148

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

though the delay time of the OLEMS is considerable compared to traditional DEMS techniques (10-15 seconds), we believe that it does not significantly influence trends observed in our results.

In the positive going scan, poisoning CO is oxidized to CO2, which results in decrease in the mass 60/44 ratio on all surfaces. In the return scan, competitive adsorption between (bi)sulfate and methanol reduces the number of free ensemble sites and, thus, favors the formation of soluble intermediates, as can be seen by the increase in the m/z 60:44 ratio between 0.85 and 0.6 V vs. RHE. Fig. 8.5 demonstrates that, in sulfuric acid, Pt(111) has the highest methylformate:CO2 ratio at high potentials (> 0.6 V vs. RHE) and, thus, favors the direct methanol oxidation pathway. At lower potentials CO is stable on the surface and only methylformate (or CO2 formed in the direct oxidation pathway) can be detected.

Interestingly, the maxima in the m/z 60:44 ratio for Pt(110) and Pt(100) lie at ca. 0.45-0.5 V, while Pt(111) shows a maximum at ca. 0.25-0.3 V vs. RHE in the negative going scan. Below 0.45 V (bi)sulfate starts to desorb from the (111) surface, which may result in an increase in the number of available sites for methanol decomposition. Methanol adsorbing on these sites can either be converted to COads or to soluble intermediates, which results in an increase in the methylformate:CO2 ratio.

In the high potential region, the m/z 60:44 ratio decreases when going from Pt(111) to Pt(110) and Pt(100) as a result of the increasing activity towards the indirect pathway in that order. In the negative going scan, even though the (100) surface shows the highest amount of m/z 60, the relative importance of the direct pathway activity on this surface is lower than that of Pt(110) (as is apparent from Table 8.1).

In perchloric acid, the oxidation characteristics of the surfaces change markedly (see Fig. 8.6). In this medium, the order of the relative selectivity towards the direct oxidation pathway is Pt(111)<Pt(100)<Pt(110). The decrease in the mass ratio for Pt(111) in the negative going scan can be ascribed to a slower decrease of the CO2 signal compared to the m/z 60 signal, as is visible in the corresponding MSCV (Fig. 8.3b), which may be due to CO2 formation by oxidation of soluble intermediates at low potentials. Despite the fact that the calculated values for Pt(111) and Pt(100) presented in Table 8.1 are quite similar, Fig. 8.6 demonstrates that Pt(100) produces relatively more soluble intermediates at lower potentials than Pt(111).

8.3.2.2. Methanol Oxidation on stepped Pt[n(111)×(111)] Type Electrodes

The CVs of the stepped electrodes in sulfuric acid (Fig. 8.2a) display all the characteristics of CVs recorded previously in 25 mM MeOH at 50 mV⋅s-1 (see Chapter 7) and show the expected increase in the maximum current density for increasing step density.[22] As the scan rate used here is 2 mV⋅s-1 the hydrogen region is nearly completely blocked by CO or other MeOH decomposition products.[22] As was anticipated, the total methanol oxidation activity in perchloric acid is higher than in sulfuric acid (compare Fig. 8.2a to Fig. 8.4a).

149

Chapter 8

The MSCVs presented in Fig. 8.2b-d and Fig. 8.4b-d indicate that apart from the expected increase in the CO2 signal, the methylformate signal also increases with the step density (see also Table 8.1). It was mentioned in the previous section that on well-defined Pt(111) in sulfuric acid only a small quantity of methylformate can be detected even at the relatively high methanol concentrations used in this study, but the ratio between methylformate and CO2 is relatively high compared to the stepped surfaces. In the presence of (bi)sulfate a comparison of the methylformate/CO2 ratios for Pt(554) and Pt(111) shows that introducing steps in the (111) plane results in an increase in the prominence of the indirect pathway. This observation may be rationalized by disruption of the (ordered) anion layer due to the presence of the step sites. As the step density is increased further from Pt(554) to Pt(553) in both electrolytes, the mass 60/44 ratios increase, which strongly suggests that steps themselves catalyze the formation of intermediates during the methanol oxidation. Based on experiments with polycrystalline, dispersed nano-particle and blackened Pt electrodes, numerous authors have already suggested that steps and crystalline defects catalyze the direct pathway.[19-21] DFT calculations using a (211) surface indicated that the prevalence of the direct oxidation pathway towards positive potentials may be due to preferred initial O-H cleavage over defect sites.[30] Note, however, that the steps on Pt(211) are of (100) orientation, which is precisely the surface which shows the lowest preference towards the direct oxidation pathway.

In perchloric acid, the amount of CO2 recorded on Pt(111) is higher relative to the amount of methylformate than in sulfuric acid media in agreement with the observations of Batista et al.[31, 32] As the step density is increased, the calculated methylformate:CO2 ratio steadily increases, illustrating that, in the absence of specifically adsorbing anions, steps catalyze the direct oxidation pathway of methanol to intermediates such as formic acid and formaldehyde.

The potential dependent m/z 60:44 ratios for the stepped surfaces in sulfuric and perchloric acid are shown in Fig. 8.7 and 8.8, respectively. In the positive going scan at potentials higher than 0.6 V vs. RHE in the presence of (bi)sulfate, Pt(111) has the highest and Pt(553) the lowest m/z 60:44 ratio, indicating that the influence of the indirect oxidation pathway increases with the step density. In the return scan, the results indicate that the selectivity towards the direct oxidation pathway increases in the order Pt(554)<Pt(111)<Pt(553). In perchloric acid the lower adsorption strength of perchlorate results in a change in this order to Pt(111)<Pt(554)<Pt(553). In general, compared to sulfuric acid media, in the perchloric acid electrolyte the relative importance of the indirect pathway in the positive going scan and the direct oxidation pathway in the negative going scan are enhanced.

These findings illustrate the varying effect of anion adsorption on the oxidation characteristics at steps and terraces. On (111) terraces, anion adsorption suppresses the formation of CO in the indirect pathway. On step sites, however, the strongly adsorbing (bi)sulfate anion seems to inhibit the direct pathway compared to perchlorate. It is possible that (bi)sulfate indirectly blocks the steps, or hinders the approach of methanol

150

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

to the steps, which by themselves catalyze the direct oxidation pathway, resulting in less methylformate being formed. In perchloric acid methanol is allowed easy access to the reactive step, which leads to more soluble intermediates being formed. Finally, reducing the terrace width results in a decrease in the role of the indirect oxidation pathway.

Further hints on the role of steps and defects in the formation of intermediates during the electrooxidation of methanol can be obtained from the position of the maxima of the mass signals. In the positive going scan the potential of the m/z 60 maximum in sulfuric acid media lies considerably lower than the m/z 44 signal maximum, signifying that intermediates can already be formed at potentials well below the oxidation potential of adsorbed CO.[26] Moreover, in the negative going scan the methylformate maximum recorded on Pt(554) in 0.5 M H2SO4 lies at ca. 0.54 V and at ca. 0.52 V vs. RHE on Pt(553), and in HClO4 at ca. 0.48 V and 0.40 V, respectively, indicating that on surfaces

0.0 0.2 0.4 0.6 0.8 1.0

-2.00

0.00

2.00

4.00

6.00

Pt(553)

Pt(554)Pt(111)

∆ ra

tio m

ass

60:4

4 si

gnal

s *1

0-3

E / V vs. RHE

0.0 0.2 0.4 0.6 0.8 1.0

-12.00

-8.00

-4.00

0.00

4.00

8.00

12.00

Pt(553)

Pt(554)

Pt(111)

∆ ra

tio m

ass

60:4

4 si

gnal

*10

-3

E / V vs. RHE

Figure 8.7. Potential dependent plot of the m/z 60:44 ratio measured for Pt(111) (solid line), Pt(554) (dashed line) and Pt(553) (dotted line) recorded in 0.5 M MeOH and 0.5 M H2SO4 at a scan rate of 2 mV⋅s-1.

Figure 8.8. Potential dependent plot of the m/z 60:44 ratio measured for Pt(111) (solid line), Pt(554) (dashed line) and Pt(553) (dotted line) recorded in 0.5 M MeOH and 0.5 M HClO4 at a scan rate of 2 mV⋅s-1.

151

Chapter 8

with a higher step density the direct pathway is active at lower potentials. The delay in the m/z 60 signal may also be attributed to an accumulation of methylformate in the electrolyte near the tip. However, the differences in the potential of the m/z 60 maximum and the maximum current density are approximately 115 mV for Pt(554) and 150 mV for Pt(553) in 0.5 M H2SO4, which, at a scan rate of 2 mV⋅s-1, implies delay times of 57 and 75 seconds, respectively. In perchlorate containing electrolytes, these delay times are even higher. As these delay times are considerably longer than previously reported typical delay times of the OLEMS system,(see ref. [48]) we believe that they are not due to slow diffusion of methylformate from the electrode to mass spectrometer, but rather hint at the formation of methylformate at potentials considerably lower than the maximum in the current density in the negative going scan.

Finally, it is noteworthy to point out that the oxidation current densities on Pt(554) and Pt(553) are higher than on Pt(110). Apparently stepped surfaces of [n(111)×(111)] orientation are more active than the [110] steps themselves. This may indicate that, the combination of a step site and a neighboring terrace site is required in order to decompose methanol to CO. If only step or terrace adsorption sites are available, methanol preferably decomposes to soluble intermediates

8.3.3. Methanol Oxidation Scheme

In this section, we would like to draw up a detailed scheme for methanol oxidation, based on the results presented here as well as previously published results. This scheme 8.1 incorporates the formation of soluble intermediates such as formaldehyde and formic acid, but also assists in understanding the structure sensitivity of the reaction and the effects of anion adsorption. Scheme 8.1 is essentially an extension of the scheme suggested recently by Cao et al.,[30] incorporating as an additional feature the presumed pathways for the formation of formaldehyde and formic acid.

As in the scheme of Cao et al.,[30] it is suggested that the decision between the direct and indirect pathways is made at the initial dehydrogenation step, i.e. the breaking of the O-H or C-H bond, respectively. The dehydrogenation step of methanol to hydroxymethyl, reaction (1), was found to be the rate determining step on Pt(111) and Pt(110),[13, 71] while on Pt(100) reaction (2) was assumed to be rate determining.[13] Our results indicate that (111) terraces in the absence of strongly adsorbing anions (i.e. (bi)sulfate), but also the Pt(100) surface, favor reaction (1). The presence of an H:C:O species has been extensively discussed in the literature.[1, 8, 10, 72-76] Whether this H:C:O species is HCO formed in reaction (3) or COH formed in reaction (4), is at present still unclear. However, it is generally accepted that at low potentials ultimately COads is formed, which acts as a surface poison.[1, 4-7] Depending on the coverage and potential, linearly, bridge or multifold-bonded CO can be formed, with the most prevalent configuration being linearly bonded CO.[55, 77]

152

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

HO

C

H3COHsol

OH C

O

CH2

O

CH3

C

O

CO2

H2C OH

HC OH

HCOOHsol + 2H+ + 2e-

O O

C

H

CO2

(8)(1)

(2) (9)

(3) (4)

(5) (6)

(7) (15)

?

H2COsol H2C(OH)2

+ H2O

(10)

(11)

(12)

(13)

(16)

HCOOCH3, sol

(14)

H3COHsol

Scheme 8.1. Schematic representation of the parallel pathway mechanism for methanol oxidation on platinum electrodes. The reactions involving the formation of CO constitute the indirect pathway, while the reactions through formic acid and formaldehyde constitute the direct pathway.

Under UHV conditions the dehydrogenation reaction (8) to methoxy is known to occur readily.[78-82] Under electrochemical conditions, our results indicate that O-H bond scission occur preferably on Pt(111) in sulfuric acid media and on steps of (110) orientation in the absence of (bi)sulfate. On Pt(110), reaction (8) also seems relatively important, regardless of the electrolyte. Calculations by Cao et al. indicate that (100) type steps also catalyze the O-H bond scission, even though the Pt(100) surface itself seems to favor C-H bond scission.[30] Further dehydrogenation of H3COads leads to H2COads (9), which may desorb as formaldehyde. This reaction is depicted in the animation (which is

153

Chapter 8

visible by flipping rapidly through the pages of this thesis) in the top right- and left-hand corner of this thesis. Whether or not H2COads reacts to a H:C:O species is not clear. Once formed, formaldehyde is rapidly and almost completely hydrated to methylene glycol, reaction (10).[51, 52] As adsorbed CO was found during oxidation of formaldehyde on platinum, Miki et al. suggested that methylene glycol dissociates to COads,[54] although Olivi et al. proposed a reaction involving non-hydrated formaldehyde.[83] Olivi et al. also reported that the Pt(100) surface is more active for the H2C(OH)2 oxidation than the Pt(110) surface.[70] This would agree well with our OLEMS data. If Pt(100) is more active for the oxidation of methylene glycol (formed during the oxidation of methanol) to CO2 than Pt(110), less formic acid would be produced on this surface and, hence, a lower ratio of methylformate/CO2 should be recorded, as indeed is the case.

Finally, our scheme explains why Osawa and co-workers observed the formation of formate on platinum surfaces during the electrooxidation of methanol, formic acid and formaldehyde (methanol: reaction (8)-(12), formic acid: reaction (13) and formaldehyde: reaction (10)+(11) and (10)+(12)+(13)).[54-56] Furthermore, they proposed that formate may react to either CO2 or COads (reaction (15) and (16), respectively). The formation of formic acid may occur via desorption of formate from the surface, or by direct electrooxidation of H2C(OH)2, and ultimately leads to the production of methylformate by a reaction with methanol, reaction (14).

We believe that scheme 8.1 presents a fairly complete and consistent picture of the methanol oxidation mechanism on platinum. The reaction path following (1) to (7) would constitute what is commonly referred to as the indirect methanol oxidation pathway, while the series of reactions from (8) to (15) represent the direct methanol oxidation pathway.

8.4. Conclusions

In this chapter we have investigated the selectivity and structure sensitivity of the methanol oxidation pathways on basal planes and stepped platinum single crystal electrodes by monitoring the mass fractions of CO2 (m/z 44) and methylformate (m/z 60, formed from CH3OH + HCOOH) in 0.5 M H2SO4 and 0.5 M HClO4 using On-Line Electrochemical Mass Spectrometry. We can summarize our main conclusions as follows:

• Contrary to previously published literature, in sulfuric and perchloric acid the (maximum) reactivity for the methanol oxidation reaction increases in the order Pt(111)<Pt(110)<Pt(100).

• In sulfuric acid media, the oxidation of methanol on Pt(111) shows a preference for the direct oxidation pathway, while in perchloric acid media the indirect pathway through COads is favored. These observations may be explained by assuming that strongly adsorbing anions reduce the amount of available “ensemble sites” necessary for dissociative adsorption of methanol via C-H bond cleavage, which normally would lead to COads formation. If only single

154

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

adsorption sites (or insufficient ensemble sites) are available for the adsorption of methanol, O-H bond scission is favored, leading to the formation of soluble intermediates. Therefore, in the presence of weakly adsorbing anions the indirect oxidation pathway is favored, as is the case for perchlorate.

• In the absence of specifically adsorbed anions, steps of (110) orientation catalyze the formation of soluble intermediates. In sulfuric acid, introducing steps in the (111) terraces leads to an increase in the relative rate of the indirect oxidation pathway, presumably as a result of disruption of the anion adlayer. However, with the exception of Pt(111), the relative importance of the direct oxidation pathway is always higher in perchloric acid compared to sulfuric acid.

• As well-ordered Pt(110) has a lower activity than the stepped surfaces Pt(554) and Pt(553) in both electrolytes, methanol appears to decompose preferably at step sites, which are directly neighbored by a (111) terrace site. This combination, i.e. step plus terrace site, seems a particular favorable “ensemble site” for methanol decomposition. Finally, based on the literature and the OLEMS data a detailed scheme of the

methanol oxidation mechanism was presented. The scheme incorporates a parallel pathway, the formation of soluble intermediates like formic acid and formaldehyde, the selectivity and structure sensitivity of the formation of these intermediates, the effect of anion adsorption, as well as the occurrence of surface adsorbed intermediates as detected by IR spectroscopy.

Acknowledgement: This research was supported by the Netherlands Foundation for Scientific Research (NWO).

References

[1] R. Parsons and T. VanderNoot, J. Electroanal. Chem. 257 (1988) 9. [2] A. Hamnett, Catal. Today. 38 (1997) 445. [3] J. Lipkowski and P. N. Ross, Electrocatalysis: Frontiers in Electrochemistry, Wiley, New York,

1998. [4] T. D. Jarvi, S. Sriramulu, and E. M. Stuve, J. Phys. Chem. B 101 (1997) 3649. [5] A. Hamnett, Comp. Chem. Kin. (1999) 635. [6] N. M. Markovic and P. N. Ross, Surf. Sci. Rep. 45 (2002) 117. [7] T. Iwasita, Electrochim. Acta. 47 (2002) 3663. [8] J. Willsau, O. Wolter, and J. Heitbaum, J. Electroanal. Chem. Inter. Electrochem. 185 (1985) 163. [9] T. Iwasita, W. Vielstich, and E. Santos, J. Electroanal. Chem. Inter. Electrochem. 229 (1987) 367. [10] S. Wilhelm, T. Iwasita, and W. Vielstich, J. Electroanal. Chem. Inter. Electrochem. 238 (1987)

383. [11] M. I. Lopes, I. Fonseca, P. Olivi, B. Beden, F. Hahn, J. M. Leger, and C. Lamy, J. Electroanal.

Chem. 346 (1993) 415. [12] H. Gerischer, C. W. Tobias, and Editors, Advances in Electrochemical Science and Engineering,

Vol. 1, 1990. [13] E. Herrero, K. Franaszczuk, and A. Wieckowski, J. Phys. Chem. B 98 (1994) 5074. [14] E. Herrero, W. Chrzanowski, and A. Wieckowski, J. Phys. Chem. 99 (1995) 10423. [15] G. Q. Lu, W. Chrzanowski, and A. Wieckowski, J. Phys. Chem. B 104 (2000) 5566. [16] S. Sriramulu, T. D. Jarvi, and E. M. Stuve, Electrochim. Acta. 44 (1998) 1127. [17] S. Sriramulu, T. D. Jarvi, and E. M. Stuve, J. Electroanal. Chem. 467 (1999) 132.

0.85 V

155

Chapter 8

[18] W. Vielstich and X. H. Xia, J. Phys. Chem. 99 (1995) 10421. [19] H. Wang, T. Loffler, and H. Baltruschat, J. Appl. Electrochem. 31 (2001) 759. [20] H. Wang and H. Baltruschat, Proc. - Electrochem. Soc. 2001-4 (2001) 50. [21] H. Wang, C. Wingender, H. Baltruschat, M. Lopez, and M. T. Reetz, J. Electroanal. Chem. 509

(2001) 163. [22] T. H. M. Housmans and M. T. M. Koper, J. Phys. Chem. B 107 (2003) 8557. [23] S. Gilman, J. Phys. Chem. 68 (1964) 70. [24] H. Massong, S. Tillmann, T. Langkau, E. A. Abd El Meguid, and H. Baltruschat, Electrochim.

Acta. 44 (1998) 1379. [25] W. Akemann, K. A. Friedrich, U. Linke, and U. Stimming, Surf. Sci. 402-404 (1998) 571. [26] N. P. Lebedeva, M. T. M. Koper, E. Herrero, J. M. Feliu, and R. A. van Santen, J. Electroanal.

Chem. 487 (2000) 37. [27] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Electroanal. Chem. 524-

525 (2002) 242. [28] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Phys. Chem. 106 (2002)

12938. [29] N. P. Lebedeva, A. Rodes, J. M. Feliu, M. T. M. Koper, and R. A. van Santen, J. Phys. Chem. B

106 (2002) 9863. [30] D. Cao, G. Q. Lu, A. Wieckowski, S. A. Wasileski, and M. Neurock, J. Phys. Chem. B 109 (2005)

11622. [31] E. A. Batista, G. R. P. Malpass, A. J. Motheo, and T. Iwasita, Electrochem. Comm. 5 (2003) 843. [32] E. A. Batista, G. R. P. Malpass, A. J. Motheo, and T. Iwasita, J. Electroanal. Chem. 571 (2004)

273. [33] S. Park, Y. Xie, and M. J. Weaver, Langmuir 18 (2002) 5792. [34] A. V. Tripkovic and K. D. Popovic, Electrochim. Acta 41 (1996) 2385. [35] Z. Jusys, J. Kaiser, and R. J. Behm, Electrochim. Acta. 47 (2002) 3693. [36] A. V. Tripkovic, K. D. Popovic, B. N. Grgur, B. Blizanac, P. N. Ross, and N. M. Markovic,

Electrochim. Acta. 47 (2002) 3707. [37] J. Shin and C. Korzeniewski, J. Phys. Chem. B 99 (1995) 3419. [38] R. R. Bruckenstein and J. Gadde, J. Am. Chem. Soc. 93 (1971) 793. [39] O. Wolter and J. Heitbaum, Ber. Bunsen-Ges. 88 (1984) 2. [40] O. Wolter and J. Heitbaum, Ber. Bunsen-Ges. 88 (1984) 6. [41] T. Hartung and H. Baltruschat, Langmuir 6 (1990) 953. [42] H. Baltruschat and U. Schmiemann, Ber. Bunsen-Ges. 97 (1993) 452. [43] H. Baltruschat, J. Am. Soc. Mass Spectrometry 15 (2004) 1693. [44] Z. Jusys, H. Massong, and H. Baltruschat, J. Electrochem. Soc. 146 (1999) 1093. [45] H. Kita and H.-W. Lei, J. Electroanal. Chem. 388 (1995) 167. [46] K. Jambunathan and A. C. Hillier, J. Electrochem. Soc. 150 (2003) E312. [47] K. Jambunathan, S. Jayaraman, and A. C. Hillier, Langmuir 20 (2004) 1856. [48] A. H. Wonders, T. H. M. Housmans, V. Rosca, and M. T. M. Koper, J. Appl. Electrochem.

submitted (2005) [49] C. Korzeniewski and C. L. Childers, J. Phys. Chem. B 102 (1998) 489. [50] C. L. Childers, H. Huang, and C. Korzeniewski, Langmuir 15 (1999) 786. [51] J. P. Guthrie, Can. J. Chem. 53 (1975) 898. [52] J. G. M. Winkelman, O. K. Voorwinde, M. Ottens, A. A. C. M. Beenackers, and L. P. B. M.

Janssen, Chem. Eng. Sci. 57 (2002) 4067. [53] S.-G. Sun, G.-Q. Lu, and Z.-W. Tian, J. Electroanal. Chem. 393 (1995) 97. [54] A. Miki, S. Ye, T. Senzaki, and M. Osawa, J. Electroanal. Chem. 563 (2004) 23. [55] Y. X. Chen, A. Miki, S. Ye, H. Sakai, and M. Osawa, J. Am. Chem. Soc. 125 (2003) 3680. [56] A. Miki, S. Ye, and M. Osawa, Chem. Commun. (2002) 1500. [57] J. Clavilier, D. Armand, S. G. Sun, and M. Petit, J. Electroanal. Chem. 205 (1986) 267. [58] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, Electrochem.Commun. 2

(2000) 487. [59] A. Al-Akl, G. A. Attard, R. Price, and B. Timothy, J. Electroanal. Chem. 467 (1999) 60. [60] A. M. Funtikov, U. Stimming, and R. Vogel, J. Electroanal. Chem. 428 (1997) 147. [61] M. T. M. Koper and J. J. Lukkien, J. Electroanal. Chem. 485 (2000) 161. [62] J. Clavilier, K. El Achi, and A. Rodes, J. Electroanal. Chem. 272 (1989) 253. [63] J. Clavilier and A. Rodes, J. Electroanal. Chem. 348 (1993) 247. [64] J. Clavilier, K. El Achi, M. Petit, A. Rodes, and M. A. Zamakhchari, J. Electroanal. Chem. 295

(1990) 333.

156

Structure sensitivity of MeOH electrooxidation pathways on Pt: an OLEMS study

[65] J. Clavilier, K. El Achi, and A. Rodes, Chem. Phys. 141 (1990) 1. [66] Y. Gao, H. Tsuji, H. Hattori, and H. Kita, J. Electroanal. Chem. 372 (1994) 195. [67] X. H. Xia, T. Iwasita, F. Ge, and W. Vielstich, Electrochim. Acta 41 (1996) 711. [68] N. M. Markovic, B. N. Grgur, C. A. Lucas, and P. N. Ross, Surf. Sci. 384 (1997) L805. [69] T. D. Jarvi, S. Sriramulu, and E. M. Stuve, Coll Surf. A 134 (1998) 145. [70] P. Olivi, L. O. S. Bulhoes, J. M. Leger, F. Hahn, B. Beden, and C. Lamy, Electrochim. Acta. 41

(1996) 927. [71] K. Franaszczuk, E. Herrero, P. Zelenay, A. Wieckowski, J. Wang, and R. I. Masel, J. Phys. Chem.

B 96 (1992) 8509. [72] T. Iwasita, W. Vielstich, and E. Santos, J. Electroanal. Chem. 229 (1987) 367. [73] J. Willsau and J. Heitbaum, Electrochim. Acta. 31 (1986) 943. [74] J. Willsau and J. Heitbaum, J. Electroanal. Chem. Inter. Electrochem. 185 (1985) 181. [75] W. Vielstich, P. A. Christensen, S. A. Weeks, and A. Hamnett, J. Electroanal. Chem. Inter.

Electrochem. 242 (1988) 327. [76] S. Wilhelm, W. Vielstich, H. W. Buschmann, and T. Iwasita, J. Electroanal. Chem. Inter.

Electrochem. 229 (1987) 377. [77] B. Beden, C. Lamy, A. Bewick, and K. Kunimatsu, J. Electroanal. Chem. Inter. Electrochem. 121

(1981) 343. [78] J. L. Davis and M. A. Barteau, Surf. Sci. 187 (1987) 387. [79] N. Kizhakevariam and E. M. Stuve, Surf. Sci. 286 (1993) 246. [80] J. Wang and R. I. Masel, Surf. Sci. 243 (1991) 199. [81] J. Wang and R. I. Masel, J. Vac. Sci. Technol. A 9 (1991) 1879. [82] B. A. Sexton, Surf. Sci. 102 (1981) 271. [83] P. Olivi, L. O. S. Bulhoes, J. M. Leger, F. Hahn, B. Beden, and C. Lamy, J. Electroanal. Chem.

370 (1994) 241.

157

Chapter 8

158

0.0 0.2 0.4 0.6 0.8

-300

-200

-100

0

100

200

300

Supporting information

j / µ

A cm

-2

E / V vs. RHE

0.5 M H2SO4 0.5 M HClO

4

0.0 0.2 0.4 0.6 0.8

-400

-300

-200

-100

0

100

200

300

400

j / µ

A cm

-2E / V vs. RHE

0.5 M H2SO

4 0.5 M HClO4

0.0 0.2 0.4 0.6 0.8-150

-100

-50

0

50

100

j / µ

A cm

-2

E / V vs. RHE

0.5 M H2SO4 0.5 M HClO4

0.0 0.2 0.4 0.6 0.8

-200

-100

0

100

200

j / µ

A cm

-2

E / V vs RHE

0.5 M H2SO4 0.5 M HClO4

0.0 0.2 0.4 0.6 0.8-200

-150

-100

-50

0

50

100

150

200

j / µ

A cm

-2

E / V vs. RHE

0.5 M H2SO4 0.5 M HClO4

(a) Pt(111)

(d) Pt(554) (c) Pt(100)

(b) Pt(110)

(e) Pt(553)

S8.1. Blank cyclic voltammograms of Pt(111) (a), Pt(110) (b), Pt(100) (c), Pt(554) (d) and Pt(553) (e) in 0.5 M H2SO4 (solid line) and 0.5 M HClO4 (dashed line) at a scan rate of 50 mV⋅s-1.

The electrooxidation of small organic molecules on platinum nanoparticles supported on gold: particle size vs. particle shape effect Abstract

The electrocatalytic properties of small platinum nanoparticles were investigated for the oxidation of CO, methanol and formic acid using voltammetry, chronoamperometry and Surface Enhanced Raman spectroscopy. The particles were generated by galvanostatic deposition of platinum on a polished gold surface from an H2PtCl6 containing electrolyte, and ranged between 10-20 nm in diameter for low platinum surface concentrations, 10-120 nm for medium concentrations, and full Pt monolayers for high concentrations. CO stripping and bulk CO oxidation experiments on the particles up to 120 nm in diameter displayed a pronounced “particle size effect”. The CO oxidation current-time transients show a current decay for low platinum coverages and a current maximum for medium and high coverages. These results were also observed in the literature for particles of 2-5 nm size and agglomerates of these particles. The similarities between the literature and our results, despite large differences in particle diameter, suggest that particle morphology is more important that actual size. Surface Enhanced Raman Spectroscopy data obtained for the oxidation of CO on the Pt-modified Au electrodes corroborates this conclusion. A difference in the ratio between CO adsorbed in linear- and bridge-bonded positions on the Pt nanoparticles of different sizes demonstrates the influence of the surface morphology. The oxidation activity of methanol was found to decrease with the particle size, while the formic acid oxidation rate increases. Again a “particle size effect” is observed for particles of up to ca. 120 nm in diameter, which is much larger than the particles for which a similar effect was reported in the literature. This emphasizes the previously made suggestion that “particle size” is not as important as particle shape. The particle shape effect for the methanol oxidation reaction can be explained by a reduction in available “ensemble sites” and a reduction in the mobility of CO formed by decomposition of methanol. As formic acid does not require Pt ensemble sites decreasing the particle size, and thus, the relative number of defects, increases the reaction rate.

This chapter is published as F.J.E. Scheijen, S. Höppener, G.L. Beltramo, T.H.M. Housmans,M.T.M. Koper, in preparation

Chapter 9

9.1. Introduction

Particles of nanoscale diameter are of considerable scientific interest, due to the different, often unexpected, catalytic properties compared to bulk electrode materials. Moreover, they are commonly used in technical electrodes. Therefore, understanding of their electrochemical properties is an important issue in the development of efficient fuel cell catalysts. Especially the electrocatalysis of small organic molecules, such as carbon monoxide (CO), methanol (MeOH) and formic acid (HCOOH), is an interesting field of study, as these reactions play an important role in fuel cell catalysis [1, 2] and they display pronounced size-dependent catalytic properties.[3-9] Due to the fact that platinum is one of the most active dehydrogenation catalysts, Pt nanoparticles are of particular significance.

In a recent paper, Friedrich et al.[9] reported a distinct particle size dependence for the electrooxidation of carbon monoxide on colloidal platinum nanoparticles supported on polycrystalline gold. They found that the CO oxidation potential on particles with an average diameter of 3 nm is shifted to higher potentials (300-500 mV) when compared to polycrystalline platinum electrodes. When the diameter of the colloidal particles is increased from 3 to 10-16 nm, the oxidation potential was found to decreases again. It was suggested that the origin of this “particle size effect” lies in the geometrical structure of the particles (and to a lesser extent in the electronic properties).

Similar effects were observed by Maillard et al. for platinum nanoparticles ranging from 1 to 4 nm supported on glassy carbon (GC).[10] They ascribed the size effect to a limited mobility of COads on particles smaller than 2 nm. A particle size dependent diffusion coefficient was assumed, which is low for small particles (<2 nm) and higher for larger particles (>3 nm). Although Arenz et al.[11] also obtained similar results for 1-5 nm Pt particles supported on carbon, they attributed the difference in CO oxidation potential for varying particle sizes to the number of defects present on the surface of a particle. As defects are able to dissociate water to form OHads more easily, the higher number of defects on larger particles results in a shift of the CO oxidation peak to lower potentials.

For the oxidation of methanol, Park et al. reported a particle size dependence for carbon supported particles in the range of 2-9 nm.[8] On particles larger than 4 nm in diameter, the recorded methanol oxidation current densities were found to approach those of polycrystalline Pt electrodes. The decreasing MeOH oxidation rate for particles smaller than 4 nm was ascribed to a decrease in contiguous Pt terrace sites. This particle size dependence of the methanol oxidation reaction was also observed by Frelink et al.[6] However, they measured a more or less constant reaction rate for particles larger than 4.5 nm. Two possible explanations for the decrease in activity with decreasing particle size were given: 1) on smaller particles the coverage of OHads increases, thus blocking more empty sites, resulting in a decrease in the methanol decomposition rate, or 2) smaller particles have fewer preferential adsorption sites for methanol. In addition to these findings, Cherstiouk et al.[7] observed lower methanol oxidation rates for Pt particles in

160

Electrooxidation of small organic molecules on Pt nanoparticles supported on Au

the range of 1.5-3 nm compared to polycrystalline Pt, which they ascribed to stronger poisoning of the nanoparticles as compared to polycrystalline surfaces due to slower oxidation of CO. Despite numerous attempts to explain the origins of the “particle size effect” on these electrooxidation reactions, it is still unclear whether the observed phenomena are a consequence of electronic or geometric effects.

The effect of the particle size in the electrooxidation of formic acid was also investigated by Park et al.[8] In contrast to methanol, the oxidation rate of formic acid on small particles was found to be considerably higher than on larger particles. It was suggested that formic acid, unlike methanol, does not require Pt ensemble site to react to CO2 and, thus, in combination with a lower poisoning rate, higher oxidation rates are observed. Moreover, unlike the oxidation of CO and methanol, formic acid does not need an additional oxygen atom (i.e. no reaction with surface oxygen containing species, like OHads, is necessary) to form carbon dioxide.

Apart from an interesting particle size effect, the three molecules discussed above are also interesting to study from the perspective of the methanol oxidation scheme.(see also Chapters 7 and 8) Methanol is one of the most promising fuels for a fuel cell, as it can be catalytically oxidized on a platinum surface, yielding CO2 and six electrons per methanol molecule.

CH3OH + H2O → CO2 + 6H+ + 6e- (9.1) This reaction has a favorable thermodynamic potential of 0.04 V vs. RHE and can, theoretically, allow fuel cell power outputs close to that of a hydrogen-based fuel cell. Unfortunately, formation of surface poisoning species and soluble intermediates, lower the overall efficiency of the fuel cell. Carbon monoxide has been identified as the primary poisoning species in many different studies,[10, 12-15] while formic acid and formaldehyde are formed as intermediate products.[16-19] (See the overall methanol oxidation scheme as proposed in Chapter 8, Scheme 8.1) It is generally accepted that the oxidation of methanol proceeds through a dual pathway, namely a direct pathway involving the formation of intermediate species such as formic acid and formaldehyde, and an indirect pathway via adsorbed CO. Again, the effects of the particle size on the selectivity and activity of the direct and indirect pathways are not known.

From this perspective it is interesting to study the mechanism and kinetics of the oxidation reaction of these three molecules on platinum nanoparticles. The goal in this chapter is to gain deeper understanding of the anomalous catalytic properties of nanoparticles for the oxidation of carbon monoxide, methanol and formic acid. As CO acts as a neutral probe toward the surface reactivity and structure sensitivity this reaction will be used as a model reaction for studying the particle size effect. The oxidation of methanol and formic acid will be discussed in light of the findings presented for CO. The Pt nanoparticles will be generated by galvanostatic deposition from a platinum containing solution on a polycrystalline gold support and characterized by cyclic voltammetry and Atomic Force Microscopy (AFM).[9] The influence of the particle size on the catalytic activity towards the oxidation of CO, MeOH and HCOOH will be investigated using cyclic voltammetry, chronoamperometry and Surface-Enhanced Raman Spectroscopy.

161

Chapter 9

9.2. Experimental Setup

All measurements were performed in a standard three-electrode cell, which was cleaned by boiling in a 1:1 mixture of concentrated sulfuric acid and nitric acid, followed by boiling twice with Milli-Q water (Millipore Milli-Q gradient A10 system, 18.2 MΩ cm, 3 ppb total organic carbon) prior to each experiment. The counter electrode consisted of a coiled platinum wire, and the reference electrode consisted of a mercury-mercury sulfate electrode (MMSE: Hg|Hg2SO4|K2SO4 (sat)) connected via Luggin capillary.

The working electrode was a gold disk of 5 mm in diameter embedded in a Teflon sheath, which was mechanically polished with alumina (up to 0.3 µm), rinsed and treated ultrasonically in Milli-Q water before use. Prior to each experiment the electrode was cleaned by applying twenty-five potential sweep oxidation and reduction cycles from 0.055 V to 1.805 V vs. RHE at a scan rate of 200 mV⋅s-1.

The blank electrolyte consisted of 0.5 M H2SO4 and was prepared with concentrated sulfuric acid (Merck, “Suprapur”) and Milli-Q water. The methanol and formic acid containing electrolytes consisted of the blank electrolyte and 0.1 M methanol (Merck, pro analysi) and 0.1 M formic acid (Merck, extra pure), respectively. Argon (Hoekloos, N50) was used to deoxygenate all solutions.

The AFM used for the characterization of the surfaces was a Multimode Nanoscope IIIa from Digital Instruments (DI), California. For the images a standard Si tip from Mikromasch (NSC 35, with a typical force constant of 7.5 N m-1) was used in the tapping mode.

Surface-Enhanced Raman Spectroscopy (SERS) measurements were performed using a DILOR T64000 spectrograph with a holographic grating of 600 gr⋅mm-1. The slit and pinhole of the system were both set at 100 µm and a CCD camera with 1024x256 pixels was used as detector. A 633 nm excitation line from a He-Ne laser was used with a power of 13 mW on the sample. The microscope objective was an Olimpus 50X, which was not immersed in the electrolyte. The laser spot on the electrode surface had a diameter of ~5µm. A notch filter was used to filter the SERS signal produced inside the fiber optics before reaching the sample and the excitation line after scattering from the sample. With this configuration a resolution of 1.5 cm-1 can be obtained.

The electrode in the SERS experiments was roughened by applying a succession of potential sweep oxidation and reduction cycles in 0.1 M KCl at 500 mV⋅s-1 from 1.25 V to –0.25 V vs. the Ag/AgCl electrode, with waiting times of 1.3 seconds at 1.25 V and 30 seconds at –0.25 V.[20] The electrochemical cell had an optical quartz window parallel to the electrode surface. Measurements were performed at room temperature (22 ºC) with a computer-controlled Autolab potentiostat (PGSTAT12). All potentials reported here have been converted to the RHE scale.

162

Electrooxidation of small organic molecules on Pt nanoparticles supported on Au

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

-400

-350

-300

-250

-200

-150

-100

-50

0

50

100

j / µ

A cm

-2

E / V vs. RHE

Figure 9.1. Blank cyclic voltammogram of polycrystalline gold in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

9.3. Results and Discussion

9.3.1. System Cleanliness and Surface Structure

The system cleanliness and structure of the gold electrode was checked by measuring a blank cyclic voltammogram (BCV) in 0.5M H2SO4 at a scan rate of 20 mV⋅s-1. A typical voltammogram of the polished electrode is shown in Fig. 9.1. The BCV displays the characteristically low double layer charge and typical surface oxidation region, which starts at ca. 1.35V vs. RHE.

9.3.2. Platinum Deposition and Atomic Force Microscopy Imaging

The platinum nanoparticles were generated by galvanostatic deposition of platinum on the gold substrate from a 0.02 M H2PtCl6 containing solution. Three surfaces with low, medium and high platinum concentrations were generated by applying a constant current and changing the deposition time. The applied current was 50 µA, while the deposition times were 5 and 40 seconds for the medium and high Pt coverages, respectively. For deposition times shorter than five seconds the reproducibility of the results decreases. Therefore, the surface with the lowest platinum concentration was generated by applying a current of 5 µA for 5 seconds.

The generated surfaces were characterized by cyclic voltammetry and AFM imaging. The corresponding BCVs and AFM images are presented in Fig. 9.2 and Fig. 9.3, respectively. Fig. 9.2 shows that increasing the amount of platinum by increasing the deposition time, results in an increase in the charge in the low potential region (between 0.05 V and 0.3 V vs. RHE), which can be ascribed to hydrogen ad- and desorption on the generated platinum particles. The feature starting at 0.85 V in the positive going scan can be ascribed to oxidation of the deposited platinum. The corresponding oxide reduction peak is located at 0.67 V vs. RHE in the negative going scan. Oxidation and reduction of

163

Chapter 9

0.0 0.4 0.8 1.2 1.6-200

-150

-100

-50

0

50

100

j / µ

A cm

-2

E / V vs. RHE

Low Medium High

0.0 0.4 0.8 1.2 1.6-40

-20

0

20

40

60

80

j / µ

A c

m-2

E / V vs. RHE

Figure 9.2. Blank cyclic voltammograms of the platinum modified gold surfaces in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1. The inset shows a CV of gold with a very high Pt concentration.

the gold surface occurs at 1.35 V in the positive going scan and 1.17 V in the negative going scan, respectively. As the platinum concentration is increased, the Pt oxidation/reduction charge increases, while the gold oxidation/reduction charge decreases due to masking of the substrate surface. Similar results were obtained by Friedrich et al. in HClO4 by depositing colloidal platinum particles on polished gold electrodes.[9] At high platinum concentrations the shape of the surface oxidation region starts to resemble that of polycrystalline Pt. Deposition of very large quantities of platinum results in CVs similar to those given by Möller et al.[21] for platinum-gold alloys (see the inset in Fig. 9.2).

The coverage of the gold supported platinum particles can be calculated by a method introduced by Friedrich et al., which is based on the real surface area of the gold electrode and the area of the Pt particles.[9] The real surface area of the gold electrode, AAu

0, is easily determined by measuring the charge of the gold oxide reduction peak in a blank cyclic voltammogram recorded from 0.05 to 1.7 V vs. RHE at a scan rate of 20 mV⋅s-1, which forms a gold oxide layer that should have a charge of ca. 380 µC⋅cm-2.[22, 23] The determination of the Pt nanoparticle area is less straightforward as geometrical structure and the size dispersion of the particles is unknown. However, if the electrochemically active area (ECA) of the Pt particles is known, a quantity analogous to the coverage, ΓPt, can be calculated using the real surface area of the gold electrode.

0Au

PtPt A

A=Γ (Eq. 9.1)

An estimate of the ECA of the platinum particles can be determined from the Pt oxide reduction charge obtained from the BCVs of the Pt modified Au electrodes (recorded under the same conditions as suggested by Angerstein-Kozlowska et al.) and relating this value to the corresponding value found for polycrystalline Pt of 440

164

Electrooxidation of small organic molecules on Pt nanoparticles supported on Au

Figure 9.3. Atomic force microscopy images of the Pt-Au surfaces, with a) low b) medium and c) high platinum concentrations

(c)

(a) (b)

µC⋅cm-2.[24] The calculated coverages for the surfaces with “low”, “medium” and “high” platinum concentrations are 0.28; 0.36 and 0.49, respectively.

As the AFM equipment requires relatively flat and smooth surfaces, the characterization of the nanoparticles was performed using glass-supported gold plates

165

Chapter 9

supplied by ArrendeeTM as substrate in the platinum deposition experiment. Prior to the platinum deposition experiment, the gold plate was flame-annealed and carefully cooled down to room temperature in an Argon gas stream. Although this procedure results in large atomically flat Au(111) planes, it otherwise does not affect the formation of the platinum particles, as is apparent from a comparison of the CVs of the thus generated samples to the CVs shown in Fig. 9.2.

The resulting AFM images for the low, medium and high platinum surface concentrations can be found in Fig. 9.3a,b and c, respectively. Fig. 9.3a shows the surface morphology of Pt-Au electrode with low Pt coverage (θPt=0.28) and demonstrates the presence of particles with an average diameter of 10-20 nm. It is possible that smaller particles are also present on the surface. However, these particles cannot be detected due to the relatively low sensitivity of the AFM setup. At higher platinum surface concentrations (see Fig. 9.3b and c) the particle size distribution increases to 10-120 nm for medium Pt loadings and to multilayer-like structures for the high platinum loadings.

Generally, the AFM images clearly show that platinum does not grow epitaxially but rather forms islands on the gold substrate in a Volmer-Weber type growth. For bi-metallic systems (in thermodynamic equilibrium), a Volmer-Weber growth mode is expected when the sum of the relative free surface energies (in this case ∆γ = γPt + γAu-Pt – γAu) of the pure metal (γPt), the pure substrate (γAu) and the interface (γAu-Pt) is smaller than zero.[25-27]

9.3.3. Carbon Monoxide Oxidation

The effect of the gold supported platinum nanoparticles on the electrooxidation of CO was investigated by CO adlayer stripping and bulk CO oxidation experiments.

9.3.3.1. Surface Stability

Fig. 9.4 demonstrates the effect of changes in the surface morphology induced by repetitive cycling to high potentials on the continuous oxidation of CO. Cycling up to 1.2 V vs. RHE results in a decrease of maximum current density for the oxidation of CO, which, in the absence of surface contamination, indicates a decrease in available platinum surface sites. This decrease in available adsorption sites may be due to agglomeration of the platinum particles or embedding/alloying of platinum with the gold substrate.[28] In order to reduce the influence of morphology changes on our experiments, only a single scan up to 1.2 V is recorded for each measurement.

Fig. 9.5 illustrates that the platinum deposition rate also influences the morphology of the formed particles and, therefore, affects the electrooxidation reaction. The two surfaces investigated in Fig. 9.5 were generated by keeping the total deposition charge constant (5000 µC⋅cm-2 per geometrical surface area) but changing the ratio between the applied current and the deposition time. The surface resulting in the first

166

Electrooxidation of small organic molecules on Pt nanoparticles supported on Au

0.6 0.8 1.0 1.20

400

800

1200

j / µ

A cm

-2

E / V vs. RHE

50 µA cm-2 - 100 s 500 µA cm-2 - 10 s

Figure 9.5. Influence of the platinum deposition rate on the CO electrooxidation characteristics of a Pt modified Au electrode with a high Pt coverage recorded in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

curve (solid thin line) was generated by a current of 50 µA⋅cm-2 applied for 100 seconds, while the second surface producing the dashed curve (dashed line) was generated by 500 µA⋅cm-2 for 10 seconds. The figure clearly shows that increasing the deposition rate results in a decrease in the CO oxidation potential. According to the groups of Markoviç and Stimming, this suggests a stronger particle size effect for particles generated by slow Pt deposition, i.e. the generated particles are assumed to be smaller in size.[9-11]

Interestingly, Arenz et al.[11] proposed that the oxidative stripping of CO is mainly controlled by the number of defect sites, which may serve as an active center for OH adsorption, rather than the mobility of CO. Using transmission electron microscopy (TEM), they found that decreasing the particle size leads to a concomitant decrease in the number of defects. Moreover, as it is well known that CO adsorbs and oxidizes preferentially at step sites,[29-31] the negative potential shift recorded on surface

0.6 0.8 1.0 1.20

50

100

150

j / µ

A cm

-2

E / V vs. RHE

Scan 1 Scan 3 Scan 5

Figure 9.4. Influence of repetitive cycling on the CO electrooxidation characteristics of a Pt-Au electrode with a “medium” Pt coverage recorded in a CO saturated solution of 0.5 M H2SO4 at a scan rate of 20 mV⋅s–1.

167

Chapter 9

generated by fast platinum deposition may also be ascribed to an increase in the number of surface defects.

9.3.3.2. Cyclic Voltammetry: Saturated CO Adlayer Stripping

rodes at a scan rate of 20m

electrode are ca. 10-

by Friedric

0.6 0.8 1.0 1.2

0

20

40

60

80

j / µ

A cm

-2

E / v vs. RHE

Low Medium High

Figure 9.6. CO-adlayer stripping voltammograms on the Pt modified Au electrodes in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

Fig 9.6 displays CO stripping CVs recorded on the Pt-Au electV⋅s-1. At low platinum coverages a peak is observed at ca. 1.07 V vs. RHE. This

feature remains as the platinum surface concentration is increased. On the Pt-Au electrode with 10-120 nm size particles another feature appears at 0.9 V vs. RHE appears, which becomes the dominant feature at high platinum coverages.

As our AFM analysis indicates that the particles on the low Pt-Au20 nm (or smaller) in diameter, we ascribe the peak located at 1.07 V to oxidative

removal of CO adsorbed on these small particles. For the medium platinum surface concentrations, in addition to the 10-20 nm size particles, Pt particles with a diameter of ca. 50-120 nm were found, which we believe are responsible for the appearance of the second feature at 0.9 V vs. RHE. At even higher platinum coverages the surface displays oxidation characteristics typical for polycrystalline platinum surfaces,[32] which is in agreement with the presence of multiple platinum adlayers (see Fig. 9.3c). Interestingly, Friedrich et al. also obtained similar CVs for agglomerates of well-defined colloidal nano-particles on gold. For particles with an average diameter of 3 nm, they observed a peak at 1.0 V vs. RHE (at coverages of θPt = 0.05 and θPt = 0.24). At θPt = 0.39 a second peak at 0.87 V vs. RHE was observed, which, as the CV displays polycrystalline platinum characteristics, was attributed to CO oxidizing on the nano-particle agglomerates. In addition, on small Pt nanoparticles with a mean particle size of 1.9 nm supported on glassy carbon, Maillard et al. also report a CO stripping peak at 0.99 V.

It is noteworthy that, although our particles are much larger than those used h et al. and Maillard et al., the distinct “particle size effect” is still present. This

suggests that perhaps not the size of the particle matters, but rather the shape or

168

Electrooxidation of small organic molecules on Pt nanoparticles supported on Au

morphology (as was already implied by Friedrich et al. and Arenz et al.[9, 11]). Furthermore, the fact that most of these studies were performed on well-defined particles, while our particles are irregular in shape (and size), yet both systems display the characteristic “particle size effect”, also suggests that the size of the particles is not as important as is generally presumed.

9.3.3.3. Cyclic Voltammetry: Bulk CO Oxidation

d in a CO saturated solution (bubbli

On polycrystalline gold CO oxidizes between ca. 0.8 V and 1.3 V vs. RHE, after which

.3.3.4. Chronoamperometry

Potential-step experiments provided a more detailed picture of the kinetics of the CO oxidation reaction on the gold supported platinum nanoparticles. The final potentials

0.0 0.4 0.8 1.2 1.6-150

-100

-50

0

50

100

150

200

250

j / µ

A cm

-2

E / V vs. RHE

Poly-Au Low Medium High

Figure 9.7. Bulk CO oxidation voltammograms on bare gold and Pt modified Au electrodes in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

Bulk CO oxidation experiments were performeng CO for 15 minutes through the blank electrolyte). The obtained CO oxidation

CVs on bare polycrystalline Au and the Pt-modified Au electrodes are presented in Fig. 9.7.

the formation of gold oxides inhibits the reaction. Due to continuous readsorption of CO from the electrolyte, the oxidation features recorded on the Pt-modified electrodes are shifted towards higher potentials compared to CO-adlayer stripping. This positive shift is well-known for Pt electrodes and can be as high as 300 mV.[30, 32] In general, increasing the amount of Pt on the surface results in an overall increase of the current density, signifying a higher number of reactive sites. For the Pt-Au surface with 20-50 nm small particles an oxidation peak is observed at 1.1 V, which shifts to lower potentials as the particle size (and number of particles) is increased. At high Pt coverages, the main CO oxidation peak is located at 0.96 V. However, even at these high coverages the feature ascribed to oxidation on small particles is still visible at 1.1 V.

9

169

Chapter 9

170

Chapter 9

170

under investigation were 0.7, 0.85 and 1 V vs. RHE. The resulting current-time curves for Au electrodes with low, medium and high Pt surface concentrations are shown in Fig. 9.8a-c, respectively. All the transients recorded on the surface with the low platinum coverage display a current decay (see Fig. 9.8a), while the surfaces with medium and high Pt coverages show a peak in the transient recorded at 0.85 V (see Fig 9.8b and c). For the highest amount of platinum, a current maximum was also found at 0.7 V vs.

5 10 15 20 250

25

50

75

j / µ

A cm

-2

t / s

0.7 V 0.85 V 1.0 V

0 25 50 75 1000

5

10

15

20

j / µ

A cm

-2

t / s

0.7 V 0.85 V 1.09 V

0 25 50 75 1000

20

40

60

j / µ

A cm

-2

t / s

0.7 V 0.85 V 1.0 V

0 50 100 1500

2

4

6

j / µ

A c

m-2

t / s

0.7 V

Figure 9.8. Current-time transients for saturated CO adlayers on the three Pt-Au electrodes with (a) low, (b) medium and (c) high platinum coverage. Step potentials are listed in the figure. The inset in Fig 9.8c shows a zoom of the transient recorded at 0.7 V vs. RHE.

(a)

(b)

(c)

Electrooxidation of small organic molecules on Pt nanoparticles supported on Au

RHE. A

on these results they concluded that the oxidation of CO on small particles follows an Eley-Rideal type mechanism, while a Langmuir-Hinshelwood type mechanism applies for the larger agglomerates. However, in contrast to these observations, Maillard et al. reported that chronoamperometric transients recorded on 1.9-3.1 nm size particles do in fact exhibit current maxima in the potential range under investigating, indicating that a Langmuir-Hinshelwood type mechanism seems more appropriate.[10]

From this perspective, it is puzzling that our ill-defined “small particles” (of 10-20 nm), which are even larger than the agglomerates used by Friedrich et al., display a current decrease rather than a current maximum. Perhaps the high size and shape dispersion of our particles leads to an overlap of the transient behavior of multiple morphologies and, thus, in an overall decaying transient. More research is needed to clarify these findings.

9.3.3.5. Surface Enhanced Raman Spectroscopy

Fig. 9.9 shows SERS spectra of the Pt-Au (medium coverage) electrode at

1700 1800 1900 2000 2100 2200 2300

t potentials higher than 1 V, all three electrodes display a current decay, which seems identical for each surface. Considering the relatively high final potential applied, the formation of oxides on the nanoparticles is the most likely cause of this decay.

Potential-step experiments performed by Friedrich et al. on nanoparticles of ca. 3 nm and agglomerates of these particles (ca. 16 nm in diameter) also showed a current decay on the small particles and a current maximum on the larger agglomerates.[9] Based

1.305 V

1.105 V

1.005 V

0.905 V

0.805 V

0.605 V

0.405 V

0.205 V

Inte

nsity

/a.u

.

wavelength /cm-1

300 400 500 600 700 800

1.305 V

1.105 V

1.005 V

0.905 V

0.805 V

0.605 V

0.405 V

0.205 V

Inte

nsity

/a.u

.

wavelength / cm-1

Figure 9.9. Low (a) and high frequency range (b) SER spectra obtained for a Au-Pt (Medium) electrode in 0.5 M H2SO4 at the indicated potentials.

(a) (b)

171

Chapter 9

different potentials in sulfuric acid saturated with CO. Two spectral regions are found to exhibit some important changes, as the potential is made more positive. One, comprising the low frequencies (Fig. 9.9a), shows three bands. The band centered at 380 cm-1 can be assigned to the stretching mode of CO adsorbed on a platinum bridge site. This mode blue shifts with a Stark tuning rate of -12cm-1/V. The band at 460 cm-1 corresponds to the stretch cm-1/V. Both modes disappear at potential higher than 1 V for medium platinum coverage. A band centered at ca. 580 cm-1 appears as the potential exceeds 1.3 V, coinciding with the Au surface reduction peak in the voltammogram. Weaver et al. attributed this band, which has a small but significant positive Stark tuning slope (ca. 10 cm-1/V), to the combined vibrations of Au-O and Au-OH surface species.[36, 37] On the basis of other techniques, other authors have also concluded that the layer formed on gold above 1.2 V is composed of both surface oxide and surface hydroxide.[38, 39] Tian’s group observed a band at 570 cm-1 on a pure platinum substrate[34] and Weaver’s group a similar band on a platinum film on gold.[35] Both attributed the band to the Pt-O bond. Unfortunately, differentiation of these three surface species is not straightforward.

A second spectral region is defined by the 1700-2300 cm-1 window (Fig. 9.9b), and contains three peaks. The band at 1890 cm-1, with a Stark tuning slope of 40 cm-1⋅V-

1, can be attributed to the CO stretching mode on platinum bridge sites.[34, 35, 40]. This band is not detectable for the gold electrode with the high platinum coverage. A second band at 2050 cm-1 can be assigned to the atop bound CO on platinum [33-37, 40] and it shows a Stark tuning of 50 cm-1⋅V-1. A third band at 2120 cm-1 is ascribable to the stretching frequency of CO adsorbed atop on gold. It red shifts with an approximate Stark tuning slope of 40 cm-1⋅V-1.[40, 41]

The bands described above are observed for all the generated platinum covered surfaces and all display similar characteristics. The CO-bands are presents at potentials negative with respect to the oxidation peak in the voltammograms and the intensities are

Figure 9.10. Inte bonded Pt-CO recorded on low, gh (b) frequency range.

0.0 0.2 0.4 0.6 0.8 1.0 1.2

0.2

0.4

0.6

0.8

1.0

mode of CO adsorbed atop on platinum,[33-35] and has a Stark tuning rate of -16

nsity ratio between the SERS intensities of bridge and linearly medium and high platinum covered surfaces in the low (a) and hi

E / V vs. RHE E / V vs. RHE

Low Medium High

ratio

(v38

0/v45

0)

vPt-COBridge / v

Pt-COa-top

0.0 0.2 0.4 0.6 0.8 1.0 1.2

0.0

0.1

0.2

0.3

0.4

0.5

vPt-COBridge

/ vPt-COa-top

Low Medium High

ratio

(v18

90/v

2060

)

(a) (b)

172

Electrooxidation of small organic molecules on Pt nanoparticles supported on Au

almost constant if they involve Pt-CO bonds. In the case of the Au-CO mode the band intensities are decreasing at more positive potentials, analogous to the oxidation of CO on pure gold electrodes.[41] The band involving OH species on the surface (590 cm-1 band) is only present at more positive potentials.

Interestingly, comparison of spectra recorded on low, medium and high platinum covered surfaces reveals a marked change in the ratio of the intensities of the linear and bridge bonded Pt-CO peaks. Fig. 9.10 a and b show the bridge vs. linear bonded Pt-CO ratios recorded on low, medium and high platinum covered Au surfaces for the low and high fre

le morphology is more important than the particle size. Precise interpretation of this data is not trivial, as many unknown factors, e.g. the CO coverage, can als ile we can ascribe a peak in the SERS spectra to bridge or linearly bonded CO, the exact nature to the binding site is not known.

9.3.4. Methanol and Formic Acid Oxidation

The effect of the particle size and morphology was also studied for the oxidation of methanol and formic acid. The reactions were investigated using cyclic voltammetry and chronoamperometry.

9.3.4.1. Cyclic Voltammetry

The CVs for the methanol and formic acid oxidation on the Pt-Au electrodes can be found in Fig. 9.11 and 9.12, respectively. The inset in Fig. 9.11 demonstrates that bare gold is virtually inactive towards the oxidation of methanol. However, deposition of even a small amount of Pt on the gold substrate immediately leads to the appearance of a peak at 0.84 V vs. RHE. As the amount of Pt is increased the total charge of the MeOH oxidation peak increases, while the onset of the reaction decreases from 0.65 V to 0.6 V and 0.55 V when going from low to medium and high Pt concentrations, respectively.

While gold is inactive towards the oxidation of methanol, it can oxidize formic acid at sufficiently high potentials, especially in the absence of strongly adsorbing anions (see Fig. 9.12).[42-45] The onset of the HCOOH oxidation on a polished gold electrode lies at ca. 0.8 V. Deposition of small amounts of platinum results in the appearance of an oxpo increased. At high platinum coverages a

quency range, respectively. The shown ratios are obtained by dividing the area of the peak ascribed to linearly bonded CO by the area of the peak ascribed to bridge bonded CO. Apparently, CO seems to favor a bridge-bonded position on smaller particles, while the Pt-atop position is preferred for larger particles and polycrystalline platinum surfaces. Although these results are by no means conclusive, they tentatively suggest that the partic

o influence the bridge-atop ratio. Moreover, wh

Despite these reservations we believe that the presented results provide further evidence for the predominant influence of the particle morphology on their electrochemical characteristics.

idation peak at 0.58 V vs. RHE in the positive going scan, which shifts to more sitive potentials as the platinum coverage is

173

Chapter 9

second feature at 0.89 V becomes visible. A similar feature was also observed for polycrystalline platinum electrodes,[46] which leads us to conclude that this peak is most likely due to oxidation of formic acid on large particles and platinum overlayers. The peak at 0.58 V can be ascribed to the oxidation of HCOOH to CO2 on Pt sites, which are unblocked by preadsorbed poisoni

hanol oxidation on the three Pt modified Au surfaces in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

acid oxidation on the three Pt modified Ausurfaces in 0.5 M H2SO4 at a scan rate o20 mV⋅s-1.

0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75-300

-250

-200

-150

-100

-50

0

50

100

0.0 0.5 1.0 1.5 2.0

-400

-300

-200

-100

0

100

j / µ

A cm

-2

E / V vs. RHE

Blank Poly-Au MeOH Poly-Au

E / V vs. RHE

j / µ

A cm

-2

Low Medium High

0.0 0.5 1.0 1.5 2.0

0

250

500

750

1000

1250

1500

1750

j / µ

A cm

-2

E / V vs. RHE

Au - blank Au - HCOOH Low Medium High

Figure 9.11. Cyclic voltammograms of the Figure 9.12. Cyclic voltammograms of thmet

e formic

f

ng species such as COads.[46, 47] Compared to polycry

was estimated by integration of the Pt oxide reduction peak. The resulting graphs are presented in Fig. 9.13 and 9.14. Although

0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1

0

10

20

30

40

50

60

70

stalline gold, the onset of the formic acid oxidation on Pt modified Au electrodes is shifted to considerably more negative potentials (at ca. 0.2 V vs. RHE).[46] In the negative going scan, the reaction proceeds on a surface that has been oxidatively stripped of pre-adsorbed blocking species. Therefore, higher current densities are obtained.

In order to compare the methanol and formic acid oxidation activities of the platinum modified surfaces, the CVs were normalized for the amount of platinum deposited. The amount of surface-active platinum

j / µ

A cm

-2 P

t

E / V vs. RHE

Low Medium High

amount of Pt in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

Figure 9.13. Cyclic voltammograms of the methanol oxidation activity for the different Pt concentrations, normalized for the

174

Electrooxidation of small organic molecules on Pt nanoparticles supported on Au

the platinum oxide reduction ch

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0

200

400

600

800

1000

0.2 0.4 0.6 0.8 1.0 1.2

0

100

200

300

400

500

j / µ

A cm

-2 P

t

E / V vs. RHE

Low Medium High

E / V vs. RHE

j / µ

A cm

-2 P

t

Low Medium High

(a) (b)

Figure 9.14. Cyclic voltammograms of the formic acid oxidation activity for the different Pt concentrations, normalized for the amount of Pt, with a) the positive going scan and b) the negative scan in 0.5 M H2SO4 at a scan rate of 20 mV⋅s-1.

arge is only a rough estimate of the total platinum coverage, Fig. 9.13 clearly shows that larger Pt particles are more active towards the methan tive surface area). Apparently, the oxidation of methanol suffers from the same particle size effect as carbon monoxide. Fig. 9.14a a

ith the lowest Pt coverage. As the amount of platinum is increased, the current maximum decreases, which implies that smaller particles are more active t than larger polycrystalline-like structures. It is also likely t

particlenm in

ol oxidation than smaller particles (of equal ac

nd b show similarly corrected plots for the oxidation of HCOOH in the positive and negative going scan, respectively. In the positive going scan the current density is highest for the surface w

owards formic acid oxidationhat surface poisoning plays an important role on the large particles. In the return

scan, however, the oxidation activity increases with the amount of platinum deposited. As the potential decreases below 0.3 V the order of the oxidation activity is again inverted, with the low Pt-Au surface now being more active than the higher coverages. These results suggest that larger particles are more active towards the oxidation of formic acid, but also poison faster.

Park et al.[8] explained the decreasing methanol reaction rate for a decreasing size by a loss of adjacent available Pt terrace sites on nanoparticles smaller than 4 diameter. They suggested that the dissociation of methanol to reactive

intermediates and chemisorbed CO requires an ensemble of catalytically active terrace sites. Thus, the decreasing oxidation activity for decreasing particle size can be attributed to a drop in the number of Pt “ensemble sites” available to the adsorption and decomposition of methanol. The fact that we can still observe an effect of the particle morphology on the reaction characteristics even for particles with a diameter up to ca. 120 nm, again suggests that the shape and morphology, rather than actual particle size play a predominant role in the reaction kinetics.

175

Chapter 9

Interestingly, from a study of the MeOH oxidation reaction on stepped platinum electrodes of [n(111)×(111)] orientation in Chapter 7, we concluded that methanol preferably decomposes on step sites rather than terrace sites.[48] Although these findings seem to contradict to those of Park et al.,[8] the OLEMS results presented in Chapter 8 show that the Pt(110) surface is less active towards methanol oxidation than the stepped surfaces Pt(554) and Pt(553) (of [(n-1)(111)×(110)] orientation).[49] Based on these results, we suggested that sites with both a step and terrace site may be particularly favorable “ensemble sites” for dissociative adsorption of methanol. This theory may be used to explain why particles larger than 3-4 nm in diameter can still display a pronounced “particle size effect”. If the particles contain a large number of defects and relatively few contiguous terrace sites, the methanol oxidation rate may be lower due to form tidefi a. to decrease the overall surface mobility of CO, thus enhancing the particle shape effect.

the potential (with maxima at 0.8 and 0.6 V vs. RHE for MeOH and HCOOH, respectively) until the onset of surface oxidation, after which the current decreases again. As the obtained graphs do not provide any new information in addition to that presented in Chapter 7, the corresponding graphs are not presented in this chapter.

9.4. Conclusion

The electrooxidation characteristics of carbon monoxide, methanol and formic acid on Pt particles of varying size were investigated by voltammetry and chronoamperometry. The particles were generated by galvanostatic deposition of platinum on polished polycrystalline gold electrodes. Three surfaces with low, medium and high Pt coverages were generated by galvanostatic deposition of platinum on polished polycry try and Atomic Force Microscop coverages consisted of

ation of soluble intermediates. Such a situaned particles with an average diameter of cour larger irregularly shaped structures. M

on is apparently easily satisfied on well-2-5 nm,[6, 8-11, 50] but may also apply oreover, the presence of defects may

Therefore, in agreement with other authors, we conclude that the particle shape indeed seems more important than the actual particle size.[9, 11] As formic acid does not require an ensemble of platinum sites, the reaction rate is found to increases for decreasing particle size (and, thus, increasing number of defects).

9.3.4.2. Chronoamperometry

The kinetics of the methanol and formic acid oxidation on nanoparticles generated for medium coverages, were also investigated using potential-step experiments. The steady-state current recorded at 300 seconds (obtained analogous to Fig. 7.6 in Chapter 7, ref. [48]) was found to follow the shape and position of the oxidation peak recorded in the cyclic voltammetry. The current densities recorded in the methanol as well as the formic acid containing electrolyte were found to increase with

stalline gold electrodes and characterized by voltammey. The electrodes with low and medium platinum

176

Electrooxidation of small organic molecules on Pt nanoparticles supported on Au

small ill-defined particles with diameters ranging from 10-20 and 10-120 nm, respectively. At higher platinum coverages Pt overlayers are formed.

Based on CO stripping voltam try and bulk CO oxidation experiments we can conclude the following:

1. CO stripping and bulk CO oxidation experiments on irregularly shaped particles with an average diameter of 10-20 nm display a pronounced particle size effect, while Friedrich et al. observe polycrystalline platinum like oxidation characteristics for 16nm large agglomerates of well-ordered particles.[51] Therefore, not the size, but the shape of the particles seems crucial to the oxidation characteristics.

2. Current-time transients recorded on these small 10-20 nm size particles showed a current decay, which can be explained by assuming an Eley-Rideal type reaction mechanism, or by assuming that overlapping signals of a Langmuir-Hinshelwood

n ood type mechanism is found.

ained by a reduction in available “ensemble sites” and a

Refere

me

type reaction on particles of varying size produce an overall current decay. Olarger particles a Langmuir-Hinshelw

3. Surface Enhanced Raman Spectroscopy experiments indicated a difference in the ratio between CO bonded linearly and in bridge configuration on the small, medium and large Pt nano-particles, indeed suggesting a dominant role of the surface morphology, rather than particle size.

4. The oxidation activity of methanol was found to decrease with the particle size, while the formic acid oxidation rate increases. Again a “particle size effect” is observed for nanoparticles of ca. 120 nm in diameter, which is much larger than the particles for which a particle size effect has been reported in the literature. This emphasizes the previously made suggestion that “particle size” is not as important as particle shape. The particle shape effect for the methanol oxidation reaction can be explreduction in the mobility of CO formed by decomposition of methanol. As formic acid does not require Pt ensemble sites decreasing the particle size, and thus, the relative number of defects, increases the reaction rate. In general, based on our results, we conclude that the “particle size effect”,

reported in the literature should be critically reviewed. A “particle shape effect” or “morphology” in many cases may be a more appropriate term. Acknowledgement: This research was supported by the Netherlands Foundation for Scientific Research (NWO).

nces

[1] M. S. Wilson, F. H. Garzon, K. E. Sickafus, and S. Gottesfeld, J. Electrochem. Soc. 140 (1993) 2872.

[2] R. Parsons and T. VanderNoot, J. Electroanal. Chem. 257 (1988) 9. [3] M. L. Sattler and P. N. Ross, Ultramicroscopy 20 (1986) 21. [4] L. J. Bregoli, Electrochim. Acta. 23 (1978) 489.

0.90 V

177

Chapter 9

178

. Electroanal. Chem. 233 (1987) 211. [21] . Möller and P. C. Pistorius, J. Electroanal. Chem. 570 (2004) 243. [22] . Angerstein-Kozlowska, B. E. Conway, A. Hamelin, and L. Stoicoviciu, Electrochim. Acta. 31

[23] nway, A. Hamelin, and L. Stoicoviciu, J. Electroanal. Chem.

[5] Y. Takasu, N. Ohashi, X. G. Zhang, Y. Murakami, H. Minagawa, S. Sato, and K. Yahikozawa, Electrochim. Acta. 41 (1996) 2595.

[6] T. Frelink, W. Visscher, and J. A. R. van Veen, J. Electroanal. Chem. 382 (1995) 65. [7] O. V. Cherstiouk, P. A. Simonov, and E. R. Savinova, Electrochim. Acta. 48 (2003) 3851. [8] S. Park, Y. Xie, and M. J. Weaver, Langmuir 18 (2002) 5792. [9] K. A. Friedrich, F. Henglein, U. Stimming, and W. Unkauf, Electrochim. Acta. 45 (2000) 3283. [10] F. Maillard, M. Eikerling, O. V. Cherstiouk, S. Schreier, E. Savinova, and U. Stimming, Faraday

Discuss. 125 (2003) 357. [11] M. Arenz, K. J. J. Mayrhofer, V. Stamenkovic, B. B. Blizanac, T. Tomoyuki, P. N. Ross, and N.

M. Markovic, J. Am. Chem. Soc. (2005) submitted. [12] N. M. Markovic and P. N. Ross, Surf. Sci. Rep. 45 (2002) 117. [13] T. Iwasita, Electrochim. Acta. 47 (2002) 3663. [14] A. Hamnett, Comp. Chem. Kin. (1999) 635. [15] T. D. Jarvi, S. Sriramulu, and E. M. Stuve, J. Phys. Chem. B 101 (1997) 3649. [16] M. Shibata and S. Motoo, J. Electroanal. Chem. 209 (1986) 151. [17] H. Wang, C. Wingender, H. Baltruschat, M. Lopez, and M. T. Reetz, J. Electroanal. Chem. 509

(2001) 163. [18] H. Wang, T. Loffler, and H. Baltruschat, J. Appl. Electrochem. 31 (2001) 759. [19] K.-I. Ota, Y. Nakagawa, and M. Takahashi, J. Electroanal. Chem. 179 (1984) 179. [20] P. Gao, D. Gosztola, L. W. H. Leung, and M. J. Weaver, J

HH(1986) 1051. H. Angerstein-Kozlowska, B. E. CoInter. Electrochem. 228 (1987) 429.

[24] D. V. Heyd and D. A. Harrington, J. Electroanal. Chem. 335 (1992) 19. [25] E. Bauer, Z. Krist. 110 (1958) 395. [26] E. Bauer and J. H. Van der Merwe, Phys. Rev. B 33 (1986) 3657. [27] A. Zangwill, Physics at Surfaces, Cambridge Univ. Press, Cambridge, 1988. [28] M. O. Pedersen, S. Helveg, A. Ruban, I. Stensgaard, E. Laegsgaard, J. K. Norskov, and F.

Besenbacher, Surf. Sci. 426 (1999) 395. [29] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, Electrochem.Commun. 2

(2000) 487. [30] N. P. Lebedeva, M. T. M. Koper, E. Herrero, J. M. Feliu, and R. A. van Santen, J. Electroanal.

Chem. 487 (2000) 37. [31] N. P. Lebedeva, M. T. M. Koper, J. M. Feliu, and R. A. van Santen, J. Phys. Chem. 106 (2002)

12938. [32] H. A. Gasteiger, N. M. Markovic, and P. N. Ross, Jr., J. Phys. Chem. 99 (1995) 8290. [33] L. Gland and E. B. Kollin, Surf. Sci. 151 (1985) 260. [34] . Q. Tian, B. Ren, and B. W. Mao, J. Phys. Chem. B 101 (1997) 1338. [35] L. W. H. Leung and M. J. Weaver, J. Am. Chem. Soc. 109 (1987) 5113.

o and M. J. Weaver, J. Electroanal. Chem. Interfacial Electrochem. 209 (1986) 377. [37] Y. Zhang, X. Gao, and M. J. Weaver, J. Phys. Chem. 97 (1993) 8656. [38]

c. submitted (2005) lectroanal. Chem. 554-

englein, U. Stimming, and W. Unkauf, Colloids and Surfaces, A: Engineering Aspects 134 (1998) 193.

J.Z

[36] J. Desilvestr

G. Tremiliosi-Filho, L. H. Dall'Antonia, and G. Jerkiewicz, J. Electroanal. Chem. 422 (1997) 149. [39] S. J. Xia and V. I. Birss, J. Electroanal. Chem. 500 (2001) 562. [40] S. Zou, C. T. Williams, E. K. Y. Chen, and M. J. Weaver, J. Am. Chem. Soc. 120 (1998) 3811. [41] G. L. Beltramo, T. E. Shubina, and M. T. M. Koper, Chem Phys Chem In press (2005) [42] G. Crepy, C. Lamy, and S. Maximovitch, J. Electroanal. Chem. 54 (1974) 161. [43] A. Hamelin, Y. Ho, S. C. Chang, X. Gao, and M. J. Weaver, Langmuir 8 (1992) 975. [44] Y. Zhang and M. J. Weaver, Langmuir 9 (1993) 1397. [45] S.-l. Chen, B.-l. Wu, and C.-s. Cha, J. Electroanal. Chem. 431 (1997) 243. [46] G.-Q. Lu, A. Crown, and A. Wieckowski, J. Phys. Chem. B 103 (1999) 9700. [47] A. Capon and R. Parsons, J. Electroanal. Chem. 44 (1973) 239. [48] T. H. M. Housmans and M. T. M. Koper, J. Phys. Chem. B 107 (2003) 8557. [49] T. H. M. Housmans, A. H. Wonders, and M. T. M. Koper, J. Am. Chem. So[50] O. V. Cherstiouk, P. A. Simonov, V. I. Zaikovskii, and E. R. Savinova, J. E

555 (2003) 241. [51] K. A. Friedrich, F. H

Physicochemical and

Summary

This dissertation contains fundamental research on the electrooxidation of small organic molecules on noble metal single crystal, nanoscale and bi-metallic surfaces. The goal of the research is to provide deeper insight into the kinetics and structure-sensitivity of these oxidation reactions on different catalysts and, thus, further the development of low temperature fuel cells like the Direct Methanol Fuel cells (DMFC) and hydrogen/oxygen fuel cells. The mechanism and kinetics are investigated using electrochemical techniques like cyclic voltammetry and chronoamperometry, Surface-Enhanced Raman Spectroscopy and mass spectrometry techniques like On-line Electrochemical Mass Spectrometry (OLEMS). The surfaces under investigation were characterized using voltammetry and techniques such as Scanning Tunneling Microscopy (STM) and Atomic Force Microscopy (AFM).

The electrooxidation of carbon monoxide, a common fuel cell catalyst poison, on stepped rhodium single crystal surfaces in sulfuric acid containing electrolyte displays a pronounced structure sensitivity. Unlike platinum surfaces, the low mobility of CO on Rh[n(111)×(111)] type electrodes together with the fact that OH formation is faster on steps than on terrace sites, results in the appearance of two distinguishable features in both the voltammetry (Chapter 2) as well as the chronoamperometry (Chapter 3). The peak visible at low potentials in the voltammetry, and at short times in the chronoamperometry is ascribed to CO reacting at or near the step, while the second peak at higher potentials and longer times is attributed to CO stripping from terrace sites. Similar to platinum surfaces, the main oxidation wave in the current-time transients is preceded by an apparently zeroth-order process, which is ascribed to a Langmuir-Hinshelwood type reaction between COads and OHads with no effective freeing of sites for OH adsorption due to relaxation of the CO adlayer (Chapter 3). The low surface mobility of CO is explained in terms of strong adsorption of (bi)sulfate anions and the ease of oxidizability of the rhodium surfaces and the nucleation and growth model is suggested as best analytical model for the reaction kinetics.

Reducing the anion adsorption strength by changing from sulfuric acid media to perchloric acid not only increases the total rate but also changes the dynamics of the CO electrooxidation reaction (Chapter 4). As the surface diffusion rate of CO is higher in the presence of perchlorate than (bi)sulfate, the reaction kinetics can best be described by the mean-field approximation. Results of Dynamic Monte Carlo simulations indeed show that in case of a low CO surface mobility the reaction nucleates at the steps and grows over the terraces, resulting in two clearly distinguishable peaks in the voltammetry and chronoamperometry. When the diffusion rate of CO is increased, the model produces results similar to the voltammetric profiles obtained for analogous single crystal platinum electrodes (Chapter 5).

179

180

Adding a monolayer of platinum to a Rh(111) surface considerably reduces the overpotential needed to oxidatively strip a saturated CO adlayer, thus making the surface more “CO-tolerant”.(Chapter 6) This effect was explained in terms of the bi-functional theory and electronic effects. The bi-functional mechanism is based on the idea that sites on the more oxophilic metal act as adsorption centers for oxygen-containing species. The electronic effect is caused by a contraction of the Pt-Pt atom distance in the Pt overlayer due to the smaller lattice constant of the rhodium substrate, which leads to a considerable decrease in the CO binding strength. Even though these systems display a pronounced effect towards the oxidation of CO, no discernable effects on the oxidation of methanol and formic acid are visible. Further addition of platinum leads to the formation of islands of multi-atomic height, which were found to display a pronounced “particle size effect”. As the amount of platinum is increased and the islands grow in height and diameter, the oxidation potential of CO increases to potentials higher than those of polycrystalline platinum. Further increasing the particle size reduced the CO oxidation overpotential again.

A complementary study of the effect of particle size on the electrooxidation of CO, methanol and formic acid indicated that the shape of the particles, rather than their actual size, is responsible for this so-called “particle size effect” (Chapter 9). The total reactivity is determined by the crystalline morphology of the particles. Measurements on stepped Pt[n(111)×(111)] type single crystal electrodes show that the (111) terraces are less active towards the oxidation of methanol than the steps (Chapter 7) and that the decomposition of methanol to adsorbed CO occurs preferably at the (110) steps. As the methanol oxidation reaction on well-ordered Pt(110) results in relatively high amounts of soluble intermediates like formic acid and formaldehyde, methanol probably requires an ensemble of both step and terrace sites to decompose to CO (Chapter 8). Therefore, surfaces with higher step densities favor the indirect oxidation pathway of methanol involving the formation and subsequent oxidation of CO to CO2. In case of only single adsorption sites, methanol preferably decomposes to soluble intermediates.

Samenvatting

Dit proefschrift betreft fundamenteel onderzoek naar de electro-oxidatieve eigenschappen van kleine organische moleculen aan edel metaal eenkristal oppervlakken, nanoschaal deeltjes en bi-metallische oppervlakken. Het doel van het onderzoek is het verkrijgen van inzicht in de kinetiek en structuurafhankelijkheid van deze oxidatie reacties op verschillende katalysatoren om zo de ontwikkeling van lage temperatuur brandstofcellen zoals de Direct Methanol Brandstof Cel (DMFC) en waterstof/zuurstof brandstofcellen te bevorderen. Het mechanisme en de kinetiek van deze reacties worden onderzocht met elektrochemische technieken zoals voltammetrie en chronoamperometrie, Surface Enhanced Raman Spectroscopy, en massa spectrometrie technieken zoals On-Line Electrochemical Mass Spectrometry (OLEMS). De te onderzoeken oppervlakken zijn gekarakteriseerd met behulp van voltammetrie en technieken zoals Scanning Tunneling Microscopy (STM) en Atomic Force Microscopy (AFM).

De electrooxidatie van koolmonoxide, een brandstofcelkatalysator vergiftigend species, aan gestapte rhodium eenkristal elektroden in zwavelzuur houdend elektrolyt vertoont een duidelijke structuur gevoeligheid. In tegenstelling tot platina oppervlakken veroorzaakt de lage oppervlakte mobiliteit van CO aan [n(111)×(111)] type rhodium elektroden, en het feit dat OH formatie sneller is aan stappen dan op de terrassen, twee duidelijk onderscheidbare kenmerken in de voltammetrie (Hoofdstuk 2) en de chronoamperometrie. (Hoofdstuk 3) De piek, zichtbaar bij lage potentialen in de voltammetrie en bij korte tijden in de chronoamperometrie, is toegeschreven aan CO die reageert bij of aan de stap, terwijl de tweede piek, bij hogere potentialen en latere tijden, toegeschreven is aan oxidatief strippen van CO aan terras-sites. Net zoals aan platina wordt de hoofdoxidatiegolf in de stroom-tijd curven voorafgegaan door een schijnbaar nulde-orde proces, welke toegeschreven wordt aan een Langmuir-Hinshelwood type reactie tussen COads en OHads in afwezigheid van effectieve toename in het aantal vrije sites voor de adsorptie van OH als gevolg van relaxatie van de CO adlaag.(Hoofdstuk 3) De lage oppervlakte mobiliteit van CO is verklaard in termen van sterke adsorptie van (bi)sulfaat en het gemak waarmee de rhodium oppervlakken kunnen oxideren en een nucleatie-en-groei model wordt gesuggereerd als beste model voor de reactie kinetiek.

Verlaging van de adsorptiekracht van het anion door van zwavelzuur houdende media naar perchloorzuur te wisselen, resulteert niet alleen in een verhoging van de totale snelheid, maar ook in een verandering in de dynamiek van de CO oxidatie reactie.(Hoofdstuk 4) Omdat de diffusiesnelheid van CO aan het oppervlak in aanwezigheid van perchlorate hoger is dan in aanwezigheid van (bi)sulfaat, kan de reactiekinetiek het beste beschreven worden door de gemiddeld-veld benadering. Resultaten van dynamische Monte Carlo simulaties laten inderdaad zien dat in het geval van lage CO mobiliteit de reactienucleatie aan de steppen plaats vindt en over de

181

182

terrassen heen groeit, hetgeen resulteert in twee duidelijk onderscheidbare pieken in de voltammetrie en chronoamperometrie. Wanneer de diffusiesnelheid van CO verhoogd wordt, produceert het model resultaten welke gelijk zijn aan de voltammetrische profielen gemeten aan analoge platina eenkristal elektroden.(Hoofdstuk 5)

Toevoeging van een platina monolaag op een Rh(111) oppervlak verlaagt de overpotentiaal welke nodig is om een verzadigde CO adlaag oxidatief te strippen aanzienlijk, wat erop wijst dat het oppervlak “CO-toleranter” geworden is.(Hoofdstuk 6) Dit effect wordt uitgelegd in termen van de bi-functionele theorie en aan de hand van elektronische effecten. De bi-functionele theorie is gebaseerd op het idee dat sites van een oxofiel metaal zich gedragen als adsorptie centra voor zuurstofhoudende species. Het elektronische effect wordt veroorzaakt door verkleining van de Pt-Pt atoom afstand in de Pt adlaag als gevolg van de kleinere rooster constante van het rhodium substraat, wat een aanzienlijke verlaging van de CO bindingssterkte tot gevolg heeft. Ondanks het feit dat deze systemen een aanzienlijk effect op de CO oxidatie laten zien, kan geen effect op de oxidatie van methanol en mierenzuur waargenomen worden. Toevoeging van meer platina leidt tot vorming van Pt eilanden van multi-atomaire hoogte. Deze eilanden vertonen een opmerkelijk “deeltjesgrootte effect”. Bij toenemende hoeveelheid platina groeien de eilanden in hoogte en diameter, waarbij het oxidatiepotentiaal van CO toe neemt tot potentialen die hoger zijn dan die gemeten aan polykristallijn platina. Boven een bepaalde deeltjesgrootte neemt het CO oxidatiepotentiaal weer af.

Een aanvullende studie naar het effect van deeltjes grootte op de electrooxidatie van CO, methanol en mierenzuur toont aan dat de vorm van de deeltjes, meer dan de daadwerkelijke grootte, verantwoordelijk is voor dit “deeltjesgrootte effect” (Hoofdstuk 9). De totale reactiviteit wordt bepaald door de kristallijne morfologie van de deeltjes. Metingen aan gestapte Pt[n(111)×(111)] type eenkristal elektroden laten zien dat de terrassen met (111) oriëntatie minder actief zijn voor de oxidatie van methanol (Hoofdstuk 7) dan de stappen en dat de decompositie van methanol naar geadsorbeerd CO bij voorkeur plaats vindt aan de (110) stappen. Omdat de oxidatie reactie van methanol aan goed gedefinieerd Pt(110) relatief grote hoeveelheden oplosbare intermediaire producten zoals mierenzuur en formaldehyde produceert, heeft methanol waarschijnlijk een combinatie van terras en stapsites nodig om tot CO te ontbinden.(Hoofdstuk 8) Om deze reden heeft het indirecte oxidatiepad van methanol waarbij CO gevormd wordt dat vervolgens tot CO2 kan worden geoxideerd, de voorkeur voor oppervlakken met toenemende stappen dichtheid. Als alleen enkele adsorptie sites beschikbaar zijn, heeft methanol ontleding de voorkeur voor oplosbare intermediaire species.

List of Publications

1. T.H.M. Housmans and M.T.M. Koper, “Methanol oxidation on stepped Pt[n(111)×(111)] electrodes: a chronoamperometric study”, J. Phys. Chem. B 107 (2003) 8557.

2. T.H.M. Housmans, J.M. Feliu, and M.T.M. Koper, “CO oxidation on stepped Rh[n(111)×(111)] single crystal electrodes: a voltammetric study”, J. Electroanal. Chem. 572 (2004) 79.

3. T.H.M. Housmans and M.T.M. Koper, “CO oxidation on stepped Rh[n(111)×(111)] single crystal electrodes: a chronoamperometric study”, J. Electroanal. Chem. 575 (2005) 39.

4. T.H.M. Housmans and M.T.M. Koper, “CO oxidation on stepped Rh[n(111)×(111)] single crystal electrodes: anion effects on CO surface mobility”, Electrochem. Comm. 7 (2005) 581.

5. T.H.M. Housmans, J.M. Feliu, R.Gomez, and M.T.M. Koper, “CO oxidation on Pt modified Rh(111) electrodes”, ChemPhysChem. 6 (2005) 1522.

6. T.H.M. Housmans and M.T.M. Koper, “CO oxidation on stepped single crystal electrodes – a Dynamic Monte Carlo study”, in preparation (2005)

7. T.H.M. Housmans, A.H. Wonders, and M.T.M. Koper, “Structure sensitivity of the methanol electrooxidation pathways on platinum: an On-Line Electrochemical Mass Spectrometry study”, J. Phys. Chem. B, submitted (2005)

8. F.J.E. Scheijen, S Höppener, G.L. Beltramo, T.H.M. Housmans, and M.T.M. Koper, “The electrooxidation of small organic molecules on platinum nanoparticles supported on gold: particle size versus particle shape effect”, in preparation (2005)

9. A.H. Wonders, T.H.M. Housmans, V. Rosca, and M.T.M. Koper, J. Appl. Electrochem. submitted (2005)

183

Curriculum Vitae

Tom Housmans was born on the 13th of April 1978 in Sittard, which lies in Limburg, the southernmost province of the Netherlands. After the elementary school in Broeksittard, he went to the Atheneum of the secondary school “het Bisschoppelijk College Sittard” also better known as “ ’t Kleesj”, where he finished in 1996. After the athenaeum he started studying Chemical Engineering and Chemistry at the Eindhoven University of Technology. Having finished his bachelors, Tom specialized in topics such as chemical reaction engineering, electrochemistry, corrosion, surface chemistry, quantum chemistry and biochemistry. (Although the subject of his internship, "The construction of polymer-varistor composites" at ABB corp. research in Baden Dättwil Switzerland, deviated a little from that specialization.) In 2001, within the prescribed time-span of 5 years, Tom obtained his Master of Science degree on the subject of "Electromachining with sulfonic acid-alcohol electrolytes" under the supervision of Dr. L.J.J. Janssen. Following his graduation, he joined Marc Koper's electrocatalysis group as a PhD student, where he worked on the subject of "Electrooxidation of small organic molecules on single crystal and bi-metallic electrodes”. The results of this study are presented in this thesis.

Tom Housmans werd geboren op 13 April 1978 in Sittard, wat in Limburg, de

zuidelijkste provincie van Nederland, ligt. Na de basisschool in Broeksittard is hij naar het Atheneum van de middelbare school “het Bisschoppelijk College Sittard”, ook bekent als “ ’t Kleesj”, gegaan, waar hij in 1996 slaagde. Na het atheneum begon hij aan de opleiding Scheikundige Technologie aan de Technische Universiteit Eindhoven. Na het behalen van zijn bachelorgraad, heeft Tom zich gespecialiseerd in onderwerpen zoals chemische reactoren, elektrochemie, corrosie, oppervlakte chemie, kwantum chemie en biochemie. (Alhoewel het onderwerp van zijn stage, “The construction of polymer-varistor composites” bij ABB corp. research in Baden Dättwil Zwitserland, iets afweek van deze specialisatie.) In 2001, binnen de voorgeschreven 5 jaar, heeft Tom zijn graad als Master of Science behaald aan het onderwerp “Electromachining with sulfonic acid-alcohol electrolytes” onder supervisie van Dr. L.J.J. Janssen. Na zijn afstuderen, is hij aangenomen als promovendus in de elektrokatalyse groep van Marc Koper, waar hij gewerkt heeft aan het onderwerp “Electrooxidatie van kleine organische moleculen aan eenkristal en bi-metallische elektroden”. De resultaten van dit onderzoek staan vermeld in dit proefschrift.

184

Acknowledgements

After having been a Ph.D. student for four years, there are of course many people who deserve a special word of thanks. The first and foremost of these people are Prof. Dr. R.A. van Santen and Prof. Dr. M.T.M. Koper for allowing me this position. Marc, apart from the stimulating scientific discussions, I would like to thank you for the great freedom you have allowed me in my work and in writing our articles.

Next up are all the current and former members of the SKA group for generating a pleasant working environment and generally being nice colleagues! Although I was never a frequent visitor of the coffee breaks upstairs, I nevertheless enjoyed your company and count many among you as friends.

My close colleagues from the electrochemistry and battery group deserve some special attention. Victor and Thijs, having shared an office with me for many years, I would like to express my sincere thanks for all the scientific and non-scientific discussions, the flexible working attitude, but mostly for your tolerance of my (continuous) chatter. Ad, I think I don’t have to tell you how important your help has been to my thesis. The fruits of our efficient and, above all, pleasant cooperation are illustrated in our publication (Chapter 8). Peter V., if left to its own, the incomprehensible protocols and large amounts of randomly generated data on our cluster would most likely have grown out to become the first true AI in the world. Therefore, I am glad that I could always rely on you to sort out any problem I might have with this machine and force it to run calculations, instead of pondering the meaning of life, the universe and everything. Then there is the impromptu “electrochemistry” coffee break group, which consisted of my dear friends, Gabi and Afifa as core members, and Frank, Tanya, Brian, Dennis, Freek and other, less frequently visiting SKA members. Guys, no matter how dreary and boring some single crystal experiments might get, I could always look forward to an amusing and relaxing evening coffee break. Your companionship always ensured that I came to work with a smile, but also that I left with one.

I would like to thank as well the “battery group” of Peter Notten. Peter, Martin, Martine, Alexander, Rogier, Chang, Paul, Afifa, and any other member whom I may have forgotten, it was a pleasure sharing a lab with you. Although most of you are delocalized between Philips and the TU/e, your presence added color to life at the university.

Next I would like to thank the entire Electrochemistry group of the University of Alicante, and specifically Prof. Dr. Juan Feliu and Prof. Dr. Rorberto Gómez. Juan, my stay at the University of Alicante was a truly enlightening experience. Moreover, without your gift of several high-quality single crystals I would not have been able to obtain the results presented in this thesis. Roberto, your help and experiences regarding (Pt modified) rhodium electrodes and STM imaging has been of paramount importance to my research. To the other members of this group: Paolo, Nuria, Antonio, Javier (1),

185

186

Javier (2), Loles, Victor, and many, many others. Thanks for the great fiestas and good coffee!

Having a set of beautiful single crystal electrodes does not automatically mean that you get good results. In order to use them properly, you first need to learn “the way of the single crystal”. Fortunately, I had the privilege of being taught by the best single crystal experts the Netherlands had/has to offer. Guillermo and Natalia, thank you for showing me what truly clean glassware is and for pointing out the differences between simply setting fire to an expensive electrode, and flame annealing it. Furthermore, I would like to thank Chrétien for his assistance with the Monte Carlo simulations and Stephanie for the detailed AFM analysis of our samples. One can’t help but wonder what this thesis would have looked like without these people.

Friends are invaluable to any PhD student. You can never have too many friends. However, Duy, sometimes you almost made me think that one is enough … almost. I will never forget all the work you have done for my wedding as well as for this thesis (Yes dear reader, most of the beautiful artwork in this thesis, like the cover and the animations at the side of the pages, are Duy’s). Someday I will repay you, that I promise!

The help of all these people has been very important to my work and to me personally. However, all this would have been infinitely more difficult without the continued support of my family. Mom and dad, Ingrid, John and Robert, I think all of you know exactly what I mean when I say, “thank you for everything!” But I would like to extend the same thanks to my family-in-law, Detti, Harry, Hendrik and Ilona, who have become just as important as my own family over the past years.

Finally, to my lovely wife Verena: How deep is the gratitude of a man to his wife? I do not know the answer to this question, but I do know that you have added meaning to this research, to my life and to everything I do. For that, your support and love I am eternally grateful.