“Wm" uni-I". 5'4. ’
12' .1)»:me ‘, I 15“”“w '- :1"! .15 “
‘i?“§j‘¥“ff3&t§’5f ‘ 3%: r

- «2% W"

.e _ "a =_..
~. 44.143 ’0

2:39 saa;«a-rm.-t MW,.-.-.- e. -,.~.- 1

9A ‘ v

gi-tfkifi'gf‘

' @4ng-
. a“: ‘

avg-.1911:

4-”

“y

’73.. m2:
" L“? .1

1. jg: ’93???“
. q

(293*
yr‘}
1 ' {‘3‘ ,
.‘ xii-5“" 3 l
I ’ ‘
.

; ' ‘5‘
la 3 J: ‘
{emit};
. ' 'V‘ {.2 9"

'. 33‘

a

 

it)... .

(ea:

3.1 1!: flu}! \

{if "Rift?
$1.436th .‘4.
» www.-
{535$ ‘91:;

1 ,v —
’1 51:1} Ll 7“ 9 ' A ’ 0 fi ‘ V , év "fig‘v
v. -. ‘ , ‘- ,- z ,
“'1 :'.‘i‘ . n. n o ‘ V9 1“ _ 1_ P I; f,. .) " ' _ .3. p. 'gg‘é-fig‘é’t'
“ ‘ ‘ l 4 " f‘ ' h- V ‘ '
. ”031) ‘ . ‘ r. I, 'y, .
m. ,(L 5,. , . .. .,__ '_
. "\ a ~ A .
‘ he ' 1 ‘s. * (I, .
A... N .3: ‘ .4.
'N E‘%“<;d' I I. ”f“
r Whig. ‘ !
'n on «gt
0—

' . . A
:3: «J v 1 "a
34% W93:

 

11483::

 

!i9'.!9!!"! !!!'!!!!'!’!!!!!!I

0409 8663

 

 

 

 

 

 

 

 

 

 

 

 

MS U is an Affirmative Action/Equal Opportunity Institution

 
    

  

LIBRARY
Michiga“ 5"“
Unlverslty

    
   
     
    

This is to certify that the

dissertation entitled

"Substrate Morphology's Influence on the
Overlayer Structure and Oscillator Strength"

presented by

Hong Wang

has been accepted towards fulfillment
of the requirements for

Ph. D. degree in Physics

fiyfic

' I Major professor
R. Tobin

DateianuanLLlQSL

0-12771

 

———-————v—-~———
*—
——— —— -___,_— ———...

OE ll RETURN BOXtomnavothb chockomnom your record.
TO AVOID FINES Mum on or More duo duo.

DATE DUE DATE DUE DATE DUE

      

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

SUBSTRATE MORPHOLOGY'S INFLUENCE ON THE
OVERLAYER STRUCTURE AND OSCILLATOR
STRENGTH.

By

Hong Wang

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY

Department of Physics and Astronomy
and Center for Fundamental Materials Research

1995

ABSTRACT

SUBSTRATE MORPHOLOGY'S INFLUNECE ON THE
OVERLAYER STRUCTURE AND OSCILLATOR
STRENGTH

By

Hong Wang

I used infi'ared spectroscopy (IRS) and electron energy loss spectroscopy (EELS)
to probe the influence of substrate morphology on the interaction strength among
coadsorbates and hence on the overlayer structure and on the oscillator strength. We
compared the coadsorption behavior of H and CO on both steps and terraces of the
Pt(335) surface, and compared our results with previous studies on similar surfaces. We
also compared the cross section and Stark tuning rate of edge and terrace CO.

Our infrared spectroscopy study of coadsorption of H and edge CO on Pt(335)
show that along the step edges of the Pt(335) surface, coadsorbed H and edge CO actually
mixed together. In contrast, on Pt(l 1 l) and Pt(112) surfaces, coadsorbed CO and H
segregate into islands. We proposed an overlayer structure model to explain our data, in
which adsorbed H continuously shifts CO from atop to bridge binding. The different
results on Pt(112) and Pt(335) mean that the interaction strength among the coadsorbates
changes with the terrace width.

With EELS, we directly verified the proposed CO site shift. We also surprisingly
found that coadsorbed H produced no observable effect on the HREEL spectra of terrace
CO.

With IRS and EELS, we found that edge atop CO has twice the cross section of
terrace atop CO, and that edge atop CO's Stark tuning rate is also twice that of terrace

atop CO. We explained these and several previous results with an electrostatic model.

This model also partly accounts for the much smaller difference found on surfaces with
much wider terraces.

Our data show that the screening of IR and static fields is different, whether by
changing coadsorbate coverage or by changing the substrate sites. This is not explained
by the standard dipole-dipole coupling model. We also found that the Stark tuning rate
measured in electrochemical cells is 3 times larger than our data if standard models of
electrochemical double layers are used. Coadsorption of H also produces different effects
in the two environments. These results require much better understanding of how the
adsorbate responds to the applied fields.

ACKNOWLEDGMENTS

I am grateful to Professor Roger G. Tobin, my thesis adviser, for his support,
encouragement, and efforts to provide me with a complete training in scientific research,
and especially for his patience. I learned greatly from both his scientific insight and his
way of life.

I am also grateful to Drs. David K. Lambert and Galen B. Fisher, my advisors in
General Motors Research and Development Center, for their support for my work as well
as for my career. I have benefited a great deal fi'om their very different yet individually
successful styles of research.

Drs. Tobin, Lambert and F isher's contribution and their spirits are reflected
throughout this work. Without their help, it is impossible for me to carry out this
research.

I would also like to thank Craig L. DiMaggio, for teaching me how to run many
instruments carefully and correctly, and for his help in carrying out many experiments.
For my 3 years in GM research, many people there have given me candid advice on both
research and many personal issues. I would like to thank particularly Drs. P. C. Wang, D.
Y. Wang, L. Green, D. L. Partin, C. M. Thrush, T. Perry, Y. T. Cheng, D. N. Belton and
T. E. Moylan.

I enjoyed the friendship and support from other members of Roger's group, Dr.
C.Chung, Dr. J. S. Luo, K. C. Lin, D. E. Kuhl, J. Gemmell, E. Krastev. I thank Dennis,
John and Ati especially for their caring about my future. I would like to thank John
partcularly for correcting the English for a large portion of this dissertation.

I would like to thank other Chinese students here in the Physics Department for
their support, especially, Q. Yang, Q. P. Zhu, J. D. Chen, B.Y. Chen, Dr. Y. Cai, Dr.
W.Q. Zhong. I would also like to thank P. H. Liu, C. C. Wang, P. C. Liao, Y. L. Yeh, S.

iv

Liu, and many other friends who make the five years at Michigan State a very memorable
period of my life.

All that I accomplished is given by my parents, who, even in the most difficult
years, taught me the importance of education and knowledge. They have always been my
mental support and the origin of my strength to face the challenges. They have given me
the most parents can give.

Finally, I want to thank my wife, Mei, the source of my happiness. She warms
my heart with her smile. Without her love, I could not have done nearly as much.

This work is partially supported by The Petroleum Research Fund, administered
by the American Chemical Society, and the National Science Foundation under Grant #
DMR-9201077. It is also partially supported by General Motors.

TABLE OF CONTENTS

LIST OF FIGURES ix
LIST OF TABLES xv
1. INTRODUCTION 1
References 10
2. EXPERIMENTAL TECHNIQUES 12
2.1 Scattering meachanisms and selection rules for RAIRS and HREELS 14
2.2 HREELS 15
2.2.1 Operation of the HREELS system 17
2.3 Infiared spectroscopy 18
2.3.1 RAIRS with polarization modulation 21

2.3.1.1 Quantitative analysis of polarization modulated RAIRS 23

2.3.2 Electroreflectance Vibrational Spectroscopy (EVS) 25

2.3.2.1 Quantitative analysis of EVS 26

2.3.3 Determination of the Stark tuning rate 27

2.3.4 Measurement of the E-field 29

2.4 TPD 30

2.5 Sample preparation and characterization 33

References 48
3. COADSORPTION OF HYDROGEN AND CO ON Pt(335): STRUCTURE

AND VIBRATIONAL STARK EFFECT 50

1. Introduction 50

2. Experiment 51

3. Experimental results 54

3.1 TPD 54

3.2 ir spectra 56

vi

4. Structural model of the C0 + H overlayer 56

5. Vibrational Stark effect and coadsorbates 61
5.1 Background 61
5.2 Our experiment 66
5.3 Comparison with electrochemical experiments 69
6. Summary 70
References 83
4. H-CO INTERACTION ON THE TERRACES AND STEP EDGES OF THE
Pt(335) SURFACE 89
1. Introduction 89
2. Experiment 90
3. Results and discussion 91
3.1 Coadsorption of H and CO on the step edge 91
3.2 Coadsorption of H and CO on the terrace 94
4. Summary 98
References 107
5. VIBRATIONAL INTENSITY AND STARK TUNING RATE OF EDGE
AND TERRACE CO ON Pt(335) 110
1. Introduction 1 10
2.Experiment 112
3. Results 113
3.1.RAIRS and EVS 113
3.2. HREELS 114
3.3. Analysis of the atop intensity 115
4. Discussion 1 16
5. Conclusion 123

References 130

vii

6. CONCLUSIONS 1 32

References 136

LIST OF FIGURES

Chapter 1
Figure 1-1. 8
Side view of the Pt(335) surface.
Figure 1-2 9
Atop and bridge CO on Pt(l 1 1).

Charter 2
Figure 2-1. 35

Dipole moment perpendicular to the metal surface is reinforced, dipole moment
parallel to the surface is screened.

Figure 2-2. 36
Block diagram of an electron energy loss spectrometer (Luo [30]).

Figure 2-3. 37
Schematic diagram of the system for high resolution electron energy loss
spectroscopy in this work (B. A. Sexton [34]).

Figure 2-4. 38
The surface electric field 117120 in (a), and the quantity (E/EmZ/cose in (b) for
platinum at 2100 cm"1 ( s = -375 - 200 i) as a function of the incidence angle 9 (
Bradshaw and Schweizer [2]).

Figure 2-5. 39
The process of a RAIR spectrum: comparing the scans with clean sample and with
adsorbate (CO) (Luo [3 0]).

Figure 2-6 40
Optical set up for RAIRS and EVS showing detector D, sample S, and electrode
E, polarizers P1 and P2 (Lambert [27]).

Figure 2-7 41
Schematic diagram of computer-controlled wavemeter for laser frequency
calibration. Back-to-back hollow corner cube reflectors are mounted on a ball
slide translation stage driven by a stepping motor. The stage moves freely except
near the end of its travel where a beaded chain coupling becomes rigid and
reverses its motion. (Evans and Lambert [21])

Figure 2-8. 42
Schematic of the system used for polarization modulated reflection absorption
infrared spectroscopy (RAIRS) (Luo [30]).

Figure 2-9 43
Process of measuring the EV spectrum by modulating the E—field applied to the
surface (Luo [3 0]).

Figure 2-10 44
Schematic of the system used for electroreflectance vibrational spectroscopy
(EVS) (Luo [30]).

Figure 2-1 1 45
Schematic of the system used for temperature programmed desorption (TPD)
(L110 [3 0])-

Figure 2-12 46
TPD spectrum for saturation coverage of NO on Pt(335). The highest peak is
from edge NO and the rest are from terrace NO. All the terrace NO stays at
identical sites before heating. ( Wang et al. [32])

Figure 2-13 47
C(4x2) structure of 0.5 ML CO on Pt(l l 1) (Luo [30]).

Chapter 3
Figure 3-1. 73

a) TPD spectra obtained by desorbing CO fi'om the Pt(335)surface (without H).
The CO dosages used to prepare (a)-(d) were 20, 1.5, 1.0 and 0.5 L, respectively.
b) Fit of two Gaussians (one fi'om edge CO and the other from terrace CO) to the
20 L TPD spectrum.

Figure 3-2. 74
TPD spectra obtained by desorbing H; from the Pt(335) surface (without CO).
From top to bottom the dosages used to prepare the surface were 40, 20, 10, 3,
1.5, 0.8, 0.5, 0.3 and 0.1 L.

Figure 3-3. 75
TPD spectra obtained by desorbing H; from the Pt(335) surface (with and without
CO). In a) 9co = 0, 0" = 0.35 ML; in b) 9co = 0, 0" = 0.25 ML; in c) 000 =
0.05 ML, 0“ = 0.31 ML; and in d) 000 = 0.16 ML, 0“ = 0.30 ML. The
hydrogen doses were a) 0.5 L, b) 0.3 L, c) 0.8 L and d) 4.5 L.

Figure 3-4. 76
Measured H occupancy of edge sites vs H2 dosage with various pre-coverages of
CO. The data with O, A and I were obtained by pre-dosing with 0.5, 1.0 and 1.5

L of CO, respectively.

Figure 3-5. 77
RAIR and EVS spectra of CO coadsorbed with H on Pt(335). The CO coverage
was 0.16 ML for all ofthe spectra. From top to bottom the spectra are for OH =
0.06, 0.10, 0.16, 0.18, 0.21, 0.29, and 0.06 ML. The lowermost spectrum
was obtained after the sample had been heated to 420 K to desorb the H but leave
the CO in place.

Figure 3-6. 78
Data showing the efl'ect of coadsorbed H on the resonant frequency of the C-0
stretch vibration at atop sites, for CO and H coadsorbed on Pt(335).

Figure 3-7. 79

Data showing the effect of coadsorbed H on integrated intensity S of the C-0
stretch vibration at atop sites, for CO and H coadsorbed on Pt(335).

Figure 3-8. 80
Data showing the effect of coadsorbed H on the Stark tuning rate (dv/dE) of the
C-0 stretch vibration at atop sites, for CO and H coadsorbed on Pt(335).

Figure 3-9. 81
Data showing the effect of coadsorbed H on the integrated intensity S for atop CO
on Pt(335). With each CO coverage, the data have been normalized by So, the
value of S at the lowest H coverage. The data with 0.06, 0.12 and 0.16 ML of C0
are represented by O, A and I, respectively.

Figure 3-10. 82
Data showing the effect of coadsorbed H on the screening factors ( O ) 7,, and (

O ) 74, for CO on Pt(335). We assume that 1,, = 1 and 14., = l at the lowest
coverage, and that ykoc W and ydcoc (dv/dE).

Chapter 4
Figure 4-1. 99
TPD spectra obtained by desorbing H; from the Pt(335) surface (without CO).
From top to bottom, the relative dosages used to prepare the surface were 10, 5,
2.5, 1, 0.5, 0.25, and 0.1. The absolute dosages are uncertain because a closer was
used, but a relative dose of l is approximately 1x 10'6 torr s.
Figure 4-2. 100
BEL spectra of edge CO on Pt(335) as a function of H coverage. The CO
coverages in (a) and (b) are 0.07 and 0.13 ML, respectively. The spectra are
arranged so H coverage increases up the page. In (a) the H coverages are 0.02,
0.10, 014,018, 0.22, and 0.24 ML; in (b) 0 (em = 0 also),0.01, 0.06, 0.10, 0.13,
0.22, and 0.40 ML.

Figure 4-3. 101
(a) Ratio Ila/In of bridge to total (bridge +atop) single-loss EELS intensity as a
fimction of H coverage, for the spectra in Fig. 4-2. (h) Ratio IB/IE where 1.; is the
elastic EELS intensity, as a function of H coverage, for the spectra in Fig. 4-2. 9
Figure 4-4. 102
EEL spectra at various CO coverages after predosing with 0.25 ML H to block the
edge sites. The spectra are arranged so CO coverage increases up the page: Geo =
0.05, 0.08, 013,019, 0.28, and 0.36 ML.

Figure 4-5. 103
Ratio IB/Im as a function of CO coverage, for terrace CO, with the edge saturated A
with H (for the EEL spectra in Fig. 44). Also shown is the bridge CO coverage,
calculated assuming that the EELS cross section of terrace atop CO is a factor 1.8
times that of terrace atop CO.

Figure 4-6. 104
EEL spectra of 0.05 ML terrace CO as a function of postdosed H coverage. The
spectra are arranged so 0“ increases up the page: OH = 0.24, 0.46, 0.70, and
0.70ML.

Figure 4-7. 105
Ratio IB/Im. from EELS intensity for terrace CO, as a function of H coverage, for
three CO coverages: A 0.05 ML, 0 0.13 ML, and I 0.19 ML.

Figure 4-8. 106
TPD spectra obtained by desorbing a saturation coverage H; from the Pt(335)
surface and by desorbing 0.48 ML of H on clean surface. In (a) 9co = 0 ML, 0“ =

1.0 ML; in (b) 9co = 0.05 ML, 03 = 0.69 ML; in (c) 900 = 0.13 ML, 0" = 0.55
ML; and in (d)9co = 0.19 ML, 0“ = 0.45 ML; (e) 9co = 0 ML, 0“ = 0.48 ML.

Chapter 5

Figure 5-1. 125
RAIR and EV spectra for 0.16 ML CO on a Pt(335) surface precovered with 0.72
ML of H, before and after annealing the sample at 420 K. Upon H desorption, the
CO moves to edge sites.

Figure 5-2. 126
RAIR and EV spectra for 0.14 ML CO on a Pt(335) surface predosed with 0.1 L,
before and after annealing the sample at 260 K. Upon annealing, the CO moves
from terrace sites to edge sites, while the 0 moves to terrace sites.

Figure 5-3. 127
HREEL spectra for 0.16, 0.13, 0.08, 0.05 ML CO ( from top to bottom) on a
Pt(335) surface precovered with 0.25 ML of H, before and after annealing at
420 K.

Figure 5-4. 128
Calculated B field normal to the average surface plane, as a function of fiactional
distance across the terrace. The field is normalized to the field on a flat surface,
and is calculated along a line one-half step height above the terrace, as shown in
the inset. The dotted curve represents a surface with narrow terraces similar to
Pt(335); the solid curve represents a surface with much wider terraces, similar to
that used in Refs. 8 and 9 of Chapter 5.

Figure 5-5. 129
EV and RAIR spectrum for 3.5 L (0.26 ML) CO on clean surface. Strong EV
features are seen corresponding to both of the peaks in the RAIR spectrum,
indicating that edge and terrace CO have comparable Stark tuning rates. The
zero-crossings in the EV spectrum occur at the same frequencies as the peaks in
the RAIR spectrum. These results are in contrast to those reported in Ref. 4 of
Chapter 5.

LIST OF TABLES

Table I. 72
Summary of the ir spectra with only CO on Pt(335). Here 9co is the CO
coverage, v is the frequency of the peak in the ir spectrum, (AR)/R is the
maximum CO-induced reflectivity change in the RAIR spectrum,
S = ](AR)/ Rdv, and (dv/dE) is the Stark tuning rate in terms of the externally
applied E-field.

Table H. 124
Ratios of vibrational intensity, vibrational cross section, and Stark tuning rate of
edge atop CO compared to terrace atop CO. The intensity and Stark tuning rate
ratios are determined directly from the experimental data. The cross section ratios
include corrections for loss of CO during annealing and for migration between

bridge and atop sites, as discussed in the text.

Chapter 1

Introduction

This work deals with chemisorption and coadsorption of several simple
molecules and atoms on a stepped Pt(335) surface and how their interactions change over
site differences of atomic scale. This work demonstrates that the existence of monatomic
steps profoundly influences the adsorption and coadsorption behavior of adsorbates on
the surface. Not only the adsorption on step edges, but the adsorption on the terraces also
differs fiom that on a flat surface.

Chemisorption of gas molecules and atoms on metal surfaces is a very important
issue in catalysis, coating and anti-corroding [1]. Transition metals, like Pt, are among
the most commonly used catalysts. The real catalysts used in applications are
. polycrystalline and have a large concentration of steps, defects and kinks. A study on a
regularly stepped surface like (335) is an important step toward understanding real
situations. One can compare the different behavior of adsorbate on terraces and steps
and even compare them with flat surfaces gaining insight into the chemical and physical
processes that happen on real catalyst surfaces. Meanwhile, more than one kind of
species is present on the catalyst under real conditions. A coadsorption study should also
be very beneficial in understanding the interaction between the adsorbed species on the
surface and how the interactions change with site differences at the atomic scale.

The Pt(335) crystal, Pt(s)[4(1 11)x(100)] in step-terrace notation, is shown in Fig.
1-1. The (335) surface consists of 4 atom wide (111) terraces, separated by monatomic
(100) steps.

The essential tool in this study is vibrational spectroscopy. Vibrational
spectroscopy is based on the fact that the constituent atoms of a molecule, solid, or

combination of the two execute multidimensional, quasiperiodic motion over potential

energy surfaces that are determined by the electronic/chemical state of the system, when
the energy content of the motion is substantially less than the electronic or chemical
dissociation energies of the bound system [2]. The low-lying vibrationally excited
eigenstates associated with energies in the ~ 50 - 250 meV ( 400 - 2000 cm'1 ) range
involve light atoms such as H, C, N, and O which are of obvious chemical importance.
Transitions between these states give rise to a spectrum which is a characteristic signature
of both the chemical species being interrogated and its local environment. The
corresponding frequency of the quasiperiodic motion of the (usually heavy) atoms
comprising the substrate is usually much less than the intramolecular frequencies.
Consequently the vibrational modes of the admolecules retain much of their free space
character, which permits species identification. The small deviations from free space
behavior, such as in frequencies and line widths, carry information about the local
environment such as bond site or molecular orientation and the characteristics of the
adsorbate-substrate interaction[2].

Adsorption on the surfaces can be atomic and molecular and can have many
different bonding situations. For the species in this study, CO stays in molecular form on
Pt(335), hydrogen dissociates on this surface, and oxygen stays in molecular form below
150 K but dissociates above 200 K [15]. Because CO adsorption is most heavily studied
and hence most clearly understood, I will use it as an example to illustrate the general
concepts of adsorption. Let us first look at CO on Pt(] 1 1), a flat surface. Studies have
shown that CO binds vertically on this surface with the C atom at the bottom [3]. (It is
relatively easy to understand, since CO donates charges to the metal, but oxygen is
unlikely to do this.) At low coverage, all the CO stay straight on top of a Pt atom. At
high coverage CO can occupy two-fold bridge sites[16], as shown in Fig. 1-2. The
binding energy for atop CO is slightly higher than that of bridge CO, 1 kcal/mol (0.043
eV). When a step is introduced, as on our sample, the binding energy at the step is

significantly larger, because of the lower coordination number for the substrate atom

there. So the adsorption takes place first at the steps and then on the terraces. On
Pt(335), both edge and terrace CO can have atop and bridge bondings. Experiment has
also found that CO does not tilt away from the surface normal by more than 10° [4].

Spectroscopy techniques used in this study were polarization modulated reflection
absorption infrared spectroscopy (RAIRS), electoreflectance vibrational spectroscopy
(EVS) and high resolution electron energy loss spectroscopy (HREELS). Temperature
programmed desorption spectrometry (TPD) was also extensively used. Detailed
discussion about the techniques will be given in Chapter 2.

In this work, I studied adsorption of CO, and coadsorption of H and CO, 0 and
CO, on Pt(335) with the above mentioned experimental techniques, with the focus on
comparison between edge and terrace sites. The existence of the steps profoundly
changes not only the morphology of the surface but also the surface electronic states.
Consequently, adsorption and coadsorption on step sites and terrace sites show different
character as compared with each other. I performed several experiments to investigate
the difference between steps and terraces and compare them with a flat surface. First, I
successfully showed that the interaction between coadsorbed species on step edges is
different from that on flat surfaces. Furthermore, this interaction is modulated by the
terrace width; and the interaction on the terraces is also different from that on flat surface.
Second, I demonstrated that the vibrational cross section of the same species is different
for step and terrace sites, and that the difference is mostly fi'om the field enhancement on
the step and screening on the terraces. The difference between the chemical environment
of the two sites is very small.

The general motivation of this work comes from many previous studies that
demonstrate differences between step and terrace sites. In particular, before our work, all
coadsorption studies of CO and hydrogen on single crystal Pt surfaces have found them
to be strongly repulsive: CO and hydrogen segregate on both Pt(l 1 1) [5,6] and along the
step edges of Pt(] 12) [7]. When segregation happens, CO is compressed together by

coadsorbed hydrogen, thus, the hydrogen-induced change in the spectrum for CO is
similar to the change caused by increased CO coverage on a clean surface. On Pt(l 1 l),
with increasing CO coverage, CO first occupies atop sites and then bridge sites. The
vibrational fi'equency also increases with CO coverage as a result of stronger coupling
between the CO molecules[16]. Segregation of CO and H on Pt(] 11) is evidenced by the
CO site shifi, frequency shift and IR peak shape changes [6] and by thermal energy atom
scattering (TEAS) and low energy electron diffraction (LEED) [5]. Edge CO on Pt(112)
is studied by electron stimulated desorption ion angular distribution (ESDIAD), which
probes the direction of the adsorbate’s axis on the surface [7]. Along the step edges of
Pt(112), edge CO exhibits complex tilting angles with increasing coverage. Segregation
of edge CO and H on Pt(112) is evidenced by the CO pattern change with increasing H
coverage, which follows the same pattern change as the CO density increases [7].
Previous experimental evidence and theoretical calculations have shown that H-CO
interaction on the surface can only be mostly indirect, or through-metal. This kind of
interaction depends strongly on the perturbation of the charge density of the substrate
metal by the adsorbates. It is conceivable that the terrace width can influence the strength
of this kind of interaction by curbing off the charge density perturbation at the steps.

I performed a RAIRS and EVS study on coadsorption of edge CO and H on
Pt(335), which is structurally similar to Pt(112) and Pt(lll) [17]. The results are
presented in Chapter 3. My results clearly show that even though the interaction between
edge CO and coadsorbed H is repulsive, the strength of this repulsion is weaker compared
to the CO—CO, and H-H repulsion. Edge CO and H mix together within the hydrogen
region, and my experimental evidence implies that CO stays only on bridge sites within
such islands, while outside such islands CO is unperturbed by H and stays on atop sites.
This result, compared with the (112) result, proves that the interactions between edge
coadsorbates are influenced by the terrace width.

I performed HREELS experiments to directly verify the site shifting of edge CO
by the coadsorbed H [18]. I also studied the coadsorption of terrace CO and H with
HREELS [18]. The results are presented in Chapter 4. I found that the site shifting of
CO is accompanied by a significant change in bridge CO's cross section. The
experiments also show H did not produce any observable effect to the terrace CO,
different fi'om both the flat surface result and the edge CO result. These results show that
the interactions between coadsorbates are much more complex than previous believed. A
subtle change in surface morphology can radically change the formation of the adsorbed
layer.

Another direct motivation of this work is the controversy over whether there is
significant E-field enhancement at the step edges. One possible difi‘erence between the
edge and terrace species is that they may have different cross sections, which may have
physical or chemical origins. A previous IR study of CO on Pt(335) has shown that the
increase of IR intensity with CO coverage is 2.7 times slower when terrace CO appears
than for edge CO[8]. Greenler et a1. explained this with a classical electrodynamics
calculation of E-field enhancement at the step edges [9]. On the other hand, Lambert and
Tobin compared the cross section of edge CO on Pt(335) with CO on Pt(] 1 l) and found
them to be similar[10], which seems to contradict Greenler et al.'s explanation.
Furthermore, Reutt-Robey et al., with a time-resolved IR study, reported that the cross
section of edge CO and terrace CO is the same within 5% on Pt(S)[28(111)x(110)] and
that put a serious question on whether there is any difference on edge and terrace CO's
cross sections[11,12].

In most of the above mentioned comparisions, terrace CO coexists with edge CO.
Strong intensity coupling between the two makes it difficult to separate the contribution
to the total intensity fiom edge and terrace CO. I performed both RAIRS and EVS as
well as HREELS experiments to compare the cross section of terrace and edge CO
directly, on Pt(335) [19]. The results are presented in Chapter 5. The terrace CO is

produced by blocking the step edges with H or O at low temperature. After heating the
overlayer to a certain higher temperature, terrace CO will move to edge sites. The
questions I wanted to answer were first whether there is difference in edge and terrace
CO's cross section, second, if there is, whether it is from chemical origin or from physical
origin?

I found clearly that edge CO has a cross section twice as large as terrace CO on
Pt(335). I also performed classical E—field calculations to show that the difference in
cross section mostly originates from field strength difference at the two sites, which
consists of both the enhancement at the edge and screening on the terrace. My
calculation also shows that screening on the terraces decrease rapidly with the increasing
of the terrace width. This explains half the diference between our result and the result on
Pt(s)[28(11 1)x(110)].

This work is also significant in reconfirming the previous finding of different
screening of static and IR fields by coadsorbates[13], which is not in agreement with the
present understanding of surface electrodynamics. Furthermore, I find that the screening
by the metal of the static and IR fields is also different, which again disagrees with the
current understanding of electrodynamics. These findings reveal that the current
understanding of surface electron response to applied E-field and the depolarization effect
by coadsorbates is still very incomplete.

This work is also important in providing a comparison between the response of
adsorbates to applied static E-field in ultrahigh vacuum (UHV) and in an electrochemical
environment. Lambert [14] has measured the Stark tuning rate of CO on a flat Ni surface
in vacuum and it is in agreement with current models of the double layer between metal
and electrolyte in the electrochemical cell. However, from the first direct comparison
between the same adsorbate on same substrate in UHV and electrochemical cell, CO on
Pt(] 1 1) [13], the Stark timing rate measured in UHV is a factor of two smaller than that

from the electrochemical measurement.

My measurement of CO’s Stark tuning rate on Pt(335) provides another direct
comparison with the electrochemical experiment [20]. I also found that the Stark tuning
rate from the UHV measurement is much too small compared to that fi'om
electrochemical measurements to be explained by current double layer models. The
coadsorbed H also produces different results in the two environments. In UHV the CO’s
Stark tuning rate is not influenced by H while in the electrochemical experiment, CO’s
Stark tuning rate goes to zero in the classical hydrogen region. These profound
differences demonstrate that the current understanding of the double layer between the

metal and the electrolyte needs significant correction.

\ 39000099
QW:::::::::

 

Figure 1-1. Side view of the Pt(335) surface.

W06)
©0000
.0001 D -

Atop CO Bridge CO

Figure 1-2. Atop and bridge CO on Pt(l 1 1).

10

References
l. G. A. Somorjai, Chemistry in Two Dimensions: Surfaces, (Cornell University Press,
Ithaca, New York, 1981).

2. J. W. Gadzuk, in Vibrational Spectroscopy of Molecules on Surfaces, edited by J .T.
Yates, Jr. and T. E. Madey, (Plenum, New York, 1987).

3. P. Hoffman, S. R. Bare, N. V. Richardson, and D. A. King, Solid State Commun. 42,
645 (1982).

M. Trenary, S. L Tang, R. J. Sirnonson, and F. R. McFeely, Surf. Sci. 124, 555 (1983).
4. J. S. Somers, T. Linder, M. Surman, A. M. Bradshaw, G. P. Williams, C. F.
McConville, and D. P. Woodruff, Surf. Sci. 183, 576 (1987).

5. S. L. Bemasek, K. Lenz, B. Poelsema, and G. Comsa, Surf. Sci. 183, L319 (1987).
6. D. Hoge, M. Tiishaus, and A. M. Bradshaw, Surf. Sci. 207, L935 (1988).

7. M. A. Henderson and J. T. Yates, Jr., Surf. Sci. 268, 189 (1992).

8. B. E. Hayden, K. Kretzschmar, A. M. Bradshaw, and R. G. Greenler, Surf. Sci. 149,
394 (1985).

9. R. G. Greenler, J. A. Dudek, and D. E. Beck, Surf. Sci. 145, L453 (1984).

10. D. K. Lambert, and R. G. Tobin, Surf. Sci. 232, 149 (1990).

11. J. E. Reutt-Robey, Y. J. Chabal, D. J. Doren, and S. B. Christrnan, J. Vac. Sci.
Technol. A7, 2227 (1989).

12. J. E. Reutt-Robey, Y. J. Chabal, D. J. Doren, and S. B. Christman, J. Chem. Phys. 93,
9113 (1990).

13. J. S. Luo, R. G. Tobin, and D. K. Lambert, Chem. Phys. Lett. 204, 445 (1993).

14. D. K. Lambert, J. Chem. Phys. 89, 3847 (1988).

15. H. Wang, R. G. Tobin, G. B. Fisher, C. L. DiMaggio, D. K. Lambert, unpublished.
16. B. E. Hayden and A.M. Bradshaw, Surf. Sci. 125, 787 (1983).

17. H. Wang, R. G. Tobin, and D. K. Lambert, J. Chem. Phys. 101, 4277 (1994).

ll

18. H. Wang, R. G. Tobin, D. K. Lambert, G. B. Fisher, C. L. DiMaggio, submitted to

Surface Science.

19. H. Wang, R. G. Tobin, D. K. Lambert, G. B. Fisher, C. L. DiMaggio, to be submitted

to J. Chem. Phys
20. C. S. Kim, W. J. Tomquist, and C. Korzeniewski, J. Phys. Chem. 97, 6484 (1993)..

12

Chapter 2

Experimental Techniques

Spectroscopy techniques used in this work include two forms of IR spectroscopy,
RAIRS and EVS, as well as HREELS. TPD is also extensively used in this work.
RAIRS, HREELS and TPD are among the most commonly used techniques in surface
science study. Review articles on RAIRS, HREELS are readily available [1,2,3,4,5,6,7].
There are also several articles about the quantitative analysis of TPD[8,9]. The EVS
system used is the only one in the world that can measure the first order Stark effect of
adsorbed molecule in vacuum. It was built by Dr. D. K. Lambert at GM R&D center.

In HREELS, a monochromatic electron beam with energy of several eV is
focused onto the surface at a relatively large angle, between 45-7 0°, and the energy
distribution of the outcoming beam is analyzed. The electrons lose energy due to long
range dipole scattering from the adsorbed molecules and excitation of the vibrational
modes of the adsorbed molecules. By analyzing the energy loss spectrum, one can then
gain information about the surface layer.

HREELS has several advantages. Among them are high sensitivity, wide
dynamic range, and relatively quick data taking. Also, because the technique is quite
mature, good HREEL spectrometers are commercially available. HREELS also has
several disadvantages, the most evident of these are, low resolution, usually between 40-
80 cm'1 in routine operations; less reliable absolute intensity due largely to work function
variation caused by difference in the surface layer and ordering of the adlayer; and the
requirement of vacuum. (By trading off sensitivity, recent advances by Ibach [10] in
HREELS design have achieved 7.9 cm'1 resolution. Commercially available instruments

have also recently arrived at the same level [11].) Because of the low resolution in our

l3

HREELS, step and terrace species usually can not be distinguished as the frequency
difference is only in the order of 20 cm].

RAIRS is also a very widely used technique. In RAIRS, an IR beam is directed
onto the sample at a near grazing angle. The interaction between the IR field and the
vibration modes of adsorbates on the surface results in absorption of IR intensity at
certain fi'equencies. By analyze the intensity distribution as a function of frequency, one
can gain much information about the surface layer. In addition to high sensitivity,
RAIRS has very high resolution, typically 1 cm]. High resolution is necessary in site
assignment, in studying frequency shifts as the result of interaction between the
adsorbates, which are usually of 10 cm'1 order, and in line shape studies which are
particularly important in understanding the dynamics behind vibrational modes. Since
RAIRS uses photons as the probe instead of electrons as in the HREELS, vacuum is not
required for the technique itself. This significantly widens the applicable area to include
applied studies. The biggest limitation to the current RAIRS technique is the lower
frequency limit which can be reached. It is very diffith to apply RAIRS to fiequencies
lower than 800 cm'1 mainly due to the low intensity of the source and very high ambient
noise. There has been much progress in this area, synchrotron sources which are several
orders brighter than the conventional sources [12] can be used and ambient noise can be
reduced by reducing the temperature of the whole system [13].

TPD has been widely used in studying kinetic parameters, like activation energy,
pre-exponential factor, as well as in adsorbate coverage measurement and site assignment
for admolecules desorbing from the surface. In TPD, the sample is heated from a low
temperature to a higher temperature, and the partial pressure increases for the molecules
of interest in the vacuum chamber are monitored as a function of the sample temperature.
The various chemisorbed phases can be distinguished by the order in which they desorb.
Population of individual phases can be deduced fi'om the integration of the particular
peak in the spectrum. It is dangerous to draw conclusions based on TPD alone, because

14

TPD is a technique that changes the surface condition, unlike RAIRS or HREELS.
When admolecules of one chemisorbed phase begin to desorb, all the admolecules in
phases with lower activation energy have already desorbed and the sample temperature is
high enough for the admolecules in this phase to rearrange. So in many cases, TPD can
only be used as a reference with RAIRS or EELS required to supply crucially needed
information.

In this Chapter, I will discuss in detail the principles of the three techniques and
the actual experimental set up in my work. I will also discuss briefly the sample
preparation and characterization at the end. Because of the tmique nature of the
polarization modulated RAIRS and EVS, discussion concerning them will be lengthier
than that of HREELS and TPD.

2.1 Scattering mechanisms and selection rules for RAIRS and HREELS

When an adsorbate chemisorbs on a metal surface, there is usually some charge
transfer 8e between the adsorbate and the substrate. The dipole moment due to the
transferred charge and its image in the substrate is [6]

u = 28e(So + q(t)) . (2.1.1)

where S0 is the equilibrium location of the static charge centroid from the image plane
and q is a possible oscillatory small displacement about this equilibrium point. Since
energy transfer will not occur in a static field, the static portion of 1.1 does not contribute
to the interaction with an electron or photon. Only the time varying part of u, the
dynamic dipole-moment does. When the adsorbate vibrates on the surface, an oscillatory
field will be set up in the vacuum above the surface. The time varying field interacts with
an incoming photon or electron, resulting in the absorption of the photon or the energy
loss of the electron. This is the basic scattering mechanism behind RAIRS or HREELS.

Because of the high mobility of electrons inside metal, the component of the E
field parallel to the surface is almost perfectly screened, as discussed later in detail in
section 2.3. The response of the metal surface that screens the parallel field will also

15

screen out any dynamic dipole moment appearing on the adsorbate in a direction parallel
to the surface. This is illustrated in Figs. 2-1 and 2-4. The combination of the two
screening effects, which are actually the same physically, results in only ‘perpendicular’
vibrations being observed in RAIRS or HREELS.

There are exceptions for this rule, under special circumstances. In EELS,
electrons can lose energy through “impact scattering”, i. e. direct scattering off the ion
cores of the adsorbates. The incident electron feels the full atomic details of the
adsorbates, and can excite dipole forbidden modes. The scattered wave is diffuse rather
than directed, as in dipole scattering, which makes it easier to observe this kind of
scattering at off specular angles. Examples of impact scattering are given in ref. [14]

Recent IR studies have also observed formally dipole-forbidden vibrational modes
[15]. The origin of this effect is explained by Persson and Volokitin based on the “surface
resistivity” concept [16]. Scattering of the electron from the adsorbates results in a broad
band absorption in the IR light [17]. When the IR frequency to coincides with the
resonance frequency (00 of the parallel adsorbate vibrations, the molecules move in
resonance with the collective drift motion of the electrons; hence the additional sm'face
resistivity vanishes and the IR reflectivity reaches the original value of the clean surface.
this results in an anti-absorption peak which is observed at the frequency mo of the
molecular vibration parallel to the surface (fi'ustrated translation or rotation).

Certain parallel vibrations with dynamic dipole perpendicular to the surface can
also be observed in HREELS and RAIRS. For example, molecular oxygen lies down on
the Pt(l 1 1) surface: when the 0-0 stretch mode is active, charge transfers back and forth
between the substrate and the molecule. This produces a perpendicular dipole moment.
This mode has been observed in both HREELS [36] and RAIRS [13].

2.2 HREELS

HREELS was one of the main probes in this study. HREELS was used to confirm

the site shift of CO by coadsorbed H as the bridge CO vibrational frequency was out of

16

the active range of the diode laser used in the IR study. HREELS is also used in studying
the different cross sections of terrace and edge CO, and coadsorption of terrace H and
terrace CO.

Low energy electrons, with energy of several eV, are the probe in HREELS. The
beam energy used was always 2.257 eV in this study. As an adsorbate on the surface
vibrates, it modulates the electric dipole moment of its environment in a time-dependent
fashion. An electron in the vacuum above the crystal senses a long-ranged dipole electric
field, and that produces small angle scattering typically substantially more intense than
the scattering with large deflection angles. One observes a ‘lobe’ of inelastically
scattered electrons sharply peaked about the specular direction [5].

Schematic diagrams of the HREELS instrument used in this study are shown in
Fig. 2-2 and 2-3. Two 127° cylindrical deflector analyzers (CDA) are used in this
system, one as the monochromator, the other as the energy analyzer. Since the resolution
requirement of the electron beam is in the order of several meV ( lmeV z 8 cm'1 ), a
monochromator must be used capable of selecting an electron beam with a very narrow
energy distribution. A lens system is used after the selector to allow the electron energy
at the sample to be independently chosen fiom the pass energy of the selector. For the
vibrational measurement, we are interested in the electrons coherently reflected from the
surface and events in which essentially no momentum is transferred are confined to a
small cone about the specularly reflected beam. The beam is then retarded and focused
into a dispersive energy analyzer.

Sensitivity and resolution are the most important requirements for a surface
science analysis technique. Sensitivity is important because the adlayer we want to study
is usually less than a monolayer, so the total amount of particle responding to the probe is
small. With a cross section close to atomic dimensions [3], HREELS has quite high
sensitivity, and is able to study a surface layer in 0.1% monolayer order . The other

important index is the resolution, which is absolutely necessary in site and species

17

assignment. High resolution is especially required in line shape studies. The typical
resolution of HREELS is between 40 to 80 cm'l, by trading off sensitivity, 8 cm'1
resolution has been achieved [10]. The ultimate resolution of a HREELS system is
decided by the resolution of the monochromator and the energy analyzer. Several
analyzer designs have been used, including a 42° cylindrical mirror analyzer (CMA), a
180° concentric hemispherical analyzer (CHA) and a 127° cylindrical deflector analyzer
(CDA). The relative merits of each of them is discussed by Avery [3] and by Ibach and
Mills [5].

Since the kinetic energy of the electron beam in the selector is as low as several
hundred meV, any magnetic field present can potentially destroy the sensitivity and the
resolution of the HREELS system. OFHC copper was used to make the bulk parts of the
system and the earth's field is screened by a cylindrical layer of mu-metal 0.014 in. thick
surrounding the instrument. The electron filament heater leads are also twisted to provide
a noninductive winding [34]. The whole system is also degaussed every time when it is
taken out of vacuum.

2.2.1 Operation of the HREELS system

Tuning of HREELS is an art. There are more than 25 adjustables controlling the
optics. I have found the best process is finding the elastic peak first and adjusting every
variable (except the bias voltage between the sample and the HREELS instrument) to
optimize the peak. The same process is repeated several times until I get a symmetric,
strong (> 2 x104 counts/sec) elastic peak. The first tuning can take as much as 1 to 2
hours. Usually within the day, if the work function of the surface has not radically
changed, the tuning of subsequent scans is easy, usually taking less than 1 min., as long
as I return the sample back to the same position.

Maj or re-tuning is necessary if the work function has changed or major change in
overlayer ordering occurs. Some experimentalists advocate compensation by changing

the bias voltage on the sample. I, however, have not found that method as reliable as re-

l8

tuning, in agreement with Avery [3] and several GM researchers[18]. Besides, changing
the bias voltage can potentially destroy the consistency of HREELS and sometimes even
cause the reflection of the electron beam without inelastic surface interaction. By
keeping the bias voltage unchanged, the tuning is quite reproducible.

Quantitative analysis of the HREELS intensity is usually done by measuring the
height of the inelastic peak and comparing it with the height of the elastic peak. The
rationale behind this practice is that HREELS resolution is significantly wider than the
intrinsic line width of the vibrational mode. I found that the loss peaks were almost
always wider than the elastic peak and integrated both the loss and elastic peaks over
frequency in this study. This is the common practice in infrared spectroscopy for
obtaining the integrated intensity. Examples of HREEL spectra can be seen in Chapters
4 and 5.

2.3 Infrared spectroscopy

Interaction between the electromagnetic field of infiared radiation and the
oscillating dipole associated with a particular normal mode excites the vibration of
admolecules on the surface. The excitation manifests itself in the absorption of the
radiation. This is the basis of reflection absorption infrared spectroscopy (RAIRS). In
RAIRS, the adlayer we are interested consists of only ~10” molecules (atoms),
significantly less than the number of molecules in a bulk sample or in a traditional high
surface area sample in the transmission IR experiment. Therefore, certain special
experimental conditions are necessary in order to observe the small absorption, at times
smaller than 0.1%.

There have been many theoretical considerations and review articles [1,2]
pertaining to such experimental conditions. The most important consideration is that
only the p-polarized component ( with the polarization in the incidence plane, while the
s-polarization is perpendicular to the incidence plane.) of the IR beam is able to interact

with the adsorbate and such interaction is most strong at near-grazing incidence [19].

19

The dependence of absorption strength on incidence angle was first considered by
Greenler [20] for reflection from a clean and highly reflective surface with classical
electrodynamics. It is then straightforward to calculate the strength of the field on the
surface with the Fresnel equations. In Fig. 2-4, the dependence of the electric field
strength E (normalized to the strength of the incoming beam) on the incidence angle 0 is
shown on a platinum surface at 2100 cm'l. The popolarized light is further split into two
components in Fig 2-4, Epi and Ep// , perpendicular and parallel to the surface,
respectively. The strength of the p-polarized light increases with 0 and falls rapidly to
zero at 90° after a maximum at about 86°. The more important quantity is the total
luminosity on the surface, EZ/cos(9), because the number of molecules with which the
incoming incident beam can interact is proportional to 1/cos(0). The total absorption is
then given by AR cc (EPDZ/COS(9), because both E, and Ep// are much smaller than Epi,
especially at high incidence angles, as can be clearly seen in Fig. 2-4. This, as a result of
the high conductivity of the metal and the boundary conduction on the surface, is a well
known result and also decides the dipole selection role in the RAIRS and HREELS
experiments, as discussed in section 2.1 . AR is also plotted in Fig. 2-4; it is very clear
that in order to get high sensitivity, the experiment has to be done at near-grazing angle.
Two scans are necessary in the conventional RAIRS, one is on the clean surface,
and the other on the surface covered with adsorbates. The reflectance change is then
obtained by subtracting the first spectrum from the second one, as illustrated in Fig. 2-5.
The absorption signal can be as small as 0.1%. Because of the subtraction involved, how
small a signal can be detected crucially depends on the stability of the system. There are
three obvious noise sources in conventional RAIRS. The first is the true noise associated
with short term fluctuations, like shot noise of the source, Johnson noise of the detector
and noise from other electronics instruments. This kind of noise can be reduced by
averaging over a longer period of time. The instability of the system over longer time

scales, for example, the temperature fluctuation of the thermal source or a drift in the

20

optical alignment, can significantly change the look of a spectrum. In my RAIRS, the
sensitivity was really limited by these kinds of instabilities of the system.

The optical system used in both RAIRS and EVS is shown in Fig. 2-6. The diode
laser was used as the IR source. Not all the components shown are in place during
RAIRS and EVS. Beam splitter B and mirror M3 are in place only during calibration of
the laser with the wavemeter and are moved during RAIRS and EVS. Mirror M6 is in
place only during alignment, when the beam cross section profile can be measured with
the pinhole A2. With M6 removed, a visible laser beam, collinear with the IR beam, is
used to check the focusing of the IR beam on the sample and measure the angle of
incidence. The angle of incidence is about 86°.

The diode laser is stepped through a set of predetermined laser currents and heat
sink temperatures. The current is between 0.05 to 0.1 A. The heat sink temperature is
slowly raised from about 70 to 100 K. The frequency of each single mode output is
calibrated by the wavemeter [21]. The wavemeter, shown in Fig. 2-7, is operated open
loop under computer control and is essentially a variable path length Michelson
interferometer in which the two separated beams reflect from back-to-back corner cube
reflectors on a translation stage. The beam splitter, B, is used to combine the IR beam
and the visible beam from a 0.6328 um He-Ne laser into a single collinear beam incident
to the wavemeter. The beam splitter in the wavemeter is ZnSe coated for use at the
Brewster angle in the 4-12 pm range and at 0.6328 um. Separated IR and visible light
interferograms are recorded by a HngTe detector and a silicon photodiode. The
interferograms are fed into a counter-timer to determine their frequency ratio. The
frequency of the diode laser is determined as [21]

co,,=m,,"—rxfe— (2.3.1)

"IR fvls .
Where, fIR/fvis is the measured ratio of fiinges from the IR laser to fiinges from the He-

Ne laser; nvis and “IR are the refiaction index of air at the He-Ne laser and the IR

21

fi'equencies; and (DIR and (Dvis are the frequency of the IR and visible beam, respectively.
The largest systematic error observed by Lambert is 0.034 cm'1 [22].

The diode laser used in this study is a stripe-geometry double-heterostructure
diode laser, grown by MBE on a PbTe substrate with
Pb099335Eu090015Te0993318e090019 active region.The laser can be tuned from about
1750 to 2050 cm"1 with gaps of about 2 cm‘1 with several mW power.

2.3.1 RAIRS with polarization modulation

Since only the p—polarized component of the IR radiation interacts with adsorbate
on the surface, in principle the absorption spectrum can be seen with polarization
modulation in which we compare the reflectance of the s and p-polarized light from the
surface. The experimental set up for this polarization modulated RAIRS is shown
schematically in Fig. 2-8. A photo-elastic modulator (PEM) is used for polarization
modulation. A PEM operates [23] by applying an oscillating stress to a transparent
material (zinc selenide), which causes the difference in optical path between light
polarized parallel and perpendicular to the stress direction to likewise oscillate. The light
intensity transmitted by the PEM and a subsequent polarizer is then modulated by an
amount proportional to the difference between incident intensity of p- and s- polarized
light. The polarization modulation technique has been used previously by several other
groups [24,25] and is discussed fully by [24].

In Fig. 2-8, lock-in A gives the difference between the p- and s- polarized lights,
(19 - 1,), lock-in C gives the total light intensity ( Ip +1, ). The only difference between
1,, and I, should be from the absorption of II, by the adsorbate on the sample. (Ip - I, )/ ( Ip
+ I, ) gives the spectrum of the adsorbate on the surface because fluctuations in the
source intensity and attenuation by gas phase molecules in the optical path are all
canceled out. A spurious signal has been observed from lock-in A, possibly from
ambient radiation modulated also by the PEM. This signal is removed by lock-in B,

referenced at the chopper frequency.

22

In principle, only one scan is necessary to get the vibrational spectrum when
polarization modulation is used. In practice, I still need to subtract the spectrum of the
clean surface from the spectrum of the adsorbate covered surface, due to the variation of
the polarization of the IR ray at different fiequencies. Since polarization modulated
RAIRS measures the difference in intensity between p- and s-polarized light, anything
which changes this difference will appear in a single spectrum, and should be avoided.
Sources of such irreproducibility include change in the polarization state of the light from
the laser and the focal point of the light on the sample. We use two polarizers, P1 and P2
in the optical path before the UHV chamber to minimize the change in the polarization
state of the light incident on the sample. The maximum variation is reduced to 1.8° when
the polarization of the incident light is changed by 90° [22]. The PEM is oriented with its
stress axis 45° fi'om the direction of p-polarization. I also tried to keep the sample
position unchanged. Unfortunately, changing sample position is sometimes unavoidable.
The sample has to be moved up and down in taking TPD and in dosing. The sample
position is also believed to move slightly in the heating and cooling process.
Irreproducibility was the major obstacle in this system in getting good RAIRS spectrum.

Another major noise source in the polarization modulated RAIRS is fi'om a F abry-
Perot effect. The IR signal varies with IR fiequency in a cycle of l cm'l. This effect is
apparently caused by a component of 5 mm/n thick, where n is the reflective index. We
have not been able to locate this component. This noise is not very serious in the baseline
region while it is very serious in the peak region because of the large slope of the profile
there.

In optical alignment, we first set P1, P2 to pass only the p-polarized light. Then
the PEM stress amplitude was adjusted so that the detector wave form, as monitored on
an oscilloscope, was nearly sinusoidal at twice the stress oscillation frequency. The lock-
in A, referenced to twice the stress oscillation frequency of the PEM, is adjusted in phase

to give maximum signal. During RAIRS, polarizers P1 and P2 are rotated to null the

23

signal fiom the lock-in A, so that both s- and p-polarized light are incident onto the
sample with equal intensity. In principle, the PEM stress amplitude should be varied to
keep the optical phase modulation unchanged during the course of a scan. In practice, the
stress amplitude has been kept constant and the resulting variation in the optical phase
modulation is quite small because only a very small frequency range is scanned.
2.3.1.1 Quantitative analysis of polarization modulated RAIRS

Quantitative analysis method for polarization modulated RAIRS has been
developed by Lambert [22]. The material in this section is largely based on refs. 22 and
30.

The objective of RAIRS is to measure the effect of adsorbate on the reflectance of

p-polarized light, Rp- The corresponding physical quantity can be denoted as [22]
AR R p withCO — R p withoutCO
R ' RpwithoutCO

 

(2.3.2)

First consider the effect of the PEM and P3 on the transmitted light. Let Is ( I], )be
the transmitted intensity when the PEM is turned off and P3 is set to pass only the s-
polarized (p-polarized) light. Let A be the ellipsometric phase difference between the
optical E-fields of the s- and p—polarized light incident on the PEM. Let ¢(t) be the
optical phase difference induced by the PEM to the component of transmitted light
polarized along and perpendicular to the stress axis. The PEM is assumed to be oriented
with its stress axis 45° from the direction of p-polarization. The intensity I(t) transmitted
with P3 set to pass p-polarized light is [22]
19, + I,

2 . I. ,1’ c.4¢(r)]+JZIZI “(Manson (2.3.3)

with polarizers P1 and P2 set to pass only the p—polarized light, Eq 2.3.1 becomes

1+(t)=lp ligzm

Let fM be the stress oscillation frequency of the PEM. The wave form of 1+(t) is most

I(1')=

 

(2.3.4)

closely sinusoidal at 2 fM if the stress amplitude of the PEM is chosen so [22]

24

Mt) =1rcos(21th t) (2.3.5)

Setting the reference phase of the lock-in A (at 2fM ) to obtain maximum signal and P1
and P2 to pass only p-polarized light makes the lock-in only sensitive to the cos(41rfM t)
Fourier component of the signal, since the orthogonal reference phase gives

j9‘cos[n cos(x)]sin(2x)dx = 0 (2.3.6)

With the lock-in phase and PEM amplitude set this way, the lock-in output is
independent of A even if P; and P2 are rotated to some other angle because

[' sin[n cos(x)]cos(2x)br = 0 (2.3.7)
Consequently, the lock-in A output is proportional to Ip-Is, the difference in reflectivity
between p- and s-polarized light incident on the PEM.

Four scans are necessary to get the quantitative RAIRS spectrum. Two of them
are for calibration purposes and are not sensitive to the surface condition. The essential
scans include that of the clean surface and of the adsorbate—covered surface.

Let V(2fM) be the rms voltage measured by the lock-in B. Similarly, let V(fc) be
the rms voltage measured by lock-in C, referenced at the chopper frequency fc- Before
beginning, with the laser operated near the center of the fi'equency range, polarizers P1
and P2 are set to null V(2fM ). During the scan, both V(2fM) and V(fc) are measured.
At optical frequency v, for the scan on clean surface, define Q1(v) to be the ratio,

Q,(v)= V(2fM)/ V(f9.) (2.3.8)

From the measured V(2fM) and V029) taken during the scan of the adsorbate covered
surface, we can similarly define Q2(v).

Polarizers P1 and P2 are set to pass only the p—polarized light in the calibration
scans. During one of them, A, both V(2fM) and V(fc) are measured as in the surface
sensitive scans and Q A(v) is defined similar to Q1(v) . During the other scan, B, the
PEM is turned off, and only V(fc) is measured. QB(v) is defined as the ratio of V(fc)

measured during the two calibration scans,

25

Q.(V) = V.(fc) / Va(fc)° (2.3.9)
The quantitative polarization modulated RAIR spectrum is then
Q - Q
Sm= ’ ‘. (2.3.10)
QAQB

The above formalism has been shown to give AR/R correctly in ref [22].

2.3.2 Electroreflectance Vibrational Spectroscopy (EVS)

In RAIRS, the surface reflectance change induced by the adsorbate is measured.
In EVS, the surface reflectance change induced by an applied E-field is measured. By
comparing the two measurements, one can measure the Stark shift of the vibrational
mode of adsorbate on the surface. There should be no difference in reflectance outside
the absorption region, since only the vibrational mode of the adsorbate responds to the
applied E-field. Consequently, EVS has zero background.

The primary effect of the applied E-field is to shift the frequency of the
vibrational mode by a value on the order of 10'3 cm'1 , under my experimental
conditions. Because the shift is so small, EVS is basically a derivative technique, and
slow changes in the RAIRS background will not show up. Only the absorption peaks with
narrow line shape are observable. The Stark tuning rate dv/dE, where v is the vibrational
frequency and E is the applied field, for adsorbed molecules can be deduced from the
comparison of RAIRS and EVS. The principle of EVS is illustrated in Fig. 2-9.

It is particularly important that a diode laser optical source is used in EVS.
Because of the small fi'equency shift, the fi'actional change in the reflected signal intensity
is on the order of 10'6 for CO on Ni or Pt, which are among the strongest lines studied in
surface science. A conventional IR source ( thermal black body source) is not bright
enough to achieve this level of sensitivity.

The experimental setup for EVS is shown schematically in Fig. 2-10. The E-field

is produced by an oscillating voltage across the gap of about 0.5 mm between the sample

26

and the facing electrode. The strength of the field is about 3x104 V/cm. The breakdown
field in vacuum is about 105 V/cm [26]. Light reflected from the sample is modulated
both by the mechanical chopper and the applied E—field. Lock-ins A and B are referenced
to the oscillating field and to the chopper, respectively. The ratio of the two lock-in
outputs is proportional to the reflectivity modulation induced by the applied field. The
polarizers P1 and P2 are set to pass only the p-polarized light during EVS.

During the alignment, light from the laser is focused to the point on the sample
where the E-field is strongest. As the first step in alignment, a visible laser beam,
collinear with the IR beam, is used to determine the angle of incidence and to make sure
that the IR beam is correctly focused.
2.3.2.1 Quantitative analysis of EVS

Quantitative analysis method of EVS has been developed by Lambert [22]. The
material in this section is largely based on refs.22 and 30. The effect of the applied E-

field to the surface reflectivity, Rp, of p-polarized light can be described as [22]
rms variation of RI, caused by E

E _ RP without E

 

(2.3.11)

Let fig be the frequency of the ac electric field applied to the surface, 100 kHz in
this study. The rms voltage V013) measured by lock-in A, referenced at fE, is

V(f.) = BI(f.)D(f.)T(f.)cos(6.). (2.3.12)
where B is the time average of the fraction of incident power transmitted by the
mechanical chopper. 101;) is the rms modulation at frequency f; of optical power incident
on the detector when the beam is not blocked by the chopper and D(fE) is detector
responsivity defined as (rms output voltage)/(rms optical power modulation), T(fE) is the
voltage transfer function of the circuitry between the detector and the lock-in amplifier,
and 85 is the difference between the lock-in reference phase and the phase that would give

the maximum output.

27

Let fc be the fi'equency at which the mechanical chopper interrupts the light. The
rrns voltage V(fc) measured by lock-in B, referenced at fc is [22]

V(fc)= [OGDQC)T(fC)cos(6 C). (2.3.13)
Here Io is the power incident on the detector when the light is not blocked by the chopper,
G = (rrns optical power modulation at fc caused by the chopper)/Io, 8C is the deviation of
the lock-in reference phase from the phase would give the maximum output. From the

above equations,

_1({.)_ Gocr.)rcr.)cos(a.r(r.) (2,9,4,
10 BB“ E )T (fr )°°s(5 E Vac?)

For the chopper used in this study, B=0.50, G=£ = 0.45 [22]. D(f) has been measured
at

 

by Lambert to be well fit by [22]

DE, A
) f1§+f2 ( )

where to 90% confidence, 169 kHz < fD < 220 kHz, here fl) is the detector roll-off
frequency. In this work, D(fC)/D(/i3) = 1.278 :1: 0.044. The ratio T(fC)/T(fi.;) is measured
by replacing the detector with an attenuator with the same output impedance and
comparing the nns voltage input with the rms voltage output of the network at fc and fig.
The measurement gives T(fc)/T(fi.;) = 4.4 x 107, accurate to within 5%. 65 and So can be
measured by comparing the signal output with the phase set as for the spectrum, with the
signal output with phase changed by 90°. In most situations, cos(5C)/cos(5p9 ) z 1. Then
the equation (2.3.14) can be effectively rewritten as

s, = 5.1x 10'7— V05) (2.3.16)

V(fc)
Examples of EV spectra can be seen in Chapter 3.
2.3.3 Determination of the Stark tuning rate
The Stark tuning rate can be deduced fi'orn the comparison of RAIR and EV
spectra. Let A(v, E = 0) be the quantity measured in RAIRS,

28

A196’)
R(V)
let AE be the quantity measured in EVS. The relationship between A5 and A(v, E) is [27]
A9 _ dA(v,E)_ aA(v,o) dvo + aA(v,E)
(12)" dB ' av dE 615'

A(v,E= o)= (2.3.17)

 

 

(2.3.18)

where Va is the resonant frequency of the adsorbate. The first term is the change in A(v,
E = 0) due to a vibrational frequency shift caused by the applied field, the second term is
the change in the intensity of A(v, E = 0) caused by the applied field. In the small E limit,
the above equation can be evaluated at E = 0. It is easier to compare the two terms in

equation (2.3.18) by integrating it over frequency:

[45(V'VV'_(dv,

a 9 9
(E) — 71E.)590A(V’E‘0)+5§JVA(V,E)JV IM. (2.3.19)

For CO on Ni and Pd, at fiequencies near resonance, the second term in equation (2.3.19)

has been shown to be 50 times smaller than the first term [27,28]. So the direct effect of

 

the applied E-field to the RAIRS intensity is negligible. Taking away the second term,
(2.3.18) changes to

[Agdv'zA(v,E=o %)(E)=%(E)(?E°). (2.3.20)

 

it is then clear that the RAIR spectrum is proportional to the integration of the EV

spectrum. The ratio between them is the product of the nns E-field and the Stark tuning

 

rate (:25) . Knowing the rrns field, it is then straightforward to calculate the Stark

tuning rate fiom the peak heights of the RAIR and integrated EV spectra. A second
method can also be used: integration of both side of equation (2.3.20) means the
integrated intensity of the RAIR spectrum multiplied by the Stark tuning rate and the rrns
applied field equates to the double integration over the EV spectrum.

In principle, either method should give the same Stark tuning rate. In practice,
variation in the baseline of the RAIR spectrum, and the fact that the baseline of the

integrated EV spectrum is usually asymmetric due to missing data points in the laser gaps

29

near the peak region, introduces error into the determination of the peak heights and
integrated intensities. Both methods have been used here to calculated the Stark tuning
rate. The reported results are always the average of the two methods.
2.3.4 Measurement of the E-field

The rms applied E-field must be known in order to determine the Stark tuning
rate. The field is produced by applying high voltage across the gap between the sample
and a counterelectrode. The rms voltage is measured during each EVS scan. To measure
<lv>, one must know several other quantities. The electric field strength varies with the
position on the sample face, as does the intensity of the beam. In the good approximation
of linear response, the reflectivity modulation caused by the actual electric field
contribution is the same that would be produced by an uniform "effective field". This

"effective field", <lv> can be written as [27]
] EIdA

(E ) W (2.3.21)

Where I is the power incident on the detector per unit reflecting area of the sample, and E
is the externally applied field. Both I and E are functions of position on the sample.

The laser intensity distribution, 1, can be measured by moving the focus of the
light outside the UHV chamber. One then compares light detected after focusing
through a 500 um diameter pinhole with that detected with the pinhole removed. In the
approximation of a Gaussian beam and geometrical optics, the fraction of incident power
transmitted through the pinhole can be used to calculate the intensity distribution on the
sample. Two other methods have been used previously [27], and both give similar
results.

The E-field distribution on the sample was calculated using the method of images
[29]. In doing so, the sample surface was approximated by an infinite plane. The gap
between the sample and the counterelectrode, which is necessary in this calculation, is

determined by measurement of a series of capacitances between the two using a

30

capacitance bridge. Detailed discussion of the capacitance measurement is given by Luo
[30].

2.4 TPD

TPD was used in this study primarily to determine relative coverages of
adsorbates on the surface, and the distribution among different chemisorbed phases
(sites). I have also used TPD to study the mechanisms and products of certain chemical
reactions. TPD can also supply information about the desorption activation energy and
preexponential factor of desorption. Quantitative analysis methods of TPD are given in
several reports[8,9], and the discussion here is largely based on them.

In TPD, the sample covered with an adlayer is heated fi'om a low temperature to a
higher temperature, usually but not necessarily at a constant heating rate. In this study,
the heating rate is always 10 K/sec. As the sample temperature increases, the desorption
rate increases and that causes the pressure in the chamber to increase. The adsorption of
admolecules on the surface in vacuum is not under equilibrium, as desorbed gases are
pumped away but no more adsorption occurs. The desorption flux is a product of
desorption rate, which is a function of the sample temperature, and the population of that
particular chemisorbed phase. The desorption rate increases with the sample temperature
while the coverage decreases with the sample temperature, as a result, a peak in the
partial pressure appears at certain temperature, and the activation energy can be
calculated from the profile. I will give the formalism in detail later in this section.

The experimental setup for TPD experiments is shown schematically in Fig. 2-11.
For the experiments performed in the IR chamber, the sample was not moved fiom the IR
position, which is approximately at the center of the chamber and facing roughly 100°
away from the mass spectrometer. For the experiments performed in the HREELS
chamber, the sample was moved to face the mass spectrometer at a distance of about 1 cm

away from the aperture of the nozzle covering the mass spectrometer. Due to the

31

proximity of the sample and the mass spectrometer, and because the desorbed gases go
into the nozzle, which has much smaller volume than the whole vacuum chamber, the
sensitivity of TPD is greatly enhanced. Since the sample is facing a small aperture, the
desorption fiom the heater wire, edge and back of the crystal is not picked up by the mass
spectrometer. The reason why this scheme was not adopted in the IR experiment is
because the baseline of the IR spectrum was extremely sensitive to the sample position
change. This is a major problem in RAIRS, which was discussed fully in section 2.3.1.

It has been shown [31] for a system with constant pumping speed, the relationship

between the partial pressure profile and the desorption rate is

 

P —P 611’
d": V [ V “'+ a], (2.4.1)
(17’, Arr, [31: an;
and
(II) .. _
dn =v0 n exp( Ed/kTs), (2.4.2)

 

47?. l3
where n is the molecular concentration on the surface, T, and T8 are the temperatures of
the surface and the gas, V is the volume of the system, A is the area of adsorption, ch
and P,y are the equilibrium partial pressure and the instantaneous pressure of the system,
B is the heating rate and 1: is the pumping time constant, B, is the desorption activation
energy, and v0“) is the prexponential factor for a desorption of order m. The order of
desorption depends upon the limiting process of desorption. For example, CO adsorbs on
Pt in molecular form, the desorption is first order, while adsorbed hydrogen on Pt(l 1 l) is
in atomic form, the two H atoms have to come together before they can desorb, this
desorption is second order.
Define

P = P,y - Peq, (2.4.3)
which is plotted vs. T, as the desorption spectrum. Combining the above two equations

we get,

32

 

dP +=v£ vg~>n~ exp(-E, /kT9) P
” +5": Bno m,

AkT n
where PM = g 0

(2.4.4)

 

, is the maximum pressure observed during a desorption

measurement in a closed system (1: = infinity) for an initial surface coverage of no. The
parameters of desorption like Ed, v0, order of desorption m and relative population of
admolecules originally on the surface can be obtained by fitting the experimental
spectrum with this equation. The second term in the left hand of (2.4.4) is usually much
larger than the first term and the latter can be neglected. In this case, 57" cc P , and the

3

amount of admolecules desorbed is then proportional to [:1 P6”; , where T1 and T2 are

the starting and stopping sample temperatures, respectively.
If there is more than one chemisorbed phase, then equation (2.4.4) changes to

_9__ Z vt7’n.'exp(—E.. W.)
W I31 1 final

 

P], (2.4.5)

where i is the label for each individual phase. From the resolved phases we can then
usually assign sites for the molecules. Differences between the activation energies,
molecules changing phase during the desorption process, the pumping speed of the
system and the heating rate can all influence whether two different chemisorbed phases
are resolved in TPD. The system parameters are almost unchangeable, so the most
important factors are the activation energy and site movement. For example, atop and
bridge CO cannot be distinguished in TPD. The activation energy difference between
the two is about 1 kcal/mol (l kcal/mol = 0.043 eV), and some bridge CO first changes to
atop CO before desorbing. On the other hand, edge and terrace CO on Pt(335) can be
clearly resolved, as their activation energies differ by about 6 kcal/mol and terrace CO
doesn’t move to edge sites before desorbing ( because all the edge sites are still

occupied).

33

In Fig. 2-12, I have plotted a TPD spectrum for saturation coverage of NO on Pt
(335). The dotted lines are fit with Gaussian peaks, which give almost identical relative
concentrations to equation (2.4.5). Oxygen and CO were used to verify that the NO peak
near 220° C (500 K) is from the edge. When the edge sites were covered with oxygen or
CO before exposing the sample to NO, the NO desorption peak near 220° C disappears
while all three other peaks remain. HREELS experiments also found that the terrace
peaks do not correspond individually with different NO sites. Actually all the terrace NO
stays at the same sites before the desorption begins [32].

TPD alone cannot decide the absolute coverage of adsorbate on the surface. In
practice, Low Energy Electron Diffraction (LEED) and other techniques are usually used
to obtain absolute coverage information at certain particular coverages. The TPD spectra
from these coverages are then used as references. By comparing other TPD spectra with
a reference spectrum, one can then infer the absolute coverage of other situation. For CO
on Pt(] 1 1), a (J3 + J3)?30° LEED pattern is observed for the maximum atop CO
coverage of 0.33 ML, and a c(4x2) pattern is observed for the saturation coverage of 0.50
ML [33] which is shown in Fig. 2-13. All other coverages can be derived by comparing
the TPD spectrum with the TPD spectrum fiom one of these two coverages.

Another method of getting absolute coverage is to use the absolute coverage
information of another adsorbate and calibrate the mass spectrometer sensitivity between
the two through one or more chemical reaction. This method involves larger errors than
the above method.

2.5 Sample preparation and characterization

The Pt(335) sample was mounted in two separate UHV chambers in the IR and
HREELS experiments. The surface was oriented to within 05° from the (335) direction,
verified by Laue X-ray diffraction. In both chambers, the sample was spot welded to two

Ta wires, which were used for heating and cooling. Auger spectroscopy was used to

34

check the surface cleanness. Common contaminants found on the sample were C, Ca,
and O. C was usually removed by heating the sample in an oxygen environment.
Chemisorbed oxygen could be removed by heating the sample to about 1100 K. Ca and
oxide were removed by Ar ion sputtering at 300 K. After sputtering and oxygen
treatments, the sample had to be annealed at high temperature to retain the surface
morphology and remove small amounts of oxide. Care was taken so that C and Ca
contaminants in the bulk do not move to the surface during the annealing. The balance
between heating and oxygen treatment is very delicate. Sample cleaning is more of an art
than a standard procedure. Detailed procedures used in the IR and HREELS experiment
will be discussed in Chapters 3 and 4.

Dosing of the sample was usually done by leaking gas into the chamber through a
controlled valve (leak valve), raising the chamber pressure. It is important to keep the
chamber pressure low and the dosing time short, especially when dealing with species for
which the pumping speed is low. A disadvantage of dosing this way is that other species
adsorbed on the chamber walls may exchange with species we are interested in and result
in an overlayer composition different from that expected. One way of reducing dosing
pressure and time and avoiding the exchange problem is to use a doser which can
produce much higher pressure in a much smaller volume than the total chamber voltune.
This way, residual pressure will not be high and with the sample very close to the doser,
the exchange is minimal. The exposure can also be measured much more accurately this
way because the time needed to move the sample close to and away from the doser is on
the order of 1 second, much shorter than the time constant for the pressure to become
stabilized, which is in the order of 1 minute. Detailed calculations for different doser
designs are given in ref. 35. In the IR experiments, CO and hydrogen were dosed by back
filling the chamber. Oxygen was dosed through a diffusive doser which has an
enhancement factor of about 20. In the HREELS experiments, all the gases were dosed
through individual closers with enhancement factors close to 100.

35

 

NHFf/H/HHD’H/

Figure 2-1. Dipole moment perpendicular to the metal surface is reinforced,
dipole moment parallel to the surface is screened.

36

   

 

Figure 2-2. Block diagram of an electron energy loss spectrometer (Luo [30]).

37

?“ ELECTRON cegecron

ANALYZER

 

 

 

II I
E MAGNETIC
; SHIELD

 

 

 

 

 

    

 

 

 

INPUT
.ILENSES

 

(CRYSTAL

EXIT
“LENSES

 

ELECTRON SOURCE MONOCHROMATOR

Figure 2-3. Schematic diagram of the system for high resolution electron energy

loss spectroscopy in this work (B. A. Sexton [34]).

38

 

 

 

 

 

 

 

 

2,, (a) 99
2100 cm

1.5-1 5p),

i.0«_.,_- - _ _ for! ,
.\\

0.5 ‘-

fi ES"€.\-
I 1
(b)

30-1

20-

104
V l

o 30 so 90

Angle ct incidence,6ldeq

Figure 24 The surface electric field E/Eo in (a), and the quantity (E/Emz/cosB in
(b) for platinum at 2100 cm-1 ( e = -375 - 200 i) as a function of the incidence
angle 0 ( Bradshaw and Schweizer [2]).

39

 

 

 

 

 

 

 

 

 

g —— Sample with CO
'3 ----- Clean sample
a
C
0)
{z
I—
<1)
3%
Frequency
Subtract the intensity of
surface without CO from
with CO.
U
u
a _
i 0 '
“25
Q
G
O
'3
8
"B
:4
Frequency

Figure 2-5. The process of a RAIR spectrum: comparing the scans- with clean
sample and with adsorbate (CO) (Luo [3 0]).

40

 

 

 

 

OAPt

Closed Cycle

Retngerator

 

 

 

Optical system

Figure 2-6 Optical set up for RAIRS and EVS showing detector D, sample S, and
electrode E, polarizers P1 and P2 (Lambert [27]).

41

 

 

 

 

HP9825 GEE ' ‘
Computer ' He-Ne

Laser

 

 

Scope
M HngTe L]—
Detector

! ZnSe 0 f
A V r- 1
BS Si Counted @2
B S Timer
ZnSe o
r ‘ f

35 Si
Detector

OAP

 

 

 

 

 

 

 

 

 

Stopping Motor

Figure 2-7 Schematic diagram of computer-controlled wavemeter for laser
frequency calibration. Back-to-back hollow corner cube reflectors are mounted
on a ball slide translation stage driven by a stepping motor. The stage moves
freely except near the end of its travel where a beaded chain coupling becomes

rigid and reverses its motion. (Evans and Lambert [21])

42

 

Electrode
P1 P2 Detector

Diode Laser
Sample
Chopper in
(A) (B)

 

 

 

 

 

 

 

Figure 2-8. Schematic of the system used for polarization modulated reflection
absorption infrared spectroscopy (RAIRS) (Luo [30]).

43

 

AR/ R

/2’\

With E0 Without E0

 

 

 

Frequency

Subtract the intensity of

E0 forum 50-0.

 

5:
O

 

 

 

 

 

Frequency

Figure 2-9 Process of measuring the EV spectrum by modulating the E-field
applied to the Stu-face (Luo [30]).

Electrode

 

PEM P3

P1 p2 Detector

 

 

 

Diode Laser
Sample

 

Chopper
out

100 KHz
High Voltage
AC source

 

ref

 

Figure 2-10 Schematic of the system used for electroreflectance vibrational
spectroscopy (EVS) (Luo [30]).

45

Mass-
Spectrometer

\_ _/

Sample
Heater '——’WW‘——W

 

 

 

 

 

 

 

I Thermo-
Power L__®__ \/ couple
Supply “’ /\ Signal ‘

Temperature ___.
Controller

 

 

 

 

 

 

 

 

Figure 2-11 Schematic of the system used for temperature programmed
desorption (TPD) (L110 [30]).

46

 

 

 

 

 

0 100 200 300 400 500 500 700
T (K)

Figure 2-12 TPD spectrum for saturation coverage of NO on Pt(335). The
highest peak is from edge NO and the rest are from terrace NO. All the terrace
NO stays at identical sites before heating. ( Wang et a1. [32])

47

Atop CO

 

ll
A
V V

 

 

Y
A

 

 

Y Y

I

 

 

\

C(4x2) Bridge CO

Figure 2-13 c(4x2) structure of 0.5 ML CO on Pt(l 11) (Luo [30]).

48

Reference

1. B. E. Hayden, in Vibrational Spectroscopy of Molecules on Surfaces, edited by J .T.
Yates, Jr. and T. E. Madey (Plenum, New York, 1987).

2. A. M. Bradshaw and E. Schweizer, in Spectroscopy of Surfaces, edited by R. J. H.
Clark and R. E. Hester ( John Wiley & Sons, New York, 1988).

3. N. R. Avery, in Vibrational Spectroscopy of Molecules on Surfaces, edited by J .T.
Yates, Jr. and T. E. Madey (Plenum, New York, 1987).

4. M. A. Chesters and N. Sheppard, in Spectroscopy of Surfaces, edited by R. J. H. Clark
and R. E. Hester ( John Wiley & Sons, New York, 1988).

5. H. Ibach and D. L. Mills, Electron Energy Loss Spectroscopy and Surface Vibrations
(Academic, New York, 1982).

6. J. W. Gadzuk, in Vibrational Spectroscopy of Molecules on Surfaces, edited by J .T.
Yates, Jr. and T. E. Madey (Plenum, New York, 1987).

7. P. Hollins and J. Pritchard, Prog. Surf. Sci. 19, 275 (1985).

8. P. A. Redhead, Vacuum 12, 203 (1962).

9. C. -M. Chan, and W. H. Weinberg, Appl. Surf. Sci. 1, 377 (1978).

10. G. Kisters, J. G. Chen, S. Lehwald, H. Ibach, Surf. Sci. 245, 65 (1991).

11. Vacuum Science Instruments, Delta 0.5 UREELS

12. For example, see C. J. Hirschmugl, G. P. Williams, P. M. Hoffmann and Y. J. Chabal,
Phys. Rev. Lett. 65, 480 (1990) and ref. 15.

13. Ph. D. thesis, Chilhee Chung, Michigan State University, 1993.

14. W. Ho, R. F. Willis, and E. W. Plummet, Phys. Rev. Lett. 40, 1463 (1978).

15. C. J. Hirschmugl, Y. J. Chabal, F. M. Hoffmann, and G, P. Williams, J. Vac. Sci.
Technol. A12, 2229 (1994).

16. B. N. J. Persson, and A. I. Volokitin, Surf. Sci. 310, 314 (1994).

17. K. C. Lin, R. G. Tobin, P. Dumas, Phys. Rev. B49, 17 273 (1994).

18. C. L. DiMaggio, T. E. Moylan, private communication.

49

19. S. A. Francis and A. H. Ellison, J. Opt. Soc. Amer. 49, 131 (1959).

20. R. G. Greenler, J. Chem. Phys. 44, 310 (1966).

21. W. J. Evans, and D. K. Lambert, Appl. Opt. 25, 2867 (1986).

22. D. K. Lambert, Appl. Opt. 27, 3744 (1988).

23. K. W. Hipps, and G. A. Crosby, J. Phys. Chem. 83, 555 (1979).

24. W. G. Golden, D. S. Dunn, and J. Overend, J. Catalysis 71, 395 (1981).

25. L. F. Sutcu, J. L. Wragg, and H. W. White, Phys. Rev. B41, 8164 (1990).

26. P. A. Chatterton, "Vacuum Breakdown," in Electrical Breakdown of Gases, J. M.
Meek and J. D. Craggs, Eds. (Wiley, New York, 1978), Chap. 2.

27. D. K. Lambert, J. Chem. Phys. 89, 3847 (1988).

28. D. K. Lambert, J. Chem. Phys. 94, 6237 (1991).

29. W. R. Smythe, Static and Dynamic Electricity, 3rd ed. ( McGraw-Hill, New York,
1968), pp. 131,132.

30. J. S. Luo, Ph. D. thesis, Michigan State University, 1992.

31. L. D. Schmidt, Catal. Rev.-Sci. Eng. 9, 115 (1974).

32. H. Wang, R. G. Tobin, G. B. Fisher, C. L. DiMaggio, and D. K. Lambert,
unpublished

33. H. Steininger, S. Lehwald, and H. Ibach, Surf. Sci. 123, 264 (1982).

34. B. A. Sexton, J. Vac. Sci. Technol. 16, 1033 (1979).

35. DE. Kuhl, and R. G. Tobin, Review of Scientific Instruments, to be published, and
references therein.

36. J. L. Gland, B. A. Sexton, and G. B. Fisher, Surf. Sci. 95, 587(1980).

50

Chapter 3

Coadsorption of hydrogen and CO on Pt(335): structure and
vibrational Stark effect

1. Introduction

The material presented in this Chapter is largly based on our paper published in
the Journal of Chemical Physics [1 12].

Important applications involve CO and H coadsorbed on Pt surfaces fi'om a
gaseous ambient, [1,2] and these have stimulated a variety of experimental studies in
vacuum. Our experiment studies CO and H coadsorbed on Pt(335) using it spectroscopy
(electroreflectance and polarization modulation), temperature programmed desorption
(TPD), and low energy electron diffraction (LEED). As shown in Fig.1-l of Chapter 1,
the Pt(335) surface consists of (111) terraces, four atoms wide, separated by monatomic
(100) steps: Pt(S)[4(111) x (100)] in step-terrace notation. Adsorbed hydrogen
dissociates on Pt surfaces. [3 ,4]

One motivation to study a highly stepped surface like Pt(335) is to understand the
polycrystalline surfaces used in applications. Both CO and H preferentially bind at a step
edge. At the low CO coverages discussed here, CO occupies only edge sites. We are
also interested in how CO's response to electrostatic and ir fields is changed by
coadsorbed H. Our data for CO and H on Pt(335) in vacuum are compared with
spectroelectrochemical data obtained by Kim et al. [5,6] for CO and H on Pt(335) in
aqueous electrolyte.

There have been previous studies of CO on Pt(335) in vacuum, [7-14] but we are
not aware of any with coadsorbed H. However, studies of CO coadsorbed with H on
Pt(112) [15] and Pt(997) [16] in vacuum have been reported. Both surfaces are vicinal
to (111) and differ from Pt(335) mainly in terrace width. In step terrace notation Pt(112)
and Pt(997) are Pt(S) [3(111) x (100)] and Pt(S) [9(111) x (111)], respectively. Bridge

51

CO coexists with atop CO on Pt(335) [10-13] and on Pt(997) [16] over a wide range of
CO coverage, but bridge CO is found on Pt(112) only near saturation CO coverage. [17]
In the present work we find another difl'erence: on Pt(112), H causes low-coverage CO to
phase separate into one-dimensional islands along the step edge, but on Pt(335) H and
CO form a mixed phase along the step edge.

This paper is organized as follows. We first discuss the experiment and our
results. Next, a structural model for CO and H on the step edge is proposed that explains
our observations. This is followed by a discussion of how E-field and coadsorbates affect
CO's vibrational spectrum. We consider both the vibrational Stark effect and chemical
explanations. We also compare our observations with previous electrochemical data.

2. Experiment

Our experiments were carried out in an ultrahigh vacuum (IJHV) chamber with a base
pressure of 2x10'10 torr. The sample was spot-welded to two Ta wires, which were also
used for heating and cooling. The sample temperature could be controlled between 100
and 1400 K. The sample was cleaned by cycles of Ar ion bombardment, reacting at 1000
K with 2x10'8 torr Oz, and annealing at about 1300 K. The sample's cleanliness was
checked by Auger spectroscopy before any ir spectra were taken. Also, to minimize the
adsorption of residual hydrogen on the sample surface, both the cold trap at the bottom of
the UHV system and the reservoir of the sample manipulator were filled with liquid N;
before the sample was allowed to cool below 300 K. Cryopumping by the cold surfaces
reduced the H2 residual gas pressure by about a factor three. The sample temperature was
kept at 105--110 K during dosing and data taking. The sample was dosed with CO or H;
by simply leaking the gas into the chamber.

Detailed descriptions of the spectroscopy techniques are given elsewhere. [18] A
single lead-salt diode laser, with a spectral range of 1947--2022 cm'l, was used as the ir
source for both reflection-absorption ir spectroscopy (RAIRS) and electroreflectance
vibrational spectroscopy (EVS). We used 13C180 for the experiment so the C = O

52

stretch mode of the atop CO fell within the tuning range of the laser; the frequencies
characteristic of bridge-bonded CO were not accessible. The RAIR spectra were obtained
using a photoelastic modulator to modulate the polarization of the light. The measured
quantity in RAIRS is the fractional change AR/R in p-polarized reflectivity (actually the
fractional change in the difference between p- and s-polarized reflectivity) induced by the
adsorbate.

The EV spectra were taken by applying an oscillating (100 kHz) high voltage
between the sample and a spherical counter electrode, which created an oscillating
electrostatic field normal to the surface. The measured quantity in EVS is 3;, the rms
amplitude of the induced oscillation in reflectivity to p-polarized light, normalized by
reflectivity. To interpret EVS spectra quantitatively, the applied field must be known.
The applied field depends on the applied potential and the sample-to-electrode distance.
The sample-to-electrode distance was determined by measuring the three-terminal
capacitance between them.

The data discussed here were obtained on three different days, each with a fixed
CO coverage. The angle of incidence of the light on the Pt(335) crystal was the same for
both RAIRS and EVS on a given day. On the three days it was 864°, 864°, and 859°,
and the rms static E-field at the surface was <E>=(3.1 :l: 0.2), (2.6 :l: 0.2), and (3.0 i
0.2) x 104 V/cm, ordered by increasing CO coverage. These values of <E> are the
average, weighted by the intensity of the focused ir beam, over the illuminated area of the
sample. [19]

The CO overlayer was prepared by dosing the sample at 105 K, annealing at 420
K to remove H adsorbed fi'om the background, and cooling back to 105 K; this procedure
removed more than 95% of the H fiom the surface, while desorbing approximately 8% of
the CO. During H2 dosing the sample temperature was 103--110 K and it was kept in this
range until the ir spectra with the highest H coverage had been taken. One EV and one
RAIR spectrum were measured for each H coverage; each pair of spectra took about 90

53

minutes. Afier the ir spectra were completed at the highest H2 dosage the sample was
heated to 420 K while the partial pressures of H2 and CO were monitored, to desorb the H
without removing CO. The sample was then cooled back to 103--110 K and the final
RAIR and EV spectra were taken.

We also performed experiments in which the sample was annealed to 198 K for
10 minutes after H; dosing. No significant difference was found between the ir spectra of
the annealed and unannealed layers. This shows that at the coverages studied, both CO
and H are sufficiently mobile at ~100 K to reach their equilibrium configuration, in
agreement with Luo et al. [11] The experiments of Henderson and Yates [15] with CO
and H on Pt(112) were done at 100 K.

The CO and final H coverages were detemiined with TPD after the last H2 dose,
referenced to the coverages obtained by dosing to saturation with CO or H; alone at 100--
110 K. The saturation coverage of CO on Pt(335) is 0.63 ML. [9] (Here 1 ML is the
coverage with an adsorbate on each surface atom of Pt.) The saturation coverage of H on
Pt(S)[9(111) x (111)] is 1.0 ML. [20] We assume that the saturation coverage of H on
Pt(335) is also 1.0 ML. The CO coverages on the three days were 0.06, 0.12 and 0.16
ML.

The other H coverages studied with it were determined in separate experiments by
repeating the CO and H2 dosing sequences and performing TPD for each dosage. The
background exposure to H2 (approximately 0.1 L) that took place during a pair of it
spectra (one EVS and one RAIRS) was accounted for in replicating the coverages. The
final coverages obtained by repeating the dosing schedule agreed with the post-ir
coverages. For example, with 1.5 L CO, the post-ir TPD gave 9C0 = 0.16 and 0n = 0.30
ML. Afier the repeated dosing schedule, TPD gave Goo = 0.15 and 0n = 0.29 ML. Here
000 and 0" are the coverages of CO and H, respectively.

We also used LEED to search for possible reconstruction of the Pt(335) surface
or the formation of ordered overlayers that might be caused by H and CO adsorption. No

54

change from the clean surface was detected in the LEED pattern. In contrast, Pt(100)
[21] and Pt(110) [22] do reconstruct.

3. Experimental results

3.1 TPD

Examples of TPD spectra for only CO on Pt(335) are shown in Fig. 3-1a. The
initial CO coverages were saturation and the three 9co studied with ir. The high
temperature TPD peak is from CO at step edges. [7] The low temperature peak is from
CO on the terrace. With saturation 900, our TPD spectrum taken at 10 K/s has peaks at
416 and 518 K, in agreement with previous studies. [7,11] To resolve the TPD curve
into an edge peak and a terrace peak they are modeled as Gaussians, as shown in Fig. 3-
1b. At saturation the edge peak is 40% of the total, in good agreement with Luo et al.
[11] who found 43% edge CO. The TPD spectra in Fig. 3-2a show that at the CO
coverages used for the ir spectra, without coadsorbed H, all of the CO was on the edge;
none was on the terrace.

TPD spectra for H alone on clean Pt(335) are shown in Fig. 3-2. These spectra
were also taken at 10 K/s. At low coverage, curve (a), there is only a single peak at 395
K. With increasing 0“, this high temperature peak stays fixed and a second peak appears
at about 315 K. The second peak shifts to lower temperature with increasing 0“. At high
0“ (0.8-l.0 ML) a third peak appears as a shoulder at about 250 K. At saturation On,
the fi-action of the total area under the low, intermediate, and high temperature peaks is
0.35, 0.39, and 0.26, respectively. With stepped Pt, the H2 TPD peak at highest
temperature is from chemisorbed H at edge sites. [23] Our data suggest that at
saturation, 1/4 of the H is at edge sites, consistent with OH = 1.0 ML.

There have been previous TPD studies of H on stepped single-crystal Pt surfaces:
Pt(S)[3(111) x (100)], [15,24] Pt(S)[6(111) x (100)], [25] Pt(S)[6(111) x (111)], [25] and
Pt(S)[9(111) x (111)]. [20] Surfaces with (100) oriented steps give H2 TPD spectra
with a high temperature peak that dominates at low 0“ and a lower temperature peak

55

with a shoulder, that increases in area with increasing OH. Surfaces with (111) oriented
steps give H2 TPD spectra that are more difficult to separate into a ‘ ‘step" and ‘ ‘terrace"
contribution if the surface is well annealed. For example, the [32 peak and B] shoulder
seen for high-0H desorption fi'om Pt(S)[9(111) x (111)] [20] are very similar to the [32
peak and [3, shoulder seen for high-03 desorption fiom Pt(] 1 l). [3] For Pt(] 1 1)
Christmann et al. [3] have argued that the peak and shoulder both come from the same
state-"the peak is distorted because the binding energy of H varies with 0“.

It is interesting to compare the H2 TPD spectra of Pt(S)[n(111) x (100)] surfaces
among Refs. 15, 20, 24, 25, and our experiments. The peak desorption temperatures
seen in Refs. 20, 24, 25, and our experiments are generally consistent; the temperatures
seen in Ref. 15 are lower. In Refs. 20, 24, and 25 the TPD peak of H; from Pt(] 11) stays
at about 330 K even though the heating rates ranged fiom 10 K/s to 82 K/s. With low OH
on Pt(S)[n(111) x (100)], our experiments and those in Refs. 20, 24, and 25 consistently
find a H; TPD peak at 400-430 K. As 0" is increased, and H begins to occupy (111)
terrace sites, a second H2 peak appears at about 300 K. In contrast, Ref. 15 finds that for
low 0“ on Pt(S)[3(111) x (100)], the H2 TPD peak is at 309 K, even lower than for H on
Pt(111)in Refs. 20, 24, and 25. The same surface was studied in Ref. 24; at low 0" the
temperature of the H2 TPD peak was 120 K higher than in Ref. 15. Only a small part of
the discrepancy can be explained by the different heating rates used in the two
experiments (3.9 K/s in Ref. 15 and 67 K/s in Ref. 24); for a first order TPD peak, the
peak temperature should shifi about 35 K. [26]

As shown in Fig. 3-3, the TPD curves of H on Pt(335) are significantly changed
by predosing with CO. As initial 090 increases, the (in that results fi'om a given H2
dosage decreases. Figure 3-4, which shows edge site occupancy by H vs H2 dosage for
three values of 000, further illustrates this point. (The mass spectrometer signals of both

CO and H; were sampled during these desorptions.) The interaction between coadsorbed

56

H and CO is repulsive since increasing 9co monotonically reduces the temperature of the
H2 TPD peaks.

The data in Fig. 3-3 show that at constant On, an increase in predosed CO
transfers H from edge sites to terrace sites (compare curves a and d ). The TPD data
taken in conjunction with ir spectroscopy were also analyzed to determine the occupancy
of edge sites by CO and H. For all three CO coverages studied, after saturation with H,
the total occupancy of CO and H at edge sites was 1.1i 0.1. Each CO at an edge site
blocks one H from adsorbing at the edge.

3.2 ir Spectra

One set of EV and RAIR spectra (with em = 0.16 ML) is shown in Fig. 3-5. The
spectra taken at the other two CO coverages are similar. The resonant C-O stretch
vibrational frequency vs total OH is shown in Fig. 3-6, and the integrated RAIR intensity
S is shown in Fig. 3-7. Both the EV and RAIR spectra were used to determine the
resonant frequency. Smooth cubic splines were first interpolated through the data. Since
the EV spectra are proportional to d(AR/R)/dv, where (AR/R) is the RAIR signal and v
is optical fi'equency, they were next integrated. The plotted peak frequency is the average
fi'om the RAIR and the integrated EV spectra.

The Stark tuning rate (dv/dE) was determined by comparing the RAIR spectra
with the integrated EV spectra. Two methods were used: comparison of peak heights
and comparison of peak areas. The measured Stark tuning rate (dv/dE) vs 0;; is plotted
in Fig. 3-8 for each 000. The data show that (dv/dE) is independent of 0“, but
decreases with increasing 9C0- The scatter in the data in Figs.3-6--3-8 comes largely
from interpolation errors. Mode hops in the diode laser’ 8 tuning curve leave gaps about 2
cm'1 wide that are later filled by interpolation. When the spectrum has important
structure in a gap---a peak for example-«some information is lost. This is especially
serious for EVS.

4. Structural model of the C0 + H overlayer

57

Let us first recall what is known about CO adsorption on clean Pt(335). It is well
established that as CO's coverage builds up it occupies sites on the step edge first. Edge
CO has a thermal desorption peak near 520 K (at 10 K/s) and an atop v (for 12C160)
between 2065 and 2080 cm'l. As 9co continues to increase, terrace CO appears near 0.20
ML. Terrace CO has a thermal desorption peak near 420 K (at 10 K/s) and an atop v
between 2085 and 2100 cm'l. At saturation(0.63 ML) all of the edge sites are occupied
by atop CO, but on the terrace bridge and atop bonded CO coexist. [5,6,10-13] Bridge
CO at the edge is also present in an intermediate coverage range. A comprehensive
model of CO buildup on Pt(335) was proposed by Luo et al. [11]

The experimental evidence fi'om previous studies of CO and H coadsorption on Pt
surfaces in vacuum [15,16,27-3 8] points convincingly toward a strongly repulsive CO-H
interaction despite early claims [27,29,30] to the contrary. On Pt(l 1 1), even though the
CO-CO interaction is repulsive, CO is pushed into islands of high density pure CO as OH
increases. [35-37] On Pt(112), Henderson and Yates observed similar behavior. [15]
They used electron-stimulated desorption ion angular distribution (ESDIAD) to monitor
edge CO. The same sequence of structures was observed as 0;; was increased at fixed 900
as when 9co was increased with 0" = 0. This shows that H and CO form segregated
one-dimensional islands along the step edge. As 0“ increases the CO is compressed.

Suppose that coadsorbed H and CO on Pt(335) also formed separate islands. An
increase in 0" should have the same effect on CO's ir spectrum as an increase in em at
0n = 0. As SH increases we would expect to observe: (1) an increase in CO's resonant
vibrational fiequency v, (2) little or no change in S, and (3) a reduction in (dv/dE).
Instead we see (Figs. 3-7--3-9) almost no change in v, a strong reduction in S, ultimately
to zero, and little or no change in (dv/dE). Clearly, something different occurs on
Pt(335), despite the strong structural similarity to Pt(112).

Our results are consistent with the following model: (1) Each CO blocks one H
adsorption site. (2) H adsorbed at the edge forms compact one—dimensional islands of

58

mixed H and CO. Within the H/CO islands, atop CO shifis to an adjacent edge bridge
site. (3) Atop CO outside the H/CO islands is unaffected by 0“.

On polycrystalline Pt there is previous evidence that both a mixed phase and
islands occur. [33] The coadsorption of H and CO on single-crystal surfaces has been
reviewed by White [31] and again by White and Akhter. [39] A mixed phase of H and
CO has previously been observed on relatively open surfaces like Ni(100), [40,41]
Ni(110), [42,43] Fe(100), [44] and Rh(100) [45] near saturation coverage, but not on
close-packed surfaces or at low coverage on any single-crystal surfaces. Our observation
of a mixed phase for CO and H coadsorbed at a step edge, while unexpected, is generally
consistent with previous experience. Sites at the edge are in an open environment. Also,
even though the total coverage is below saturation the local coverage at the edge is still
high; even at our lowest coverage 1/4 of the edge sites are occupied by CO.

Assumption (1) follows from our TPD measurements (Sec. 3.1) which show that
at saturation 0“ there is one adsorbate (H or CO) per edge atom for all three 9co-

Assumptions (2) and (3) explain the ir spectra. Figure 3-7 shows that increasing
0“ strongly reduces the ir intensity of atop CO. Our model explains this: the CO is being
shifted to an ir-silent site. The limited timing range of our laser did not allow us to see
the CO at bridge sites in the present experiment, but a subsequent electron energy loss
spectroscopy (EELS) experiment [46] has confirmed that its coverage does increase with
03. The same efl‘ect has been observed in electrochemical experiments with coadsorbed
CO and H on Pt(335), [5,6] and similarly in vacuum experiments on Rh(100) [45] and
Ni(110). [42] Conversion of atop CO on the edge to bridge CO on the edge is also
plausible on energetic grounds: with low 900 on Pt(] 11), atop CO is only 0.45 kcal/mol
more strongly bound than bridge CO. [20]

Other explanations that we have considered for the disappearance of CO from the
ir spectrum as H is coadsorbed on Pt(335) are less plausible. The CO does not move to

atop sites on the terrace. Our ir spectra show that the terrace atop 9co < 0.008 ML in the

59

range of total 9co and 0H discussed here. The inability of H to displace CO fi'om edge to
terrace sites is consistent with the known binding energies of CO and H at the two sites:
edge CO is 5--8 kcal/mol more strongly bound than terrace CO [14,15] but edge H is
only 3 kcal/mol more strongly bound than terrace H. [23]

It is conceivable, but unlikely, that H causes a nearby CO to tilt nearly parallel to
the surface (at least 77° from the surface normal to account for the observed loss of
intensity). Tilted CO is commonly observed on stepped surfaces. [15,47-49] However,
an explicit search [8] for tilted CO on Pt(335) found it to be vertical to within 10°. Even
for CO on Pt(112), where CO does tilt, [15] the maximum tilt angle is only 20°.

Strong screening of the field at C0 adsorption sites within islands is also unlikely.
Generally, H adsorbs inside the image plane on metal surfaces. [50] On Pt(111) an
explicit calculation shows that H is adsorbed at 3-fold hollow sites 0.95A above the top
most Pt layer. [51] For CO on Pt(] 11) the distance from the topmost Pt layer to the
center of the CO molecule is 2.43A from LEED. [52] There are several ways to estimate
where the image plane is on the Pt(] 11) surface. [53] They all suggest that the center of
the CO bond is outside the image plane. We believe that edge atop CO on Pt(335) is a
comparable distance from the topmost Pt atoms. Thus the center of the CO bond is at
least 1 A above the H layer. Both experiment and theory suggest that coadsorbed H does
not significantly screen the local E-field at the, CO adsorption site.

Reduction of CO's vibrational polarizability or, by nearby H can be ruled out.
Since S is reduced fi'om its original value by at least a factor 20, and since (1,, ac (e*)2, the
dynamic dipole e" would need to be reduced by at least a factor 4.5 to explain the data.
However, the measured (dv/dE) is expected to be proportional to e*, so (dv/dE)
should drop by a factor 4.5 with increasing 91+ Instead, Fig. 3-8 shows that (dv/dE)
does not change by more than about 20%.

Figures 3-6 and 3-8, which show the resonant frequency and (dv/dE) for the atop
CO that remains ir-active, demonstrate that this CO is unaffected by coadsorbed H,

60

except at the very highest 03. In Sec. 5.2 we estimate the 0“ induced change in the local
density of ir-active atop CO to be at most 0.02 ML. Since the total 900 on the edge
remains constant with increasing 0", the average CO-CO distance in the mixed phase is
about the same as in the pure phase. This observation indicates that the mixed H/CO
islands are compact. A uniform or random distribution of H atoms would lead to a
gradual decrease of the frequency and a gradual increase in (dv/dE) as the average
distance between ir-active COs increased. Even though CO and H compete for sites, and
the pairwise interactions are repulsive, the equilibrium state on the step edge has two
phases-«compact mixed islands and unaffected pure CO regions---with the same CO
density in both.

Figure 3-9 provides further evidence that the local CO density is not affected by
H. It shows S as a function of OH, normalized to S at the lowest OH. For all three 0CD, a
given 0" eliminates the same fiaction of the initial intensity, regardless of the initial 9co-
If CO were expelled from growing H islands, more complicated behavior would be
expected: one slope at low 0” as the pure CO phase is compressed and a different, 9co -
dependent slope at high 0“ as CO is incorporated into the growing H islands.

It is surprising that a mixed CO/H phase occurs on Pt(S) [4(111) x (100)] but that
an island phase occurs on Pt(S) [3(111) x (100)] and on Pt(111). As discussed by White
and Akhter, [39] in a situation with only pair interactions, a mixed phase between species
A and B is energetically favored over an island phase if

1
3AB<'2‘ (3M + 833) (3‘1)

Here SM; is the interaction energy between A and B, and similarly for SM and 333.
Since H is inside the image plane, we do not expect its electrostatic interactions to be
important. Generally, the strongest interaction between H and a coadsorbate is mediated

by conduction electrons in the metal.

61

One explanation for the difference between the surfaces with three-atom and four-
atom wide terraces is that molecules on adjacent terraces interact significantly, and this
interaction changes the inequality in Eq.(3-1). In an extended Hiickel calculation for CO
and H on Rh(l 1 1), Ruckenstein and Halachev [54] showed that the through-metal
interaction has a different dependence on separation for the CO-CO, H-H, and CO-H
interactions. The length scale associated with the difference is a few lattice spacings.
Although the computed energy difference is small, it could explain the qualitative change
in behavior in going from three to four lattice spacings between edges.

Another explanation for the difference is that the local electronic structure near
the edge is different on the two surfaces (without adsorbates), and this alters the
interactions between adsorbates along an individual edge. A jellium calculation [55]
shows that a single step has an associated dipole. The E-field from a line dipole decays
as l/d2 where d is the distance fiom the line. The induced surface charge at the nearest
step consequently decreases by about a factor of two on going from Pt(112) to Pt(335). It
is plausible that the extra charge could affect the CO-CO, H-H, and CO-H interactions
differently, and this could change the equilibrium structure fiom a segregated phase to a
mixed phase.

5. Vibrational Stark effect and coadsorbates
5.1 Backgron

Our experiments directly measure the vibrational Stark effect: the effect of a
static E-field on a molecule's vibrational spectrum. The Stark effect with externally
applied E-field has also been studied theoretically. Quantum mechanics has been used to
express (dv/dE) for a molecule on a surface in terms of the molecule's dipole moment
and potential energy frmctions. [19,56] (Here E is the externally applied electrostatic
field.) The molecular properties needed for the calculation are measurable. There have
also been ab initio calculations of (dv/dE) for a single molecule on a surface [57-60] or

62

in spatially uniform E-field. [61-67] In the limit of low adsorbate coverage the
measurement-based and ab initio calculations of (dv/dE) agree, and both have
successfully predicted the directly measured (dv/dE). [10,53] With saturation CO
coverage on Ni(100) good agreement was also found. [19] With high coverage CO on
Pt(111) and Pt(335), however, our previous experiments have found discrepancies
between theoretical prediction and experiment. [9,10,53]

There are diverse experiments in which a change in static E-field affects v.
Examples include the ‘ ‘chemical" shift Avchcm vs adsorbate coverage for a homogeneous
layer, Av induced by a coadsorbate, and Av induced by varying the substrate electrode's
potential in an electrochemical cell. One of our motivations is to examine how well the
vibrational Stark effect explains such data.

In many experiments Av is proportional to the change in local static field Em :

dv
A = — AE . 3-2

For Eq.(3-2) to be useful, however, (dv/dEloc) must be relatively insensitive to the
environment, or at least the changes must be theoretically understood. For a single CO
molecule on a metal surface, for example, theory predicts that (dv/dE) is approximately
proportional to the dipole moment derivative e* so some account must be taken of the
molecular environment. [68] Since e“ can be estimated from EELS or it intensities, it
is relatively straightforward to take its variation into consideration. If other molecular
properties of CO varied strongly with environment they would be more difficult to
account for. There is evidence fiom an EELS experiment [14] that the important
molecular properties of CO at edge and terrace sites on Pt(335) are similar, lending
support to the usefulness of Eq.(3-2).

Coadsorption experiments provide strong evidence that CO's (dv/dEloc) is
relatively insensitive to the local chemical environment. Typically, experiments do find a

linear correlation between v and the estimated change in static E10, caused by a given

63

coadsorbate. [69-72] This is seen, for example, in studies that correlate the coadsorbate-
induced fiequency shift with the change in work function Ad). For CO on Pt(111) both
the chemical shift [73] Avchcm and the CO induced change in work function, [74-76] A4) ,
have been measured vs 9co- (Experimentally Avamm is determined by varying the
isotopic composition of the overlayer at constant total 9co- ) At a temperature of 125 K,
the measured Avchem decreases fiom 0 at 9co = 0 to -10 cm'1 at 9co = 0.33 and then
increases back to 0 at 0 = 0.5. Similarly, at 130 K the measured A4) decreases fi'om 0
at 9co = 0 to -0.295 :1: 0.038 eV at Geo z 0.33 ML and then increases back to 0 at 900 z
0.5 ML. (The 000 at which All» returns to zero is 0.50 ML in Ref. 75 but 0.40 ML in
Ref. 76.) Both Avflmn and Adi have the same functional form vs 9co- The
proportionality constant between them is (av/do ) = 34 a 4 cm" /eV.

Other studies have given similar results. A study [72] of CO on Ni(l 11) found
that v has the same linear dependence on A4» for coadsorbed O, CO, and Xe. The
observed proportionality constant was (dv/dtb ) = 35 cm'l/eV. A linear correlation of
(dv/d¢ ) ~ 45 cm'l/eV between Ab and v was found by Yamarnoto and Nanba [77] for
CO on Ag coadsorbed with Xe, Kr, O, and C1.

Electrochemical experiments with CO on Pt(111) measure a similar quantity. [78-
88] Here, the potential (I) of the Pt electrode relative to a reference electrode is
controlled directly while v is measured with it spectroscopy. Again, the data show Av cc
Ad). The measured (dv/dfb) depends on the solvent and solute of the electrolyte and
on 900- In aqueous 0.1 M HClO4, with 9co = 0.1 and 0.65 ML, (dv/d<D) = 45 and 30
cm'lN, respectively, for the spectral line of atop CO. The close quantitative agreement
between (dv/dCD) for the chemical shift of CO on Pt(111) in vacuum and (dv/dCD) for
CO at a Pt(111) electrode in an aqueous electrolyte suggests that they have a common
origin.

Electrochemical experiments have also tried to distinguish between AID and local
E-field as the controlling variable for Av by varying either the solvent [86,87 ,89-94] or

64

the solute [79,95-98] in the electrolyte. The experiments suggest that the controlling
variable is the potential drop across the CO molecule---the local E-field---not just (D of
the electrode.

However, other experiments suggest that (dv/dEloc) varies significantly with
local environment, and that Eq.(2), which focuses on Bloc, is less predictive than
explicitly considering the chemical change of the molecule. In a study of CO on Rh(l 11)
coadsorbed with Na, benzene, fluorobenzene, and ethylidyne, for example, Mate et al.
[99] found that with equal concentrations of CO and the coadsorbate, Av was
proportional to the coadsorbate's dipole moment. They also tried to estimate Bloc at the
CO adsorption site. The correlation between Av and AEloc was much worse than the
correlation between Av and the coadsorbate's dipole moment.

A previous comparison [53] of the effect of em on S and (dv/dE), for CO on
Pt(l 1 1), also called into question the assrunptions involved in using Em for predictive
purposes. As we discuss in Sec.5.2, S at Goo (7i, e*)2, where 7,, is an effective
screening factor, and (dv/dE) oc Yon, where he is the screening factor for the dc field.
Within dipole coupling theory we expect no ~ 7,, , so (S/Oco)1/2 and (dv/dE) should
have the same dependence on em even if e "' varies with coverage. Experimentally [53]
this is untrue for CO on Pt(] 1 1). A similar comparison for CO on Pt(335) is discussed in
Sec. 5.2.

Comparison between the Stark effect measured in vacuum and electrochemical
data also raises doubts about the transferability of (dv/dEloc) in Eq.(2) from one situation
to another. The vacuum experiments directly measure (dv/dE). At low 9co in vacuum,
screening (by both the adsorbates and the metal's conduction electrons) is expected to be
negligible so (dv/dE) should approach (dv/dEloc). For CO on Pt(] 1 1) in the limit of low
9co in vacuum, [53] (dv/dE) = 75 i: 9 cm'1 /(V/A).

In an electrochemical experiment the CO molecule is in the compact double-layer,

and a model of the double-layer is required to relate A Em to change in electrode

65

potential (D. There are many models of the double-layer. For an aqueous double-layer
in the limit of low 000, the model of Bockris et al. [100] implies that (dEm/dCD ) = 0.29
(V/A)/V. An alternative model---a CO monolayer-«implies [19] that (dEloc/dCD ) = 0.27
(V/A)N. If we assume that the applied E-field is unscreened (Yea = 1 or E = E10,, ) at
the CO molecule in the vacuum experiment then the Bockris model and the CO
monolayer model imply that (dv/d<D ) = 22 and 20 cm‘1 N, respectively, while the
measured value at low CO coverage is 45 cm'1 N. To use these models to explain the
(dv/de) observed in the electrochemical experiments we would have to assume that no
~ 0.5 at the CO adsorption site on Pt(] 11) in vacuum. This is too small to be explained
by presently accepted theory. [53] (Early analyses [19,68] that obtained better
agreement for the CO on Pt electrochemical experiments extrapolated from vacuum
measurements with CO on Ni surfaces. [19,101] )

These conflicting results lead us to consider theories that focus on chenrical
interactions between adsorbates, rather than on Elm. (There is no fundamental conflict
between these alternatives. In some situations they are tied together by the Hellman-
Feynman theorem. [102] ) One chemical theory for the effect of coadsorbates on v has
been proposed by Ueba. [103] (Similar theories [104-106] had been proposed earlier.)
Ueba starts with a Hamiltonian that includes the possibility of charge transfer between
CO molecules, and between a CO molecule and the metal. Let Av be the difference
between the observed v at 000 and the singleton frequency. It consists of two parts: Av =
Avail, + Avchem where Avail, is from dipole-dipole coupling and A vehem is from other
effects. One prediction is that A VMOC A¢ , the change in b caused by adsorbing CO
on the clean surface, consistent with the experimental results surveyed above. Another
prediction is that CO's vibrational polarizability orv increases with increasing 9co-

In the following sections we analyze our data for C0 + H on Pt(335) to see
whether it is best described by the vibrational Stark effect and Blue or by chemical
interactions. We find that the Stark effect theory accurately predicts (dv/dE) at low

66

coverage. However, the effect of coadsorbates is not readily explained by variations in
Bloc. We also find qualitative support for Ueba's prediction of a coverage-dependent
vibrational polarizability.

5.2 Our experiment

The Stark tuning rate (dv/dE) that we measure in terms of the externally applied
E-field is given in Table I. With only CO on the clean Pt(335) surface, our data
generally agree with previous UHV studies. [7-13] In particular, we find that (dv/dE) =
88 :h 9 cm'1 /(V/A) at low coverage (0.06 ML). In an earlier measurement at the same
coverage in Ref.9: (dv/dE) = 60 at 8 cm'1 /(V/A). There is also a theoretical relationship
between (dv/dE) and the measured ir cross section. [68] At this 000, theory predicts
[9] (dv/dE) = 70 i 11 cm'1 /(V/A). Theory and experiment are consistent at low CO
coverage.

Dipole-dipole coupling is expected to affect the peak frequency v, integrated
intensity S, and Stark tuning rate (dv/dE). Our data at the three CO coverages are
shown in Table I. The variation of v with 9co is largely explained by dipole-dipole
coupling.

We find that v increases with em at 70 :l: 5 cm'1 /ML. For comparison, Hayden
et al. [7] found a slope of 52 cm'1 M for atop 12C160, corresponding to 48 cm"l /ML
for 13C180, significantly smaller than we observe. Both in our experiment and in Ref. 7,
S increased nearly linearly with total 9co in this coverage range.

A linear increase of S with em at low 9co would ordinarily be expected.
However, in the present experiment not all the CO is atop CO—--a 900 dependent fraction
of it is bridge-bonded-«so the linear relation between S and 9co is really a non-linear
dependence of S on the CO population being observed. Evidence that the bridge-bonded
fraction of 000 increases with 000 comes both fiom EELS [11] and ir [12,13]
experiments. Since S cc av, the vibrational polarizability, one explanation for the data is

that or, increases with 900- The fiaction of CO at atop sites is 1.0, 0.81, and 0.77 at

67

9co = 0.06, 0.12, and 0.16 ML, respectively. (These estimates come from Ref. 11,
modified by recent evidence [46] that atop and bridge CO have the same EELS intensity
on the step edge.) As total 9co increases from 0.06 to 0.16 ML, or, would have to
increase by >20% to keep S/Oco constant.

An increase in or~ with 900 was predicted by Ueba, [103] although the largest
increase shown in his paper is only 13% at 1 ML. We are not aware of previous
experimental evidence for this effect. The ir spectra of CO on Cu(100) and Ru(100) are
well fit [107] by a dipole-dipole coupling model that assumes that orv is coverage-
independent.

Other explanations for the apparent increase in or, with 9co that we have
considered are less plausible. The inclusion of dipole screening does cause a non-linear
dependence of S on Goo, but it has the wrong sign and only increases the need for or, to
increase with 900- It is also possible that the ratio of bridge-to-atop CO is irreproducible;
the data would be explained if orv remained constant and all the CO was atop bonded in
the present experiment and in Ref. 7, only in Refs. 10-14 was part of it bridge bonded.
To support this explanation, bridge-bonded CO was looked for and not seen with it in
Ref. 7. However, data from the EELS experiments is very consistent and the same crystal
was used in the present experiment and in Refs. 9--14.

Standard models of dipole coupling, [67 ,107] together with the assumption that
the only coverage-dependent changes in (dv/dEloc) are due to changes in or, , predict:

S(ecoloc(t.)’a.(ea)e....

(flyea) act... “'(°C°)[dv)(0). (3-3)

dE or,(0) E

where ydc is the screening factor for the static field and 7,, is an effective it screening

 

factor. [107] Dipole-coupling theory [67,107] predicts 7,, e Vie for CO. (Previous
derivations have been for a system with only one CO species; with multiple species, 7,,

and m are modified, but these conclusions are still correct. [108] ) We find, however,

68

that for CO on Pt(335) ydc varies more rapidly with 9co than does yin similar to the
results of Luo et al. [67] for CO on Pt(] 1 1). Fig. 3-10 shows 7,, and “toe calculated
fiom Eq.(3), assuming (1,,(000) and 0atop/0co are constant and vi, = yd, = 1 at the
lowest coverage. If we instead take 0atop/Oco fiom Ref. 11, the discrepancy between 11,
and ydc is even more pronounced. Allowing a coverage-dependent on, will change the
values of 7,, and we in Fig. 3-10, but will not affect the disagreement between them.
These results suggest either that dipole coupling theory is inadequate for the calculation
of E100 , or that (dv/dEloc) exhibits coverage-dependence beyond that due to av.

The dependence of the C-0 stretch fi'cquency on work function, which is known
from the experiments surveyed in Sec. 5.1 and expected from Ueba's theory, [103] can be
used to analyze the small Av of CO vs 0“. In particular, Aver,em cc Ab . The maximum
H-induced Ab for Pt(S)[6(111) x (100)] is [109] 0.08 eV; we assume similar behavior
for our surface, Pt(S)[4(1 l 1) x (100)]. The H-induced Ab is then proportional to GB at
step sites and reaches 0.08 eV when all step sites are filled. With 0.06 ML CO, 21% of
the step sites are blocked fiom H occupation so the maximum H-induced Ab is 0.06
eV. If we assume that Ab from 9co and 03 have the same effect on v and use (dv/db )
= 30 cm'1 /eV, then A v = 2.0 cm'l, more than half the observed shift. Since the
measured Ab is an average over the surface, the Ab at the edge, and therefore the actual
H-induced fiequency shift, could be significantly larger. This estimate, though crude,
suggests that most, and perhaps all of the observed Av could be caused by the H-induced
Ab , rather than by changes in local 9co-

Figure 3-6 shows a slight decrease in v for 0“ above 0.3 ML for the two lowest
9co- (For 900 = 0.16 ML we were not able to reach such high 05. ) At the highest 0", v
was 1973 a: 1 cm" with 900 = 0.06 ML and 1975 a 1 cm" with em = 0.12 ML.
Taking into account the H-induced chemical shift, these fi'equencies are close to that of an
isolated atop CO molecule on the H-saturated edge (at 0" = 0 the singleton fiequency [7]

~1968 cm'l ). This is consistent with our model since at high 0" nearly all of the atop

69

CO has been shifted to bridge sites. As the population of atop CO decreases its dipole
interaction also decreases. This reduces v for the CO that remains.

Comparison of the data of (dv/dE) vs em with the data of (dv/dE) vs OH also
suggests that the local density of CO that contributes to the ir spectrum is independent of
0”. At each 9co in Fig. 3-6, a straight line was fit to the data of (dv/dE) vs OH. The fit,
expressed in terms of the effect of 9co on (dv/dE), sets a limit to the effect of 0,; on
local 9co- The estimated change in local 9co is 0--0.02 ML, consistent with the
observed CO being in the same local environment at all but the highest H coverages.

5.3 Comparison with electrochemical experiments

Our experiments measure (dv/dE) where E is externally applied in vacuum.
Electrochemical experiments measure (dv/dd> ), where (b is the potential of the sample
relative to a reference electrode. As discussed in Sec. 5.1, to explain both (dv/dE)
measured in vacuum and (dv/dCD ) measured in an aqueous electrolyte for CO on
Pt(l 1 1), the local E-field in the compact double-layer must be a factor two larger than
predicted by two different models. However, as solvent and solute are changed there is
good correlation between (dv/d<b) and the expected (dEloc/dfb) in the compact double
layer.

Our experiment also finds a discrepancy between (dv/dE) in vacuum and (dv/db
) in an aqueous electrolyte. In the electrochemical experiments of Kim et al. [5,6] with
CO on Pt(335) in 0.1 M HClO4, (dv/dCD) for atop CO was found to depend on (D and
900- At low 000, there is a sharp transition between (dv/d<D) with (D <-0.1 V and (b
>-0.1 V (versus the saturated Calomel electrode). With (D <-0.1 V, (dv/d¢) is zero.
With o >-o.1 v, (dv/d<b )~ 75 cm‘1 N. At saturation am, the transition has
disappeared and (dv/dd) ) = 33 cm'1 N over the entire potential range. A decrease in
Stark timing rate as 9co increases is seen both in the vacuum and electrochemical
experiments. However, to explain the (dv/db) seen in the electrochemical experiment at

low Goo (and for (D >-0.l V) with (dv/dE) measured in vacuum at low 9cos we would

70

need to have (dE/d<D) = 0.85 (V /A)/V. In contrast, models of the compact double layer
discussed in Sec. 5.1 suggest that (dE/d<D )~ 0.28 (V/A)/V. These two estimates of
(dE/d<D) differ by a factor 3.0.

The sharp transition of (dv/de) in the electrochenrical experiment at (I) <-0.1 V
coincides with a peak in the cyclic voltammograrn of the Pt(335) electrode. The peak is
ascribed to H adsorption at the step edge. [110] Various models of H adsorption on
single-crystal Pt have been proposed. [111] One interpretation of the electrochemical
experiment is that coadsorbed H reduces CO's (dv/d<1>) to zero. In contrast, in our
vacuum experiment coadsorbed H has no effect on CO's (dv/dE). This difference is very
surprising.

6. Summary

We have investigated the coadsorption of H and CO on the step edges of Pt(335).
In striking contrast to the similar Pt(112) surface, [15] on Pt(335) H and CO form
compact mixed H/CO islands, within which the CO occupies only bridge sites. The
nrixed islands coexist with a pure CO phase that is largely unaffected by the presence of
H. A similar mixed phase has been observed previously for coadsorbed CO and H on
polycrystalline Pt by Thrush and White. [33] Complete segregation of H and CO,
however, occurred on the structurally similar Pt(112) surface. [15] The Pt(112) surface
used in that experiment also gave, with low H coverage and no CO, an anomalously low
temperature thermal desorption peak from edge H.

The Stark tuning rate that we measure for CO on Pt(335) is consistent with earlier
measurements [9] in vacuum and with theoretical prediction, but is a factor 3.0 too
small to account for the (dv/dd) ) seen for CO on Pt(335) in the electrochemical
experiments of Kim et a1. [5,6] Also, we find that H does not affect the Stark tuning rate
of CO on Pt(335) in vacuum, but H is able to completely suppress (dv/dd) ) in the

electrochemical experiment.

71

Our evidence for an increase in the vibrational polarizability of CO with
increasing CO coverage lends qualitative support to Ueba's theory of coadsorbate effects.
[103] The small shift of CO's resonant frequency with OH is also roughly consistent
with the dependence of fi'equency on work function expected both from theory and from

previous experiments.

72

TABLE 1. Summary of the ir spectra with only CO on Pt(335). Here 9co is the CO

coverage, v is the frequency of the peak in the ir spectrum, (AR)/R is the maximum CO-
induced reflectivity change in the RAIR spectrum, S = [(AR)/ Rdv, and (dv/dE) is the

Stark tuning rate in terms of the externally applied E-field.

 

 

 

 

 

 

 

 

 

900 v (Am/R s (dv/dE)
(ML) (cm-1) (Io-2) (cm-1) [cm-l/(VIAH
0.060 1 0.002 1974.4 1: 0.5 4.8 i 0.3 0.24 :l: 0.02 88 :l: 9

0.120 :1: 0.004 1976.8 1: 0.5 7.8 d: 0.2 0.50 i 0.05 69 :l: 7

0.160 :1: 0.005 1981.8 :1: 0.5 13.8 :t 0.3 0.61 :t: 0.03 52 :l: 5

 

 

73

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

fi 1 I
(a
7.5 if
c t:
:3 3
E E
9‘3 2 .
g (a) g
E '2’
"” A
3 (C) A 8
(d)
n l a 1
100 400 700 200 400 600 800
Temperature (K) Temperature (K)

Figure 3-1. a) TPD spectra obtained by desorbing CO from the Pt(335)surface (without
H). The CO dosages used to prepare (aHd) were 20, 1.5, 1.0 and 0.5 L, respectively. b)
Fit of two Gaussians (one from edge CO and the other from terrace CO) to the 20 L TPD

spectrum.

74

 

 

 

 

H2 Signal (Arbitrary Units)

l

l

8 'f—l—f—Wr r

l

 

 

ll

 

300 500
Temperature (K)

hr

Figure 3-2. TPD spectra obtained by desorbing H; from the Pt(335) surface (without CO).
From top to bottom the dosages used to prepare the surface were 40, 20, 10, 3, 1.5, 0.8,
0.5, 0.3 and 0.1 L.

75

 

(a)
(b)
(C)

H2 Signal (Arbitrary Units)

((0

L L A_ l A L L

100 300 500
Temperature (K)

 

 

 

Figure 3-3. TPD spectra obtained by desorbing H2 from the Pt(335) surface (with and
without CO). In a) eco = 0, OH = 0.35 ML; in b) 9co = 0, OH = 0.25 ML; in c) 000 =
0.05 ML, 9“ = 0.31 ML; and in (1) 9co = 0.16 ML, 0“ = 0.30 ML. The hydrogen doses
were a) 0.5 L, b) 0.3 L, c) 0.8 L and d) 4.5 L.

76

 

 

 

 

 

I 1.0 t 1 j fl 1
>9
.0 .0
if: .0
a
a 3". “ --------------------- -‘
8 0.5 - 5"” .
O a. A
o .3,
3: '1
W M—O—‘i
0
U!
'0
ll] 0 e - . - L A J
0 1.5 3.0 45

H2 Exposure (L)

Figure 3-4. Measured H occupancy of edge sites vs H2 dosage with various pre-coverages
of CO. The data with O, A and I were obtained by pre-dosing with 0.5, 1.0 and 1.5 L of
CO, respectively.

77

 

 

AR/R
Sir/(E)

 

 

 

 

 

 

 

1 r 1 - l l
1950 1990 2030 1950 1990 2030
Frequency (cm'l) Frequency (011")

Figure 3-5. RAIR and EVS spectra of CO coadsorbed with H on Pt(335). The CO
coverage was 0.16 NIL for all of the spectra From top to bottom the spectra are for OH =
0.06, 0.10, 0.16, 0.18, 0.21, 0.29, and 0.06 ML. The lowermost spectrum was
obtained after the sample had been heated to 420 K to desorb the H but leave the CO in

place.

78

 

 

 

 

 

 

 

 

 

 

 

 

1985 v fi ~ r ‘ r ' l ' F '
am a 0.06 ML em = 0.12 ML em = 0.16 ML
1“ ‘i 1“ *
g 1980 r - r ‘ - e’ -
a J l i f f 1 b t
§ 1 ' l I
8’ 1975 ’- ¥ " " l " " "‘
it { J
n L a l - l g L _L I n I
1.9700 0.25 0.5 o 0.25 0.5 o 0.25 0.5
an (ML) 93 (ML) 98 (ML)

Figure 3-6. Data showing the effect of coadsorbed H on the resonant frequency of the C-
O stretch vibration at atop sites, for CO and H coadsorbed on Pt(335).

79

v ' v '

 

 

 

 

 

 

 

 

 

 

 

 

‘ l ‘ l ' l ' 1
ago 8 0.00 ML 0co = 0.12 ML_ _{9co = 0.16 MI:
0.6 b d P '
r l i l h f
T ‘ i
E - d — -
3 0 3 - a i r
m -l
. f l a .- . i a r- .
i , , 0 t
0
O - l - n l - 1 , l -
0 0.25 0.5 0 0.25 0.5 0 0.25 0.5
on (ML) 98 (ML) 93 (ML)

Figure 3-7. Data showing the effect of coadsorbed H on integrated intensity S of the C-0
stretch vibration at atop sites, for CO and H coadsorbed on Pt(335).

80

 

 

 

oco=o'.06 ML 0co=({12 ML oco=o'.rc ML
:80? * "' ’- ‘l " 'l
.4:

\
>60? " P } 1‘ r- -
' l
5" l lg
gm- - - - ,1 {I -
3;

I

 

 

 

 

 

 

 

 

 

20 l I
0 0.25 0.5 0 0.25 0.5 0 0.25 0.5
91! (ML) 98 (ML) 08 (ML)

Figrn'e 3-8. Data showing the efi‘ect of coadsorbed H on the Stark tuning rate (dv/dE) of
the C-0 stretch vibration at atop sites, for CO and H coadsorbed on Pt(335).

81

 

”it ' ‘ ' '-

s/su
P
U!
I
1.4
r—o—I

 

 

 

° 0-2 0.4 A 0.6
93 (M L)
Figure 3-9. Data showing the effect of coadsorbed H on the integrated intensity S for atop
CO on Pt(335). With each CO coverage, the data have been normalized by So, the
value of S at the lowest H coverage. The data with 0.06, 0.12 and 0.16 ML of C0 are
represented by O, A and I, respectively.

82

 

 

 

 

1.1 r 2" r
t. 4;; ....................... ‘
i3
l. “ a
E 0.9 / ““
g r 7dc ‘ \ .1
b 0.7 - '
tn
0.5 ‘ ‘ J
0 0.1 0.2
9co (ML)

Figure 3-10. Data showing the efl‘ect of coadsorbed H on the screening factors ( O ) vi,
and( O ) 7..., for CO on Pt(335). We assume that 7,, = 1 and 74, = 1 at the lowest
coverage, and that no: ,[S / 0‘, and Rex (dv/dE).

83

References
l. K. C. Taylor, in Catalysis Science and Technology, edited by J. R. Anderson and M.
Boudart (Springer-Verlag, Berlin, 1984), Vol. 5, Chap. 2, pp. 119--l70.
2. G. Baier, V. Schfile, and A. Vogel, Appl. Phys. A 57, 51 (1993).
3. K. Christmann, G. Ertl, and T. Pignet, Surf. Sci. 54, 365 (1976).
4. K. Christmann, Surf. Sci. Rep. 9, 1 (1988).
5. C. S. Kim, W. J. Tomquist, and C. Korzeniewski, J .Phys. Chem. 97, 6484 (1993).
6. C. S. Kim, C. Korzeniewski, and W. J. Tomquist, J. Chem. Phys. 100, 628 (1994).
7. B. E. Hayden, K. Kretzschmar, A. M. Bradshaw, and R. G. Greenler, Surf. Sci. 149,
394 (1985).
8. J. S. Somers, T. Lindner, M. Surman, A. M. Bradshaw, G. P. Williams, C. F.
McConville, and D. P. Woodruff, Surf. Sci. 183, 576 (1987).
9. D. K. Lambert and R. G. Tobin, Surf. Sci. 232, 149 (1990). Recent unpublished data
shows that (dv/dE) for terrace CO on Pt(335) is actually larger than we saw here.
10. J. S. Luo, R. G. Tobin, D. K. Lambert, F. T. Wagner, and T. E. Moylan, J. Electron
Spectrosc. Relat. Phenom. 54/55, 469 (1990).
11. J. S. Luo, R. G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio, Surf. Sci.
274, 53 (1992).
12. J. Xu, P. Henriksen, and J. T. Yates, Jr., J. Chem. Phys. 97, 5250 (1992)
13. J. Xu and J. T. Yates, Jr., J. Chem. Phys. 99, 725 (1993).
14. J. S. Luo, R. G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio, J. Chem.
Phys. 99, 1347 (1993).
15. M. A. Henderson and J. T. Yates, Jr., Surf. Sci. 268, 189 (1992).
16. E. Hahn, A. Frike, H. Rider, and K. Kern, Surf. Sci. 297, 19 (1993).
17. J. T. Yates, Jr., private communication.
18. D. K. Lambert, Appl. Opt. 27, 3744 (1988).
19. D. K. Lambert, J. Chem. Phys. 89, 3847 (1988).

84

20. K. Christrrrann and G. Ertl, Surf. Sci. 60, 365 (1976).

21. B. Kldtzer and E. Bechtold, Surf. Sci. 295, 374 (1993).

22. C. S. Shem, Surf. Sci. 264, 171 (1992).

23. B. Poelsema, G. Mechtersheimer, and G. Comsa, Surf. Sci. 111, 519 (1981).

24. K. E. Lu and R. R. Rye, Surf. Sci. 45, 677 (1974).

25. D. M. Collins and W. E. Spicer, Surf. Sci. 69, 85 (1977).

26. C.-M. Chan and W. H. Weinberg, Appl. Surf. Sci. 1, 377 (1978).

27. V. H. Baldwin, Jr. and J. B. Hudson, J. Vac. Sci. Technol. 8, 49 (1971).

28. K. Kawasaki, T. Kodama, H. Miki, and T. Kioka, Surf. Sci. '64, 349 (1977).

29. J. H. Craig, Jr., Surf. Sci. 111, L695 (1981).

30. J. H. Craig, Jr., Appl. Surf. Sci. 10, 315 (1982).

31. J. M. White, J. Phys. Chem. 87, 915 (1983).

32. D. E. Peebles, J. R. Creighton, D. N. Belton, and J. M. White, J. Catal. 80, 482
(1983).

33. K. A. Thrush and J. M. White, Appl. Surf. Sci. 24, 108 (1985).

34. E. G. Seebauer and L. D. Schmidt, Chem. Phys. Lett. 123, 129 (1986).

35. S. L. Bemasek, K. Lenz, B. Poelsema, and G. Comsa, Surf. Sci. 183, L319 (1987).
36. K. Lenz, B. Poelsema, S. L. Bemasek, and G. Comsa, Surf. Sci. 189/190,
431hfilbreak (1987).

37. D. Hoge, M. Titshaus, and A. M. Bradshaw, Surf. Sci. 207, L935 (1988).

38. D. H. Parker, D. A. Fisher, J. Colbert, B. E. Koel, and J. L. Gland, Surf. Sci. 258, 75
(1991).

39. J. M. White and S. Akhter, CRC Crit. Rev. Solid State Mater. Sci. 14, 131 (1988).
40. D. W. Goodman, J. T. Yates, Jr., and T. E. Madey, Surf. Sci. 93, L135 (1980).

41. L. Westerlund, J. Jtinsson, and S. Andersson, Surf. Sci. 199, 109 (1988).

42. J. Bauhofer, M. Hock, and J. Kflppers, J. Electron Spectrosc. Relat. Phenom. 44, 55

(1987).

85

43. J. G. Love, S. Haq, and D. A. King, J. Chem. Phys. 97, 8789 (1992).

44. M. L. Burke and R. J. Madix, J. Am. Chem. Soc. 113, 1475 (1991).

45. L. J. Richter, B. A. Gurney, and W. Ho, J. Chem. Phys. 86, 477 (1987).

46. H. Wang, R. G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio,
unpublished.

, 47. T. E. Madey, J. T. Yates, Jr., A. M. Bradshaw, and F. M. Hoffinann, Surf. Sci. 89,
370 (1979).

48. P. Hofrnann, S. R. Bare, N. V. Richardson, and D. A. King, Solid State Comm. 42,
645 (1982).

49. M. D. Alvey, M. J. Dresser, J. T. Yates, Jr., Surf. Sci. 165, 447 (1986).

50. J. R. Smith, S. C. Ying, and W. Kohrr, Phys. Rev. Lett. 30, 610 (1973).

51. P. J. Feibelman and D. R. Hamann, Surf. Sci. 182, 411 (1987).

52. D. F. Ogletree, M. A. Van Hove, G. A. Somorjai, Surf. Sci. 173, 351 (1986).

53. J. S. Luo, R. G. Tobin, and D. K. Lambert, Chem. Phys. Lett. 204, 445 ( 1993).

54. E. Ruckenstein and T. Halachev, Surf. Sci. 122, 422 ( 1982).

55. M. D. Thompson and H. B. Huntington, Surf. Sci. 116, 522 (1982).

56. J. Marti and D. M. Bishop, J. Chem. Phys. 99, 3860 (1993).

57. P. S. Bagus, C. J. Nelin, W. Muller, M. R. Philpott, and H. Seki, Phys. Rev. Lett. 58,
559 (1987).

58. P. S. Bagus, C. J. Nelin, K. Hermann, and M. R. Philpott, Phys. Rev. B 36, 8169
(1987).

59. P. S. Bagus and G. Pacchioni, Surf. Sci. 236, 233 (1990).

60. M. Head-Gordon and J. C. Tully, Chem. Phys. 98, 3179 (1993).

61. C. W. Bauschlicher, Jr., Chem. Phys. Lett. 118, 307 (1985).

62. J. L. Andres, M. Duran, A. Lledés, and J. Bertran, Chem. Phys. 151, 37 (1991).
63. J. L. Andres, J. Marti, M. Duran, A. Lledés, and J. Bertran, J. Chem. Phys. 95, 3521
( 1991).

86

64. J. Marti, A. Lledés, J. Bertran, and M. Duran, J. Comp. Chem. 13, 821 (1992).
65. Z. Xu, J. T. Yates, Jr., L. C. Wang, and H. J. Kreuzer, J. Chem. Phys. 96, 1628
(1992).

66. D. M. Bishop, J. Chem. Phys. 98, 3179 (1993).

67. K. Hermansson, J. Chem. Phys. 99, 861 (1993).

68. D. K. Lambert, Solid State Comm. 51, 297 (1984).

69. C. L. Angell and P. C. Schaffer, J. Phys. Chem. 70, 1413 (1966).

70. K. J. Uram, L. Ng, and J. T. Yates, Jr., Surf. Sci. 177, 253 (1986).

71. F. M. Hoffmann, N. D. Lang, and J. K. Nerskov, Surf. Sci. 226, L48 (1990).

72. Z. Xu, M. G. Sherman, J. T. Yates, Jr., and P. R. Antoniewicz, Surf. Sci. 276, 249
(1992).

73. M. Tilshaus, E. Schweizer, P. Hollins, and A. M. Bradshaw, J. Electron Spectrosc.
Rel. Phenom. 44, 305 (1987).

74. K. Horn and J. Pritchard, J. Phys. (Paris) 38, C4--164 (1977).

75. G. Ertl, M. Neumann, and K. M. Streit, Surf. Sci. 64, 393 (1977).

76. P. R. Norton, J. W. Goodale, and E. B. Selkirk, Surf. Sci. 83, 189 (1979).

77. I. Yamamoto and T. Nanba, Surf. Sci. 202, 377 (1988).

78. F. Kitamura, M. Takeda, M. Takahashi, and M. Ito, Chem. Phys. Lett. 142, 318
(1987).

79. L.-W. H. Leung, A. Wieckowski, and M. J. Weaver, J. Phys. Chem. 92, 6985 (1988).
80. N. Furuya, S. Motoo, and K. Kunimatsu, J. Electroanal. Chem. 239, 347 (1988).
81. S.-C. Chang, L.-W. H. Leung, and M. J. Weaver, J. Phys. Chem. 93, 5341 (1989).
82. F. Kitamura, M. Takahashi, and M. Ito, Surf. Sci. 223, 493 (1989).

83. S.-C. Chang and M. J. Weaver, J. Chem. Phys. 92, 4582 (1990).

84. Y. Kinomoto, S. Watanabe, M. Takahashi, and M. Ito, Surf. Sci. 242, 538 (1991).
85. S.-C. Chang and M. J. Weaver, J. Phys. Chem. 95, 5391 (1991).

87

86. S.-C. Chang, X. Jiang, J. D. Roth, and M. J. Weaver, J. Phys. Chem. 95, 5378
(1991).

87. X. Jiang and M. J. Weaver, Surf. Sci. 275, 237 (1992).

88. M. J. Weaver, Appl. Surf. Sci. 67, 147 (1993).

89. M. R. Anderson, D. Blackwood, and S. Pons, J. Electroanal. Chem. 256, 387 (1988).
90. M. R. Anderson, D. Blackwood, T. G. Richmond, and S. Pons, J. Electroanal. Chem.
256, 397 (1988).

91. J. G. Love and A. J. McQuillan, J. Electroanal. Chem. 274, 263 (1989).

92. J. D. Roth, S.-C. Chang, and M. J. Weaver, J. Electroanal. Chem. 288, 285 (1990).
93. A. E. Russell, S. Pons, and M. R. Anderson, Chem. Phys. 141, 41 (1990).

94. A. E. Russell, D. Blackwood, M. R. Anderson, and S. Pons, J. Electroanal. Chem.
304, 219 (1991).

95. K. Ashley, M. G. Samant, H. Seki, and M. R. Philpott, J. Electroanal. Chem. 270,
349 (1989).

96. L.-W. H. Leung, S.-C. Chang, and M. J. Weaver, J. Chem. Phys. 90, 7426 (1989).
97. M. R. Anderson and J. Huang, J. Electroanal. Chem. 318, 335 (1991).

98. J. D. Roth and M. J. Weaver, Langmuir 8, 1451 (1992).

99. C. M. Mate, C.-T. Kao, and G. A. Somorjai, Surf. Sci. 206, 145 (1988).

100. J. O'M. Bockris, M. A. V. Devanatharr, and K. M[iiller, Proc. R. Soc. London Ser. A
274, 55 (1962).

101. D. K. Lambert, Phys. Rev. Lett. 50, 2106 (1983); 51, 2233 (E) (1983).

102. R. P. Feynman, Phys. Rev. 56, 340 (1939).

103. H. Ueba, Surf. Sci. 188, 421 (1987).

104. N. K. Ray and A. B. Anderson, J. Phys. Chem. 86, 4851 (1982).

105. H. S. Luftman and J. M. White, Surf. Sci. 139, 369 (1984).

106. S. Holloway and J. K. Norskov, J. Electroanal. Chem. 161, 193 (1984).

107. B. N. J. Persson and R. Ryberg, Phys. Rev. B 24, 6954 (1981).

108. H. Wang and R. G. Tobin, unpublished.

109. D. M. Collins and W. E. Spicer, Surf. Sci. 69, 114 (1977).

110. A. Rodes, K. El Achi, M. A. Zarnakhchari, and J. Clavilier, J. Electroanal. Chem.
284, 245 (1990).

111. A. Wieckowski, P. Zelenay, and K. Varga, J. Chim. Phys. 88, 1247 (1991).

112. H. Wang, R. G. Tobin, D. K. Lambert, J. Chem. Phys. 101, 4277 (1994).

89

Chapter 4

H-CO interactions on the terraces and step edges of the
Pt(335) surface

The material in this Chapter is largly based on a paper that has been submitted to
Surface Science [53].

1. Introduction

Coadsorbed H and CO on stepped Pt is an interesting model system, relevant to
such technologies as catalysts, chemical sensors and fuel cells. In this Chapter, I present
a high resolution electron energy loss spectroscopy (EELS) and temperature-programmed
desorption (TPD) investigation of CO and H on Pt(335).

As shown in Fig. 1-1, this surface is highly stepped with (111) terraces four atoms
wide. In step-terrace notation it is Pt(S)[4(111) x (100)]. In a previous experiment [1]
we used IR laser spectroscopy to study CO at step sites on Pt(335) with coadsorbed H.
These results were presented in Chapter 3. However, that experiment was limited both by
the restricted tuning range of the laser and by the range of CO coverage investigated so
only atop bonded CO along the edge could be seen. In the present experiment, EELS
allows all the CO to be seen, although it does not allow edge and terrace CO to be
distinguished spectroscopically. Also, as an alternative to having CO just at edge sites,
by blocking the edge sites with H before dosing with CO we are also able to put CO just
at terrace sites.

To our knowledge, this is the first EELS study of coadsorbed CO and H on Pt.
However, EELS has previously been used to obtain vibrational spectra of coadsorbed CO
and H on Cr(111) [2], Cu3Pt(111) [3], Fe(100) [4], Ir(110) [5], Ir(111) [5], Ni(100) [6-8],
Ni(110) [9,10], Ni(l 1 1) [6], Pd(100) [8,11,12], and Rh(100) [13,14]. The coadsorption of
CO and H on metal surfaces has been reviewed by White and Akhter[15]. Other
experiments that have studied coadsorbed CO and H on Pt are discussed in Chapter 2

90

[Ref. 1]. In vacuum, IR vibrational spectra of coadsorbed CO and H have been obtained
on Pt(111) [16] and Pt(S)[9(111) x (111)] [17]. There have also been many
electrochemical experiments that have used IR vibrational spectroscopy to study CO on
Pt in situations where H must also have been present [1,18]. In particular the
coadsorption of CO and H on Pt(335) in water has been studied [19,20].

As a matter of notation, we let 000 and 0H be the CO and H coverages,
respectively. Coverages are given in monolayers (ML), where 1 ML is l adsorbate per
surface Pt atom. Previous experiments have shown that for clean Pt(335) covered only
with CO, at saturation 9co = 0.63 NH. [21]. Experiments on a similar surface, clean
Pt(S)[9(111) x (111)], have shown that at H saturation 0“ = 1 ML [22]. We assume that
the same is true for H on Pt(335). On clean Pt(] 1 1) the saturation 03 = 0.80 ML [23].

2. Experiment

Our experiments were canied out in an ultrahigh vacuum (UHV) chamber with
base pressure of 3.5 x 10'11 torr. The sample was spot-welded to two Ta wires, which
were used for both heating and cooling. The sample temperature could be varied fiom 90
K to over 1400 K. The sample was cleaned by cycles of sputtering in Ar, cycling the
sample temperature between 570 K and 1023 K in 1.0 x 10'8 torr oxygen and annealing at
1300 K for one nrinute. The sample cleanliness was always checked by Auger
spectroscopy and EELS. The gases were dosed through individual closers, with
enhancement factors of about 100 over background dosing.

The EELS apparatus has been described elsewhere [24]. We scanned from 300 to
5000 cm'1 , which includes both the C-0 internal stretch and the C-Pt stretch vibrations.
The H-Pt stretch was too weak to be detected. All EEL spectra were measured at a
sample temperature of 90 K.

The TPD scans reported here were taken at 10 K/s with the sample facing the

mass spectrometer, as described in Chapter 2.

91

3. Results and Discussion

Our EELS and TPD results for CO alone on Pt(335) agree closely with those
reported by Luo et al. [25], and are consistent with other IR and TPD data [1,21,26]. At
low coverage, CO occupies only edge sites. At higher coverages, CO begins to occupy
terrace sites as it continues to fill the edge sites. At saturation, all of the edge sites are
occupied (one CO per edge Pt atom). In the model of Luo et al. [25] all of the CO at edge
sites is atop bonded at saturation 9coo However, on the(111) terraces there is a mixture of
atop and bridge-bonded CO. A similar CO structure occurs on Pt(l 1 1) at saturation [27].

Our TPD results for H alone on Pt(335) are shown in Fig. 4-1. As seen previously
[1], at the lowest coverage there is one peak at 395K from H at edge sites. After the edge
is saturated, H begins to fill terrace sites and a new peak appears at ~ 315 K. The only
significant difference between Fig. 4-1 and the data in Chapter 3 [Ref. 1] is that in the
present work we do not see a low temperature shoulder on the terrace peak. The low
temperature shoulder may have been from H on the back or sides of the crystal; in the
present work only the fiont face of the crystal was exposed to the mass spectrometer.
3.1 Coadsorption of H and CO on the step edge

At low Boos in equilibrirun, all of the CO is at edge sites. Figure 4-2 shows EEL
spectra at various 0“, for Geo = 0.07 and 0.13 ML. The overlayer was prepared by dosing
CO at 90 K and then annealing at 420 K for one minute, both to allow the layer to
equilibrate and to desorb any H adsorbed from the background. The sample was cooled
to 90 K and all H dosing and EEL spectra occurred at that temperature. Following each
spectrum the sample was heated to 420 K while the partial pressure of H; was monitored
to determine 0". A previous experiment [1] showed that throughout this range of 9C0 and
0“, the coverage of atop CO on the terrace < 0.008 ML.

In Fig. 4-3a we plot the ratio In llm, vs. 0" fi‘om the data in Fig. 4-2, where 13 is
the integrated single-loss peak from bridge-bonded CO and 1,0, is the total integrated
single-loss peak (both bridge and atop-bonded). The data clearly show that coadsorbed H

92

shifts CO fiom atop to bridge sites. At the highest 0" studied here the atop peak is
reduced to about 25% of its original intensity. The IR data in Ref. 1 show that the
elimination of the atop CO intensity is almost complete at higher 0".

The conversion of atop CO to bridge CO with increasing 0" that we observe
supports the model proposed in Chapter 3 [Ref. 1]. In the model, one-dimensional
islands of mixed H and CO coexist with islands of pure CO. Both phases have equal CO
concentrations. A mixed phase had previously been observed on relatively open surfaces
of Ni [7,10,28-30], Fe [31], and Rh [13,14]. Both a mixed phase and islands occur on
polycrystalline Pt [32]. The model for CO on Pt(335) explains the linear decrease of atop
CO's IR intensity with increasing 0“ as a consequence of H displacing CO from atop to
bridge sites within the mixed islands. Similar site shifting has been observed in other
experiments [10,13,19,20].

Our EELS data and the IR results in Ref. 1 both show that the effect of OH on atop
CO's vibrational fiequency is small. In the IR spectra the change was < 5 cm'l. In the
EEL spectra in Figs. 4-2a and 4-2b the atop frequency changes ~ 20 cm’l, less than the ~
60 cm" instrumental linewidth.

Since the data in Figs. 4-2a and 4-2b were acquired at constant 000, we can use
them to compare the EELS cross sections of atop and bridge CO on the step edge, and to
analyze the dependence of the cross sections on 9u- Other studies [21,26,33-35] have
examined the relative cross sections of edge atop and terrace atop CO, but to our
knowledge this is the first measurement of the relative cross sections of atop and bridge
CO on the step edge. On Pt(11 l), Mieher, Whitman and Ho [33] calibrated the
populations with low-energy electron diffraction and found the EELS cross section of
atop CO to be 1.8 times that of bridge CO. On Pt(335), however, it is immediately
apparent from Fig. 4-2 that the cross sections of edge atop and edge bridge C0 are
comparable-«at least when the bridge CO is in a H-rich environment. In both Fig. 4-2a

93

and Fig. 4-2b the bridge band intensity in the top spectrum is comparable to or greater
than the atop intensity in the bottom spectrum.

Figure 4-3b displays I‘m/IE vs 0", where 1;; is the intensity of the elastic peak. The
same qualitative behavior is found for both values of 9co- As 0,, increases, Int/IE
gradually drops, as we would expect if bridge CO has a smaller cross section than atop
CO---as on Pt(111). At higher 0", although the fraction of bridge CO continues to
increase, Lot/IE rises, and at the highest H-coverage is a factor 1.1 :l: 0.2 greater than at 0"
= 0.

The nonmonotonic behavior of Inn/IE shows that coadsorbed H has a strong effect
on the EELS cross section of edge bridge CO. Further evidence comes from the spectra
for 0.13 ML of CO with 0;; between 0.23 and 0.40 ML; in this range of H coverage 1.0.03
increases by more than a factor of two, while IB/lm, (and thus the bridge coverage) barely
changes. There is other evidence that coadsorbates affect CO's cross section. Reflection-
absorption IR spectra [1,21,26] of CO alone on Pt(335) suggest that the IR cross section
of edge atop CO increases by about 20% with increasing CO coverage. Such an increase
is qualitatively consistent with a prediction by Ueba [36]. Nevertheless it is unexpected
that H could cause a factor two increase in the cross section of edge bridge CO,
particularly since the interaction between H and CO is relatively weak. For example, H
induces only a small shift in C03 vibrational fiequency, and on stepped Pt the work
function changes due to H and C0 are comparable [37].

We have found that intensity ratios like those in Fig. 4-3 are reproducible. While
measuring the spectra in Figs.4-2a and 4-2b we did not change the settings of the EELS
system. Moreover, the spectrum at 0“ = 0.10 ML in Fig. 4-2b was measured twice-«both
before and afler that at 0.40 ML---and both ratios, of Lot/IE and IE/Itots varied by less
than10%. This consistency reinforces our belief that the increase in Inn/IE on going from
9n = 0.23 to 0.40 ML is not an artifact. The similarity of the Itot/IE at high OH to the initial

94

intensity for both CO coverages lends further support. We consider it likely, therefore,
that the enhancement of the edge bridge CO cross section by coadsorbed H is real.
3.2 Coadsorption of H and CO on the terrace

Previous comparisons between adsorbates on Pt(335) and Pt(112) have shown
that a difference of one row in terrace width has a profound effect [l,21,25,38,39]. It is
therefore of interest to examine the coadsorption of H and CO on the terrace. It is
difficult to isolate the properties of terrace CO because ordinarily edge CO is also present
[1,21,25,26]. One approach to avoid edge CO is to block the edge sites with a different
species [40].

In the present experiment, we use H to block the edge sites. We first saturate the
step edge with 0.25 ML of H, dose with CO, and then add more H. The entire experiment
is done at 90 K to prevent terrace CO fi'om exchanging with H at the step edge. The H
coverages were determined by repeating the dosing sequence and measuring TPD spectra
in a separate experiment. The CO coverages were 0.05, 0.13 and 0.19 ML.

An IR experiment performed on the same crystal in a separate chamber
demonstrates that this procedure gives terrace CO without edge CO---and in particular
that CO does not exchange with edge H at this temperature. When the sample was
predosed with 0.6 ML of H and then exposed to CO, the IR spectrum shows a peak at the
frequency characteristic of terrace CO, and none at the fiequency of edge CO. These
results will be presented in the next Chapter and reported in a separate publication[41].

Our EELS and TPD measurements, with no H added after the CO dose, confirm
that the CO is predominantly on the terrace, and also provide evidence that its behavior
with the edge saturated with H is essentially the same as with CO on the edge. Figure 4-4
shows EEL spectra as a function of ego, after predosing with 0.25 ML of H. Figure 4-5
shows IB/lw. vs Goo, together with the bridge CO coverage calculated assuming the atop
CO cross section is a factor 1.8 that of bridge CO, as on Pt(] 1 1) [33]. The bridge CO
coverage < 0.03 ML for 91:0 up to 0.16 ML, and then increases almost linearly after that.

95

This behavior is essentially identical to that reported for terrace bridge CO on Pt(335) by
Luo et al. (cf. Fig. 3 of Ref. 25), with the CO coverages offset by about 0.2 ML--
approximately the edge site coverage. We find from TPD that the saturation coverage of
our postdosed CO is 0.36 ML, in agreement with the saturation coverage of terrace CO
on the clean surface, 0.38 ML. Finally, our measured 13]th agrees with Ref. 25 if we
make some reasonable assumptions: without H all edge CO is at atop sites [25], and the
EELS cross section of terrace atop CO is 0.5 that of edge atop CO [41] and 1.8 that of
terrace bridge CO [33].

Finally, as we show below, postdosed CO responds differently to additional H
than does edge CO, providing further confirmation that at 90 K predosed H effectively
blocks CO adsorption at edge sites.

Figure 4-6 shows EEL spectra of 0.05 ML of CO on the terrace and various
amounts of postdosed H; our spectra (not shown) with em = 0.13 and 0.19 ML on the
terrace are similar. As 0“ on the terrace is increased the EEL spectra show only a slight
shift of intensity from the atop to the bridge band with the first H postdose. The observed
IB/Iw, vs 0“ is plotted in Fig. 4-7 for all three Goo, and should be contrasted with the
comparable data for edge CO in Fig. 4-3a. The small initial shift with 0.05 ML CO is
consistent with a small amount of CO at the edge. Other than that, the total intensity,
band frequencies, and line shapes are all independent of 0“. In essence coadsorbed H on
the terrace has no observable eflect on the vibrational spectrum of terrace CO, in dramatic
contrast to the site shift and intensity enhancement observed on the step edge.

The coadsorption behavior on the terrace is also very different fi'om that seen on
Pt(l 1 1), where compact islands containing pure C0 are formed [17,42,43]. With 0.23
ML CO on Pt(] 1 1), dosing with H to saturation changes the single-loss EELS intensities
for atop and bridge CO by factors of < 0.5 and > 2, respectively; the peak shapes change
significantly; and the vibrational frequencies of both atop and bridge CO shift > 20 cm'1

[17].

96

Figure 4-8 shows TPD spectra of a saturation coverage of H desorbing from
Pt(335) with varying amounts of CO and 0.48 ML H on clean surface. As the sample
temperature is increased the CO becomes mobile and replaces H at the edge sites. If there
is enough CO to fill all the edge sites then no H desorption from edge sites is observed.
Note that the peak desorption temperature of terrace H increases from 272 to 300 K as
000 increases from 0 to 0.19 ML. About 2/3 of this increase comes as 9co increases from
0 to 0.05 ML. Meanwhile, the bottom two spectra have similar H coverage but different
CO coverage, 0.19 ML and 0 ML, yet there is no difference in terrace H desorption
temperature.

In comparison, for CO and H on Pt(] 1 1), Peebles et al.[44] found that with 0.23
ML of CO and a saturation coverage of H, the peak desorption temperature was 260 K.
Without CO it was 290 K. CO shifts the H desorption temperature down by 30 K. This
is strong evidence for a repulsive interaction between coadsorbed H and CO, and is in
sharp contrast with what happens on the terraces of Pt(335). The H2 TPD data suggest
that the interaction between coadsorbed terrace CO and H is very weak, in agreement
with the EELS observation.

The difference in TPD data for CO and H on the two surfaces suggests that CO
and H are not segregated on Pt(335) terraces as they are on Pt(] 1 1). The insensitivity of
terrace CO vibrations on Pt(335) to coadsorbed H appears to support this. Our EELS data
are similar in some respects to what is seen for atop CO on the edge. With all the CO on
the edge, an increase in 0H incorporates more CO in the mixed islands, and converts that
CO fi'orn atop to bridge sites, but the remaining C0 is unaffected. In particular, even
though the H-CO interaction is repulsive, coadsorbed H on the edge does not compress
the remaining atop CO. Similarly, on the terrace, coadsorbed H does not affect the local
CO density; see the data for 0.19 ML of CO in Fig. 4-7. If H and CO were to segregate
as they do on Pt(l 11), the local CO density at saturation OH would be 0.6 CO/Pt. But a
comparisonof IB/Ito, from Fig. 4-7 with Fig. 4-5 shows that the local CO density < 0.22

97

CO/Pt. On the other hand, if the two species were to mix uniformly on an infinite Pt(l 1 1)
surface, at saturation (in each CO would have two nearest-neighbor sites occupied by H,
two occupied by CO and two empty; in the mixed phase on the step edge,where both of
the nearest-neighbor sites are occupied by H, complete conversion of atop to bridge is
observed. It is surprising that a comparable H density on the terrace produces no
discernible effect.

One explanation for our data is that the H added to the terrace goes to a subsurface
location. This has previously been suggested to explain certain experiments involving H
on Pt(111) [45,46] although that interpretation is not generally accepted [47,48]. There is
also evidence that H on Pt(110) goes to a subsurface site associated with step troughs
[49]. Some electrochemical experiments with H on Pt have also been explained with
subsurface H [50,51]. In particular, a spectroelectrochemical study [19,20] of CO and H
on Pt(335) found that CO's vibrational frequency is independent of electrode potential in
the potential range where H adsorption occurs. However, our H2 TPD spectra do not
have a extra peak above the desorption temperature of edge H, as is seen on Pd(110)
where there clearly is subsurface H [52].

A second interpretation for our data is that the H on the terrace has no effect on
the CO on the terrace, even though they are fully mixed. This interpretation calls for an
explanation of the apparent weakness of the H—CO interaction on the terrace. Also, since
the terrace sites on Pt(335) are close-packed, H and CO would not be expected to form a
mixed phase [15].

A third explanation is that H occupies sites on the surface that do not significantly
affect the CO observed in the IR spectrum. For example, the CO could build up fi'om the
outside edge of the step while the H builds up fi'om the trough. The finite step size would
tend to leave voids in the CO pattern on the side of the step near the trough where extra H
could be accommodated. Also, screening is expected to diminish the contribution to the

IR spectrum of CO at sites down in the trough. An electrostatic model, presented in the

98

next Chapter [41], that successfully explains the enhanced relative IR cross section of CO
at edge sites predicts that CO at the three atomic rows on the terrace have relative IR
cross sections of 1.0, 0.57 and 0.03, respectively, as one moves away from the step edge
toward the trough.

In our experiment, since the sample could not be annealed above 90K, the layer
may not have equilibrated completely. Luo et al. [25] found that annealing was necessary
to equilibrate the atop/bridge ratio for pure CO on Pt(335) at coverages where there was
substantial terrace site occupation, but they speculated that rearrangement of edge CO at
high coverage was the major barrier. Moreover, the coadsorption experiments on
Pt(] 1 1), which showed segregation of H and CO, were performed at 100 K. This shows
that H and C0 are mobile enough on the flat surface to rearrange at that temperature.

4. Summary

We have studied the coadsorption of H and CO on both the edge and the terrace of
Pt(335). For edge CO, we found that coadsorbed H continuously shifts CO from amp to
bridge sites, confirming the model presented in Chapter 3, proposed by Wang et al. [1].
The site shift pernrits a direct comparison between the EELS cross sections of edge
bridge and edge atop CO. The cross section of edge bridge CO in the presence of
saturation H coverage is a factor 1.1i 0.1 that of edge atop CO without H; on Pt(] 1 1), the
cross section of atop CO is a factor 1.8 that of bridge CO [33]. Coadsorbed H apparently
has a large effect on the cross section of edge bridge CO.

We studied terrace CO by first saturating the step edges with H at 90 K. The BEL
spectrum of terrace CO is not changed by increasing 0“, even to saturation. This
behavior is different from the segregation found on Pt(l 1 1) [17,42,43]. Evidently the
nature of the H-CO interaction on Pt surfaces is very sensitive to the local surface

structure.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

T V l I
A
m
fl
C
3
£5
4 A \
V
'5
C
.9
m 7, AL
a v-
I wfi v“:
1 1 4 1
100 300 500

Temperature (K)

Figure 4-1. TPD spectra obtained by desorbing H; from the Pt(335) surface (without
CO). From top to bottom, the relative dosages used to prepare the surface were 10, 5,
2.5, 1, 0.5, 0.25, and 0.1. The absolute dosages are uncertain because a closer was used,

but a relative dose of l is approximately 1x 10'6 torr s.

100

 

 

a) i b)
1894 "

2052

2035
/

1864‘ I 2033

rare. 2050
I
1361‘ , 2046
1013‘ ,2052

107k I 2957

1870\ . 2038 2063

Intensity (Arb. Units)

‘31:"

- 2054 - 2074
5‘ l A 56' ‘
Q 2058 ] 1074‘

I l
1000 2000 3000 0 1000 2000 3000
Loss Energy (cm") Loss Energy (cm")

 

 

 

 

 

 

 

 

 

 

 

.

 

 

01-

Figure 4-2. EEL spectra of edge CO on Pt(335) as a function of H coverage. The CO
coverages in (a) and (b) are 0.07 and 0.13 ML, respectively. The spectra are arranged so
H coverage increases up the page. In (a) the H coverages are 0.02, 0.10, 014,018, 0.22,
and 0.24 ML; in (b) 0 (9co = 0 also),0.01, 0.06, 0.10, 0.13, 0.22, and 0.40 ML.

IB/Itu

101

 

 

1.0 r 0.3 '

 

 

 

 

 

 

 

 

0.07 ML
0 J r
0 0.2 0.4 0 0.2 0.4
H Coverage (ML) H Coverage (ML)

Figure 4-3. (a) Ratio lull”, of bridge to total (bridge +atop) single-loss EELS intensity as
a function of H coverage, for the spectra in Fig. 4-2. (b) Ratio IB/IE where I; is the elastic
EELS intensity, as a function of H coverage, for the spectra in Fig. 4-2.

102

 

1867 / 2069

F

 

1867 ‘ 2079

\

 

F

x 1882\ - 2100

i

 

‘2090
1374

\

F

 

, 2092
1885

\

Intensity (Arb. Units)

i

 

 

58' " 1889‘ ’2079

 

 

2069
/

l U 1880‘

l J
0 1000 2000 3000

Loss Energy (cm’1)

 

 

 

 

-

 

Figure 4-4. EEL spectra at various CO coverages after predosing with 0.25 ML H to
block the edge sites. The spectra are arranged so CO coverage increases up the page: 9co
= 0.05, 0.08, 013,019, 0.28, and 0.36 ML.

103

 

 

 

 

 

r 0.2 3
E
0.4 *- 0
8'
3 a
\ 3
4‘ q 0 r u
0.2 - o
u
0
at
P.
b
0 I . 0 m
0 0.2 0.4

Total CO Coverage (ML)

Figure 4-5. Ratio lull”. as a function of CO coverage, for terrace CO, with the edge
saturated with H (for the EEL spectra in Fig. 4-4). Also shown is the bridge CO
coverage, calculated assuming that the EELS cross section of terrace atop CO is a factor

1.8 times that ofterrace atop CO.

104

 

 

 

 

 

A
a
“E
a
e'
3 1891\ [2077
=3 63- -'
5 1876s ,2081
all
5.
1891~ ’2077
J Li 1882\ ,2068

 

 

1 1 I
0 1000 2000 3000

Loss Energy (cm")

 

 

Figure 4~6. EEL spectra of 0.05 ML terrace CO as a function of postdosed H coverage.
The spectra are arranged so 9a increases up the page: 0“ = 0.24, 0.46, 0.70, and 0.70ML.

105

 

0.6 r ,

 

 

 

0.4 bA 4
It
1..
\
.5
0.2 r- -
I "a. e
O
O r J
0.2 0.4 0.6 0.8

H Coverage (ML)

Figure 4-7. Ratio IB/Im from EELS intensity for terrace CO, as a function of H coverage,
forthree CO coverages: A 0.05 ML, 0 0.13 ML, and I 0.19 ML.

106

 

 

 

 

 

 

 

 

 

 

 

 

 

Q? l
«H l
“E l
D
e
3 (a)
3
C VVJVA
.9
U)
.. (b)
I t
(c) - _
fl“) _1
(e) ' F T
100 300 500

Temperature (K)

Figure 4-8. TPD spectra obtained by desorbing a saturation coverage H; from the Pt(335)
surface and by desorbing 0.48 NIL ofH on clean surface. In (a) 000 = 0 ML, 0“ = 1.0
ML; in (b) 000 = 0.05 ML, 0“ = 0.69 ML; in (c) 000 = 0.13 ML, 0“ = 0.55 ML; and in
(d)9co = 0.19 ML, 0" = 0.45 ML; (e) 9co = 0 ML, 0" = 0.48 ML.

107

References

l. H. Wang, R. G. Tobin, and D. K. Lambert, J. Chem. Phys. 101, 4277(1994).

2. M. Nagoshi and Y. Fukuda, Appl. Surf. Sci. 60/61, 688(1992).

3. C. Becker, U. Schrbder, G. R. Castro, U. Schneider, H. Busse, R. Linke, and K.
Wandelt, Surf. Sci. 307—309, 412(1994).

4. P. B. Merrill and R. J. Madix, Surf. Sci. 271, 81(1992).

5. T. S. Marinova and D. V. Chakarov, Surf. Sci. 217, 65(1989).

6. G. E. Mitchell, J. L. Gland, and J. M. White, Surf. Sci. 131, 167(1983).

7. L. Westerlund, L. ansson, and S. Andersson, Surf. Sci. 199, 109(1988).

8. C. Nyberg, L. Westerlund, L. ansson, and S. Andersson, J. Electron Spectrosc. Relat.
Phenom. 54/55, 639(1990).

9. N. D. S. Canning and M. A. Chesters, Surf. Sci. 175, L811(1986).

10. J. Bauhofer, M. Hock, and J. Kiippers, J. Electron Spectrosc. Relat. Phenom. 44,
55(1987).

11. C. Nyberg and L. Westerlund, Surf. Sci. 256, 9(1991).

12. C. Nyberg and L. Westerlund, Chem. Phys. Lett. 185, 445(1991).

13. L. J. Richter, B. A. Gurney, and W. Ho, J. Chem. Phys. 86, 477(1987).

14. L. J. Richter, T. A. Gerrner, and W. Ho, Surf. Sci. 195, L182(1988).

15. J. M. White and S. Akhter, CRC Crit. Rev. Solid State Mater. Sci. 14, 131(1988).
16. E. Hahn, A. Ficke, H. Rbder, and K. Kern, Surf. Sci. 297, 19(1993).

17. D. Hoge, M. Titshaus, and A. M. Bradshaw, Surf. Sci. 207, L935(1988).

18. For reviews of this work see K. Ashley and S. Pons, Chem. Rev. 88, 673(1988); M.
J. Weaver, Appl. Surf. Sci. 67, 147(1993); and H. Seki, IBM J. Res. Dev. 37,
227(1993).

19. C. S. Kim, W. J. Tomquist, and C. Korzeniewski, J. Phys. Chem. 97, 6484(1993).
20. C. S. Kim, C. Korzeniewski, and W. J. Tomquist, J. Chem. Phys. 100, 6280994).
21. D. K. Lambert and R. G. Tobin, Surf. Sci. 232, 149(1990).

108

22. K. Christmann and G. Ertl, Surf. Sci. 60, 365(1976).

23. K. Christmann, G. Ertl, and T. Pignet, Surf. Sci. 54, 365(1976).

24. B. A. Sexton, J. Vac. Sci. Technol. 16, 1033(1979).

25. J. S. Luo, R. G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio, Surf. Sci.
274, 53(1992).

26. B. E. Hayden, K. Kretzschmar, A. M. Bradshaw, and R. G. Greenler, Surf. Sci. 149,
394(1985).

27. G. Ertl, M. Neumann, and K. M. Streit, Surf. Sci. 64, 393(1977).

28. D. W. Goodman, J. T. Yates, Jr., and T. E. Madey, Surf. Sci. 93, Ll35(1980).

29. J. G. Love, S. Haq, and D. A. King, J. Chem. Phys. 97, 8789(1992).

30. S. Haq, J. G. Love, and D. A. King, Surf. Sci. 275, 170(1992).

31. M. L. Burke andR. J. Madix, J. Am. Chem. Soc. 113, 1475(1991).

32. K. A. Thrush and J. M. White, Appl. Surf. Sci. 24, 108(1985).

33. W. D. Mieher, L. J. Whitman, and W. Ho, J. Chem. Phys. 91, 3228(1989).

34. J. E. Reutt-Robey, D. J. Doren, Y. J. Chabal, and S. B. Christrnan, J. Chem. Phys.
93, 9113 (1990).

35. R. G. Greenler, J. A. Dudek, and D. E. Beck, Surf. Sci. 145, L453(1984).

36. H. Ueba, Surf. Sci. 188, 421(1987).

37. D. M. Collins and W. E. Spicer, Surf. Sci. 69. 114(1977).

38. M. A. Henderson and J. T. Yates, Jr., Surf. Sci. 268, 189(1992).

39. J. Xu, P. Henriksen, and J. T. Yates, Jr., J. Chem. Phys. 97, 5250(1992).

40. A. Szabé, M. A. Henderson, and J. T. Yates, Jr., J. Chem. Phys. 96, 6191(1992).

41. H. Wang, R G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio,
unpublished.

42. S. L. Bemasek, K. Lenz, B. Poelsema, and G. Comsa, Surf. Sci. 183, L319(1987).
43. K. Lenz, B. Poelsema, S. L. Bemasek, and G. Comsa, Surf. Sci. 189/190, 431(1987).

109

44. D. E. Peebles, J. R. Creighton, D. N. Belton, and J. M. White, J. Catal. 80,
482(1983).

45. W. Eberhardt, F. Greuter, and E. W. Plummet, Phys. Rev. Lett. 46, 1085(1981).

46. C. M. Greenlief, S. Akhter, and J. M. White, J. Phys. Chem. 90, 4080(1986).

47. P. R. Norton, J. A. Davies, and T. E. Jackman, Surf. Sci. 121, 103(1982).

48. K. Christmann, Surf. Sci. Rep. 9, 1(1988).

49. E. Kirsten, G. Parschau, W. Stocker, and K. H. Rieder, Surf. Sci. 231, L183(1990).
50. A. Wieckowski, P. Zelenay, and K. Varga, J. Chem. Phys. 88, 1247(1991).

51. I. M. Tidswell, N. M. Markovic, and P. N. Ross, Phys. Rev. Lett. 71, 1601(1993).
52. R. J. Behm, V. Penka, M.-G. Cattania, K. Christmann, and G. Ertl, J. Chem. Phys.
78, 7486 (1983).

53. H. Wang, R. G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio, submitted
to Surf. Sci.

110

Chapter 5

Vibrational Intensity and Stark Tuning Rate of Edge and
Terrace CO on Pt(335)

This Chapter is largely based on a investigation that will be submitted to the
Journal of Chemical Physics [26].
1. Introduction

Steps and other surface defects are important in heterogeneous catalysis at metal
surfaces. A surface with defects has a wider variety of sites — some more exposed to the
vacuum and some more tightly coupled to the metal — than does an atomically flat
surface. Screening of an external electric field is one measure of site diversity. In this
study we compare the vibrations of atop bonded CO at sites on the step edges and on the
flat terraces of stepped Pt and try to explain our observations in terms of the electric field.

As shown in Fig. 1-1, the surface we use is Pt(335): Pt(S)[4(111)x(100)] in step-
terrace notation. We compare the vibrational intensity of atop-bonded CO at step edge
and terrace sites using reflection-absorption infiared spectroscopy (RAIRS) and high-
resolution electron energy loss spectroscopy (HREELS). We also compare the Stark
timing rates (the change of vibrational frequency in an electrostatic field) of CO at the
two sites, using electroreflectance vibrational spectroscopy (EVS). We manipulate the
CO so it is either all on the edge or all on the terrace using coadsorbed H or O. This
allows us to compare CO at the two sites with total CO coverage 900 held constant.

To get CO on the terrace, the surface is first dosed at low temperature (near 100
K) with enough 0 or H to fill the edge sites. If the H-predosed surface is later heated to
420 K the H desorbs and the CO moves to edge sites. Heating the O-predosed surface to

260 K causes the O to move to the terrace while the CO moves to the edge. Our use of

111

coadsorbed O to manipulate CO's site occupancy on stepped Pt follows Szabo et al. [1]
Also, Hahn et al. [2] have used H on stepped Pt to prevent CO adsorption.

Previous experiments have also investigated the difference in vibrational cross
section between edge and terrace CO. One approach has been to rely on the natural
sequence of site filling as CO coverage builds up on the surface. At low coverage, CO
preferentially occupies sites on the edge. At higher coverage it increasingly occupies
terrace sites [3,4,5] In an earlier RAIRS study of CO on Pt(335), Hayden et al. [3] found
that the rate of increase of integrated intensity with 9C0 at high coverage is 2.7 times
higher than at low coverage. Greenler et a1 [6] estimated the E-field distribution on the
same surface and found that at the center of the C=O bond the field at an edge site is 1.5
times greater than the average for the terrace sites, corresponding to a factor of 2.2
difference in IR cross-section. On the other hand, Lambert and Tobin found the cross-
sections of edge and terrace CO to be nearly the same, and roughly equal to that of CO on
Pt(111) [4]. Both of these analyses involve uncertainties, however. First, the terrace CO
is not studied in isolation, since edge CO is always present, and the two vibrational bands
are strongly dipole-coupled [3,4,6-8]. Moreover, both analyses assumed that all CO was
on atop sites; it has now been established [5] that there is a substantial and coverage-
dependent population of bridge-bonded CO. Inclusion of bridge CO in the analysis
would tend to reduce the cross section ratio below the value of 2.7 found by Hayden et al.

A beautiful experiment of Reutt-Robey et al. [9,10] is not subject to these
limitations. Using time-resolved IR spectroscopy and a pulsed molecular beam, they
studied CO diffusion from the terrace to the step edges of Pt(S)[28(111)x(110)]. They
found the cross-sections of edge and terrace atop CO to be equal within 5%.

Measurements of the Stark tuning rate also permit a straightforward interpretation,
since they involve the ratio of an electroreflectance spectrum to an RAIR spectrum
[11,12] and so do not require that the coverage be known. The Stark tuning rate of CO on
Pt(111) was measured by Luo et al.[13], that of edge CO on Pt(335) was measured by

112

Lambert and Tobin [4] and by Wang et al. [14] Between Pt(111) and Pt(335), the
estimated values differ by only 15% in the dilute limit, and are equal within experimental
error. The calculation of Greenler et al. [6] would predict a 30% difference. On the other
hand, Lambert and Tobin [4] found the Stark tuning rate of terrace atop CO on Pt(335) to
be at least eight times smaller than that of edge CO, while an electrochemical study by
Kim et al. [15,16] found a ratio of only 2.4.
2. Experiment

Details of the spectroscopy techniques and sample preparation procedures are
given in Chapter 2 and elsewhere [11,14,17,18]. We used a single lead-salt diode laser,
with a spectral range of 1947 to 2022 cm'l, as the IR source for both RAIRS and EVS.
The IR study used l3C130. This allowed atop CO to be seen with the laser, but not bridge
CO. Since the present experiment is only concerned with ratios of Stark tuning rates, the
normal measurement procedure [11] was simplified. Consequently, the EVS spectra we
display have arbitrary, but consistent units. The HREEL spectra went from 300 to
5000 cm“1 with 60-70 cm'1 resolution. All of our measurements were repeated several
times and were reproducible.

All spectroscopic measurements occurred with sample temperature 95 — 105 K.
In the IR experiments, CO and H2 were dosed by background filling while oxygen was
dosed by an effusive doser placed one sample diameter away from the sample. The doser
enhanced the effective pressure at the sample by a factor of about 20. In the HREELS
experiments, each gas was dosed through an individual closer, with enhancement factors
of about 100 over background dosing. The H and CO coverages were determined by
temperature-programmed desorption (TPD). For CO the saturation coverage was
assumed [4] to be 0.63 monolayer (ML); for H it was assumed [14] to be 1 ML (Here,
one ML corresponds to one adsorbate per surface Pt atom.) Our TPD results for all three
adsorbates are in agreement with previous measurements on stepped Pt [1,3,4,5,18,19].
The CO coverage was kept below 0.2 ML to avoid populating terrace sites [5]. The 02

113

dosage was chosen to just saturate the edge sites (as seen with TPD). On Pt(335), terrace
and edge 0 exhibit TPD peaks near 750 and 850 K, respectively [18].

For the H coadsorption experiments, the surface was first dosed with H2 near 100
K, (0.72 ML for the IR experiments; 0.25 ML for HREELS) and then with CO. Infrared
(RAIR and EV) or EEL spectra were next measured, and the sample was heated to 420 K.
A TPD spectrum taken as the sample temperature was raised showed that this desorbed
all the H, but 95% of the CO remained. After the sample cooled back to 100 K, one more
set of IR or EEL spectra was acquired. Finally the sample was heated enough to desorb
all the CO. During this desorption CO coverage was determined with TPD.

The procedure for the O coadsorption was similar. The initial 02 dose was 0.1 L
(1 L =10‘6 torr sec) and the sample temperature was 190 K. This saturated the edge sites
with O and ensured that the 02 all dissociated [18]. Next, 3.0 L of CO was dosed at 150
K, giving a CO coverage of 0.19 ML. This overlayer was studied. To get the CO to
migrate to edge sites, the overlayer was annealed for five minutes at 260 K. On Pt(112)
[1], terrace CO and edge 0 switch position at 230 K. On Pt(335) we observed a partial
switch at 230 K but it was not complete until 260 K. On Pt(335), the switching procedure
caused about 15% of the CO to react with 0. As the switch took place, the C02 signal
showed a minor peak at 180 K, with about 10% of the main peak's area, and some
desorption at 260 K as a precursor to the main COz desorption peak at 320 K.
3. Results

3.1. RAIRS and EVS

Figure 5-1 shows RAIR and EV spectra of 0.16 ML CO on a sample precovered
with 0.72 ML of H, and again after heating to 420 K to desorb the H. Desorbing the H
decreases the CO band's peak frequency 1) and increases its intensity. With H, v
=1995 cm"; after the H desorbs, v=1984 cm'l. It is well established [3,4,14] that on
Pt(335), for 13C130 on the edge, 1975<v<1985 cm'1 and on the terrace 1987<v

114

<2000 cm'1 . We therefore attribute the frequency shift to the movement of CO from
terrace to edge sites. Annealing increases the integrated area of the RAIRS band a factor
of 1.6 i 0.2. Any loss of CO during annealing would reduce the intensity. We estimate
that <0.003 ML of background 13C130 adsorbs during the annealing and cooling.

The EV spectra are proportional to d(AR/R)/dv where AR/R is the RAIRS signal
and v is the frequency. The Stark tuning rate (dv/dE) is proportional to the ratio of the
integrated EV spectrum to the RAIR spectrum, and can be estimated by comparing either
peak heights or integrated areas. The values given here are obtained from the average of
the two methods. We find, from the data in Fig. 5-1, that desorbing the H and shifting the
CO from terrace to edge sites increases the Stark tuning rate by a factor of 2.0 :i: 0.2.

Figure 5-2 shows RAIR and EV Spectra for 0.19 ML CO adsorbed on a surface
predosed with 0.1 L 02, before and after annealing at 260 K to move the CO to edge
sites. As in the H-coadsorption experiment, annealing reduces v from 2004 to 1986 cm],
and increases the intensity. Again, we attribute the shift to CO movement from terrace to
edge sites. The unusually high terrace-CO frequency is explained by the effect that O on
the step edge has on the work firnction, as discussed below. For the data in Fig. 5-2 we
find that annealing increases the intensity is by a factor of 1.4 :l: 0.2 and increases the
Stark tuning rate by a factor of 2.0 :i: 0.2. The smaller intensity enhancement with O as

opposed to H is attributable largely to loss of CO during annealing (see below).

3.2. HREELS

Figure 5-3 shows a set of EEL spectra (with 0CQ=0.05, 0.08, 0.13 and 0.16 ML)
on a Pt(335) surface precovered with 0.24 ML of H at 95 K. The spectra in Fig. 5-3a
were taken before annealing, in Fig. 5-3b after annealing to 420 K to desorb the H. As
with the IR spectra, H-desorption tends to decrease v and increase the intensity of the
atop CO band. For 0.05, 0.08, 0.13 and 0.16 ML, Av = 8, 0, 32, and 27 cm'l,

respectively. The apparent absence of a shift at 0.08 ML could be due to uncertainty in

115

identifying the peak position; the spectrum before annealing is noisier than most and the
atop peak exhibits an unusual and probably spurious asymmetry. The spectra also show
that H affects the intensity of the band due to bridge-bonded CO. We account for this as

we estimate the change in atop CO's cross section.

3.3. Analysis of the atop intensity

Table II summarizes our experimental data. The intensity ratios are obtained by
directly comparing the integrated intensity of the atop CO band before and after
annealing. Of greater interest is the cross section ratio, which represents the ratio of
integrated intensity per C0 for edge CO compared to terrace CO. The two ratios differ
because the amount of atop CO changes upon annealing. First, some CO is lost during
the anneal; for the IR experiments we estimate this loss at 5% for H coadsorption and
15% for O coadsorption; for EELS the loss was less than 5%. Second, annealing causes
CO to migrate from atop to bridge sites, and from terrace to edge sites. This effect is
visible in the EEL spectra of Fig. 5-3.

To determine the relative populations of bridge and atop CO fi'om the EEL
spectra, we need the cross section ratio between atop and bridge CO. In Table II we give
our best estimate for this quantity and its uncertainty (including both random and
systematic error). For the IR experiments we use the bridge CO correction determined
from the EELS data at 9C0 = 0.16 ML.

Before annealing, the bridge CO is presumed to be on the H-saturated step edge
— terrace bridge sites are not occupied at these coverages [5,18]. Previous EELS
measurements have shown that the cross sections of edge bridge and edge atop C0 are
equal [18]. Any bridge CO on the terrace would, by analogy with Pt(111) [20], be
expected to have an EELS cross section 1.8 times smaller than that of terrace atop CO. It
is clear fi'om TPD that all the CO after annealing is on the step edge, but without H, so

116

again the bridge CO cross section is between 1.0 and 1.8 times smaller than the atop CO
cross section. Since the amount of bridge CO is relatively small, the value of the atop
cross section ratio is rather insensitive to our assumption about the bridge CO cross
section, as reflected in the uncertainties in Table II.

All the data in Table 11 show that the vibrational cross section of edge CO is larger
than that of terrace CO. We estimate that the vibrational cross section of edge atop CO is
2.0 :1: 0.2 times that of terrace atop CO.

The Stark tuning rate data in the last column of Table II are unaffected by loss of
CO or migration between sites. To get (dv/dE), a peak in the integrated electroreflectance
spectrum is divided by a peak in the RAIR spectrum. Since both of these are proportional
to the atop CO coverage, the coverage cancels out. We find that the Stark tuning rate of
edge atop CO is also 2.0 i 0.2 times that of terrace atop CO.

4. Discussion

We have developed an electrostatic screening model that explains much of our
data. Before describing our model, we consider alternative explanations.

One possibility is that CO is chemically different at edge and terrace sites. A
previous EELS experiment [21] that looked at the coverage dependence of overtone
intensities was unable to find any chemical difference between edge and terrace CO.

Since coadsorbates are used to control CO's binding site, it is possible that the
effects we observe are caused by a direct interaction of H or O with CO. Coadsorbed H
and CO on Pt(335) has been studied with RAIRS, HREELS and TPD [14,18]. Edge CO
is strongly affected by coadsorbed H. In the present work, however, the H is all gone
when the CO is at edge sites. The vibrational properties of terrace C0 are completely
insensitive [18] to the presence of terrace H. The effect of edge H on terrace CO is also
expected to be negligible. We believe that H desorption affects CO's vibrational intensity
and Stark tuning rate primarily through the CO's change in adsorption sites.

117

We have less information about the effects of coadsorbed 0, but the strong
similarity we see between the effects of H and O is consistent with a shift of CO from
terrace to edge sites also being the most important effect. As seen in Fig. 5-2a,
coadsorbed 0 does cause an unusually high v for terrace CO, but this is easily explained.
It is well established [14] that a change in work function b induces a proportional Av for
CO, as discussed in Chapter 3, with a tuning rate (dv/db) = 34 :i: 4 cheV. Collins and
Spicer [22] found that filling the edge sites with O on Pt(S)[6(111)x(100)] causes a 0.4
eV increase in b, while saturating the terrace sites with 0 increases b by about 0.1 eV,
and filling the edge sites with CO increases b by only 0.03 eV. With the step edge
saturated with O, we would expect an O-induced Av of +13 cm'1 for the terrace CO. The
observed Av is 10 :i: 2 cm’1 (the estimated v for no 0 takes 9C0 into account). After
annealing, with O on terrace sites, the expected Av for the edge CO is only 3 cm'l, the
observed Av = 3 :i: 2 cm“.

Our electrostatic model is similar to a model for the Pt(335) surface described by
Greenler et al.[6]. We know that the step height between (111) terraces is 2.26 A. For
CO on Pt(111) Ogletree et al. [23] used LEED to show that center of the CO bond is 2.43
A outside the top layer of Pt atoms. However, the electrical surface is at the image plane,
not at the outside layer of Pt atoms. On Pt(100), spectroscopic studies [24] of image
potential-induced image states have shown that the image plane is 1.05 A outside the
outermost Pt atoms. This suggests that on Pt the center of the CO bond is approximately
1.38 A outside the image plane. In our model we represent the stepped surface by an
ideal ‘conductor with sharp, perfectly rectangular steps. The electrostatic field is applied
by an equipotential plane, parallel to the average (335) surface and 100 step-heights
away. We consider the field component normal to the average surface plane. The field is
evaluated at points along a line parallel to the terrace, and half a step height (1.13 A)
above it as shown in the inset in Fig. 5-4. The results are not strongly sensitive to the
height assumed. Greenler et al. [6] studied the enhancement dependence on the height of

118

the center of the C=O bond to the image plane; from 0.9 to 1.23 A, or fiom 1.23 to 1.4 A,
the result only changes by 5%. Our model ignores smoothing of the electron density,
variation of the field over the spatial extent of the CO and possible tilts of the molecules.
On the other hand, an explicit study found that CO does not tilt by more than 10° on
Pt(335) [30]. Greenler et al. also considered the effects of tilting in their model. A tilting
of 10° is only going to make a difference of 5%. The abrupt metal-vacuum interface and
sharp step edge used in the model tend to overestimate the field enhancement near the
edge.

Figure 5-4 shows the calculated magnitude of the electric field normal to the
average surface plane as a function of distance from the step edge, for two surfaces, with
ratios of terrace-width to step-height of 3.87 and 29.2, which simulate Pt(335) and the
surface used by Reutt-Robey et al. [9,10], respectively. The field is normalized to the
field on a flat surface, so a value smaller than 1 means screening, bigger than one means
enhancement.

For the (335) surface, the field at the step edge is enhanced by 15% relative to the
flat surface, while the field in the center of the terrace is suppressed by 12%; thus the
effects of " lightning rod" enhancement at the edge and screening on the terrace are
comparable. The field at the edge is 1.3 times larger than at the center of the terrace, in
good agreement with the value of 1.31 found in Ref. 6. This difference in field between
edge and terrace sites implies a factor of 1.5 — 2 difference in vibrational cross section, in
good agreement with our observations. Also, a factor of 1.3 difference is expected
between the IR cross sections of edge CO on Pt(335) and CO on Pt(111); this is
consistent with the observation of Lambert and Tobin [4] that these two cross sections are
roughly equal. Our field calculation also agrees with the observed difference (~15%) in
Stark tuning rate between edge CO on Pt(335) [4,13], and CO on Pt(111) [12]. In the
dilute limit these tuning rates were found to be 8.8 :t 0.9 cm'l/(V/A) and 7.6 :i:

1.6 cm'1/(V/A), respectively.

119

For the surface with wide terraces, the calculated screening on the terrace is much
smaller than it is for the (335) surface. At the height of a CO molecule, the field at the
step edge and at the center of the terrace differ by 18%, which corresponds to 1.4 times
difference in the cross section. Because the field at the step of this surface varies rapidly
around the step edge, it is conceivable that if we round off the sharp step edge in our
model, the enhancement will be reduced. It is possible that the field induced cross
section on that surface is only 1.2, that is still bigger than the limit set by Reutt-Robey et
al. Another possibility is that they have edge bridge CO that is not observed in their
spectrum, but at the very low coverages they used, it is quite unlikely. Nevertheless, our
calculation shows that the cross section difference between edge and terrace CO
decreases with increasing terrace width.

The electrostatic model explains: (I) The enhancement of the IR and EEL cross
sections, and the Stark tuning rate, of edge CO relative to terrace CO on Pt(335) (this
work); (2) The small difference in cross section and Stark tuning rate between edge CO
on Pt(335) and CO on Pt(111) [4,13,14]. The model also partly explains the lack of any
significant cross section enhancement for edge CO on a surface with 28-atom-wide
terraces [9,10]. The strong dependence of the cross section enhancement on terrace width
is particularly strong evidence for an electrostatic mechanism, since a chemical
mechanism would be expected to have shorter range.

Despite the model's success, not all of the Stark tuning rate results are explained.
Our EVS data (see Figs. 5-1 and 5-2 and Table 11) show that the Stark tuning rate of edge
CO is a factor of 2 greater than that of terrace CO — a ratio as large as the enhancement
of the IR intensity. The electrochemical experiments of Kim et al. [15,16] also indicate a
tuning rate enhancement of at least a factor of 2; with aqueous electrolyte they found a
tuning rate (with electrode potential) of 75-80 cm“N at low coverage (edge CO) and
33 cm'1/V at high coverage (both edge and terrace CO); in contrast, Ref. 25 shows that on
Pt(111) at low CO coverage the tuning rate with aqueous electrolyte is 40-44 cm‘1/V.

120

(Discrepancies in absolute Stark tuning rate between electrochemical and vacuum studies
are discussed in Refs. 13 and 14.)

Since the Stark shift of adsorbed molecules is linear in the applied field while the
IR absorption is quadratic, a simple electrostatic model would predict that the tuning rate
ratio should be equal to the square root of the cross section ratio, i. e. smaller, not equal as
we find. Our results therefore suggest a difference in screening between the static and IR
fields. Similar differences have been noted previously. Both Luo et al. [13] and Wang et
al. [14] have presented evidence that the suppression of the Stark tuning rate by
coadsorbates is significantly stronger than would be predicted from the suppression of the
IR intensity due to dipole screening.

One possible explanation for the discrepancy in observed ratios lies in the fact that
the local electric field varies strongly with position at a metal surface. The field used to
calculate it absorption or the Stark effect should actually be a weighted average over the
molecule, perhaps including even nearby regions of the metal. The field is averaged in
different ways for ir absorption and the Stark effect. Instead of interacting with the local
field at the center of the C-0 bond, the molecule interacts with the local field at the
positions of the C and 0 nuclei. It has been shown theoretically that ir absorption really
does depend on the local field at the nuclear positions in the actual molecule. [27,28] This
makes sense since we are looking at molecular vibrations that occur as the nuclei move.
The nuclei move because an electric field acts on the nuclear charge. most theoretical
treatments, however, consider the local field at the nuclear positions with the molecule
removed. For the vibrational Stark effect, an argument that relates the field with the
molecule removed to the molecule's response is given by Lambert [29]. Presumably the
vibrational Stark effect could also be expressed in terms of the local field at the nuclei.
As an extreme example, suppose that the local field at the C and 0 nuclei contribute
equally, both for the vibrational Stark effect and for ir absorption. Then our data would

be consistent with equal screening at all of the 0 sites and at the C site on the edge, but

121

complete screening at the C site on the terrace. Since screening, whether by a nearby step
or by neighboring adsorbates, can affect not only the magnitude but also the spatial
variation of the field, it is possible for screening to have different effects on these
different properties, as we observe.

In order for C nucleus to be completely screened, it has to be inside the image
plane. There are several ways to estimate the relative position of the C nucleus and the
image position on Pt(111). For atop CO, a LEED study [23] found that the C-Pt bond
length is 1.85 A, and the C-0 bond length is 1.15A.

The jellium edge on Pt(111) is 1.13 A outside the topmost Pt layer. By linear
extrapolation from Lang and Kohn's calculation [31] (for rs > 2 an.) to 1.45 a.u., the
radius of a Pt atom containing one valence electron, the image plane is 0.89 A outside the
jellium edge. So the C nucleus is 0.17 A inside the image plane while the O nucleus is
0.98 A outside the image plane. It is easily conceivable that the image plane will be
closer to the metal at the step site. Screening of an external electric field at A1(100) [32]
and Ag(100) [33] has been calculated by the surface embedding method. The screening
charge is found lay on top of the surface atoms, which means the effective image plane
position is lower in the open area of the surface. The difference on the flat Ag(100)
surface is in 0.23 A at 1.85 A outside the topmost atom layer. Moreover, a calculation
based on the surface states energies [24] have found that image plane moves closer to the
metal by just going from (111) to (100) faces. The difference is in the order of 0.1 to 0.2
A for Ni, Ag, and Au. So it is possible that at the step site, the C nucleus is just outside
the image plane. That can explain our data for same enhancement ratio for both IR cross
section and the Stark tuning rate.

On the other hand, there are several quite different estimates about the position of
the image plane. One comes from fits of the standard model to RAIR spectra of CO on
Pt(111). The fit is best if the center of the C=O bond is 1.1 A outside the image plane,
which means that the C nucleus is 0.52 A outside the image plane. Ref. 24 gives that on

 

122

Pt(100) the image plane is 1.05 A outside the topmost Pt atoms, while the distance should
be larger on Pt(111) by about 0.1 to 0.2 A. This means that the C nucleus is about 0.6 to
0.7 A outside the image plane. If this is true, then there should be no big difference in the
level of screening of the applied field at terrace or edge site at C nucleus other than what
is calculated in our model.

We want to point out that even if the first estimation is true and the big difference
in CO's Stark tuning rate at the two sites is caused by the change in C nuclei's relative
position to the image plane, that still would not explain the different screening of IR and
DC field on Pt(111) [13] and Pt(335) [14], because in those experiments all the CO stays
in equivalent sites.

Finally we discuss an unresolved experimental discrepancy. The Stark tuning rate
of terrace atop CO measured in this study, as well as in the electrochemical work of Kim
et al. [15,16] is much larger than reported by Lambert and Tobin [4], who used the same
apparatus and the same crystal as we did here. Figure 5-5 shows our RAIR and EV
spectra for 0.26 ML CO on Pt(335). At a comparable coverage Lambert and Tobin's
RAIR spectrum looked similar, with edge and terrace CO peaks of comparable intensity.
Their EV spectrum, however, showed a strong EVS peak corresponding to edge CO, but
no EVS signal from terrace CO. In fact they observed a small peak at the terrace CO
vibrational frequency, where a zero-crossing would ordinarily be expected. They
concluded that the Stark tuning rate of terrace CO is at least a factor of eight smaller than
that of edge CO. Our EV spectrum shows EVS peaks of comparable size for both CO
species, and we find only a factor of two difference in Stark tuning rates.

We have no firm explanation for this discrepancy. Both Lambert and Tobin's
result and ours were reproduced many times. We investigated coadsorption with O and
H, as well as C contamination, but were unable to reproduce Lambert and Tobin's results.
We offer two observations: the sample was repolished between the two sets of

experiments, and Lambert and Tobin observed Sn contamination — although the Sn

123

concentration was below the limit of Auger detection for their final experiments. We
note that the EELS experiments of Luo et al. [21], which were aimed at explaining the
large difference in Stark tuning rate between edge and terrace CO, were performed before

the crystal was repolished.

5. Conclusion

We have compared the vibrational cross section and the Stark tuning rate for
terrace and edge atop CO on Pt(335) using RAIRS, EVS and HREELS. The CO
adsorption site was controlled by coadsorption of H and O. The cross section of edge
atop CO is 2.0 d: 0.2 times greater than that of terrace atop CO and the ratio of the Stark
tuning rates is also 2.0 i 0.2. The cross section ratio agrees senriquantitatively with a
classical E-field calculation. The model is able to partly account for the much smaller
difference in cross sections observed by Reutt-Robey et al. [9,10] on a Pt surface with
much wider terraces. We conclude that there is little chemical difference between edge
and terrace atop CO, and that the difference in E-field strength between edge and terrace
sites largely accounts for the variation in vibrational intensity and Stark tuning rate.

The ratio of the Stark tuning rates at the two sites is larger than would be expected
fi'om a simple model and the observed ratio of IR intensities. This discrepancy is
consistent with other experiments that have found a significant difference in the screening
of static and IR fields [13,14]. It is possible that IR intensity and Stark tuning rate
measurements are probing different aspects of the E—field distribution on the surface.

Our determination that the Stark tuning rate for terrace atop CO on Pt(335) is only
two times smaller than that of edge atop CO is in agreement with an electrochemical
study [15,16], but contradicts the previous experiments of Lambert and Tobin [4]. This
difference remains unexplained, but suggests that the Stark tuning rate of terrace CO may

be sensitive to surface preparation.

124

Table II. Ratios of vibrational intensity, vibrational cross section, and Stark tuning
rate of edge atop CO compared to terrace atop CO. The intensity and Stark tuning rate
ratios are determined directly from the experimental data. The cross section ratios
include corrections for loss of CO during annealing and for migration between bridge and

atop sites, as discussed in the text.

 

Experiment Intensity Cross section Stark tuning rate

ratio ratio ratio

 

IR — H coadsorption

900 = 0.16 1.6 :t 0.2 2.1 i 0.3 2.0 i 0.2
IR — O coadsorption

0“, = 0.19 1.4 :l: 0.2 2.0 d: 0.4 2.0 i 0.2
HREELS — H coadsorption

000 = 0.05 2.3 1.4 :l: 0.4
em = 0.08 2.6 1.8 r 0.4
900 = 0.13 2.2 2.3 r 0.3
0CO=0.16 1.7 21:02

 

Average: 2.0 i 0.2 2.0 :1: 0.2

 

125

 

 

  
 
 

 

 

 

 

 

 

 

 

 

 

RAIRS Before H desorption EVS Before H desorption
A _
1'3
'5 Afler H
S W e .
<1 ' 5
a?
1...
i
i
1950 1970 1990 2010 1950 1970 1990 2010
Frequency (cm‘i) Frequency (cm")

Figure 5-1. RAIR and EV spectra for 0.16 ML CO on a Pt(335) surface precovered with
0.72 ML of H, before and after annealing the sample at 420 K. Upon H desorption, the

CO moves to edge sites.

126

 

 

 

 

 

 

 

 

 

 

LRAIRS Before 0 exchange EVS Before 0 excihangeJ
i- d
A
s r l
’E
3 cl
5 e l .
3 “ '
V
J: " . -‘
After 0 A
. exchange .
. I
t’_L_..__11 A I . _1._.J. 1.1—W
1950 1970 1990 2010 1950 1970 1
Frequency (cm“) Frequency (cm")

Figure 5-2. RAIR and EV spectra for 0.14 ML CO on a Pt(335) surface predosed with
0.1 L 02, before and after annealing the sample at 260 K. Upon annealing, the CO

moves fi'om terrace sites to edge sites, while the 0 moves to terrace sites.

127

 

intensity (Arb. Unite)

 

 

 

 

it:

404

1070

(a)T Before anneal
\

1006

1000\

1079
\

 

T

‘i

F~
'92

intensity (Arb. Units)

’2070

F

2070

1

 

 

1000

Loss Energy (end)

 

I T I

(b) After 420 K anneal

 

 

 

 

 

 

 

 

 

 

OP
g

Figtu'e 5-3. HREEL spectra for 0.16, 0.13, 0.08, 0.05 ML CO ( from top to bottom) on a
Pt(335) surface precovered with 0.25 ML of H, before and after annealing at 420 K.

 

128

 

1.2 - pt(s)[28(111) x (100)]

 

 

 

 

 

 

 

 

 

 

u?
-e
O
.5
'6
E ,, ”L
2 0.4 r E31"; d
pt(s)[4(111) x (100)] M
t 220 \" L ‘
0.00 - . . . 50 . . ‘ ‘ 100

x/L(%)

Figure 5-4. Calculated B field normal to the average surface plane, as a function of
fractional distance across the terrace. The field is normalized to the field on a flat
surface, and is calculated along a line one-half step height above the terrace, as shown in
the inset. The dotted curve represents a surface with narrow terraces similar to Pt(335);

the solid curve represents a surface with much wider terraces, similar to that used in Refs.

Sand 9.

129

 

AR/R

  

2%

 

 

%
1950 1970 1990 2010
Frequency (cm‘i)

 

l I I l

S. (Arb. Units)

 

 

 

 

1_
1950 1970 1990 2010
Frequency (cm“)
Figure 5-5. EV and RAIR spectrum for 3.5 L (0.26 ML) CO on clean surface. Strong
EV features are seen corresponding to both of the peaks in the RAIR spectrum, indicating
that edge and terrace CO have comparable Stark tuning rates. The zero-crossings in the
EV spectrum occur at the same fi-equencies as the peaks in the RAIR spectrum. These
results are in contrast to those reported in Ref. 4.

130

References

l. A. Szabb, M. A. Henderson, and J. T. Yates, Jr., J. Chem. Phys. 96, 6191(1992).

2. E. Hahn, A. Fricke, H. Rbder, and K. Kern, Surf. Sci. 297, 19 (1993).

3. B. E. Hayden, K. Kretzschmar, A. M. Bradshaw, and R. G. Greenler, Surf. Sci. 149,
394 (1985).

4. D. K. Lambert and R. G. Tobin, Surf. Sci. 232, 149 (1990).

5. J. S. Luo, R. G. Tobin, D. K. Lambert, G. B. Fisher, C. L. DiMaggio, Surf. Sci. 274, 53
(1992).

6. R. G. Greenler, J. A. Dudek, and D. E. Beck, Surf. Sci. 145, L453 (1984).

7. F. M. Leibsle, R. S. Sorbello, and R. G. Greenler, Surf. Sci. 179, 101 (1987).

8. R. K. Brandt and R. G. Greenler, Chem. Phys. Lett. 221, 219 (1994).

9. J. E. Reutt-Robey, Y. J. Chabal, D. J. Doren, and S. B. Christrnan, J. Vac. Sci.
Technol. A 7, 2227 (1989).

10. J. E. Reutt-Robey, D. J. Doren, Y. J. Chabal, and S. B. Christrnan, J. Chem. Phys. 93,
9113 (1990).

11. D. K. Lambert, Appl. Optics 27, 3744 (1988).

12. D. K. Lambert, J. Chem. Phys. 89, 3847 (1988).

13. J. S. Luo, R. G. Tobin, and D. K. Lambert, Chem. Phys. Lett. 204, 445 (1993).

14. H. Wang, R. G. Tobin, and D. K. Lambert, J. Chem. Phys. 101, 4277 (1994).

15. C. S. Kim, W. J. Tomquist and C. Korzeniewski, J. Phys. Chem. 97, 6484 (1993).
16. C. S. Kim, C. Korzeniewski, and W. J. Tomquist, J. Chem. Phys. 100, 628 (1994).
17. B. A. Sexton, J. Vac. Sci. Technol. 16, 1033 (1979).

18. H. Wang, R. G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio,
unpublished.

19. D. M. Collins and W. E. Spicer, Surf. Sci. 69, 85 (1977).

20. W. D. Mieher, L. J. Whitman and W. Ho, J. Chem. Phys. 91, 3228 (1989).

131

21. J. S. Luo, R. G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio, J. Chem.
Phys. 99, 1347 (1993).

22. D. M. Collins and W. E. Spicer, Surf. Sci. 69, 114 (1977).

23. D. F. Ogletree, M. A. Van Hove, and G. A. Somorjai, Surf. Sci. 173, 351 (1986).

24. N. V. Smith, C. T. Chen, and M. Weinert, Phys. Rev. B 40, 7565 (1989).

25. L.-W. H. Leung, A. Wieckowski, and M. J. Weaver, J. Phys. Chem. 92, 6985 (1988).
26. H. Wang, R. G. Tobin, D. K. Lambert, G. B. Fisher, C. L. DiMaggio, to be submitted
to J. Chem. Phys.

27. P. Lazzretti, and R. Zanasi, Phys. Rev. A24, 1696 (1981).

28. P. W. Fowler and A. D. Buckingham, Chem Phys. 98, 167 (1985).

29. D. K. Lambert, Solid State Comm. 51, 297 ( 1984).

30. J. S. Sommers, T. Lindner, M. Surman, A. M. Bradshaw, G. P. Williams, C. F.
McConville, and D. P. Woodruff, Surf. Sci. 183, 576 (1987).

31. N. D. Lang, and W. Kohn, Phys Rev. B, 7, 3541 (1973).

32. J. E. Inglesfield, Surf. Sci. 188, L701 (1987).

33. G. C. Aers, J. E. Inglesfield, Surf. Sci. 217, 367 (1989).

132
Chapter 6

Conclusions

In conclusion, by using RAIRS, EVS, and HREELS techniques, I obtained several
significant results about the influence of substrate morphology on the relative interaction
strength among the adsorbates on the surface, as well as the field distribution on stepped
surfaces.

First, 1 demonstrated that coadsorbed edge CO and H do not segregate on this
surface, even though the interaction between them is still repulsive. This is in sharp
contrast to earlier results of coadsorption of H and CO on both Pt(111) and on the step
edge of Pt(l 12), two surfaces structurally similar to Pt(335). Instead, H and CO mix into
one dimensional islands along the step edges on Pt(335). Within such islands, CO is
shifted by H from atop to bridge sites.

The different overlayer structures show that the relative interaction strength is
strongly influenced by the substrate morphology. A small change in the substrate can
introduce drastic change in the overlayer structure. The difference between the (335) and
(112) results may be related to the idea of "quantum corrals" [1]. As the H-CO
interaction is mostly indirect, through metal, the strength depends strongly on the
perturbation of the substrate charge density. When we put CO and H together, how much
the changes they produce in the substrate charge density match one another is the
deciding factor of their interaction. The match can be modulated by the terrace width, as
the perturbation wave will be reflected at the next step edges.

I confirmed the proposed site shifting of CO by H with HREELS experiment. I
found that coadsorbed H continuously shifts edge CO from atop to bridge sites; this
process is almost complete. With this site shifting, I compared the cross section of edge

atop and bridge CO. Surprisingly, first, H has a big effect on the cross section of edge

133

bridge CO; second, edge bridge CO has almost the same cross section as edge atop CO,
at least in a H-rich environment. This is very different from the well established result
that on Pt(111), atop CO has a cross section 1.8 times bigger than that of bridge CO. I
also found that edge atop CO's cross section increases by about 20% with increasing CO
coverage; this is in qualitative agreement with a theory of coadsorbate effects. On the
other hand, the effect expected from theory is much smaller than 20%, let alone the huge
difference in edge bridge CO's cross section observed by us. Our results demonstrate that
the coverage-dependent cross section is much more complex than current theory predicts.

I also studied the coadsorption of terrace CO with H on Pt(335). Coadsorbed H
has no observable effect on the HREEL spectrum of terrace CO. This is very surprising,
since coadsorbed H and CO segregate on Pt(111) and Pt(112), and even on the step edges
of Pt(335) coadsorbed H has big effects on edge CO. Our TPD data also suggest that the
interaction between terrace CO and terrace H is very weak. The reason for this is not
clear. We offered several possibilities: H may go to subsurface sites, H and CO may
occupy difi‘erent rows on the surface naturally, and H may influence some of the CO that
is not observed by our spectroscopy tools. This result again demonstrates the strong
influence of substrate morphology on overlayer structure.

This study is also significant for its direct comparison of the vibrational cross
section and the Stark tuning rate for terrace and atop CO. We found that the cross section
of edge atop CO is 2.0 :l: 0.2 times greater than that of terrace atop CO and the ratio of the
Stark tuning rate is also 2.0 :i: 0.2. The cross section ratio is in qualitative agreement with
a classical E-field calculation. In contrast to previous belief that only field enhancement
at the step edges is important, this model shows that the screening of the field on the
terraces is equally significant. The model is also able to partly account for the much
smaller difference in cross sections for CO on Pt surfaces on much wider terraces
observed by others. This agreement demonstrates that there is little chemical difference

between edge and terrace CO, in agreement with a previous HREELS study [2].

134

The ratio of the Stark tuning rate for CO at the two sites is larger than would be
expected from the E-field calculation and the observed ratio of IR intensities. Our data
also show that the coverage-dependent screening of the DC and IR field is different, in
agreement with a previous finding for CO on Pt(111). It is possible that on step surfaces,
the screening and enhancement of applied fields vary significantly over the size of the CO
molecule, and the CO response in RAIRS and EVS, depends on different spatial
averages. On the other hand, this still would not explain the difference in coverage-
dependent screening, which is in contradiction with current models of depolarization
within the overlayer.

This work is also significant in providing another direct comparison of the Stark
tuning rate measurement between vacuum and electrochemical experiments. I observed
the Stark tuning rate for edge CO at low coverage to be 88 i 9 cm"1 (WA), in agreement
with a previous vacuum study, but 3 times smaller compared to electrochemical results if
conventional double layer models are used. This is similar to with a previous comparison
between UHV and electrochemical studies for CO on Pt(111). Coadsorption of H also
produces different results. In vacuum, CO's Stark tuning rate is not changed by
coadsorbed H but in an electrochemical cell it goes to zero in classical hydrogen region.
These results indicate that the electrochenrical double layer is probably more complex
than we thought. A better understanding of it can be achieved by models that can
explain the difference between the UHV and electrochemical studies.

In summary, I found and analyzed several intriguing results in the overlayer
structure, relative interaction strength among the coadsorbates, and field distribution on
the surface because of the existence of the steps. I have modeled the results, with
overlayer structure and electrostatic calculations and offered speculative explanations for
still unexplained results. These results will be interesting both as a model for practical
catalysis or in distinguishing between electrostatic and chemical effects in chemisorbed

systems. They are also very useful for theorists working on the understanding of

135

chemisorption and interactions among coadsorbates, as well as on the understandings the
complex response of metal surface to applied electric fields, and the understanding of
electrochemical double layers.

Future experiments can be carried out in several directions that address the
unanswered questions and significantly enhance our understanding in the two areas. For
example, edge CO and H coadsorption could be studied on a series of samples with
various terrace width. A Detailed STM study could examine the perturbation of the local
surface density of states by adsorbates on high index surface. UHV water coadsorption
experiments have been used to model the electrochemical double layer [3], but CO’s
response to applied field has not been probed under such conditions. An EVS study of
water’s influence on CO’s Stark timing rate could be very usefirl in understanding the
different results and gaining a better understanding of the electrochemical double layer.
Stark tuning rate measurement of other species would also contribute to the
understanding of how admolecules respond to applied fields. Isotope mixture
experiments could also help clarify the origin of the different screening of IR and DC
fields.

136

References

1. M. F. Cromrrrie, C. P. Lutz, and D. M. Eigler, Nature, V363, 524 (June 10, 1994).

2. J. S. Luo, R. G. Tobin, D. K. Lambert, G. B. Fisher, and C. L. DiMaggio, J. Chem.
Phys. 99, 1347 (1993).

3. F. T. Wagner, in " The Structure of Electrified Interfaces, " Vol. 2 in the series
"Frontiers of Electrochemistry," edited by J. Lipkowski and P. N. Ross, Jr., VCH
Publishers.

"illilllllllllliltill“