mmvyvv—u—u

 

 

 

 

 

 

 

 

   

TU‘KV—‘— _—fi

 

w—v qu
” ‘ .2'

| u

 

 

“m
. -rm 1‘
l ‘ ‘A ‘h t
.
r I
V A

 

       
     

         

 

 

       

 

   

 

   

   

 

     

   

 

   

   

             

 

 

' -‘ . ‘ . _ ' . z ' . . ‘ v. ‘ V .
’ , A ' ' . V ' ‘ A. '
' ‘ ‘ h‘ ' H . f h . ‘ .Z ‘ ‘ >
I - ‘v V ' I V ‘ ‘ . ‘ C". ' _ . .
r , , , , ‘ . . . : v . . . t. ,
- ‘ , ‘ r V *. . - , , _ ‘t , ‘ ‘ ‘
. 4 _ ‘ _ ‘ 7‘ _: . _.
’ " ' : . , ‘ '. I -* < " ' z ‘ A ,
. ; . . \ ‘ ‘ ‘ . ‘ —. .
a . , ' - - . ~ 1 '.
. ‘ - .' A ’ , j ’ ‘ ‘
- . . .7 V ‘ . , ‘
\ ~ 7 . ' . i ‘ V .'. _ - A , - . ,-
x . . . ‘ . .. , _,. ‘ A A A I .
I V ‘ ‘ . ’ h < ' r 1 .
. ‘ ~ ' ‘ " V . ' ' " 1 '
‘ I ' 2 ‘ ‘ A ' ' ‘ >
! ‘ 7' I ‘ . ’ ~ ‘ "' u ’ ' , ‘ ‘ .‘v‘
.' . . , ‘ . . A - _ . - . .
, ., . _ . . L 7 r. . 7 ‘. .
s:.; h I H .‘ . , , >‘ ‘ ;>‘ ‘ . In .1
v . . A . ‘. , ‘ . . . . . . ‘ .'. -.
‘ ' ' . ‘1 * - " ‘ ‘u f _ ‘ A
- ‘ . . ‘ ~ .v- ‘
n . .. 4‘ ,V‘ . . . ‘ , , . ;. . v
; . , . ‘ _ , , .7... .....‘. .. — org—3411‘ 5-54
..‘ 1,“. . . - . - . .‘u ..n '1 .‘v u ‘ nu . . . ‘»;..‘....1-a ' -».u .r-- ‘ ‘ ‘ u. n .V. ‘ . ...... ..y. 7 n .‘ - .o: ‘ 7 . ‘

 

 

VERSITY LIBRARIES

LIBRARY |\llllll\llllllllllllll W. \\ \fll ll ..

Michigan State
University

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

dissertation entitled

Electron Dynamics as Observed by Adsorbate—Induced
Broadband Infrared Reflectance Change Measurements

presented by i

Dennis Eugene Kuhl

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Physics

em:

Roger G. Tobin

Major professor

 

 

 

Date August 7, 1996

 

MS U i: an Affirmative Action/Equal Opportunity Institution 0- 12771

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Is An Affirmative MIoNEquol Opportunity Institution
l Wanna-m

 

———-—~—~r._

ELECTRON DYNAMICS
AS OBSERVED BY ADSORBATE-INDUCED BROADBAND INFRARED
REFLECTANCE CHANGE MEASUREMENTS
By

Dennis Eugene Kuhl

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

1996

ABSTRACT

ELECTRON DYNAMICS
As OBSERVED BY ADSORBATE-INDUCED BROADBAND INFRARED
REFLECTANCE CHANGE MEASUREMENTS
By

Dennis Eugene Kuhl

In some cases gas adsorption on clean metal surfaces induces a negative
broadband change in the metal’s infrared (IR) reflectance AR/ R which is large compared
to the adsorbates’ dielectric properties. A model has been proposed. which explains
AR/ R in terms of diffuse conduction electron scattering by the adsorbates and relates
AR/ R to a wide variety of physical phenomena including a resistivity change. In this
study I experimentally test the model by measuring AR/ R as a function of frequency,
temperature, and coverage for CO on Pt(l 11). This marks the first time the scattering
model has been applied to a transition metal. A second set of experiments tests the
predicted linear relationship between AR/ R and tAp , the resistivity change (times the
sample’s thickness), by measuring both effects simultaneously for O and formate
adsorption on an epitaxial Cu(100) thin film.

The frequency dependence of AR/ R for CO on Pt(111) is at least qualitatively

consistent with the theoretical prediction.

The magnitude of AR/R at 2500-2800 cm'I is smaller at 90 K than at room
temperature by about 30%. This result appears surprising, since the scattering model
predicts that |AR| should be larger at the lower temperature. The clean surface
reflectance R, however, is also larger at low temperature, with the result that a decrease in
IAR/Rl at these frequencies is consistent with the theory.

The magnitude of the reflectance change peaks at a coverage of about 0.33 ML,
and decreases toward saturation. The slope of the curve at low coverage corresponds to a
scattering cross section per CO on the order of 1 A2. The presence of a peak in the
coverage dependence could be attributable either to partial ordering in the overlayer or to
changes in CO's electronic structure.

Simultaneous measurements of AR/ R and tAp confirm linearity for 0 adsorption
on Cu(100), although there is a large discrepancy between the predicted and measured
slopes. The linearity test for formate adsorption on an O predosed Cu surface was
inconclusive. As the formate replaces the O on the Cu( 100) surface, AR/ R appears to
return very close to zero. tAp , however does not return nearly as close to zero. This
difference, along with the possible lack of a linear relationship for formate, could indicate

that the scattering model does not apply to formate adsorption on Cu(100).

‘B.N.J. Persson and AI. Volokitin, Surf. Sci. 310 (1994) 314 .

to my Savior

iv

ACKNOWLEDGMENTS

I want to first thank my wife, Lesley, for her support and encouragement
throughout this Ph.D. project. She has been a wellspring of faith and love, as well as an
example of perseverance. Her patience during the many days I was away completing this
dissertation is greatly appreciated. Knowing that her prayers and thoughts were always
with me provided me with powerful incentive to complete this long-term project.

I owe a great debt of gratitude to my parents, Dale and Shelva Kuhl, for the firm
foundation they provided me. It was only through their sacrifices that I was able to attend
college and pursue the career in science of which I’d always dreamed. Moreover, their
love, faith, and prayers have given me an anchor to depend on all my life.

While thanking family members for their contributions to my efforts, I would be
remiss if I didn’t mention my brother and sister, Douglas Kuhl and Stephanie Kuhl
Lowman. I can always count on Doug to provide a respite from academic matters,
usually via a golf game. Furthermore, he is always willing to help however he can, from
moving my things to letting me stay over at his house in Canton. Steph and her husband
Walt deserve recognition especially for their willingness to provide me with a place to

stay on numerous occasions during the process of writing this dissertation, along with

occasional meals. It was always relaxing to go to their home at the end of a frustrating
day of writing.

Two church congregations have contributed to the pursuit of my goals over the
years. I was raised in the congregation at Moreland Christian Church, and it was through
them that my system of values and spiritual foundation was built. Upon coming to
graduate school I was very fortunate to have the opportunity to participate in the
congregation at University Christian Church. They helped me to grow spiritually during
the pursuit of my Ph.D.

Numerous graduate students and friends have played a role in these efforts. First
my Science Theatre friends: Bill Abbett, Dave Bercik, Danielle Casavant, Joy Conrad,
Jennifer Discenna, Erik Hendrickson, Gerd Kortemeyer, Jeff Kriessler, Rod Lambert,
Normand Mousseau, and Steve Snyder. Several friends provided me with a place to stay
during the writing process while I made my home out of town: Larry and An Heimann
(and Alex and Mark!), Jeff Schubert (the most comfortable sofal), Jeff Kriessler (too bad
the organ wasn’t finishedl), Carl and Patricia Hoff (and Scott!).

This work was supported in part by the Petroleum Research Fund, The Center for
Fundamental Materials Research, and the NSF under Grant No. DMR-9201077 and
DMR-9400417 (MRSEC). Helpful communications with CT. Campbell and M.
I-Iugenschmidt (through Campbell) contributed to the closer modeling in Chapter 3.
R. Naik graciously shared her expertise in the technique of growing the thin films.

I want to thank Prof. Don Jacobs of the College of Wooster for encouraging me to

pursue graduate school. The example he set of a teacher-scholar continues to be an

vi

inspiration to me. Prof. Jerry Cowen of Michigan State University was also a valuable
source of encouragement.

My coworkers in our research group deserve special citation. The UHV and IR
apparati were assembled by Chilhee Chung. Keng-Ching “Kathy” Lin patiently
instructed me in the operation of the system, and pioneered a great deal of the work
contained herein. Hong Wang was always a source of encouragement. Larry Voice
contributed to the thin film work and shared my love of junk food. Evstatin “Ati”
Krastev developed the evaporation system, characterized the thin films, and ably took
over the experiments when I moved on.

I have saved for last the individual who deserves the greatest thanks, my advisor
Prof. Roger Tobin. Without Roger, my pursuit of a Ph.D. would have gone nowhere.
His creativity, knowledge, and expertise made this project possible. He was always
willing to field any type of question, and could always be counted on to help me out of a
jam. The patience of which he was capable was particularly impressive when an
expensive piece of equipment fell victim to my inexperience. In many ways he set for me

a high standard toward which to strive as a graduate student, and in the years to come.

vii

TABLE OF CONTENTS

Chapter 1. Introduction

Section l-l. Motivation for Studying Adsorbate-Induced
Broadband Reflectance Changes

Section 1-2. Adsorbate-Induced Broadband IR
Reflectance Change in Metals

Section 1-3. Adsorbate-Induced Broadband AR/R
and Resistivity Change Ap in Metals

References

Chapter 2. Apparatus and Experimental Technique

Section 2-1. Ultrahigh Vacuum Techniques and Cleaning
the Pt(111) Surface

Section 2-2. Reflection-Absorption Infrared Spectroscopy
Section 2-3. RAIRS Measurement Technique

Section 2-4. Thin Film Growth

Section 2-5. Adsorbate-Induced Resistivity Change Measurement

References

viii

18

26

29

29

35

39

44

50

53

Chapter 3. Capillary and Effusive Gas Dosers 56

Section 3-1. Theoretical basis for closer modeling 58
Section _3-2. Calculations 62
Section 3-3. Results and discussion 64
References 71

Chapter 4. Infrared Reflectance Change Induced by CO Adsorption on Pt(111) 72

Section 4-1. Frequency Dependence 74
Section 4-2. Temperature Dependence 78
Section 4-3. Coverage Dependence 82
Section 4-4. Conclusions 91
References 92

Chapter 5. Changes in the DC Resistivity and the Infrared Reflectance
Induced by Oxygen and Forrnate Adsorption on Cu(100) 95

Section 5-1. 0 Adsorption on a Cu(100) Thin Film 96

Section 5-2. Forrnate Adsorption on an O-Predosed Cu(100) Thin Film 105

References 109

Chapter 6. Conclusions 111

References 115

ix

LIST OF TABLES

Table 1-1. Comparison of predicted and measured reflectance changes
AR/ R by Lin, Tobin, Dumas, Hirschmugl, and Williams in
Reference 17. The quantity na is the adsorbate coverage.

The values of tAp come from the literature.

Table 1-2. Comparison of predicted and measured values of the ratio
AR/Ap by Hein and Schumacher, reproduced from Reference 22.

Table 3-1. Comparison of circular effusive doser arrays with various
numbers of holes. All designs except those with eight holes comprise
a single center hole surrounded by a ring of equally spaced holes at
d/D = 0.8. The eight-hole arrays have only the ring of holes, without a
center hole. It can be seen that increasing the number of holes from
9 to 17 has little effect. The center hole improves the enhancement
factor slightly; it improves the uniformity for em, = 60° and

degrades it for 0",“, = 40°, but in both cases the effects are small.

Table 4-1. Bulk material parameters for Pt. Frequencies in s'1 have been
divided by 21m to convert to cm".

Table 4-2. Scattering cross section per adsorbate for
several adsorption systems.

23

64

73

90

LIST OF FIGURES

Figure 1-1. C=O internal stretch measured by RAIRS for 0.41 ML CO
adsorbed on Pt(111). This plot is the fractional difference of
a scan of a CO covered surface and a scan of a clean surface.

Figure 1-2. Broadband reflectance change induced by 0.25 ML 0 adsorbed
on Cu(100). This is a synchrotron-based RAIRS study performed
by Lin, Tobin, and Dumas, reproduced from Reference 18. The
gaps in the data and the scatter at high frequencies are artifacts
of the polyethylene windows on the UHV chamber.

Figure 1-3. Slab model for near-surface conductivity. l is the elastic
mean free path, as and 0'3 are the near-surface and bulk
conductivities respectively.

Figure 2-1. A model of our Pt(111) surface. From Reference 2 by KC. Lin.

Figure 2-2. AES for a clean Pt surface. The vertical axis is the
derivative of the number of electrons detected at a particular
energy, with respect to energy.

Figure 2-3. TPD spectrum from a clean Pt(111) surface following
02 dosing at 90 K.

Figure 2-4. The layout of the optical path. Reproduced from
Reference 6 by C. Chung.

Figure 2-5. An example of a reflectance time scan. The raw data are
shown in (a), while the data in (b) are divided by the initial value
with a linear baseline removed. The background pressure during
the CO dose was 8.0 x 10'10 Torr.

xi

19

31

33

34

36

41

Figure 2-6. Dependence of C=O stretch peak frequency on coverage
for atop CO on Pt(111). From Ref. 23 by J .S. Luo, R.G. Tobin,
and UK. Lambert. 43

Figure 2-7. A typical resistivity time scan. The data are divided by
the initial value with a background slope subtracted.
Background pressure during the dose was 5x10'6 Torr. 51

Figure 3-1: Origin of the cos3(0) term in the calculation of
flux per unit sample area. 60

Figure 3-2: Normalized flux per unit sample area versus the squared distance
from the sample's center for various sample to closer distances, indicated
by 9mm: = arctan(D/2€). d/D is the ratio of the doser's radius to the

sample's radius. Shown are plots for a capillary array of d/D = 2.0 and
1.0, a centered circular nine hole effusive array of d/D = 1.0 and 0.5,
and a single effusive doser. 66

Figure 3-3: Area-weighted fractional standard deviation 0' versus the fraction
of emitted molecules directly impinging on the sample f,- for various

sample to closer distances, indicated by Omar = arctan(D/2€). d/D is the

ratio of the doser's radius to the sample's radius. The data shown are for a
capillary array of d/D = 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0; a centered
circular nine hole effusive array ofd/D = 0.3, 0.5, 0.65, 0.8, 1.0,

1.35, 1.7, 2.0; and a single effusive emitter. 68

Figure 4-1. Fractional reflectance change AR/ R as a function of frequency,
for 0.33 ML CO on Pt(111) at room temperature. The lines represent
fits of the Persson-Volokitin model1 to the data, with the specularity
parameter p as the only adjustable parameter. The dotted line represents
the best fit using DC parameters, and the solid line the best fit
using optical parameters (see Table 4-1). 76

Figure 4-2. Dependence of the reflectance on Z/ 8 predicted for adsorption on
Pt by the Persson-Volokitin scattering theory, for the indicated values of
(II/(o, . The vertical scale for (a) and (c) is arbitrary. The small arrows
indicate the room temperature (RT) and 90 K values of l/ 5 for Pt. DC
values of the bulk parameters have been used (see Table 4-1).
(a) Absolute reflectance change AR. (b) Clean-surface reflectance R.
(c) Fractional reflectance change AR/ R. 81

xii

Figure 4-3. (a) Fractional reflectance change AR/ R as a function of coverage,
for CO on Pt(111) at 315 K. Data are shown for two IR frequencies,
2500 and 2800 cm". (b) Work function change for CO on Pt(111),
reproduced from References 34 (Norton, et al., triangles) and 35
(Ertl, et _aI., circles). 86

Figure 5-1. Time scans of (a) the reflected IR intensity (i.e. voltage output
from a lock-in amplifier) at 2800 cm'1 and (b) the voltage drop across
the sample (i.e. . voltage output from a lock-in amplifier) with an
applied current. Both 50 L doses involved background pressures
of 5 x 10'7 Torr for 100 s. 98

Figure 5-2. Time scans of (a) the broadband reflectance change AR/ R at
2800 cm'1 and (b) the fractional change in resistivity Ap/ p . Each
curve has a linear drift removed. Both 50 L doses involved
background pressures of 5 x 10'7 Torr for 100 s. 99

Figure 5-3. AR/ R versus tAp for 02 adsorption on a Cu(100) film. The inset
is an enlargement of the lower right hand corner. The solid line is a
linear fit to the data, which are taken from about the 200 s to 500 S
region of Figures 5-2(a) and 5-2(b). 102

Figure 54. AR/ R versus tAp for formate adsorbtion on a Cu(100) film. The

inset is an enlargement of the upper left hand corner. The data are taken
from about the 500 s to 620 8 region of Figures 5-2(a) and 5-2(b). 107

xiii

Chapter 1

Introduction

l-l. Motivation for Studying Adsorbate-Induced Broadband Reflectance Changes
The study of the physics and chemistry of solid surfaces and interfaces, referred to
as “surface science” due to its interdisciplinary nature, is of great importance both

2 The multibillion-dollar fields of lubrication and

technologically and scientifically."
sensor technology inherently involve surface properties. An understanding of surface
processes such as catalysis, corrosion, oxidation, and crystal growth is essential to many
diverse industries. On the other hand, most scientific work geared toward developing a
basic understanding of the properties of condensed matter, both theoretical and
experimental, tends to focus on bulk properties. This is due in part to the intrinsic
difficulty of dealing with surfaces. In the words of Wolfgang Pauli:2 “The surface was
invented by the devil.” Although the physical laws that describe interactions at surfaces
are essentially the same as those for the bulk, the broken symmetry in the direction
normal to the surface and the different environment of atoms at the surface greatly
complicate theoretical treatments. Nevertheless, those same features lead to interesting
new phenomena. In addition to the theoretical constraints, experimental research on

surfaces has been historically inhibited by the difficulty of obtaining, characterizing, and

maintaining clean surfaces for study.

2

While the thermodynamics of surfaces were nearly completely described by
Gibbs3 in 1877, the complexity of performing calculations and the inability to
experimentally test theoretical predictions limited progress in surface science for the next
80-90 years. In the last 30 years, however, several technological advances have Spurred a
renaissance in surface science research. The three most significant advances involve
vacuum technology, electron diffraction, and high-speed computers.2 First, through
advances related to the space program, it became possible to commercially develop ultra-
high vacuum (UHV) chambers in which one could obtain and maintain clean surfaces
through cleaving or cleaning procedures. Second, knowledge of the submonolayer
chemical composition of surfaces could be obtained through electron spectroscopy.
Third, the availability of high-speed computers permitted numerical solutions to surface
science problems.

One of the most heavily studied areas of surface science is the interaction of gas
atoms or molecules with metal surfaces, called adsorption. Surface vibrational
spectroscopy (SVS) has been employed with great success to probe the vibrational
resonances within the adsorbed molecule or between the molecule (or atom) and the
substrate (i.e. the metal surface). SVS includes such techniques as electron energy loss
spectroscopy (EELS), surface enhanced Rarnan spectroscopy (SERS), thermal energy
atom scattering (TEAS), and reflection-absorption infrared Spectroscopy (RAIRS),
among others.‘1 Studies utilizing these techniques have yielded a great deal of
information regarding the fascinating complexities of the coupling and energy transfer

between the vibrational and electronic states of the adsorbate and those of the substrate.

0.01 . I . I fl I e r

 

0.00
-0.01
-0.02

g -0.03

-0.04

 

0.41 ML
-0.05 ‘ ~
CO/Pt(l 1 1)

I

I

-0.06 ..

 

 

-0.07 1 41 I 1 4 I A l A
2050 2070 2090 2110 2130 2150

frequency (cm'l)

 

Figure 1-1. C=O internal stretch measured by RAIRS for 0.41 ML CO adsorbed on
Pt(111). This plot is the fractional difference of a scan of a CO covered surface and a
scan of a clean surface.

Figure 1-1 shows an example of a RAIRS measurement of the internal stretching
mode of CO adsorbed on a Pt(1 1 1) surface. The coverage is 0.41 monolayer (ML) CO.
First a scan is conducted through the spectral region with the Pt surface clean. A second
scan is conducted through the same spectral region after the surface is dosed with CO. In
Figure 1-1 we represent the data in terms of the fractional reflectance change AR/ R by

subtracting the “dosed scan” from the “clean scan,” and dividing by the “clean scan.”

4

Information about numerous physical processes can be gained by measuring quantities
such as the peak frequency, the peak width, and the peak intensity while varying specific
parameters, forexample the temperature or coverage.

While using RAIRS to study such vibrational modes provides information about
the coupling and energy transfer between the adsorbate and the substrate, using the
technique to measure broadband changes in the reflectance allows the exploration of
issues related to electron dynamics in the near-surface region of the substrate. In order to
motivate the connection between the adsorbate-induced broadband change in a metal’s
reflectance and electron dynamics, it is useful to discuss the current understanding of the
effect of adsorbates on electron transport in metals.

Even without strong energetic coupling adsorbates can profoundly affect electron
dynamics in the near-surface region of a metal. An isolated adsorbate on an otherwise
perfect surface breaks the translational symmetry parallel to the surface and sets up a
static impurity potential from which conduction electrons can scatter without conserving
the parallel component of momentum. In other words, the presence of an adsorbate
changes the surface scattering of conduction electrons from specular to diffuse. These
scattering events increase the electrical resistivity near the surface, with observable and
sometimes dramatic consequences.

It has been known for many years that even a monolayer of adsorbed gas can
increase by more than 20% the resistance of a metal film hundreds of atomic layers
thick.5'7 Indeed, measurement of this physical property forms the basis for an important
class of chemical sensors. It was recognized by Holstein8 and othersg'” as long ago as

1948 that diffuse scattering of conduction electrons from surfaces would necessarily also

5
reduce the infrared reflectance of the surface. Only recently, however, have such
broadband changes in the IR reflectance been measurable.

2 5 '3 0 I o .
have revrved Interest In

Recent experiments12 '24 and theoretical developments
these scattering effects as they relate to adsorbate-substrate dynamics. A number of
careful infrared studies on single crystals in UHV have revealed adsorbate-induced
decreases (typically ~l%) in the broadband reflectance AR/ R of metal surfaces. The

dielectric properties of the adsorbate layer itself would be expected to produce an
increase in the reflectance, but at least two orders of magnitude smaller than the observed
effects.31 The observed reflectance changes, therefore, result from changes in the
electronic properties of the substrate caused by the binding of the adsorbates.

Figure 1-2 is an example of one of the recent measurements of a broadband
change in the IR reflectance. It is reproduced from Reference 18, a synchrotron-based
study of 0 adsorption on Cu(100) by Lin, Tobin, and Dumas. Figure 1-2 shows the data
for 0.25 ML 0 on a Cu(100) surface. The data is collected by first performing a scan
with a clean surface, followed by a scan with a dosed surface. As in Figure 1-1, Figure 1-
2 shows the fractional reflectance change AR/ R by subtracting the dosed scan from the
clean scan and dividing by the clean scan. The gaps in the data and the scatter at high
frequencies result from the use of polyethylene windows on the UHV chamber, which
have their highest transmittance in the low-frequency regime. Note the characteristic
shape of the frequency dependence. Also note that the size of the effect, ~ 1%, is

considerably smaller than the peak in Figure 1-1.

 

 

 

0.000 _ -
-0.003 .. :, -
g -0.006 ..'.. “23' "if?
. 3. 51:31.5! : .
-0.009 Tr-" ._ s
- to..,, .
"a: 2‘: .‘1
-0.012 - I—
9 = 0.25 ML
-0015 + 3 4

 

 

 

O 400 800 1200 1600 2000
Frequency (cm-1 )

Figure 1-2. Broadband reflectance change induced by 0.25 ML 0 adsorbed on Cu(100).

This is a synchrotron-based RAIRS study performed by Lin, Tobin, and Dumas,

reproduced from Reference 18. The gaps in the data and the scatter at high frequencies
are artifacts of the polyethylene windows on the UHV chamber.

It is worth remarking in passing that measurements of broadband changes in
surface reflectance are far more demanding experimentally than the more common
observations of relatively sharp vibrational bands. The broadband measurements demand
that the absolute intensity be held constant (or independently measured) to an accuracy
considerably better than 1% between the measurement on the clean surface and that on
the adsorbate-covered surface. Such stability is not easily achieved. Vibrational

measurements, on the other hand, are relatively insensitive to changes in signal level so

7

long as the background remains relatively smooth. Additionally, investigation of the
frequency dependence of the broadband AR/ R requires the ability to probe very low
frequencies. Few RAIRS experiments have successfully yielded information below

1000 cm".

Persson and Volokitinzs'29

have recently proposed a model based on these ideas of
conduction electron scattering which accounts for the adsorbate-induced change in the
broadband IR reflectance, AR/ R , of metals. Using the model, they develop a relationship
between AR/ R and the adsorbate-induced change in resistivity, Ap. Furthermore, within

the model they relate26’28'29

the adsorbate-induced AR/ R and Ap to the observation of
antiabsorption resonances with RAIRS, the damping of hindered vibrational modes, and
atomic scale friction. While it remains possible that electron scattering does account for
all these effects, it has been shown17 that the elements of the model which relate AR/ R
and Ap do not rely on the connections to the other phenomena treated. Regardless of the
validity of the connections to the other effects, experimentally testing the predicted
relationship between AR/ R and Ap , along with predictions regarding just the broadband
AR/ R , can help establish whether or not scattering is the dominant mechanism. This
knowledge will greatly contribute to our understanding of the interactions of gas
molecules and atoms with metal surfaces.

1 have used RAIRS to perform a series of measurements of the broadband AR/ R
which test the validity of the Persson-Volokitin scattering model. In Chapter 4,

frequency-, temperature-, and coverage-dependent measurements of AR/ R for CO

adsorption on a Pt(1 1 1) single crystal are presented and the results are compared to the

8

model’s predictions. In Chapter 5, simultaneous measurements of the adsorbate-induced
AR/ R and Ap for O and formate adsorption on an in situ grown epitaxial Cu(100) film
are presented. That data is used to test the predicted relationship between AR/ R and Ap.

The remainder of Chapter 1 introduces the relevant parts of the Persson-Volokitin
scattering model. The expression for AR/ R which comes out of the model is presented
in Section 1-2. Recent measurements of AR/ R to which the model has been applied are
discussed. In Section 1-3 the expression relating AR/ R and Ap is developed, along with

a discussion of attempts to experimentally test it.

1-2. Adsorbate-Induced Broadband IR Reflectance Change in Metals

The primary motivation for the development of the Persson-Volokitin model”29
of conduction electron scattering from adsorbates comes from several remarkable
observations of rather strong antiabsorption resonances associated with hindered
rotational and translational modes of adsorbates on metal surfaces. Antiabsorption
resonances are spectral features which exhibit an apparent increase of the reflectance on
resonance relative to nearby frequencies, as opposed to the usual decrease due to the
absorption of light by a resonant mode. In 1988 Reutt, Chabal, and Christrnanl4 reported
measurements of formally dipole-forbidden low-frequency frustrated translations of H on
W000) and H on Mo(100). In 1990 Hirschmugl, Williams, Hoffrnann, and Chabal”
reported measurements of the dipole-forbidden frustrated rotation of CO on Cu(100) at

low-frequency. It is noteworthy that both experiments revealed a broadband absorption

of IR light.

9

The dipole selection rule for IR detection of vibrational modes comes from the
fact that the component of the incident IR electric field parallel to the surface, E", is
strongly screened inside the metal. Since E” is continuous at the surface, the electric field
on the vacuum Side is almost entirely perpendicular to the surface. Therefore, only
modes with a dynamic dipole moment normal to the surface are observable with
RAIRS.32

Persson’s idea regarding the observation of dipole-forbidden modes in the form of
antiabsorption resonances was that even though E” is strongly screened inside the metal,
its finite penetration can induce an oscillating drift current. The carriers of this drifi
current scatter diffusively from the adsorbates, causing an increase in the surface
resistivity of the metal. A consequence of the resistivity increase is a decrease in the
metal’s broadband reflectance. When the frequency of the incident IR light matches the
frequency of a hindered parallel mode of the adsorbate, however, the molecules move in
resonance with the drift motion of the electrons. Thus the additional surface resistivity
vanishes and the IR reflectance returns to its clean surface value. This model has done a
good job accounting for the measurement of the frustrated rotation of CO on Cu(100) an
Cu(11 1).19 The elements of the model which are tested here are the predictions regarding
the broadband decrease in the IR reflectance.

The Persson-Volokitin scattering model has its origins in the work of Fuchs,33

10,11 .
on the anomalous skin effect. The

Holstein} Reuter and Sondheimer,9 and Dingle
anomalous skin effect is a surface property concerning electron transport and requiring

nonlocal corrections to the classical Fresnel formulae. The classical Fresnel formulae,

which depend on the bulk dielectric function e , often provide an adequate description of

10

the reflectance of a metal’s surface using a local form for e . A local description is valid

when the condition
0) >> y—F— 1-1
8

is met, where vF is the Fermi velocity (i.e. the maximum velocity of a conduction

electron), and 6 is the classical skin depth. Equation 1-1 requires that the incident E-
field does not vary appreciably over the distance an electron travels during one cycle of

the incident field. The distance an electron travels in the bulk between elastic scattering

events is given by the elastic mean free path l, which takes the form
I = th B , 1-2

where r B is the bulk scattering time. For most metals, 5 is on the order of hundreds of

A and l is at least comparable to 6. In some cases, such as Cu, 6 is much larger than 8 .

For low frequencies, where the condition given in Equation 1-1 is not met, a local
description of e , and consequently the reflectance, is no longer valid. This is because an
electron traveling into the near-surface region will experience a force from the penetrating
E-field and propagate ballistically into the bulk. This is called the anomalous skin effect

and the resulting transfer of energy from the incident light to the metal causes the metal’s

ll

reflectance to be less than unity. Moreover, it accounts for an intrinsic frequency-
dependence in the reflectance.

In addition to making nonlocal corrections to the Fresnel formulae, Holstein
studied the general case where some fraction of the conduction electrons, given by the
quantity (l-p), scatter diffusively from surface imperfections. The parameter p is know
as the Fuchs specularity parameter and represents the probability of specular scattering.
Persson adapted this treatment, interpreting adsorbates as surface imperfections.

The scattering model of Persson and Volokitin incorporates a number of

Simplifying assumptions:

1. The metal is treated as a free electron (Drude) gas described by an electron density n

and a bulk scattering time 13, or, equivalently, a plasma frequency 03p and a mean-

free-path l.

2. The adsorbate affects only the scattering rate, represented for the most part simply
through a change in the Fuchs specularity parameter33 p, which represents the
probability of specular scattering from the surface; (1 — p) is the probability of diffuse

scattering.

3. The adsorbates form a dilute, disordered layer, so that scattering from the individual
adsorbates is independent. The scattering probability can then also be described by a

scattering cross section per adsorbate, 2:

12
(1- p) = n02 1-3
where no is the surface density of adsorbates.

With these assumptions the fractional reflectance change AR/R for p-polarized light can

be written:28

(£)=__3_VL(1-__’L).f[m g] 1.4

R 4c c050 (ll—1’5

where 0 is the angle of incidence of the light. Note that V; within a Drude model
depends only on n, the conduction electron density. The frequency dependence is a
universal function28 f that depends on the frequency 0) relative to 001 and the ratio of the

electron mean-free-path l to the classical skin depth 8 = c/mp. (01 is a characteristic

 

rolloff frequency given by
v 0)
(0' = F p , 1-5
c

Near 001 an electron travels a distance comparable to the skin depth during one cycle of

the electromagnetic field; (01 therefore specifies the frequency range below which

nonlocal electrodynarnic effects are important.”28

13

At sufficiently high frequencies f is of order unity; in the approximation that the
clean surface reflectance is one it approaches unity exactly, but this is not always a good
assumption. It, is important to note that the parameters that enter into f do not involve
properties of the adsorbate; the frequency dependence depends only on the metal, while
only the magnitude of the reflectance change depends on the adsorbate. It is also worth
noting that Equation 1-4 does not require the Specific model of adsorbate vibrational
damping advanced in Reference 29.

The complete analytical expression for AR/ R is

 

 

 

 

/ ‘l
a... (myKellie/0.46)] .6
R — 2 i , -
( 1:00 p c030 8(0/(0195/5ML

where g(0)/001, 13/5) and 8((0/031, lit/54) are complex-valued functionsz8 given by

 

g=jdy(1-}17)F2(y)+mg 4mm Dd),

, y -0)2-i0)11

[yi_%) 170)]: 1-7

and

 

e=q2-££(£) 23+[[-i1]-1]1n .3 , 1-8
4B0, “I B I3 'q

where

14

o l l
F(y)=idq8(m/wra€/5rq)l+ifiy 1-9

 

and

-——-. 1-10

The parameter a is the magnitude of the reflectance change at high frequencies, which is

given by the prefactor of f in Equation 1-4.

a = 3vF(1- p)
4ccosO

In addition to predicting the frequency-dependent broadband AR/ R, Equations
1-7 to 1-11 also relate AR/ R to the damping of hindered translations and rotations

24,30,34

(observed as antiabsorption resonances). The first term in the expression for

g(c0/0)1,€/5) in Equation 1-7 essentially contains the information about AR , while the
second term gives the antiabsorption feature at 00 o. The parameter 1] is the damping
coefficient associated with the resonance. The parameter 0) 0 does not come into play
when calculating just the broadband AR/ R. Whether or not n is essential to calculating

the broadband AR/ R remains a controversial point.”37

As will be discussed in Chapter
4, it is possible to set 11 equal to zero in the above expressions and still accurately predict

the frequency-dependence of AR/ R. While that would seem to be conclusive evidence

that n is not intrinsically related to AR/ R, Persson claims that the quantity (l-p) is

15

proportional to n .25 Nevertheless, the model has been very successful in explaining the

16,19

existence and line Shapes of the antiabsorption resonances and in accounting for the

spectral Shape of the broadband reflectance.’8"9

The earliest reports on adsorbate-induced broadband reflectance changes dealt
with adsorbates on W(100), by Riffe, Hanssen and Sieversn’l3 and Reutt, Chabal and

Christrnan. '4

Electron scattering effects were considered, but the spectral Shape, which
showed an increase in the reflectance change (back toward zero) at high frequencies, led
Reutt et al. to interpret the results in terms of surface electronic states.

Persson’s scattering model was applied in the analysis of data collected by
Hirschmugl et al. for CO on Cu(100) and Cu(111).’6"9 It was found to do a good job
fitting the frequency dependence in both cases.

Another stringent test of the frequency dependence was performed by Lin, Tobin,
and Dumas.l8 Figure 1-2 shows one of their measurements of the broadband reflectance

change, together with a fit to the Persson-Volokitin model.27 The agreement with the
predicted functional form is excellent, and the values of 0), are largely, though not
completely, in agreement both among experiments and between experiment and theory.18
The model was unable to fit their data, however, for their highest coverage, 0.35 ML.
Lamont, Persson, and Williams38 measured the broadband AR/R caused by H
adsorbed on Cu(111). They found that scattering model works for low frequencies, but
AR/ R increases much more rapidly for 0) > 1000 cm'l than predicted by the model.

Following the example of Reutt et al.,'4 they tentatively interpret this behavior in terms of

adsorbate-induced surface states.

16

Thus far I have only discussed the frequency dependence predicted by the

scattering model. Because Equation 1-8 contains the ratio [/6 , and the elastic mean free
path I is a strongly temperature-dependent quantity, there is a natural temperature
dependence in AR/ R. Plots of AR , the absolute change in reflectance, which appeared
in Reference 28 and were parametrized by the ratio €/5 would lead one to believe that
for decreasing temperature (i.e. increasing [/8 ), AR/ R , the experimentally measured

quantity, should get larger. The magnitudes of both AR and R increase as the
temperature decreases, however, and a proper consideration of the clean surface
reflectance R shows that the behavior of AR/ R depends on the particular temperature and
frequency regimes in which the measurements are taken. The temperature dependence
will be further discussed in Chapter 4, Section 4-2.

The scattering model also predicts a linear dependence of AR/ R on coverage.

This can be seen by combining Equations 1-3 and 1-8 to get

a: ”FE no. [1-12]
4cc030

 

Recall that a is the magnitude of AR/ R at high frequencies and na is the number density
of adsorbates.
In their work, which predated the Persson-Volokitin scattering model, Riffe,

12,1

Hanssen, and Sievers 3 reported a coverage dependent adsorbate-induced broadband

absorption for N2, 02, CO, H2, and D2 on W(100) using measurements of the

17

attenuation of surface electromagnetic waves. With increasing coverage, the magnitude

of the attenuation coefiicient increased monotonically for N2, while a peak was observed
for both 02 and CO. H2, and D2 adsorption each showed a slightly more complicated,

nonmonotonic dependence on coverage. Pointing to earlier theoretical work,"'l "33 Riffe
et al. attributed the broadband absorption primarily to surface-reconstruction-induced
changes in the scattering of conduction electrons. They found correlations between the
structure they observed in the coverage-dependent absorption and ordered patterns
deduced from low energy electron diffraction (LEED) experiments.39 Using these
correlations, they explained the observed structure in terms of ordering in the adlayer.

For CO on Cu(100) and Cu(l 1 l), Hirschmugl et 01.15.16 observed the predicted
linear dependence up to saturation. For 0 on Cu(100), however, Lin, Tobin, and
Dumasl8 saw a roughly 20 % decrease, well outside experimental error, in AR/ R for their
highest coverage. It is interesting that this was also the coverage at which the scattering
model failed to fit the frequency dependence.

In the present work I extend the investigation of the Persson-Volokitin scattering
model to a transition metal, Pt, and explore the dependence of the reflectance change
AR/ R on frequency, temperature, and adsorbate coverage. Since it is not clear that the
Drude model should work for transition metals, this set Of experiments should help to

establish the bounds of usefulness of the Persson-Volokitin scattering model.

18

1-3. Adsorbate—Induced Broadband AR/R and Resistivity Change Ap in Metals
Following the example of Lin, Tobin, Dumas, Hirschmugl, and Williams,17 below
I develop the relationship between the adsorbate-induced fractional change in the
reflectance AR/ R and the adsorbate-induced change in resistivity Ap. The starting point
is to expand the appropriate classical Fresnel formula40 for p-polarized light using the

Feibelman d-parameter formalism.“

=__ , 1-13
c 0050

 

95 40) Im(d")
R

where 0) is the frequency of the light, c is the speed of light, and 0 is the angle of

incidence. Equation 1-13 holds when the magnitude of the complex dielectric function e

of the metal meets the condition |e| >> (1/ cos2 0)>> 1. The parameter (1" is given by
1-14

where o u is the complex conductivity parallel to the surface for a uniform electric field,

and z is the direction normal to the surface. The Im(dl) is considered to be negligible

compared to Im(d“) due to the strong refraction by the metal which makes E i very

19

small just inside the metal. From Equation 1-13 it can be seen that the problem reduces
to calculating o ll(z).

The effect of adsorbates on the electrical conductivity is felt in the near-surface
region to a depth on the order of the elastic mean free path I, defined in Equation 1-2.

Further inside the metal, the conductivity takes the bulk value 0'3. Persson proposes a

slab model29 as illustrated in Figure 1-3.

 

Figure 1-3. Slab model for near-surface conductivity. 1’ is the elastic mean free path, 0‘s
and GB are the near-surface and bulk conductivities respectively.

This model provides the following form for the surface conductivity to a depth on the

order of the mean free path 3:

o,(m)=oB(0))+Ao(m). l-15

20

This slab model provides a useful means of mapping a thin film, with which it is
straightforward to measure the adsorbate-induced resistivity (conductivity) change, onto
the near-surface region of a bulk crystal.

Using Equation 1-15 for the conductivity to a depth 2 and GB elsewhere, d” can be

calculated from Equation 1-14 and inserted into Equation 1-13 to get

 

 

ARE = 4m(t/c)lm[Ac(co)]. 1-16

If the frequency is restricted to a region where the classical skin depth 6 >> vF /0)
(i.e. the same condition as in Equation 1-1), a local description of the metal’s dieletric
response can be used. This “high frequency” condition implies that an electron does not
travel a significant distance during one cycle of the incident electric field. The Drude
model for the bulk conductivity is a useful approximation to describe the local response

in many metals:

__n__8213

_in8), 1-17

03(9)): m(

where n is the conduction electron density and m is the electron mass. Expanding A0 to

first order in the changes An and Ar gives

21

A0 ___(__(0)_ An At

70(0)) n8 —+ ”(l—ions)

1-18

 

Since the scattering model assumes that the dominant process in the conductivity
change is scattering, the first term in Equation 1-18 can be neglected. Expressing things

in terms of resistivities rather than conductivities, Equations 1-16 and 1-18 lead to

W ——

Here p(€) refers to the resistivity of a film of thickness Z and p3 refers to the bulk
resistivity. It is useful to note, however, that the resistivity of a film of arbitrary thickness
can be used because the quantity tAp(t) is independent of thickness 1. This property can

be seen in the Fuchs-Sondheimer model for thin film conductivity as well as through

5,7

experimental evidence. This, along with Equation 1-16, allows the fractional

reflectance change to be written

 

E = _ 4" AP“) = -————.[tAp(t)] 1-20
R c: BeosG p 8 —,,cp 1:0089

The only term in Equation 1-20 that refers to a thin film property rather than a bulk

property is tAp(t). This is made possible through the adoption of a slab model.

22

Furthermore Equation 1-20 predicts the relationship between AR/ R and tAp(t) to be

independent of frequency.

 

Making use of
p Blr = '£2 a 1'21
ne
we get
AR 4ne2 J
__ = _ tA . 1-22
R (me c080 [ p]

Note that this derivation avoids any assumptions regarding the damping of vibrations.

The first attempt to test this relationship was performed by Lin et al.1.] They
measured the broadband AR/ R for single coverages of CO and O on a Cu(100) single
crystal, and for CO on a Ni(100) single crystal, and compared their results to values of
AR/ R predicted by Equation 1-22. The predicted values were calculated from literature
values of tAp , measured on polycrystalline thin film samples. The results of Lin et al.
from Reference 17 are presented in Table 1-1.

As can be seen in Table 1-1, they found reasonable agreement for CO and O on
Cu(100), but a discrepancy for CO on Ni(100). There are inherent difficulties with the
comparison, however. Lin et al. point out that there exiSts a difference in
crystallographic orientation; the films are polycrystalline but believed to be primarily of

(111) orientation, while the bulk single crystals were oriented in the (100) direction.

23

Additionally, variations in sample preparation and uncertainties in surface structure and
sample thickness make the values of tAp reliable to only within about a factor of two, as

indicated by the. values of tAp for O on Cu listed on Table 1-1.

Table 1-1. Comparison of predicted and measured reflectance changes AR/ R by Lin,
Tobin, Dumas, Hirschmugl, and Williams in Reference 17. The quantity na is the
adsorbate coverage. The values of tAp come fiom the literature.

 

 

"a (AP (AR/R)ca]c (AR/ R)8Xp
system (1014 cm’z) (10'12 0 cm2) (%) (%)
CO/Cu(100) 6.1 0.60 -2.2 -1 . 1:1:0.2
-1 .1101"
O/Cu(100) 3.8 0.42 -1.5 -1.1i0.2
0.70 -2.6
CO/Ni(100) 6.4 1.47 -5.8 -0.2i0.2

 

. . . . 21
21corrected for a difference in ineldence angle

M. Hein and D. Schumacher recently sought to alleviate the problem of using
measurements on different samples by making simultaneous measurements on in situ
grown thin film samples.22 They studied 0, CO, and Cu on Cu films, and Ag on Ag
films. From their data they were able to plot the reflectance change versus the resistivity
change to test the predicted linear relationship. They found strong evidence of linearity

between tAp and AR/ R for O and CO on Cu, although they found the slope to differ

24

significantly from the predicted value. For Cu on Cu and Ag on Ag, they observed the
resistivity reach a maximum at some particular coverage, afier which it decreased. They
found that the linear relationship holds for coverages below the resistivity maximum. In
the case of Ag on Ag they found a large discrepancy between the predicted and measured
values of the slope. The results of the slope comparison of Hein and Schumacher from

Reference 42 are presented in Table 1-2.

Table 1-2. Comparison of predicted and measured values of the ratio AR/Ap by Hein
and Schumacher, reproduced from Reference 22.

 

 

(AR/AP)... (AR/49)....
system (149 ' cm)” (119 ' cm)_l
O/Cu -0.03O -0.0085
CO/Cu -0.014 -0.0077
Cu/Cu -0.010 -0.0085
Ag/Ag -0.025 -0.0082

 

In Chapter 5 I will present a test of the linearity prediction for O and formate
adsorption on a Cu( 100) thin film. This experiment differs from the Lin, et al. '7 study

and the Hein and Schumacher study22 in that the Cu(100) film is grown epitaxially on a

25

Si(100) substrate. Therefore this is the first simultaneous test of the predicted linear

relationship between AR/ R and tAp on a well characterized thin film sample.

26

References

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

2 A. Zangwill, Physics at Surfaces (Cambridge University Press, Cambridge, 1988).
31 .W. Gibbs, Collected Works, Vol. 1 (Longmans-Green, New York, 1928) 184.

4J.W. Gadzuk, in Vibrational Spectroscopy of Molecules on Surfaces, ed. J .T. Yates, Jr.
and TE. Madet (Plenum, New York, 1987).

5 P. Wissmann in: Surface Physics, ed. G. Htihler, Springer Tracts in Modern Physics Vol.
77 (Springer, New York, 1975).

6D. Dayal, H.-U. Finzel and P. Wissmann in: Thin Metal Films and Gas Chemisorption,
ed. P. Wissmann (Elsevier, Amsterdam, 1987).

7D. Schumacher, Surface Scattering Experiments with Conduction Electrons, ed. G.
Hdhler, Springer Tracts in Modern Physics Vol 128 (Springer, New York, 1993).

8T. Holstein, Phys. Rev. 88 (1952) 1427.

9G.E.H. Reuter and EH. Sondheimer, Proc. Roy. Soc. London, Ser. A 195 (1948) 336.
”RE. Dingle, Physica 19 (1953) 311.

”RB. Dingle, Physica 19 (1953) 729.

l2D.M. Riffe, L.M. Hanssen and A.J. Sievers, Phys. Rev. B 34 (1986) 692.

l3D.M. Riffe, L.M. Hanssen and A.J. Sievers, Surf. Sci. 176 (1986) 679.

”1.13. Reutt, Y.J. Chabal and SB. Christrnan, Phys. Rev. B 38 (1988) 3112.

15OJ. Hirschmugl, G.P. Williams, F.M. Hoffmann and Y. J. Chabal, Phys. Rev. Lett. 65
(1990) 480.

“’CJ. Hirschmugl, Y.J. Chabal, F.M. Hoffrnann and G.P. Williams, J. Vac. Sci. Technol.
A 12 (1994) 2229.

27

17KC. Lin, R.G. Tobin, P. Dumas, C.J. Hirschmugl and GP. Williams, Phys. Rev. B 48
(1993)2791.

"’K.C. Lin, R.G. Tobin and P. Dumas, Phys. Rev. B 49 (1994) 17 273 (1994); ibid. 50
(1994)17760.

I9C.J. Hirschmugl, G.P. Williams, B.N.J. Persson and A. I. Volokitin, Surf. Sci. 317
(1994) L1141.

2"DM. Riffe and A.J. Sievers, Surf. Sci. 210 (1989) L215.

2’13. Borguet, J. Dvorak and H.L. Dai, SPIE Proceedings 2125 (1994) 12.

22M. Hein and D. Schumacher, J. Phys. D: Appl. Phys. 28 (1995) 1937.

23C.L.A. Lamont, B.N.J. Persson and G.P. Williams, Chem. Phys. Lett. 243 (1995) 429.
24J. Krim, D.H. Solina and R. Chiarello, Phys. Rev. Lett. 66 (1991) 181.

25B.N.J. Persson, Chem. Phys. Lett. 197 (1992) 7.

26B.N.J. Persson, Chem. Phys. Lett. 185 (1991) 292.

27B.N.J. Persson and AI. Volokitin, J. Electron Spectrosc. Relat. Phenom. 64/65 (1993)
23.

28B.N.J. Persson and AI. Volokitin, Surf. Sci. 310 (1994) 314 .
29B.N.J. Persson, Phys. Rev. B 44 (1991) 3277.

301.8. Sokoloff, Phys. Rev. B 52 (1995) 5318.

3|R.G. Tobin, Phys. Rev. B 45 (1992) 12 110.

”Yr. Chabal, Surf. Sci. Repts. 8 (1988) 211.

33x. Fuchs, Proc. Cambridge Phil. Soc. 34 (1938) 100.
34B.N.J. Persson, 1. Chem. Phys. 98 (1993) 1659.

35R.G. Tobin, Phys. Rev. B 48 (1993) 15 468

36B.N.J. Persson, Phys. Rev. B 48 (1993) 15 471

28

37T.A. Gerrner, J .C. Stephenson, E.J. Heilweil and RR. Cavanagh, J. Chem. Phys. 101
(1994) 1704.

38CA. Lamont, B.N.J. Persson, and GP. Williams, Chem. Phys. Lett. 243 (1995) 429.
39DA. King and G. Thomas, Surf. Sci. 92 (1980) 201.
40W.L. Schaich and W. Chen, Phys. Rev. B 39, (1989) 10714.

"Pr. Feibelman, Prog. Surf. Sci. 12 (1982) 287.

Chapter 2

Apparatus and Experimental Technique

When used to study the interactions between adsorbed gases and surfaces,
reflection-absorption infrared spectroscopy (RAIRS) is typically performed on ordered
surfaces of bulk Single crystal samples. This helps to make analysis of the results
tractable; single crystal surfaces facilitate the application of theoretical results to physical
systems. The production of clean surfaces suitable for study on such samples requires
intensive cleaning procedures and ultrahigh vacuum (UHV) conditions. A second option
for the production of suitably clean, ordered surfaces is to grow epitaxial thin metal films
in situ in UHV. Both large single crystals and thin metal films have been used in this
study.

Section 2-1 describes the UHV apparatus, the Pt(1 11) single crystal sample, and
the surface cleaning techniques. Section 2-2 tells about the home-built RAIRS apparatus
while section 2-3 details the RAIRS measurement technique. Cu(100) thin film growth is
discussed in section 2-4. Finally, section 2-5 describes the resistivity change

measurement.

2-1. Ultrahigh Vacuum Techniques and Cleaning the Pt(lll) Surface
If each impinging molecule sticks to the sample surface (i.e. sticking coefficient is
one) and the surface atomic density is about 3 x 10”cm '2 (both of which are reasonable

physical assumptions in many cases, depending upon the mass of the gas, among other

29

30
variables), an exposure of 1x10'6Torr-s, defined as a Langmuir, will result in a

monolayer coverage on the sample surface. Thus, a clean metal surface at a pressure of

1x 10"°Torr will require a little more than two and one half hours to build up a
monolayer of contamination. Our experiments were typically done at a base pressure of
about 4x 10'1 1Torr.

The stainless steel UHV chamber is 12” (30.5 cm) in diameter and has a height of
25” (98.4 cm). It is rough pumped by a molecular sorption pump and a turbomolecular
pump. During normal UHV operation the chamber is pumped by ion pumps1 and by a
titanium sublimation pump (as needed). A bakeout at about 100°C for 36-48 hours is
required to achieve UHV. This is accomplished by means of a series of heater tapes
wrapped around the chamber, an internal infrared (IR) lamp, and a heater around the ion
pumps. During bakeout the chamber is wrapped in aluminum foil to more evenly
distribute the heat and the temperature is monitored by an array of thermocouples.

From bottom to top the chamber is divided into the IR/evaporator level, the
surface analysis level, and the manipulator. The IIUevaporator level will be discussed
separately. The surface analysis level includes instrumentation for ion sputtering, low
energy electron diffraction (LEED), and Auger electron Spectroscopy (AES). There is
also a residual gas analyzer which enables temperature programmed desorption (TPD).
The manipulator allows the sample to translate in the X, Y, and Z directions as well as
rotate in the azimuthal angle 4).

The sarnple used in the single crystal experiments was a 1 cm x 3 cm Pt single

crystal with a surface normal 27° from the [111] direction, as determined by Laue X-ray

31

diffraction and LEED. The lower right comer was cut away because it exhibited a greater

degree of miscut from the [111] direction. The remaining 2.7° miscut results in a roughly

In other words, there is a one-atom step about every twenty

5% density of _step sites.

2

surface Pt atoms, as shown in Figure 2-1.

 

 

 

 

 

Figure 2-1. A model of our Pt(1 1 1) surface. From Reference 2 by KC. Lin.

We do not expect the ~ 5% density of step sites to affect our results because the step sites

should be saturated with CO at the coverages studied. Moreover, the change in scattering

rate due to CO on the terrace should be independent of the presence of scattering centers

on the step edges (Mathiessen's rule). For purposes of this study, then, the sample can be

regarded as a Pt(1 l 1) crystal.

32

The sample was mounted at the end of a stainless steel tube, which could act as a
liquid nitrogen cold finger, on the manipulator. The sample was heated radiatively by a
tungsten filament and the temperature was measured by two chromel-alurnel
thermocouples spot-welded to the edge of the sample. Temperatures from ~ 90 K to
~ 1400 K were routinely accessible.

The sample could be dosed with various gases by means of a leak valve and a
multihole effusive doser array,3 which provided a flux enhancement of about 16 over
background dosing. Details of doser design are given in Chapter 3. AES was used to
determine the contaminants present on the surface and thus the necessary cleaning
protocol. For example, large amounts of Ca, which segregated from the bulk, required
cycles of Ar+ sputtering and armealing to 1175 K. This typically left a great deal of C on
the surface. Exposure to 02 at various elevated temperatures was an effective way to
remove the C, but often left a small amount of O on the surface which Was difficult to
remove. Surface 0 can usually be removed by heating to 1360 K. In cases where this
was not effective, the O was probably bonded to another surface contaminant such as Si,
which is difficult to observe with AES because the primary Si peaks overlap with strong
Pt peaks.4 In these instances it was necessary to sputter the surface again and repeat the
entire procedure. When the AES indicated no contaminants above the noise level (~ 4%
of the Pt 237 eV peak) the surface was considered clean. Figure 2-2 shows an ABS scan
that meets this criterion. Note that the positions of the peaks associated with the major

contaminants are indicated in the figure.

33

 

 

 

 

 

 

 

 

 

 

0.18 . . a - . . , . ,
0.14 - -
0.10 - 0 j
C Ca j ‘
g 0.06 r- l l
m i 1
E L
z 0.02 -
'O
. T 1 T .
-0.02 - j Pt 1 i 1 Pt Pt -
. Pt“ Pt j Pt .
-0.06 - -
. 1 t
P
_010 l 1 Pt I t. l r l r l
100 200 300 400 500

electron energy (eV)

Figure 2-2. AES for a clean Pt surface. The vertical axis is the derivative of the number
of electrons detected at a particular energy, with respect to energy.

Unfortunately, the hot filament used in AES itself tended to deposit traces of C on
the surface. Therefore, the last sequence performed consisted of exposure to oxygen at
90 K followed by TPD and flash heating to 1360 K. A secondary indication of a clean
surface was when repeated 90 K 02 closing and TPD showed that the 02 peak had reached
a maximum. A typical TPD spectrum following 90 K 02 dosing of a clean Pt(lll)
surface is shown in Figure 2-3. Note that the 02 peak dominates all other monitored

masses. H2 and a gas with a mass of 28 amu, either CO or N2, each Show small peaks at

34

low temperature. The fact that the peak occurs at low temperature indicates that it

 

 

 

 

 

 

 

 

probably corresponds to N2.
I I V l I
2.0 - -
1.6 - .
”a?
3 <— 02
'3 1.2 r -
E9
5 H20
g 0.8 L .
L CO 01' N2
0 4 - n. H2 -
t‘ ‘4
’W
0.0 a l a l a l m
0 200 400 600 800 1 000

temperature (K)

Figure 2-3. TPD spectrum from a clean Pt(1 11) surface following 02 dosing at 90 K.

As stated above, an ordered surface with its normal in the (111) direction was
indicated for our Pt sample by its hexagonal LEED pattern. Doublet spots confirmed the

roughly 5% miscut from the (111) direction calculated from Laue X-ray observations.

35

We were also able to observe two ordered patterns for CO adsorbed to the Pt(1 l 1) surface

which have been previously studied.5 At a coverage of 0 = 0.33 ML a diffuse
(«5 x J3)R30 pattern was observed. A fairly sharp c(4 x 2) pattern was observed at a

coverage of 0 = 0.5 ML.

2-2. Reflection-Absorption Infrared Spectroscopy

Figure 2-4, which is reproduced from Reference 6 by C. Chung, shows the layout
of the spectroscopic system. The instrument is a home-built apparatus with a spectral
range of ~ 300 - 3000 cm", and is designed for use in the study of low-frequency
vibrations of atoms and molecules at surfaces. While many design choices were made
with that specific mission in mind, the instrument was successfirlly utilized in this study
to measure the broadband change in reflectance, as will be described in Section 2-3.

A conventional silicon-carbide globar heated resistively to about 1300 K serves as
the IR source. The globar is situated at one focal point of an externally adjustable gold-
coated ellipsoidal mirror. The IR light from the source is intensity-modulated with a
tuning fork chopper7 which is placed at the other focal point of the ellipsoidal mirror.
The tuning fork chopper requires no lubrication and very low power, making it ideal for
operation in vacuum at low temperature. It chops the light at a frequency of 800 Hz,
which is chosen to avoid mechanical resonances in the instrumentation. The source
housing, chopper, and mirror are rigidly mounted to a copper plate which is in contact
with a liquid nitrogen reservoir. The plate is supported by three rods kinematically

mounted to the bottom of the source cryostat. The source housing, chopper, and mirror

36

are cooled to liquid-nitrogen temperature in high vacuum to reduce modulated radiation

from sources other than the globar as well as atmospheric absorption of the light.

 

FAR- IWRARED SLRFACE MCIROSCOPY SYSiEM Liquid NlUO‘lICII l.t)‘"Cd

Spectral rmge: .11!) ~ .1000 aI-l (stating petlfmltflt'l
Resolution: 1-5 ‘CI'I
SWIG 1w: 20 ~ 1200 K

 

 

 

 

   

 

 

 

-._. "1;. ;_—JV

 

Ultrmiql Vaculn Surface Analysis
Elmer uith Single Crystal Staple
on Lite-cooled manipulator

1".1' '«I.11,'tl l .!1

l1.ll 11 vi: Delel lur

 

 

 

 

 

 

 

 

 

(\ .3 E35

’Inr' V-—

1:;

 

 

\. ‘4’“ . .. i
R C t t Other surface proves. . .__. _, d
' 509’“? W05 0, low inetqy Electron Uillnlt't i011 _
Ill" lN-coaled 5’1'9'95- Auger Electron ‘upet‘llanopy f- ‘0 CI, ”4
i’OlUf l/tfl (lfld 010111le “1911111” [#330er 1011 'lpr’iil OSLOpy

Figure 2-4. The layout of the optical path. Reproduced from Reference 6 by C. Chung.

Within the same high vacuum chamber, the IR light passes through the chopper to
the transfer optics. These consist of two gold-coated, off-axis parabaloidal mirrors, the
first of which can be adjusted from outside the vacuum. The transfer optics magnify the

image of the chopper by a factor of 1.8 so that the IR light will interact with as many

37

adsorbates on the sample surface as possible. The light is incident on the sample at 86°
from the surface normal through a differentially pumped CSI window8 on the UHV
chamber. CSI is chosen due to its high transmittance down to low frequencies.

The reflected IR light passes out of the UHV chamber through a second CSI
window into a complementary transfer optics which are a mirror image of the first pair.
Thus they demagnify the IR light by a factor 1.8 and focus it onto the entrance slit of the
spectrometer.

The Spectrometer, the heart of the apparatus, is a home-built Czemy-Turner
grating spectrometer housed in high vacuum and cooled to liquid nitrogen temperature.
As with the source, the vacuum serves to reduce atmospheric absorption as well as
provide thermal insulation for the cold components. Cooling the spectrometer reduces
the background photon flux that reaches the detector. The optical components are rigidly
attached to a large aluminum plate which is in contact with a liquid nitrogen reservoir.
The optical plate sits on three invar rods kinematically mounted to the bottom of the
spectrometer shell. During normal operation the optical plate is clamped to the outer
shell for mechanical stability via three steel screws.

A variety of entrance slits for the spectrometer are mounted on an adjustable
wheel at the focus of the transfer optics, allowing the resolution of the spectrometer to be
changed externally. After the entrance slit the light encounters a gold-coated folding
mirror which directs it toward the collimating mirror. The collimating mirror is a gold-
coated, off-axis parabaloidal mirror, focused at the entrance slit. The now collimated

light next diffracts from the grating, which is a gold-coated replica on an aluminum

38

substrate.9

Three different gratings are used to cover the entire spectral range and the
system must be opened to change gratings. The diffracted light is collected and focused
by the camera mirror, which is also a gold-coated, off-axis parabaloid. Following the
camera mirror is another folding mirror which directs the light through a polarizer. The
polarizer consists of a gold wire grid with a line spacing of 2880 lines/mm on a silver
bromide substrate.10 Finally, the p-polarized component of the light focuses on the exit
slit wheel. Similar to the entrance slit wheel, the exit slit wheel is externally adjustable.
The exit slit wheel, however, includes a variety of low-pass IR filters which enable the
attenuation of higher orders of diffraction.

Upon leaving the spectrometer, the light enters a final transfer optics consisting of
a gold-coated, externally adjustable, ellipsoidal mirror. The exit slit of the spectrometer
is situated at one focal point of the ellipsoidal mirror, and the detector is situated at the
other. The transfer optics is in the same high vacuum Space as the spectrometer, the UHV
charnber-to-spectrometer transfer optics, and the detector, and has its own liquid nitrogen
reservoir.

Just before the detector in the optical path is a Winston cone. A Winston cone is a
parabolic cone designed such that any light entering the wide end within the angular
acceptance of the cone undergoes successive reflections and emerges from the narrow
end. It collects the IR light and delivers it to the small area of the detector element
without forming an image. The Winston cone has a gold-coated interior surface, an
entrance aperture of 10 m, an exit aperture of 1.0 mm diameter, and a focal ratio of 4.4.

The IR detector is an extrinsic Si:B photoconductor operated at liquid He

temperature.11 The signal fi'om the detector is amplified by a two-stage transimpedence

39

amplifier (TIA) with two switchable feedback resistors. The high and low feedback
resistors are mounted on the liquid He temperature cold plate to reduce Johnson noise and
can be chosen via a latching relay. The first stage of the amplifier uses a matched pair of
J230 JFETS mounted on the liquid He cold plate and heated to 77 K. The second stage is
a conventional AC coupled operational amplifier at room temperature with a gain which

is switchable between two and ten.

Section 2-3. RAIRS Measurement Technique

The detector is normally operated at a DC bias of 2.50 Volts. When the
spectrometer is warm, or when the grating is at zero degrees, the detector must be set to
low feedback. High feedback can be used when the spectrometer is cold because the
background photon flux is so much smaller. The chopped signal is taken to a lock-in
amplifier12 and demodulated using the chopper output as the reference. The IR intensity
data are collected from the lock-in amplifier by computer as a function of either
frequency (grating angle) or time (for a fixed grating position).

A vibrational measurement involves scanning through the spectral region of
interest once when the sample surface is clean, dosing with a gas, and scanning through
the same spectral region a second time. Absorption peaks or other spectral features due
to adsorption of the gas are found when the data are represented as a fractional difference
(i.e. the clean scan is subtracted from the scan of the closed surface, and the difference is
divided by the clean scan). It is possible to choose a particular frequency of interest and
monitor the intensity of the reflected light as a function of time while dosing with the gas.

This type of scan is referred to as a “time scan.” If a frequency away from sharp spectral

40

features is chosen, one can use a time scan to observe in real-time the change in intensity
of the broadband reflectance (at a single frequency) caused by the adsorption of a gas on a
metal surface. Both spectral scans and time scans were used in this study.

A typical data taking procedure for a broadband measurement on the Pt(1 1 1)
sample went as follows. First the sample was flashed to 975 K to clear the surface of a
passivating layer of CO. The monochromator was set to a single IR frequency and the IR
intensity was monitored for 5-30 minutes, during which interval the sample was dosed
with CO. Finally, TPD to 975 K was performed, which provided information about the
coverage of CO during the measurement and prepared the sample for the next
measurement.

Figure 2-5 shows the change in the reflected IR intensity versus time for a single
measurement. Figure 2-5(a) is the raw data, showing the voltage output of the lock-in
amplifier. In Figure 2-5(b) the data are normalized to the initial value, with a systematic
linear baseline drift subtracted. During this run the sample was closed with CO for 100
seconds beginning at about the 200th second at a background pressure of 8.0 x 10"0 Torr.
The pressure at sample was enhanced by about a factor of about 16 due to the closer (see
Chapter 3). The fractional change in reflected IR intensity for this run, AI / I , and

therefore the fractional reflectance change, AR/ R, was 0.00200 i 0.00002.

41

 

 

 

 

 

 

 

 

 

 

0.03526 . r . r, ~ . r
(a) *
0.03524 - .
3 0.03522 . ~
s. '2
.5 l i
8 1
fig 0.03520 - 1 1 -
a: 1
3 l0.42i:0.03 ML CO/Pt(l 11) l
_8 0.03518 - 2800 cm-1 i
‘13 315 x 1 1
1
0.03516 - q
dosing ‘-
0.03514 . ' . L. . 1' ~ ' .
0.0005 * I ' f; ' I I r I f
l ! ‘1
.‘ 1 ’
‘ ll 11 l I a”
0.0000 1' i . 1 ‘.Id ”H" ”i. ' Ill 1
l H 11 1 ~ l 11'] g A
1 . . l *
-0.0005 - ' ; Ti ‘
1 i
l : ‘
S -0.0010 - 1 ; 0.200%10.002% -
<1 . 1 1
04210.03 ML CO/Pt(1 11) l ?
-0.0015 - ' . -
2800 cm1 i z
. . , : 4
315 x 'L
_0.0020 b .1, Ill Alli“ I.‘ l i' i Ii
111411 l l 1111
-0.0025 . . 3
fl dosrng A‘—
0 100 200 300 400 500

time (s)

Figure 2-5. An example of a reflectance time scan. The raw data are shown in (a), while
the data in (b) are divided by the initial value with a linear baseline removed. The
background pressure during the CO dose was 8.0 x 10'[0 Torr.

42

This particular run exhibited an unusually stable baseline, which can be observed
in Figure 2-5(a) . In most cases baseline drifts of 0.2% to 2% over the course of a run
limited the accuracy. We attribute the majority of this drift to a progressive loosening of
the electrical contacts on the globar as it is cycled through temperature extremes in
normal use. Tightening the contacts tended to reduce the observed drifi, but doing so was
inconvenient because it required disassembly of the source cryostat. Even larger drift
was observed when the coldfinger was filled with liquid nitrogen due to sample drift
resulting from expansion and contraction of the coldfinger as the liquid nitrogen boiled
away. Usually the drift from both origins was quite linear which allowed us to subtract
the baseline from the measurement, and the reflectance change could be measured to an
accuracy of 0.02% or better.

It is of interest to compare the sensitivity of our measurements to that of the U4IR
beamline at the National Synchrotron Light Source (N SLS) of Brookhaven National
Laboratory, where most of the previous broadband reflectance studies have been carried

13-18

out Although the IR flux from the synchrotron is much greater than in our system,

electron beam movement and uncertainties in beam current limit the sensitivity to

broadband changes to about 0.2%,”’18

nearly an order of magnitude above our level. The
changes observed here for CO on Pt(1 l 1), which have a maximum magnitude less than
0.3%, would be extremely difficult to measure with the synchrotron system. A
significant drawback of our system, however, is that we measure only a single IR

frequency in each run, whereas the FTIR spectrometer at the NSLS obtains the entire

spectrum at once.

43

The CO coverage was determined with reference to the C=O stretching frequency
of atop-bonded CO, an example of which is shown in Chapter 1, Figure 1-1. This
absorption peak appears around 2100 cm'l and is rather sharp (i.e. a width of S 10 cm").
It has been well studied (References 19-23 provide just a small sampling of the literature

concerning the C=O stretch mode), so the dependence of the peak frequency on coverage

is well known. When measuring AR/ R via time scans at 2500 cm“1 and 2800 cm'l, we

first measured the clean-surface reflectance through the spectral region 2050 to

2150 cm'l. After the time scan, during which CO was dosed onto the sample, we again
scanned the same region to record the atop-bonded C=O stretching vibration. ‘The
saturation coverage at room temperature was assumed”25 to be 0.5 ML and the rate of
change of frequency with coverage was taken from Figure 2-6, which comes from

Reference 23 and is a compilation of published data.

 

 

 

 

2110L- r + ' _.
2‘ +01:
's -: c
u q o e e
V t. v
E "’ o
C, + a O
a +" a. x
g 53° 9.. x
11.2090—" x x —
x
to x
or
(L x x

2080 L 1

0 0.3 0.6

Total CO Coverage (ML)

Figure 2-6. Dependence of C=O stretch peak frequency on coverage for atop CO on
Pt(1 l 1). From Ref. 23 by J.S. Luo, R.G. Tobin, and D.K. Lambert.

Measurements at lower frequencies required different gratings, and the C=O stretch
region could not be measured; for these measurements a dosage vs. coverage curve
determined from the higher-frequency measurements, as well as integrated TPD signals,
were used to determine the coverage. Uncertainty in the actual room temperature
saturation coverage of our sample could introduce an overall shift in the coverage scale.
Broadband reflectance change measurements on the thin fihn samples followed a
very similar protocol. There were, however, additional considerations in timing the
operations which resulted from the steps needed to grow a new film sample for each

measurement and to make a simultaneous resistance measurement.

2-4. Thin Film Growth

As shown in Chapter 1, the scattering model predicts a linear relationship between
the adsorbate-induced IR reflectance change, which is a bulk quantity, and the adsorbate-
induced resistivity change, which is typically measured on thin film samples. In order to
examine this relationship experimentally, it is necessary to choose a sample which is
thick enough to allow reflectance measurements to probe the bulk reflectance, yet thin
enough for the change in resistance due to the presence of adsorbates to be measurable.
The relevant length scale for reflectance measurements is the classical penetration depth
5 , which for Cu at room temperature15 is 270 A. The sample should be thicker than this
to satisfy the requirement that it be “bulk-like.” The scale of the resistivity change is

related to the elastic mean free path; the adsorbate-induced fractional change in resistivity

45

decreases in proportion to the ratio of If to the thickness,26 l/ t. The elastic mean free

path for Cu at room temperature is about 1140 A. We have chosen to grow Cu thin film
samples about 500 A thick on hydrogen-terminated silicon surfaces prepared by wet
chemical etching, which have been shown to be excellent substrates for growing epitaxial
Cu(100) films.2’°3"

E.T. Krastev made a number of modifications to the UHV chamber to equip it for
thin film growth.38 A sample transfer system was added which includes a
turbomolecular-pumped loadlock, a magnetically coupled linear translator, and a custom-
designed home-built sample holder. This allows the rapid transfer of new Si substrates
into the UHV chamber. The Si substrate is mounted on a small disk which is inserted
into the UHV chamber through the loadlock via the linear translator. Three spring-loaded
posts and a keyhole locking system affix it to the sample holder. This sample holder
replaces the one used in the single crystal experiments, but attaches to the end of the same
liquid nitrogen coldfinger and manipulator. The sample substrate has Six electrical
contacts: two for a thermocouple and four for a four-probe resistance measurement. As in
the single crystal experiments, the sample is heated radiatively by a tungsten filament.
The minimum temperature is limited by the quality of thermal contact allowed by the
spring tension holding the sample disk in place. The maximum temperature is limited by
the tendency of the springs to degrade at high temperatures. Temperatures of about 110
K to 775 K can be regularly employed.

An evaporator was added38 which includes two evaporation filaments, a rotating

shutter, a crystal thickness monitor, and appropriate shielding. The thickness monitor and

46

shielding can be cooled by flowing water or liquid nitrogen. Recently, Krastev again
modified the evaporator to include a loadlock.38 It is now possible to remove the
evaporator filaments to reload the evaporation metal without breaking the UHV.

In preparation for a measurement, the (100) Si substrates (B-doped, 20—50 S) cm
resistivity) are cleaved to the appropriate size for the sample holder and then prepared for
insertion into the vacuum chamber. The Si is then degreased in an ultrasonic cleaner
using first acetone and then methanol. Before etching, electrical contacts for the four-
probe resistance measurements are deposited onto the Si(100) substrates in a high
vacuum (HV) evaporation chamber. Using an aluminum foil mask wrapped around the
Si, 100 A of Cr followed by 1500 A Ag are deposited onto the edges of the Si. The Cu
films are deposited so that they overlap the contacts on the ends.

Crucial steps in the preparation of the Si are etching for 30 - 60 seconds in a 10%
aqueous solution of HF, and pull-drying (slowly and smoothly removing the substrate
from the solution with the surface vertical, so that the liquid sheets off smoothly with no
droplets). The etching and pull-drying lead to an extremely flat and chemically inert
surface, with virtually all the Si dangling bonds terminated”42 with H. After etching, the
samples are loaded into the loadlock and pumping is begun as quickly as possible.

After transferring the Si substrate into the UHV chamber, the substrate is heated
to 700 K for 10 minutes to outgas it. AES has shown that this leads to the production of
the cleanest Cu surface; the largest remaining contaminants tend to be C and O at a level
of less than 4% of the Cu 60 eV LMM AES peak. When fihns are grown without first
outgassing the Si, the C and 0 levels can be as high as 30% - 40% of the Cu 60 eV LMM

ABS peak. In order to sustain the outgassing temperature without overheating the

47

electrical feedthroughs or springs, it is necessary to cool the coldfinger with liquid
nitrogen. The liquid nitrogen also helps bring the sample’s temperature down after the
outgassing. Since the best film growth is achieved at room temperature,36 the sample
temperature is chosen to be about room temperature by balancing the heat lost to the
coldfinger with heat added by the sample heater. Following the outgassing, it normally
takes about one hour to stabilize the temperature. Once the temperature is stabilized, the
film can be deposited.

A Significant amount of work has been done by E.T. Krastev and L.D. Voice to
characterize the crystal structure, surface morphology, and electrical conductivity of these
films as a function of deposition rate, deposition temperature, and post-deposition
annealing.36 Some of the results will-be discussed here. The best films we grow are
obtained near room temperature at growth rates of 1.0-10 A /s. Higher temperatures and
higher deposition rates produce lower quality films. Post-deposition annealing, while
expected to improve the smoothness of the surface, can only be done at relatively low
temperatures due to the formation of copper silicides at the Si-Cu interface at higher
temperatures. Temperatures up to 125°C have little effect, so post-deposition annealing is
not used in this work.

While my interest is in producing high quality films for use in
reflectance/resistance studies, the Cu(100) on Si(100) growth system is unusual and
interesting in its own right because there is a 40% mismatch between the two lattices.
This fact makes the relatively easily achieved epitaxial growth of Cu(100) on Si(100)
curious. The existence of a mixed "buffer layer" about 100 A thick between the Si and

the Cu was discovered by an earlier study.43 This buffer layer is quite likely important to

48

the epitaxial growth. Reflection high-energy electron diffraction43 (RHEED) and grazing
incidence X-ray diffraction”44 studies have shown that the Cu lattice is rotated by 45°
relative to the Si lattice, with Cu(010) parallel to Si(011). The rotation reduces the lattice
mismatch between the two materials from 40% to 6%. Even a 6% mismatch, however, is
unusually large for epitaxial growth. The strain is probably at least partially relieved by
the buffer layer.

For most of the characterization work, E.T. Krastev deposited the copper films in

a resistive evaporator with a HV base pressure 1 X 10"8 Torr, although later films were
prepared in the 10'll Torr UHV chamber. The UHV films show test results indicating
that they are no worse than the high vacuum films in terms of structure and morphology,
and are usually among the best. Thus, the high vacuum film results can be considered
essentially valid for the UHV films as well.

The film orientation in the direction perpendicular to the surface plane was
characterized by standard 0-20 X-ray diffraction (XRD). Comparing36 XRD scans of
completely disordered (i.e. powdered) bulk Cu, polycrystalline Cu films, and Cu films
grown on oxidized (i.e. unetched) Si substrates to the scans of films grown on etched Si
substrates demonstrated highly oriented growth in the direction perpendicular to the
surface, with the Cu(100) direction aligned with Si(100).

45-48
Four

The inplane orientation was determined by means of X-ray pole figures.
very distinct poles were observed which confirm that the films posses very highly
oriented cubic structure and therefore they are epitaxial. Quantitatively, more than 97%

of the copper is oriented.38

49

For the UHV films we were able to examine the surface order qualitatively using
low energy electron diffraction (LEED). At room temperature we successfully observed
the square pattern expected for a Cu(100) surface.

The surface morphology was studied in air using a commercial atomic force
microscope49 (AF M). The primary measure of surface roughness used was the root-
mean-square (RMS) deviation of the height from the mean. Scans of bare Si substrates
yielded RMS roughness values of 2.0 - 3.0 A. Typical AFM images for Cu samples
grown at deposition rates of 1.0 -2.0 A /s at room temperature exhibit a granular structure
with RMS roughness values of 10 - 20 A. Although the films are epitaxial, they are
relatively rough on the atomic scale.

Finally, L.D. Voice performed four-wire resistance measurements on selected
films during HV deposition, which allowed the fihn resistance to be continuously
monitored as a function of film thickness. He used the growth conditions which had been
demonstrated to produce the best films in terms of crystal structure and surface
morphology.36 The results indicated a high degree of scattering, either from surface

defects or internal grain boundaries, was likely. Nevertheless, for the highest thicknesses

the conductivity approaches a value of 55 (u!) - m)-l , which agrees with the value of o for

pure Cu at the deposition temperature, 54 (u!) - m)".

50

2-5. Adsorbate-Induced Resistivity Change Measurement

Once the new film is deposited, it is necessary to wait for the UHV chamber to
attain an acceptable base pressure, and for the sample to stabilize at the desired
temperature. The pressure typically reaches 1x10'10 Torr after approximately 20 minutes,
at which point the measurement is started. During the waiting period the sample must be
moved to the IR position and the spectrometer must be prepared for a measurement by
setting the frequency and maximizing the light intensity that reaches the detector. When
all systems are ready, simultaneous reflectance and resistivity time scans are begun.

The internal reference of a lock-in amplifier12 (LIA), set to 1.00 V at 1.0 KHz or
3.0 KHz, is applied to the sample. This provides approximately 1.7 mA current to the
sample due to the 600 (2 output impedance of the LIA which is large compared to the
sample’s resistance, typically 2 - 5 Q. The voltage drop across the sample is measured by
the LIA using two shielded coax cables, with the shields grounded, for leads. The LIA
reads the difference between the two leads, which helps to eliminate noise common to

both. The capacitive and inductive pickup in the leads were both measured to be about
8x10'9 V/J Hz , which is only slightly above the limit of sensitivity of the LIA, 5x10'9

V/JH—z at 1 KHz. The data is collected from the LIA by a computer using a program
similar to the one used for reflectance time scans.

As with the reflectance time scans, for resistivity time scans it is important to
measure the clean-surface value long enough to establish a stable baseline. The sample is
then dosed with the desired amount of gas by backfilling the UHV chamber. The doser is

not used in the thin fihn experiments due to Space constraints in the UHV chamber; with

51

the closer in place it is impossible to transfer a sample between the loadlock and the
sample mount. Following the dose, the resistance is monitored long enough to again

establish a stable baseline in both time scans. Figure 2-7 is an example of a resistivity

 

 

 

 

 

 

 

 

 

time scan.
0.12 ' I ‘ I ' I ‘ I ' I ' I ' I
0.08 - .
i 1
g 0.06 P _, dosing (<— -
4 500 L CO dose
004 _ 530 A Cu(100) -
300 K
0.02 - -
0.00 ~=— -
0 100 200 300 400 500 600 700 800
time (s)

Figure 2-7. A typical resistivity time scan. The data are divided by the initial value with
a background slope subtracted. Background pressure during the dose was 5x10'6 Torr.

52

In Figure 2-7, the data are divided by the initial value and a linear baseline is
subtracted. The 500 L CO dose was achieved by backfilling to a pressure of 5x10’6 Torr
for 100 5. During the dose Ap/ p quickly reaches a plateau. After the dose ends a
surprising thing happens in this run: Ap / p actually decreases. This could result from an
oversaturation of CO on the surface such that the coverage decreases due to CO leaving
the surface as the CO pressure in the UHV chamber decreases. Other possibilities include
relaxation of the CO overlayer or adsorbate-induced reconstruction of the Cu surface.
Simultaneous measurements of reflectance and resistivity will be discussed in more detail

in Chapter 5. '

53

References

lPerkin-Elmer Corp. TNBX-1000 ion pump system.

2RC. Lin, Ph.D. thesis, Michigan State University, 1995, unpublished.

3D.l~:. Kuhl and R.G. Tobin, Rev. Sci. Instr. 66 (1995) 3016.

4LE. Davis, N.C. MacDonald, P.W. Palmberg, G.E. Riach, and RE. Weber, Handbook
of Auger Electron Spectroscopy, 2nd ed. (Physical Electronics Industries, Inc., Eden
Prairie, MN, 1976).

5H. Steininger, S. Lehwald and H. Ibach, Surf. Sci. 123 (1982) 264.

6C. Chung, Ph.D. thesis, Michigan State University, 1993, unpublished.

7Multi-Scanning Systems Corp., type RC-2.

8R.G. Tobin, C. Chung and J .S. Luo, J. Vac. Sci. Technol. A 12 (1994) 264.

9Milton Roy Co.

loPerkin-Elrner Corp.

11Infrared Laboratories, Inc.

'2 Princeton Applied ResearchCorp.

l3C.J. Hirschmugl, G.P. Williams, F.M. Hoffinann and Y. J. Chabal, Phys. Rev.
Lett. 65 (1990) 480.

“or. Hirschmugl, Y.J. Chabal, F.M. Hoffrnann and GP. Williams, J. Vac. Sci.
Technol. A 12 (1994) 2229.

”KC. Lin, Ra. Tobin, P. Dumas, C.J. Hirschmugl and GP. Williams, Phys.
Rev. B 48 (1993) 2791.

16KC. Lin, R.G. Tobin and P. Dumas, Phys. Rev. B 49 (1994) 17 273 (1994);
ibid. 50 (1994) 17760.

54

17Cl Hirschmugl, G.P. Williams, B.N.J. Persson and A. I. Volokitin, Surf. Sci.
317(1994)L1141.

18C.L.A. Lamont, B.N.J. Persson and GP. Williams, Chem. Phys. Lett. 243
(1995) 429. ,

198.13. Hayden and AM. Bradshaw, Surf. Sci. 125 (1983) 787.

20M. Tushaus, E. Schweizer, P. Hollins, and AM. Bradshaw, J. Electron Spect. 44 (1987)
305.

21CW. Olsen and RI. Masel, Surf. Sci. 201 (1988) 444.

221D. Beckerle, R.R. Cavanaugh, M.P. Casassa, E.J.Heilweil, and J.C. Stephenson, J.
Chem. Phys. 95 (1991) 5403.

231s. Luo, R.G. Tobin and D.K. Lambert, Chem. Phys. Lett. 204 (1993) 445.
2“1.). Malik and M. Trenary, Surf. Sci. 214, (1989) L237.
25L.F. Sutcu, J.L. Wragg and l-I.W. White, Phys. Rev. B 41 (1990) 8164.

26P. Wissmann, in Surface Physics, edited by G. Hahler, Springer Tracts in Modern
Physics 77 (Springer, New York, 1975).

27CA Chang , Phys. Rev. B 42 (1990) 11 946.
28CA Chang, Surf. Sci. 237 (1990) L421.

29C..A. Chang, J. Appl. Phys. 68 (1990) 5893.
3"CA Chang, J. Vac. Sci. Technol. A 8 (1990) 3779.
3’C.- A. Chang, J. Vac. Sci. Technol. A 9 (1991) 98.

32Y.-T. Cheng, Y.-L. Chen, M.M. Karmarkar and W.-J. Meng, Appl. Phys. Lett. 59
(1991) 953.

33Y.-L. Chen and Y.-T. Cheng, Mat. Lett. 15 (1992) 192.
34Y.-T. Cheng and Y.-L. Chen, Appl. Phys. Lett. 60 (1992) 1951.

35R.Naik, M. Ahmad, G.L.Duifer, C.Kota, A.Poli, Ke Fang, U. Rao and J.S. Payson, J.
Magnetism and Magnetic Mat. 121 (1993) 60.

55

36E.T. Krastev, L.D. Voice, and R.G. Tobin, J. Appl. Phys. 79 (1996) 6865.

37Minsu Longiaru, R.G. Tobin, E.T. Krastev, and L.D. Voice, J. Vac. Sci. Technol., to be
published.

38E.T. Krastev, Ph.D. thesis, Michigan State University, in preparation.

39T. Takahagi, I. Nagai, I. Ishiytani, H. Kuroda and Y. Nagasawa, J. Appl. Phys. 64
(1988)3516.

40T. Takahagi, I. Ishiytani, H. Kuroda, Y. Nagasawa, H.1to and S.Wakao, J.Appl. Phys.
68 (1990) 2187.

“YJ. Chabal, G.S. Higashi, K. Raghavachari and VA. Burrows, J. Vac. Sci. Technol. A
7(1989)2104.

42P. Dumas, Y.J. Chabal and GS. Higashi, Phys. Rev. Lett. 65 (1990) 1124.

43BCDemczylr, R.Naik, o. Auner, C.Kota and U.Rao, J. Appl. Phys 75 (1994) 1956.
44CA Chang, J.C. Liu, and J. Angilello, Appl. Phys. Lett. 57 (1990) 2239.

45J.S.Kallend, U.F.Kocks, A.D.Rollett, H.-R.Wenk, Materials Science and Engineering,
A132 (1992) 1.

4(U.F.Kocks, J.S.Kallend, H.-R.Wenk, A.D.Rollett, S.I.Wright, popLA manual, Los
Alarnos National Laboratory, July 1994.

47H.-R.Wenk, Preferred Orientation in Deformed Metals and Rocks: An Introduction to
Modern Texture Analysis, Academic Press, New York, 1985.

48M Hatherley, W B Hutchinson An Introduction to Texture in Metals, The Institution of
Metallurgists, Monograph No 5.

49Digital Instruments Nanoscope III Scanning Probe Microscope.

Chapter 3

Capillary and Effusive Gas Dosers

In many surface science experiments, including those that comprise this study, it
is necessary to expose a surface to a known and uniform flux of gas. Often dosing can be
accomplished simply by filling the vacuum chamber to a uniform pressure. When
dealing with highly reactive gases, gases that are not pumped effectively, gases that
exchange rapidly on the chamber walls, and very high flux levels, however, it is
necessary to use a doser that provides a high flux at the sample while keeping the
background pressure in the chamber low. An example where a doser is needed which is
relevant to this study is 02, which is important in the Pt surface cleaning process and has
a low sticking coefficient on Pt.

Dosers are typically one of two types: effusive, in which the aperture diameter is
much larger than its length, or capillary, in which the diameter is much smaller than the
length. The effusive doser is usually a single effusive aperture, whereas the capillary
doser is commonly used either singularly (the “needle” doser) or in an array. The process
of designing and implementing a closer appropriate for our UHV system led us to
investigate the relatively unexplored possibility of arrays of effusive dosers.

The primary criteria in designing a doser are the uniformity of the flux on the
sample and the enhancement factor. The enhancement factor is defined as the ratio of the

flux at the sample to the flux on a surface elsewhere in the chamber, or equivalently the

56

57

ratio of the "effective pressure" at the sample to the rise in the background pressure. The
design of a doser inevitably involves a tradeoff between these two criteria.

In previous work, Campbell and Valonel published a theoretical analysis
comparing capillary arrays with a single effusive source. They calculated flux
distributions and enhancement factors for circular capillary arrays ranging from a single
capillary to an array with a diameter equal to the sample diameter, for a range of sample-
to-doser distances. Their results Show that a single needle closer2 gives a very large
enhancement factor but an extremely nonuniform flux distribution. They also concluded
that a single effusive source (“cosine emitter”) is in many cases as good as a capillary
array in both uniformity and enhancement factor. To obtain reasonable uniformity, a
capillary array needs to be at least as large as the sample; no results were given, however,
for arrays larger than the sample. Winkler and Yates4 considered larger arrays but
calculated only the enhancement factor, not the flux distribution.

Here we extend Campbell and Valone's calculations to larger capillary array sizes.
In the following, all dimensions are assumed to be much smaller than the mean free path
of the gas molecules. We is also show that a single effusive source is far less effective
than their curves suggest, and indeed is usually inadequate unless a very large flux
gradient across the sample is tolerable. We present flux calculations for effusive arrays
consisting of a relatively small number of effusive sources situated near the perimeter of
the sample, and Show that such arrays can be competitive with capillary arrays if the
closer is not too close to the sample. This Chapter is based on work published in

Reference 4.

58

3-1 Theoretical basis for doser modeling
The starting point is the calculation of F(0), the integral-normalized flow of

molecules per unit solid angle from a single source, at an angle 0 to the source axis:

9 3-1

 

where 1(0) is the flux in molecules-s"-sr", and Nm, is the total flow out of the source in

molecules-s". For an effusive source F (0) is

F(9)= €089 . 3-2
7!

 

and for a single straight capillary of length L and radius a it is:5

nap;

2cos0 W 2 ; cos0
F(9)= “W {(1——)R(p)+$(l-W{l—Q—pZY]+ }0<a<0c 3-3

 

 

2 2

2
F(9)=cos 0+cos0 9>9e
nsrnO 2

 

59
where p = L(tan 0)/2a, R(p) = COS'1(p) — p(1-—p2)‘/2v 0,, = tan'1(20/L), and W =
(8a/3L)(l + 8a/3L)'1. As other authors, such as Benziger and Madix, have noted,5 the
flux from a capillary of large L/a is highly collimated. For example, for Ma = 40, 97% of

the molecules emerge within a cone of half-angle BC = 29°. Little is gained by using

larger L/a ratios; the central peak becomes sharper, but the fraction of flux at larger
angles does not change. For this reason, we will follow Campbell and Valone in using
L/a = 40 throughout this work.

The highly collimated flow from a single capillary has been cited1 as
demonstrating that a single needle doser is unsatisfactory. This conclusion is correct but

probably overstated. The expressions for F(G) for 0 < BC in Equation 3-3 assume that

some molecules can go directly through the capillary from the reservoir behind it, without
hitting the walls.5 This is generally correct for capillary arrays, but a needle doser more
ofien consists of a tube with at least one bend between the open end and the gas reservoir.
Such a geometry will lead to a flux distribution less strongly peaked in the forward
direction than that given by Equation 3-3. Nevertheless it will be more directional than a
single effusive source, which we Show below is already unacceptably nonuniform for
most applications.

From a practical point of View the flux per unit solid angle is of less interest than
the integral-normalized flux per unit area of the sample, which we denote G(0). Figure
3-1 illustrates the geometry of a planar sample oriented perpendicular to the axis of the

emitter.

60

sample

 

 

 

Figure 3-1: Origin of the cos3(0) term in the calculation of flux per unit sample area.

Using this geometry, it is easy to show that the integral-normalized flux per unit area of

the sample is given by

0(a) = F(e)cos30 . 34

The extra factor of cos30 has a noticeable but modest effect on the rather directional flux

from a capillary, but a severe effect on the flux from an effusive source. In fact, it is

somewhat misleading to refer to the single effusive doser as a “cosine emitter” when one

61

considers the extra factor of c0530. The flux uniformity plots shown by Campbell and
Valone plotted 0(0) for the capillary arrays, but F (0) for the efiirsive source. The result is
a serious underestimate of the flux gradient for the effusive source. When the flux per
unit area is considered, an effusive emitter placed 1/2 the sample diameter away has a
flux at the edge of the sample that is only 25% of the flux at the center, not 71% as
Reference 1 suggests. Such nonuniformity is unacceptable for most surface science
experiments, particularly those such as infrared reflectance and temperature-prograrnmed

desorption that average over the entire sample area.

The enhancement factor E is given byI

f,S 2n

(1—sf,) kBT ’ 3'5

 

E21+
A

where A is the sample area, S is the pumping speed, k3 is the Boltzmann constant, T is the
temperature, s is the sticking coefficient and f; is the fraction of molecules leaving the

closer that directly impinge on the sample. The right-hand side of Equation 3-5 gives the
value of E if all molecules that do not immediately strike and stick to the surface are
scattered into the vacuum chamber and subsequently pumped at speed S. E can be much
larger if many of the molecules that miss the sample strike and are immediately trapped
on other surfaces — cold surfaces of the sample mount or manipulator, for example. Since

the expression for E involves quantities that are not properties of the doser, we

62

characterize the effective enhancement of the closer with fs, which is a property of the

closer geometry alone.

3-2 Calculations
The following computer modeling assumes a circular sample of diameter D
parallel to a circular doser array of diameter d and a distance i away, as shown in Figure

3-1. The doser-to-sample distance is characterized by the angle 0,“, from the center of
the closer to the edge of the sample; tan 0",“ = D/2l’. The capillary arrays consist of 5096

capillaries with L/a = 40, on a square net. Tests with larger numbers of capillaries
showed that this was a sufficiently dense grid to be effectively continuous. The flux is
calculated along a line of 50 points from the center to the edge of the sample. The
effusive arrays consist of a set of holes equally spaced on a circle of diameter d, with or
without an additional hole in the center. More complex geometries, with multiple rings
of holes, would probably improve the results when the doser and sample are very close

together, but do not provide much improvement for emax < 60°. Since the effusive arrays

we considered have four-fold, but not fill] rotational symmetry, we calculate the flux at
1254 points in a square net covering one quadrant of the sample.

At each point on the sample, the flux per unit area from the entire array, 00,,(5),

is calculated by summing the contributions from all the individual sources:

Garr(5) = 26(91') 3-6

63

where 0, is the angle from the ith source to the point I) and the sum runs over all the

sources. Each .emitted molecule is given only one chance to stick so that no multiple
collisions are included.
We characterize the uniformity of the flux distribution with the area-weighted

fractional standard deviation in flux, which is given by

 

9 3-7

 

linens-Gale
0’ = Gave

where

1 ..
Gave = '2 ij(P)dA 3-8

is the average flux, A is the sample area, and the integrals are taken across the entire
sample. Since the flux distribution functions are normalized to the .total rate at which

molecules are emitted, the fraction of gas directly hitting the sample is given by

3-9

64

where N is the total number of sources. All calculations are accurate within :l:1%;

convergence was checked by performing calculations with larger numbers of data points.

3-3 Results and discussion
Table 3-1 shows the values of f, and o for several effusive array designs with

d/D = 0.8 and em, = 40° and 60°.

Table 3-1. Comparison of circular effusive doser arrays with various numbers of
holes. All designs except those with eight holes comprise a single center hole surrounded
by a ring of equally spaced holes at d/D = 0.8. The eight-hole arrays have only the ring
of holes, without a center hole. It can be seen that increasing the number of holes from 9
to 17 has little effect. The center hole improves the enhancement factor Slightly; it
improves the uniformity for em, = 60° and degrades it for 0",“ = 40°, but in both cases

the effects are small.

 

 

 

9m, Number of holes f3 0'
40° 5 0.32 0.15
9 0.31 0.12

17 0.30 0.1 l

8 0.30 0.09

60° 5 0.56 0.25
9 0.54 0.10

17 0.53 0.08

8 0.51 0.13

65

The five, nine and l7-hole arrays include a center hole; the eight-hole arrays lack the

center hole. For 0”,“ = 60°, the inclusion of the center hole improves both the uniformity
and the enhancement modestly; for 0",” = 40° it slightly improves the enhancement at

some cost in uniformity. Increasing the number of outer holes from four (five-hole array)
to eight (nine-hole array) leads to significant improvement, but further doubling the
number of outer holes to 16 (17-hole array) has little effect. Based on these observations,
we confine our study of effusive arrays to nine-hole arrays comprising a ring of eight
holes around a central hole.

Figure 3-2 compares the flux uniformity across the sample for several choices of

doser design at different doser-to-sarnple distances, parameterized by the angle 0",“, (see
Figure 3-1 for a definition of 0",“). The plots Show the flux per unit sample area Gm(p),
normalized to the value in the center of the sample, 00,,(0), along a radial line. For the

effusive arrays the line passes directly in front of one of the perimeter holes. The data are
plotted versus the square of the distance from the center of the sample, p2, so that equal
intervals along the x-axis correspond to equal areas of the sample. This presentation
provides a more accurate visual representation of the uniformity than a plot of the flux VS.
p, which underemphasizes the significance of the region near the edge. For those
capillary array geometries that were also considered in Reference 1, the two calculations
are in excellent agreement. For the single effusive doser, however, there is strong
disagreement, since Reference 1 plotted the flux per solid angle F (0) rather than the flux
per sample area G(0). Our results Show that the single effusive doser provides reasonable

uniformity only when the doser—to-sample distance is very large (I > D).

 

 

66

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1.0 %-
l ‘5‘?--‘-""--,
0.9 ~ \‘35‘\ i * -
\ 5 \\\ t ,
h _, \:,.\c \ag‘\-:
0.8 ~ V; 0.8 -
. \ 09 f“? -\ ‘ . I“. 3:
0.7 *- ' \3\ E 0.7 s
' ‘ ‘ A ' .- s’
’5 0.8 r . 3 0.8 - r
‘6 ' i U i s:
> 0.5 ~ . > 0.5 r '3
a i i 3 04 i
c 0-4: 0m,=25° ‘ u - . -
0.3 -T - 0.3 ~
. - -- - d/D = 2.0 capillary . .
0.2 I --0-- d/D = 1.0 capillary 1 0.2 .
,i ~2— dfb =1.0 effusive , . 1
,2 —-?- d D = 0.5 effusive , 0 1 , .
0‘1 i —9— single effusive , ' L 9m“ '-" 40° ,
. A 0.0 ’ t
0”0.0 0.2 0.4 0.8 0.0 1.0 0.0 0.2 0.4 0.8 0.8 1.0
92 92
1.0 Wasn‘t-”run.”
we. ‘ ’ 1‘
“ ‘ O. t ‘
0.9 m\ V \ . ‘ .\ //‘\-11
film 1 ~ ‘9” .
0.8 r) g V. . -
s \ /O’ \ 1
\ \e—«x/ .
0.7 r- , \\ '. r
t \ X ‘ t
8 0-5 ’ E] \ ‘ 8
v , \ \ . , v
u \\ . o
> 0.5 " \\ "\~ 7 >
m * ’. R * 3
F; 0.4 1- \ ‘ u
1 (Si . <
0.3 " \ V '1
L "is 5‘
0.2 .\ \
* Bu
°" i err... = 60° \r» 4.
0.0 . m . . J
0.0 0.2 0.4 0.6 0.8 1.0
p2
Figure 3-2: Normalized flux per unit sample area versus the squared distance from the

sample's center for various sample to closer distances, indicated by 9max = arctan(D/2€).

d/D is the ratio of the doser's radius to the sample's radius. Shown are plots for a
capillary array of d/D = 2.0 and 1.0, a centered circular nine hole effusive array of d/D =
1.0 and 0.5, and a single effusive doser.

67
Not surprisingly, the greatest uniformity is provided by a large diameter (d/D = 2)
capillary array. Perhaps more surprising is the relatively poor uniformity given by a

capillary array equal in size to the sample. Even for 0mm = 75°, with the closer very close

to the sample (3 = 0.13D), 20% of the sample area receives a flux that is smaller than that

at the center by more than 20%. When the doser is farther away, 0mm < 45°, the nine-

hole effusive array with d/D = 1 provides a more uniform distribution than the d/D = 1
capillary array. As the closer and sample are moved closer together the discrete nature of
the effusive array becomes apparent in the strongly peaked flux distribution. It is likely
that greater uniformity for short doser-to-sample distances could be achieved with more
elaborate patterns of effusive emitters, but this is really the regime in which capillary
arrays afford the greatest advantage.

A complete comparison of closer designs must of course consider enhancement as
well as uniformity. Figure 3-3 plots 0, the area-weighted fractional standard deviation of

the flux, vs. f,, the fraction of emitted gas that impinges directly on the sample, for
effusive and capillary arrays of varying diameters and for different values of 0”,“. (Our
values of fs for the d/D = 1 capillary array agree with those of Campbell and Valonel and
Winkler and Yatesz) The single effusive doser represents the d/D = 0 limit of the effusive
array. In the d/D —-> 00 limit the value of a must again approach the single source value,

since the outer holes are too far away to contribute any flux; this is why the effusive array

curves have a "U" shape.

68

 

 

 

 

 

 

 

 

 

 

 

 

0.7
9 --O-- capillary array ’ .
0.6 ~ —t— 9 hole effusive array -- -
I single effusive doser I
- A smart = 400
0.5 — , .. .
., a
0.4 . —~ ~
D ° em“ = 25° .
. . .
0.3 ~ -. o
A
t . ‘
0.2 r "’ , ‘1
; f .9 d/D = 1.0
Od/D=I.0 _d/D=260 “ . _
0.1 - f . ‘ .
. A g
. x ..
’A’. ‘4‘.
0'000 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8 1.0
f, fs
. . I
2.0 : ‘1' 9| ‘
i o h i e - 76° 3
1.6 L email = 60 4- \1 max - J
. .. l“. . t
: 1
1.2 " -‘ \i\ g. .1
D ' 41 A\\ l; i
.' . ‘ 1’.‘ .
' a 1 ~\ .
0'8 F ' +5 V ‘ d/D 0 ' d
p o .‘ g l. l‘- t
a d/D = 2.0 f P .. k..\
> I} o \ “‘7 . .
0.4 i' \x f o a. r'" _,
r \ . -‘ t.
“ ‘ .. d D = 1.0
' t\t.”‘,‘d/D=r.o J / ’--O
00i -A..A.-." " 1 AA AA-""—." .
' 0.0 0.2 0.4 0.8 0.8 0.0 0.2 0.4 0.6 0.8 1.0
f, fs
Figure 3-3: Area-weighted fractional standard deviation 0 versus the fraction of

emitted molecules directly impinging on the sample f, for various sample to closer
distances, indicated by 9max = arctan(D/2l’). d/D is the ratio of the doser's radius to the

sample's radius. The data shown are for a capillary array of d/D = 0.4, 0.6, 0.8, 1.0, 1.2,
1.4, 1.6, 1.8, 2.0; a centered circular nine hole effusive array of d/D = 0.3, 0.5, 0.65, 0.8,
1.0, 1.35, 1.7, 2.0; and a single effusive emitter.

69

An ideal doser would have a small value of o and a large value off,, and would

lie in the lower right corner of the graph. Acceptable maximum values of o and

minimum values off, of course depend upon the application.

Several conclusions can be drawn from Figure 3-3:
1. Single effusive dosers and small capillary arrays (d/D << 1) are of little use unless
very large flux variations (0 > 0.2) are acceptable.

2. For any value of 0mm capillary arrays outperform effusive arrays.
3. For large doser-to-sample distances (small 0",”) dosers of any design are of marginal
usefulness, since it is virtually impossible to achieve f, greater than 0.3.

4. For small doser-to-sample distances, capillary dosers with diameters 20 - 40% larger
than the sample perform extremely well.

5. Effusive arrays perform best when the holes are placed near the perimeter of the
sample.

6. For intermediate doser-to-sample distances (0m, ~ 60°), effusive arrays perform
nearly as well as the best capillary arrays.

Under some experimental conditions these conclusions must be modified. One
reason for using a doser arises when the dosed molecules are very rapidly pumped, for
example by cold surfaces in the chamber, so that backfilling requires admitting an
excessive amount of gas. In such cases a large enhancement E can be achieved even with

a small value of 1;, and an effusive doser placed far enough from the sample to ensure

good uniformity becomes a more attractive option.

70

A second situation in which dosers are often used is when the dosed molecules
have a low sticking probability s. In this case the flux uniformity for capillary arrays
placed close to the sample will not be as good as our calculations suggest. Molecules that
do not stick to the sample are likely to scatter off the closer or its mount and reimpinge on
the sample. These multiple collisions, which are not included in our model, will tend to
enhance the relative flux at the center of the sample at the expense of the edges, leading
to a less uniform flux distribution. In such a case a larger doser-to-sample distance may
be preferable, and an efiirsive array, because of its simplicity and smaller optimal
diameter, becomes a more attractive option.

One virtue of effusive arrays is that they are easily made, and can be tailored to
unusual shapes and sizes of samples. For example, surface infrared Spectroscopy usually
requires relatively large oval or rectangular samples, and the doser must often be placed
at some distance from the sample to allow optical access. Since the infrared beam usually
probes the entire sample surface, good uniformity is essential. An effusive array with
holes placed near the edges of the sample can be an effective doser for such an
application.

The model presented here can be readily implemented using commercial software
on a personal computer, and calculations can be performed for any desired sample and
closer geometry. It is then straightforward to optimize a closer design for the demands of

any particular experimental situation.

71

References

lC.T. Campbell and S.M. Valone, J. Vac. Sci. Technol. A 3 (1985) 408.
2A. Winkler and J.T. Yates, Jr., J. Vac. Sci. Technol. A6 (1988) 2929.
3HOW. Beijerinck and NR Verster, J. Appl. Phys. 46 (1975) 2083.
4D13. Kuhl and R.G. Tobin, Rev. Sci. Instr. 66 (1995) 3016.

5). Benziger and RJ. Madix, Surf. Sci. 94 (1980) 119.

Chapter 4
Infrared Reflectance Change

Induced by CO Adsorption on Pt(111)

Measurements of the broadband IR reflectance change AR/ R were made for CO
adsorbed on a Pt(111) surface. The frequency, temperature, and CO coverage were
systematically varied in order to test the predictions of the Persson-Volokitin scattering

"2’3 The choice of the CO/Pt(111) system was made with several important

model.
considerations in mind. First, CO adsorption on Pt(1 1 1) has been extensively studied so
that phenomena such as site occupationf’s”6 coverage dependence of the C=O stretch
vibration,7 and ordering in the adlayer,8'9 among others, are well known. Second, the
scattering model is developed assuming the metal’s electrical conductivity is well-
described by the Drude model. Since Pt is a transition metal, it is not clear that the Drude
model should apply in this case. Thus, testing the scattering model’s predictions for Pt
should provide an interesting test of the extent of the model’s validity. Third,

temperature dependence enters the scattering model through the ratio of the mean free
path to the skin depth, [/8 , which will be discussed below. Pt varies over an interesting
range of [/8 at accessible temperatures.

The predictions of the scattering model rely on a knowledge of the conduction
electron density n of the substrate. Table 4-1 lists a number of material parameters for Pt

calculated from n, with n determined using two different methods. One set, labeled

72

73

“DC,” is obtained by assuming it corresponding to one electron/atom, and a value of the
bulk scattering time t B inferred from literature values of the DC resistivity.lo The other

set, labeled “optical,” is obtained from the Drude parameters that best fit optical

properties measured at room temperature.ll The value of r B at 90 K for the optical set is
estimated by assuming the ratio between 13(273 K) and 13(90 K) is the same as for the

DC parameters.lo Our major results do not depend strongly on which set of parameters is

used.

Table 4-1. Bulk material parameters for Pt. Frequencies in s'1 have been

divided by 27m to convert to cm".

 

 

03p to. 8 13' (273 K) [(273 K) [(90 K)
1cm") cm") (A) cm") (A) (A)
DC3 77,000 372 207 968 79 330
Optical” 41,500 133 380 558 91 380

 

aElectron density based on one electron/atom; scattering time based on DC resistivity.lo

bRoom temperature values from optical data,ll 90 K values calculated by assuming the
ratio between 13 (273 K) and 1:,(90 K) is the same as at DC.lo

The results discussed in this chapter were previously reported in a paper by Kuhl,

Lin, Chung, Wang, Luo, and Tobin.’2

74
4-1. Frequency Dependence

Frequency-dependence in the Persson-Volokitin scattering model takes the form

of
safe 4), 4-,
R a)

where a) is the frequency of interest, [ is the elastic mean free path, and 8 = c/a) p is the

' classical skin depth. 0) , is a characteristic rolloff frequency given by

 

0),: p, 4-2

where vF is the Fermi velocity, 0) pis the plasma frequency, and c is the speed of light.

Since [, 8 , and (al are each determined by n, a property of the substrate independent of
the adsorbate, the function j: which contains the entire frequency dependence, is a
universal function for a given substrate. The effect of the adsorbate enters through the

prefactor a which is given by

a = m , 4-3
4c cos0

75

where 0 is the angle of the incident IR radiation and p is the Fuchs specularity parameter
(i.e. the probability of specular scattering). In Equation 4-1 the universal function f is
expected to approach one at high frequencies, so, while f determines the shape, a
determines the magnitude of the absorption.

As shown in Chapter 1, the complete analytical expression for AR/ R is

 

 

 

 

( I
95_ .40 (tn/01)!)2 Re[g(a)/a)l,[/5)] 44
R - 1:2 4a) dq ’
(l—[nmpcose ImJ:>8(a)/0),,[/8,q))

where g(a)/0) l,[/8) and 8(0) /a),,[/8,q) are complex-valued functions. Near and below

001 nonlocal effects become important. Equation 4-4 has been successfully applied to a
number of experimental studies, including CO on Cu(100),1 O on Cu(100),13 and CO on
Cu(111).'4

Figure 4-1 shows my measurements of AR/ R vs. frequency, for 0 = 0.33 ML CO
on Pt(1 1 1) at room temperature. The coverage 0 = 0.33 ML was chosen because the
maximum AR/ R occurs near there (see Section 4-3) and thus AR/ R is insensitive to
coverage in that region. Additionally, larger changes are easier to measure. The data
points below 700 cm'1 represent averages of several measurements at nearby frequencies.

It is apparent that the data exhibit scatter larger than the (rather conservative)

experimental error bars. This scatter does not arise from instrumental error in individual

76

measurements, but rather from run-to-run variations in the actual reflectance change,
possibly due to impurity levels below the Auger sensitivity level. AS a result of this
scatter, and the small magnitude of the reflectance change, only a very rough comparison

with theory is possible.

 

0.000

I ' I ‘ I

   

0.33 ML CO/Pt(l 1 l)
315 K

I

-0.001

 

 

-0.002

 

 

l a l r I A l

0 500 1 000 1 500 2000 25 00 3000

-0.003

 

Frequency (cm‘l)

Figure 4-1. Fractional reflectance change AR/ R as a function of frequency, for 0.33
ML CO on Pt(1 11) at room temperature. The lines represent fits of the Persson-Volokitin
model1 to the data, with the specularity parameter p as the only adjustable parameter.
The dotted line represents the best fit using DC parameters, and the solid line the best fit
using optical parameters (see Table 4-1).

77

As seen in Chapter 1, the function g(a)/a) , , [/8) in Equation 4-4, given by

 

g = idyil - ijF’(y)+ a): 4mm [idyi-l— - L] FUJI, 4'5

. y y’ -w’-iwn y2 y4

contains information about an expected antiabsorption resonance feature associated with
a frustrated vibrational mode of CO on the Pt(111) surface at a frequency a) o and with a
damping coefficient 11 . Our measurements did not probe a sufficiently low frequency
regime to observe this feature. Furthermore, the elements of the Persson-Volokitin
scattering model which relate the broadband decrease in AR/ R to the damping of

. . . 15-17
hlndered modes remain controversral.

For calculations of just the broadband
reflectance change AR/ R, it is sufficient to use only the first term in Equation 4-5,
thereby dropping all information about the antiabsorption resonance. Note that doing this
ignores the claim by Persson that 1] cc (1 - p), which is embedded in a. Nevertheless, this
approach has been successful fitting the experimental AR/ R data in the study by Lin et
al. ’3 of O on Cu(100). Under this approach, once a)l is calculated from the conduction
electron density, Equation 4-4 contains only one adjustable parameter: the magnitude of
the asymptotic reflectance change a.

The dotted line in Figure 4-1 shows the best fit to the Persson-Volokitin model

achievable using the DC material parameters to determine a)l . The solid line shows the

best fit using the optical parameters. The data are consistent with the model, especially

78

when the optical values of the bulk parameters are used. Unfortunately, the uncertainties

both in the data and in the material parameters preclude any firm conclusions.

4-2. Temperature Dependence
Within the Persson-Volokitin scattering theory, the only parameter that is
expected to show a strong dependence on temperature is the mean-free-path [ = th B ,

which enters through the ratio [/ 8 in the luliversal function f. The scattering time ‘t 8

increases strongly with decreasing temperature, while VP and 8 are' essentially

temperature-independent. As can be seen from the DC values given in Table 4-1, for Pt [
changes by more than a factor of four going from 79 A at 273 K to 330 A at 90 K (our
measurement temperature, close to liquid nitrogen temperature). Since no optical
measurements leading to Drude parameters are available at temperatures comparable to

that of liquid nitrogen, we assumed the optical parameters scale like the DC parameters.

This gives a corresponding change in [ from 91 A at 273 K to 380 A at 90 K

Similar to Figure 5 of Reference 1, Figure 4-2 shows a number of curves
calculated from the Persson-Volokitin scattering model. The three quantities shown are
plotted against the ratio [/8, which is a scale related to the inverse of the temperature.
Each curve represents a different frequency regime, parametrized by the ratio 00/00l . The
first set of curves, Figure 4-2(a), illustrates the absolute change in reflectance AR. Note
that the vertical scale in Figure 4-2(a) is a negative scale, with zero at the top, because AR
is an absorption of the IR light. For [ / 8 greater than about two (i.e. toward low

temperatures), AR depends only very weakly on [/8, but for lower values (i.e. toward

79
higher temperatures) AR increases rapidly with increasing [ / 8 (decreasing temperature).
For Pt we estimate [ / 8 = 0.4 (0.24) at room temperature and 1.6 (1.0) at 90 K using DC
(optical) parameter values. We would therefore expect |AR| to increase by about a factor
of two when the sample is cooled from room temperature to 90 K, depending slightly on
the measurement frequency.

The experimentally measured quantity, however, is the fiactr’onal reflectance
change AR/ R, where R is the clean-surface reflectance, which also increases’ with
increasing [/ 8, as shown in Figure 4-2(b) (the details of this curve depend on 11; Figure 4-
2(b) is calculated using the DC value for Pt). This effect tends to offset or even overcome
the change in AR. In fact, as shown in Figure 4-2(c), the magnitude of the fractional
reflectance change AR/ R actually decreases with increasing [/ 8 for 00/0) 1 > 4.6. It is
worth noting that the importance of the temperature dependence of R to the ratio AR/ R
has not been previously recognized. In fact, the failure to account for this dependence

actually led Persson to an incorrect conclusion regarding the AR/ R for CO on Ni(100).l

80

Figure 4-2. Dependence of the reflectance on [/8 predicted for adsorption on
Pt by the Persson-Volokitin scattering theory,l for the indicated values of ctr/0)l .
The vertical scale for (a) and (c) is arbitrary. The small arrows indicate the room
temperature (RT) and 90 K values of [ / 8 for Pt. DC values of the bulk
parameters have been used (see Table 4-1). (a) Absolute reflectance change AR.
(b) Clean-surface reflectance R. (c) Fractional reflectance change AR/ R.

81

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. _ _ q a a q q a a _ _ _ a A a q a 4 q a q q a a T a
l l t J
\I .a 44 ) \l 4 a a.“ L2
(av . f O E .
.a 4? e o t
_ T L I o— A
.. _.
a *0. c a #0
_ c i. I 1 TI — lo.
M ... . e K x p. 2
9 . i a w A. m 0 .0
O —
~ . » r __ l
.. a4 + A. b a 40
.I. liii l l.Iu—I|‘Ii4 i ii i i i g iii .fi
6 O. l l 1 A P. 15
. .. .. .l
J r
a .a a 0
~. I m 4. A
. .. .
a 6.0 O O
L . r r ,./ i r .. lo.
.. .. I.
a a 0.9 if a 40
T e. , o E, l
I
R C D T . . l T ./‘
R I. R .
l r 1 r 5
+ a I O + I . Fir/ll 1+ + a I 9 o
. 1 .II 1 i illl liiiii lbril. 4’1, IL 0 W4 /
a a a O .a rut. k a a ,0
0 0 0. 0. o. 0.0.0. - 0 0. 0 o.
L 4 7 N I 47m L 4 7 m
> f p E p p _ p L E — p p - r p P P . h b b r W p P p o.
0 2. 4. 6. 8. 0 0. 9 8. 7. 6. 50 4. s a 6. 0.0
0 o D a mu 1.. I 0 0 0 o 00 o o 4. .... 9..

2.3 so e :5 ~52

82

In order to perform the first experimental study of the effect of changing [/ 8 for a
single adsorption system, I made a number of measurements at 90 K at frequencies of
2500 cm'1 and 2800 cm", corresponding to c0/0), values close to 7 (20) for DC (optical)
parameters. Over a range of CO coverages near 0.33 ML, the largest value of AR/R
observed was -0.20%i0.02% compared with -0.28%d:0.03% at room temperature, giving
a ratio of the 90 K to room temperature AR/ R of 0.73:0]. The theoretically predicted
ratio using DC (optical) parameter values is 0.85 (0.72). In View of the uncertainties in
both the calculation and the experiment, we regard this as reasonable agreement. It
would be of interest to perform similar measurements at 0) /0)l < 4.6 , but we have not yet
attempted them.

My measurements also permit an estimate of the ratio of the clean surface
reflectance at 90 K to that at room temperature; we find R(90 K)/R(315 K) = 1.3i0.02.
The predicted ratio, using DC (optical) parameter values, is 1.2 (1.5), so the agreement

between experiment and theory is again good.

4-3. Coverage Dependence

For a dilute, disordered overlayer, the scatterers will be independent and the
scattering of electrons will be incoherent — the scattering rates from different adsorbates
will simply add. The scattering probability can then be described by a scattering cross

section per adsorbate 2,

(l—p)= n02, 4-6

83

where It, is the surface density of adsorbates — the coverage. If the scattering cross
section per adsorbate is constant with respect to coverage, the resulting reflectance
change (or resistivity change) will vary linearly with adsorbate coverage. This can be
seen by substituting Equation 4-6 into the expression for the asymptotic value of AR/ R,

Equation 4-3, which gives

a- 3vF-2

— 4-7
4c cos0

 

a O

A linear dependence is in fact generally observed at low coveragem’m'25 For CO on
Cu(100) and Cu(111) the linear dependence persists up to saturation.22’23

For other systems, however, nonmonotonic dependences have been observed.
The first measurements of nonmonotonic coverage-dependent reflectance changes were
made by Riffe et al.2"26 They measured the change in attenuation of surface

electromagnetic waves (SEW), which is the inverse of measuring the change in IR

reflectance, but gives similar information. They studied N2, 02, CO, H2, and D2 on

W(100). The magnitude of the SEW attenuation coefficient 01 increased monotonically

with coverage for N2, while for both 02 and CO it reached a maximum and then dropped
at higher coverages. H2 and D2 showed a more complicated nonmonotonic dependence.

In a subsequent IR reflectance study of H on W(100), Riffe and Sievers correlated the

nonmonotonic coverage-dependence of the reflectance change with ordering of the

84

overlayer, at a frequency where the reflectance change was dominated by scattering rather
than electronic state effects.” This correlation is based on adlayer patterns observed in
separate low energy electron diffraction (LEED) experiments.28

Another example of a nonmonotonic coverage dependence in AR/ R was reported
by Lin, Tobin and Dumas for O on Cu(100).13 The magnitude of the reflectance change
increased monotonically up to 0.25 ML, but decreased by about 20% when the coverage
increased to 0.35 ML. While only four coverages were studied, the departure from the
low-coverage linear behavior at the highest coverage was well outside the experimental
uncertainty. Similar to Riffe et al., Lin and coworkers tentatively attributed their

observations of a nonmonotonic coverage-dependence to ordering of the overlayer. It is

well known from the literature that 0 forms a (2J2 x J2)R45° structure with a “missing

row” reconstruction (every fourth row is vacant) on Cu(100) at high coverages.”34

Using LEED, Wuttig, Franchy, and Ibach established29 that there is a first order phase
transition, at and above room temperature, from a low-coverage disordered phase into the
ordered phase at a critical coverage of BC = 0.34. Note that this is very close to the
coverage at which Lin et al. saw a drop in AR/R. Unfortunately, when they attempted
LEED observations themselves they were unable to correlate ordered LEED patterns with
the IR measurements due to temperature constraints.

The idea that ordering of the overlayer could be responsible for a nonmonotonic
coverage-dependence is based on a reduction of the extent of nonspecular scattering. A
fully ordered overlayer restores translational symmetry to the surface (the scattering from

different adsorbates is coherent) and diffuse scattering is no longer possible. Nonspecular

85

processes are allowed only if the scattering vector is a reciprocal lattice vector of the
overlayer. Since these vectors are usually different from those of the clean surface, and
since the scattering amplitudes are in any case different, the reflectance need not return to
its clean surface value.l4 Nevertheless, it can be expected to be different from the
reflectance from a disordered overlayer of the same surface density. Hirschmugl and
coworkers, on the other hand, claimM’ZZ’23 that ordering in the CO on Cu(100) and
Cu(l 1 1) systems has no effect on the coverage dependence of AR/ R. There are ordering
transitions in both of those systems and a linear coverage dependence was nevertheless
observed. Thus it was with the following questions in mind that we investigated the
coverage dependence of CO on Pt(111): ( 1) Is the coverage-dependence for CO on
Pt(1 11) monotonic or nonmonotonic? (2) If nonmonotonic, does ordering in the
overlayer explain the coverage-dependence? (3) If not ordering, what mechanism
accounts for a nonmonotonic dependence on coverage?

Figure 4-3(a) shows the fractional reflectance change AR/ R as a function of CO
coverage for CO on Pt(111) at ~ 315 K at both 2500 cm'I and 2800 cm". These
frequencies were chosen to be high enough that AR/ R is frequency-independent, and well

away from the C=O stretch vibrations at 1850 and 2100 cm". Within the Persson-
Volokitin scattering model AR/ R becomes frequency-independent for 0) >> I g‘ ,0) I. As
shown in Table 4-1, 1 3'1 corresponds to at most about 970 cm", while 0)l is

considerably lower.

86

 

 

 

 

 

 

 

 

0.0000 . . . . . . . r - r
.- ‘ j ‘1
-00005 - _f E _._ 2800 cm" : -
___=—__ ; —:-—- 2500 cm" i
r .L L ‘
. i I '
—0.0010 _ t——:—.-——+ "
l. I — i' ' ° -
r —s— r
% -00015 - _a—: T ‘
Dfi ‘” —!—— . T .
Q ' =- T 4 ‘
l I
00020 - )———I_.' ' __._t__.=‘:_:_1 ..
i. .. +r—-—-1: ' -1
i—O—_-___a__j _'__
l ' I“ '
- - r—z—l ~ ' -
0.0025 . i—T
. ' szT— .
1
00030 - (a) I r 1
1 L I r l A r l
r I r F fl 1 I r
0.00 r- ‘ -
A
004 - ' -
o
1- ‘ a
A a
T) -0.08 - ‘ a
:5
<1 ' e‘ a ‘
-0.12 . o ‘
T a
A
-0.16 - .A "
. (b) .
-0.20 1 1 ' ' ' ' ' ' ' '
0.0 0.1 0.2 0.3 0.4 0.5

CO coverage (ML)

Figure 4-3. (a) Fractional reflectance change AR/ R as a function of coverage, for CO
on Pt(1 l l) at 315 K. Data are shown for two 1R frequencies, 2500 and 2800 cm”. (b)
Work function change for CO on Pt(1 l 1), reproduced from References 35 (Norton, et al.,
triangles) and 36 (Ertl, et al., circles).

87

Strong nonmonotonic coverage-dependence was Observed for CO on Pt(1 11).
The magnitude of AR/ R increases with increasing CO coverage up to about 0 = 0.33 ML.
Above 0.33 ML the magnitude of AR/ R decreases with increasing coverage until
saturation is achieved at 0.5 ML.

It is interesting to note that 0.33 ML is the coverage at which the CO overlayer
forms a (J3 x J3)R30° ordered structure on Pt(1 l 1), as observed9 by LEED. Below

0.33 ML CO forms a disordered overlayer on Pt(1 11). At 0.5 ML the overlayer forms a
c(4 x 2) ordered structure.9 It is tempting, therefore, to attribute the decrease in [AR/RI

to ordering, following the suggestions made for H on W(100)27 and O on Cu(100).’3
Since long-range order is not achieved at room temperature, however, ordered LEED
patterns cannot be observed for CO above 300 K. It was necessary to cool the sample in
order to achieve observable LEED patterns.

In order to cool the sample, it was necessary to fill the coldfinger with liquid
nitrogen. During a complete IR measurement a sufficient quantity of liquid nitrogen
boiled away for the liquid level in the coldfinger to drop significantly. This allowed
some fraction of the coldfinger’s length to begin to warm up and lengthen. The
lengthening of the coldfmger caused minute changes in the sample’s position, which
rendered the IR measurement unusable. Thus, systematic coverage-dependent
measurements were not possible below room temperature, and we have no direct
evidence for adsorbate ordering. If ordering in the overlayer is the reason for the
nonmonotonic coverage dependence at room temperature, the effect must be assigned to

partial, short-range ordering, for which there is no direct experimental evidence.

88
An additional explanation for a nonmonotonic coverage-dependence could be that
the scattering cross section per adsorbate varies with coverage, either because different
binding sites are being occupied or due to coverage-dependent changes in the electronic
structure of the adsorbates.
It is well known that CO occupies both atop and bridging sites on Pt(1 1 1) and that

445.35
Moreover, the two

the relative populations of the two sites are coverage-dependent.
species have distinctly different electronic states and dipole moments,35 so it is entirely
plausible that they might have different scattering cross sections for conduction electrons.
Site exchange alone, however, cannot account for the drop in |AR/R| seen in Fig. 4-3(a).
The largest possible drop from this effect would occur if the layer were entirely atop-

bonded at 0.33 ML and half atop, half bridging at 0.5 ML, and if the bridging species had

zero cross section. In this case we would have:

1%(050 ML) _ 0.25 _

_ —0.76, 4-8
%(0.33 ML) 0.33 ( )

 

whereas the experimental value of the ratio is 0.43. Moreover, it is clear that bridge Sites
are occupied even at 0.33 ML and that the atop population drops very little, if at all,
between 0.33 and 0.50 ML.5'35

The final possibility that we consider is a coverage-dependent scattering cross
section caused by changes in the chemical bonding of the CO. Strong evidence for such

changes comes from measurementsms’36 of the work function change All), which show a

89

coverage dependence strikingly similar to that of AR/ R, as shown in Fig. 4-3(b) where
the work function data of References 35 and 36 are reproduced. (It should be noted,
however, that the coverage scale for Reference 35 was set by assuming that the minimum
value of Ad) occurs at 0.33 ML; in Reference 36 the coverages were determined directly
from temperature programmed desorption (TPD) and LEED.) The rapid increase in All)
between 0.33 ML and 0.4 ML or 0.5 ML is attributed"'35‘36 to depolarization of the
adsorbate bond, which could very well also be accompanied by a decrease in the
scattering cross section. We currently regard this as the most likely explanation for
Figure 4-3(a), but the ordering hypothesis cannot be ruled out.

The data in Fig. 4-3(a) can also be used to estimate a value for the cross section 2
for diffuse scattering of Pt conduction electrons from the CO adsorbates. Equation 4-6
gives a linear expression relating the asymptotic value of AR/ R to the coverage in the
form of the surface density of adsorbates n“. We measured a linear relationship for
coverages below 0.33 ML. The scattering cross section 2 can be extracted from the slope
of that line. Using DC (optical) values for the material parameters, we find 2 = 0.8 A2
(0.9 A2), for coverages below 0.33 ML. For comparison, Table 4-2 lists cross sections for
diffuse scattering of conduction electrons for a number of other adsorbate systems,
determined via measurements of the coverage-dependent adsorbate-induced change in
resistivity. Compared to other chernisorption systems, CO on Pt(111) is apparently a
rather weak scatterer. Also for comparison, the area per Pt atom on the surface is 6.6 A2

and the area per CO molecule at 0.5 ML coverage is 13.2 A2.

90

 

 

 

Table 4-2. Scattering cross section per adsorbate for several adsorption systems.
system 2 (A2) reference
chemisorption CO/Pt(1 l 1) 0.8
H/Ni(l l l) 8.2 3
CO/Ni(lll) 16.0 3
Nz/Ni(l 11) 4.8 3
CO/Cu(111) 5.8 3
O/Cu(111) 18.8 3
Ag/Ag(l 1 1) 14 37
physisorption HZO/Ag(1 l l) 0.06 2
CO/Ag(1 l l) 1.0 3
C2H4/Ag(l l l) 0.5 3
Xe/Ag(111) ~ 0.6 3
C6H6/Ag(l l l) 0.6 3
Can/Agfl 1 1) 0.6 3
CzHg/Ag(l 1 1) 0.1 3

 

91

44. Conclusions

A number of conclusions can be drawn from these measurements of the

adsorbate-induced change in the broadband IR reflectance AR/ R of Pt(1 11) as a

function of frequency, temperature, and CO coverage.

The Persson-Volokitin scattering model does a good job treating the temperature- and
coverage-dependence of CO on Pt(1 1 l), a transition metal surface. This is significant
because the the scattering model assumes that the electrical conductivity is described
by the Drude free electron model. The Drude model is not expected to apply to

transition metals.

It is important to consider the temperature dependence of the clean surface reflectance
R in addition to the temperature dependence of the absolute change in reflectance AR

when dealing with the fractional change AR/ R.

Further work must be done to understand the presence of a peak in the coverage
dependence in selected systems. For the CO/Pt(11 1) system the peak cannot be
attributed to changes in adsorption site, but could be attributable either to partial

ordering in the overlayer or to changes in CO'S electronic structure.

Broadband IR reflectance measurements can provide an accurate measure of the
scattering cross section of conduction electrons from adsorbates and can offer a new
probe of surface dynamics. Calculating and predicting the values and variations of
these cross sections presents a challenging and fertile field for theoretical

investigation.

92

References

'B.N.J. Persson and AI. Volokitin, Surf. Sci. 310 (1994) 314 .
2B.N.J. Persson, Chem. Phys. Lett. 197 (1992) 7.
3B.N.J. Persson, Phys. Rev. B 44 (1991) 3277.

4GS. Blackman, M.-L. Xu, D.F. Ogletree, M.A. Van Have, and GA. Somorjai,
Phys. Rev. Lett. 61 (1988) 2352.

5W.D. Mieher, L.J. Whitman, and W. HO, J. Chem. Phys. 91 (1988) 3228.

6IV. Nekrylova, C. French, A.N. Artsyukhovich, V.A. Ukraintsev, and 1. Harrison, Surf.
Sci. Lett. 295 (1993) L987.

7J.S. Luo, R.G. Tobin and D.K. Lambert, Chem. Phys. Lett. 204 (1993) 445.

8K. Horn and J. Pritchard, J. Phys. (Paris), Colloq. C4, Suppl. No. 10, 38 (1977)
164.

9H. Steininger, S. Lehwald and H. Ibach, Surf. Sci. 123 (1982) 264.

loJ. Bass and K.H. Fischer, in: Landolt-Bornstein Numerical Data and
Functional Relationships in Science and Technology, vol. 15a, ed. K.H.
Hellwege (Springer-Verlag, New York, 1982) 63.

llM. A. Ordal R.J. Bell, R.W. Alexander, Jr., L.L. Long, and MR. Querry, Appl. Opt. 24,
(1985) 4492.

12D. E. Kuhl, K. C. Lin, C. Chung, J. s. Luo, H. Wang, and R. G. Tobin, Chem.
Phys. 205 (1996) I, invited paper in special issue on “Surface Reaction
Dynamics.”

l3K.C. Lin, R.G. Tobin and P. Dumas, Phys. Rev. B 49 (1994) 17 273 (1994);
ibid. 50 (1994) 17760.

”or. Hirschmugl, G.P. Williams, B.N.J. Persson and A. I. Volokitin, Surf. Sci.
317 (1994) L1141.

93

15R.G. Tobin, Phys. Rev. B 48 (1993) 15 468
”B.N.J. Persson, Phys. Rev. B 48 (1993) 15 471

”TA. Germer, J .C. Stephenson, E.J. Heilweil and RR. Cavanagh, J. Chem. Phys. 101
(1994) 1704.

l8P. Wissmann in: Surface Physics, ed. G. H8hler, Springer Tracts in Modern Physics
Vol. 77 (Springer, New York, 1975).

'9D. Dayal, H.-U. Finzel and P. Wissmann in: Thin Metal Films and Gas Chemisorption,
ed. P. Wissmann (Elsevier, Amsterdam, 1987).

20D. Schumacher, Surface Scattering Experiments with Conduction Electrons, ed. G.
H8hler, Springer Tracts in Modern Physics Vol 128 (Springer, New York, 1993).

2'D.M. Riffe, L.M. Hanssen and A.J. Sievers, Surf. Sci. 176 ( 1986) 679.

22C]. Hirschmugl, G.P. Williams, F .M. Hoffmann and Y. J. Chabal, Phys. Rev.
Lett. 65 (1990) 480.

23C.J. Hirschmugl, Y.J. Chabal, F.M. Hoffmann and GP. Williams, J. Vac. Sci.
Technol. A 12 (1994) 2229.

24E. Borguet, J. Dvorak and H.L. Dai, SPIE Proceedings 2125 (1994) 12.

25M Hein and D. Schumacher, J. Phys. D: Appl. Phys. 28 (1995) 1937.

2(’D.M. Riffe, L.M. Hanssen and A.J. Sievers, Phys. Rev. B 34 ( 1986) 692.

27D.M. Riffe and A.J. Sievers, surf. Sci. 210 (1989) L215.

28D.A. King and G. Thomas, Surf. Sci. 92 (1980) 201.

29M. Wuttig, R. Franchy, and H. Ibach, Surf. Sci. 213 (1989) 103.

3"BC. Zeng, R.A. McFarlane, and K.A.R. Mitchell, Surf. Sci. 208 (1989) L7.
31l.K. Robinson, E. Vlieg, and s. Ferrer, Phys. Rev. B 42 (1990) 6954.

32MC. Asensio, M.J.Ashwin, A.L.D. Kilcoyne, D.P.Woodruff, A.W. Robinson,

Th. Lindner, J.S. Somers, D.E. Ricken, and AM. Bradshaw, Surf. Sci. 236
(1990)].

94

33Ch.Woll, R.J. Wilson, S. Chiang, H.C. Zeng, and K.A.R. Mitchell, Phys. Rev B
42 (1990) 11926.

3"R. Mayer, C.-S. Zhang, and K.G. Lynn, Phys. Rev. B 33 (1986) 8899.
35PR. Norton, J.W. Goodale and EB. Selkirk, Surf. Sci. 83 (1979) 189.
36G. Ertl, M. Neumann and KM. Streit, Surf. Sci. 64 (1977) 393.

”B.N.J. Persson, D. Schumacher, A. Otto, Chem. Phys, Lett. 178 (1991) 204.

Chapter 5
Changes in the DC Resistivity and the Infrared Reflectance

Induced by Oxygen and Formate Adsorption on Cu(100)

It has long been known that adsorbates increase the DC resistivity of thin metal
films."8 Attempts have been made to relate resistivity changes to measurements of

adsorbate-induced reflectance decreases in the visible light region?“

More recently,
sensitive measurements of the broadband absorption of IR light caused by adsorption on a
metal surface”’13 has renewed interest in the possibility of a relationship between these
two physical effects. Using a model based on conduction electron scattering, Persson,
building on earlier work, developed a linear relationship14 between the adsorbate-induced
resistivity change Ap and the adsorbate-induced broadband IR reflectance change AR/ R.

As shown in Chapter 1, following the example of Lin et al.,15 the expression

relating Ap to AR/ R is given by

 

AR_ 4ne2 )
_R—_ (mccos0 [tAp] 5-1

where n is the conduction electron density, e is the electronic charge, m is the effective
electron mass, c is the speed of light, 0 is the incidence angle of the IR radiation, and t is

the thickness of the thin film sample. The only quantity in Equation 5-1 that refers to a

95

96

thin film is tAp; the rest of the terms refer to bulk properties. This is not an
inconsistency, however, since the Persson scattering model is a slab model in which the
near surface region of a bulk sample can be mapped to a thin film. The quantity tAp has
been shown to be independent of film thickness.‘5

In this chapter I present experiments that were performed to test the linear
relationship predicted in Equation 5-1. 0 adsorption on a Cu(100) film will be discussed
in Section 5-1. Formate adsorption on the same Cu(100) film will be discussed in

Section 5-2.

5-1. 0 Adsorption on a Cu(100) Thin Film

Simultaneous measurements of the resistivity change and reflectance change were
performed while dosing in situ grown epitaxial Cu(100) films with 02 and, subsequently,
formic acid in order to test the linear relationship between AR/ R and tAp . The formic
acid dose will be discussed in the next section. The Cu(100) films were grown on etched
Si(100) substrates as described in Chapter 2, Section 2-4, and generally conform to the
characterization described therein. With the help of E.T. Krastev, a number of runs have
been performed which have established a high degree of reproducibility. For the
purposes of this discussion I will restrict myself to the single run shown in Figure 5-1.

The Cu(100) film studied in the time scan shown in Figure 5-1 was grown at
~ 40°C, 2 A/s, and a background pressure during growth of ~ 4.8 x 10''0 Torr. Its
dimensions were 1.000 cm x 1.880 cm x 507 A. The resistance, measured ex situ, was

3.26 (2, which leads to a resistivity of p = 8.79uQ-cm. The value for bulk Cu is

97

p = 1.70uQ-cm .16 The combined resistivity and reflectance measurement, described in
Chapter 2, Section 2-5, was performed with the spectrometer set at 2800 cm". The initial
UHV background pressure was 1.8 x 10'lo Torr. After about 200 s of the resistivity and
reflectance time scans had elapsed, the sample was dosed with 50 L 02 by backfilling the
UHV chamber (note that the closer was not employed) to 5.0 x 10'7 Torr for 100 3. At
about the 500 3 point, the sample was exposed to 50 L of formic acid vapor, using the
same dose of 5.0 x 10'7 Torr for 100 5. At the end of an 02 or formic acid dose at a
dosing pressure of 5 x 10'7 Torr, the UHV chamber typically reaches the low 10'9 Torr
region in under 10 seconds. After that, it requires on the order of another minute to
achieve a pressure in the mid 10'‘0 Torr region. Figure 5-1 shows the “raw” data.

During the time scans shown in Figure 5-1 the coldfinger was full of liquid
nitrogen and the sample was held at 28°C via the tungsten filament heater. Just before
the experiment began, the coldfmger was refilled. This resulted in a damped vibration of
the coldfinger which in turn affected the sample’s position, which showed up as erratic
drift in the early part of the baseline of the IR time scan. Consequently, the first 24 s of
data have been eliminated from both the IR and resistance time scans. A remnant of the
drifi can be seen in the data prior to about 150 s in Figure 5-1(a), after which the drift

throughout the rest of the time scan is essentially linear.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

98
0.02220 ' T ‘ T V I V r ' T r ' I -
. ! = L (a) .
0.02215 ~ M . : .
A, - L ‘ ‘,__.' dosing . ”l dosing ‘
5; 1501.02 ' :soLrormic
‘5 0.02210 ' l ‘ - 3CId "‘
'2' : .
2 . t
.5 0.02205 - I i
s ; I
3 . 5
3; 0.02200 A . -
“no
2 . Cu(100) film , I
02 9 2800 cm‘1 I
0. 1 5 l- 313 K A
i 4
0.02190 4 . 4 . A 1 L
0.00297 A A A A A . A A A
b i (b) 1
0.00294 - _ g l .
r Cu(100)f11m , g *
313 x 4
0.00291 - ' +
E * ‘
3 ». i , .
3 . j t ,
3 0.00288 » i . g .
g . F = f ‘
> » g i __ l ‘
0.00285 - ‘ 3 ‘ ~
t f i
’ dosing ? . dosing
0.00282 ~ 1 l 50 L o, I 50 L formic -
‘ ' acid
P ' i 4
0.00279 4 . 11 . 1 . L . 1] . ll . l i
o 100 200 300 400 500 600 700 800

time (s)

Figure 5-1. Time scans of (a) the reflected IR intensity (i.e. voltage output from a lock-
in amplifier) at 2800 cm'1 and (b) the voltage drop across the sample (i.e. . voltage output
from a lock-in amplifier) with an applied current. Both 50 L doses involved background
pressures of 5 x 10'7 Torr for 100 s.

99

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.002 A . . , . , . , - f , . C
’ . i (a) .
0.000 - - -- .................. .................. _
-0.002 .
. ‘ dosing l
g 0‘0“ l' ’3 SOLforrnic 1
§ acid ‘
-0.006 i .
: Cu(100) film ; l 1
2800 car! - .
-0.008 r 313 K | .
i i
-0.010 A . - A i A
0.06 - . . .1 A .1 . r . ,, - ,1 r , .
' ' f l (b) .
0.05 - 5 .
Cu(100) film i
313 K '
0.04 A ' .
g 0.03 - -
0.02 - . .
0'01 ' . dosing ‘ dosing J
_ SOLO; SOLformic
acid
0.00 - .
l A ll . l L

 

 

 

 

O l 00 200 300 400 500 600 700 800

time (8)

Figure 5-2. Time scans of (a) the broadband reflectance change AR/ R at 2800 cm'l and
(b) the fiactional change in resistivity Ap/ p. Each curve has a linear drift removed.
Both 50 L doses involved background pressures of 5 x 10'7 Torr for 100 s.

100

It may be noted that the vertical lines representing the beginning and ending of the
dosing do not precisely line up between Figures 5-l(a) and 5-l(b). This is due to the fact
the two data sets were collected on separate computers; the time axes begin close to but
not exactly at the same point.

The raw IR intensity and voltage drop data of Figure 5-1 is represented in Figure
5-2 as the broadband reflectance change AR/ R and the fractional change in resistivity
Ap / p respectively. A linear drift has been removed fi'om each of the curves shown in
Figure 5-2. For the IR reflectance data the slope of the curve was determined from the
data between 120 s and 202 s because this is where the baseline appears to have
stabilized, as can be observed by the consistency between the slope before and after
dosing in Figure 5-l(a). The dashed line at AR/ R = 0 in Figure 5-2(a) is provided as an
aid to the eye. For the resistivity data, the slope was determined from the data between
26 s and 202 s. The resistivity data had an extremely stable baseline, so that removing
the linear drifi had essentially a negligible effect.

It can be immediately observed from Figure 5-2 that there is a strong
correspondence between AR/ R and Ap / p . At the beginning of the dose, the broadband

IR reflectance AR/ R drops by about 0.9% while the resistivity increases by about 5.5%.
It is interesting to note that both Ap/ p and [AR/RI appear to decrease somewhat after

achieving a maximum. The value of Ap / p just before the formic acid dose is about

5.2% while AR/ R is about -0.8%. This could be related to a nonmonotonic dependence

on coverage for AR/ R that was reported by Lin et a1. 17 for O on Cu(100) (see Chapter 4,

101
Section 4-3 for a discussion of nonmonotonic dependences of AR/ R on coverage). It is
conceivable that both AR/ R and Ap/p have a nonmonotonic dependence on coverage
here, although the fact that the decrease appears to continue after the dose when the
coverage is not changing could point to some other explanation.
At room temperature 02 is known to dissociate on the Cu surface and leave an

17-25

overlayer of atomic oxygen. One might suspect that the decrease in Ap/p and

[AR/RI following the dose results from an oversaturation coverage of 0 during the dose

followed by O leaving the surface as the pressure in the chamber drops. This is not a
likely explanation for two reasons. First, the decrease clearly begins before the dosing is

completed. Second, it is known”’30

that O has a very small sticking coefficient on Cu
which gets smaller at higher coverages. We have no independent measure of coverage at
this time, but it is likely that the coverage is well below saturation, which is 0.5 ML.20

Moreover, the O that does stick does not desorb thermally at room temperature.“30

The absence of ordered patterns observable by LEED at room temperaturen’zo'25
indicates that the decrease probably cannot be attributed to ordering in the overlayer
either. Additional possible explanations include relaxation in the overlayer or adsorbate-
induced surface reconstruction.

A straightforward means of testing the linearity predicted by Equation 5-1 is to
plot AR/ R from Figure 5-2(a) versus tAp extracted from Figure 5-2(b). Note that the
slight difference in the origin of the time axes becomes significant here, however, the

sharp onset of change observed in both Figures 5-2(a) and 5-2(b) allows for a convenient

reference point. Figure 5-3 is a plot of AR/ R versus tAp for this experiment.

102

 

 

 

 

 

 

 

 

0.001
’ .\
-0 001 L .\\\
\.
\.
.\\
0003 ~ \
\\
-0.005 r - \
g -0.007 A \ \
oft... o
-0.007 ~-0.008» ._';t;
0.009 l "I... N};
-0.009 - ' ° -°'_~.
‘ 0.010 5
2.30 2.36 2.42 2.48
_0011 1 1..-2 grAihiL - : 1
-0.1 0.2 0.5 0.8 1.1 1.4 1.7 2.0 2.3 2.6
tAp (10*5 unocmz)

Figure 5-3. AR/ R versus tAp for 02 adsorption on a Cu(100) film. The inset is an

enlargement of the lower right hand comer. The solid line is a linear fit to the data, which
are taken from about the 200 s to 500 5 region of Figures 5-2(a) and 5-2(b).

Figure 5-3 clearly exhibits qualitatively the linear relationship between AR/ R and
tAp predicted by Equation 5-1. The data plotted in Figure 5-3 come from the 206 s to

500 8 region of Figure 5-2(a) and the 208 s to 502 5 region of Figure 5-2(b). Note that the
02 dose ends at about 300 seconds, so that ~ 2/3 of the data plotted in Figure 5-3 comes

from the interval in which the UHV chamber was pumping residual 02. The data appears

103

to move back up the curve a little in this region and then bunch up vertically near
p = 2.3 x 10“ uQ-cm2 ; i.e. AR/ R drifts slightly more than tAp .
Equation 5-1 also makes a quantitative prediction about the relationship between

AR/ R and tAp . The slope of the line should be given by

 

d(AR/ R) _ 4ne2
d (tAp) — -( me 0050] ° 52

The angle of incidence 0 of the IR radiation is 86° (see Chapter 2, Section 2-2). The
conduction electron density n can be calculated by assuming one conduction electron per

26'” which gives a value for Cu of n = 8.5 x 1022 cm'3. Using these values, along

atom,
with literature values‘6 of the constants e, m, and c, a predicted value of the slope can be

calculated. Equation 5-3 gives the predicted slope, and Equation 5-4 gives the slope

determined by a linear fit of the data.

 

(AR/R] = —6.1><10'°(Q-cmz)-l 5-3
tAp calc

[AW—R) =-3.7x1093:.0.1x109(o.cm2)" 5-4
tAp exp

104

There is a 155% discrepancy between the predicted and measured values. This is
similar to the findings of Hein and Schumacher,28 who reported a 72% discrepancy
between the predicted value and their experimental results for O on Cu. It is a bit
surprising that our discrepancy is even larger than theirs because their films very likely
had poor surface conditions due to the fact that they were grown on spectroscopic glass
slides and were therefore not epitaxial. The Persson scattering model is developed for a
well ordered, smooth surface.

As discussed in Chapter 1, Lin, Tobin, Dumas, Hirschmugl, and Williams”
compared synchrotron measurements of AR/ R for O on a Cu(100) bulk single crystal
with two different AR/ R predictions they calculated from two published tAp
measurements on thin films. They found discrepancies of between the calculated and
experimental values of 36% and 136%. The value that came closer to their measured
AR/R was not far outside of the experimental error, and thus they considered the
comparison to be encouraging for several reasons: (1) they compared a bulk single

crystal AR/ R to a thin film tAp , (2) the thin films studied by others were poorly
characterized, and (3) the wide variation in published tAp values indicated a poor degree
of reproducibility of the film surfaces. Nevertheless, their large discrepancies are similar
to that found in the careful simultaneous measurement of AR/ R and tAp presented here. ~

Here, in the first study of the linear relationship between AR/ R and tAp using
simultaneous reflectance and resistivity measurements on a single, well characterized,

epitaxial thin film sample, I find strong evidence in support of the linear relationship

105

predicted by the Persson scattering model. There is, however, a large discrepancy

between the experimentally measured slope and that predicted by the model.

5-2. Formate Adsorption on an O-Predosed Cu(100) Thin Film

The formate species consists of a hydrogen, a carbon, and two oxygen atoms '
(HCOO). It can be obtained through the direct decomposition of formic acid on Cu(100),
but it has been shown that predosing the surface with O greatly enhances the sticking

29 We chose to examine formate adsorption on Cu(100) due to

coefficient of formate.
earlier indications that its broadband IR AR/ R behavior departs from that expected by the
Persson scattering model.

Lin, Tobin, and Dumas performed a reflection-absorption infrared spectroscopy
(RAIRS) study of formate on a Cu(100) bulk single crystal in which they looked30 for a
broadband AR/ R. They distinctly measured a broadband absorption when they predosed
the Cu(100) surface with 0. Upon subsequent exposure to formic acid, however, the
broadband AR/R returned to zero within the ~ 0.2% sensitivity of the synchrotron-based
experiment. It was felt that a simultaneous measurement of tAp would provide useful
information about the effect of formate on the scattering of conduction electrons in the
metal.

Figures 5-1 shows reflectance and resistivity time scans of formate adsorption on
a Cu(100) thin film surface predosed with O. The size of the formic acid dose was 50 L,

consisting of 5 x 10'7 Torr for 100 s at about the 500 5 point in the time scans. Figure 5-2

presents the data of Figure 5-1 in terms of AR/ R and Ap/ p , with linear drifi removed. It

106

can be clearly seen in Figure 5-2(a) that the broadband IR reflectance returns toward its
clean surface value. There does appear to be a slight difference from AR/ R = 0, below
the sensitivity level of the synchrotron-based RAIRS apparatus. The plot of Ap/ p
reveals that the resistivity also returns in the direction of its clean surface value, although
there is a distinct offset from Ap/ p = 0.

The Persson scattering model deals with changes from a clean metal surface to a
surface with an adsorbate, whereas our formate experiment measures the change from an
O predosed surface. The reaction of the formic acid with the O replaces O adsorbates
with formate adsorbates. The actual chemistry is not well understood. Nevertheless, a
qualitative test of the linear relationship between AR/ R and tAp for formate adsorption
on Cu(100) from an O predosed surface is interesting. Figure 5-4 is a plot of AR/ R
versus tAp for the formic acid dosing region of the data shown in Figure 5-2. The data
in Figure 5-4 comes from the 500 s to 598 3 part of Figure 5-2(a), and the 504 s to 602 8
part of Figure 5-2(b). These data regions were chosen to provide the most nearly linear
plot.

The plot in Figure 5-4 does not appear particularly linear, although it’s difficult to
make a careful determination due to the fact that both AR/ R and tAp reach plateaus so
quickly. The bulk of the data falls into the upper lefi hand corner of Figure 5—4, which is
enlarged in the inset, where neither AR/ R nor tAp undergo significant change. A
greater number of data points in the region of rapid change would allow a better

synchronization of the AR/R and tAp time axes.

107

0.000 ... . ,.,

 

I

'fifir

-0.002

 

 

 

 

 

 

 

_0004 p 0.0000 . . . d
g 0.00041 1

00008: ; .. '1 O
.0006 ~ 00012: ’° '. ,. .’ .

-0.0016:

.0.0020- ‘3
'0’008 " 1.06 A 1.08 A 1.10 ‘ I

1.0 1.2 A 1i4 A 11:6 4 ITS 2.0 2:2 A 2.4

tAp (106 flocmz)
Figure 5-4. AR/R versus tAp for formate adsorbtion on a Cu(100) film. The inset is an

enlargement of the upper left hand comer. The data are taken from about the 500 s to
620 8 region of Figures 5-2(a) and 5-2(b).

As the formate replaces the O on the Cu(100) surface, the AR/ R and tAp
induced by the 0 both return in the direction of their clean-surface values. This could be
due to the removal of the O scatterers, provided the formate does not significantly induce
conduction electron scattering. The fact that AR/ R appears to return very close to zero

suggests that the formate does not scatter conduction electrons, but the fact that tAp does

108

not return nearly as close to zero suggests that formate does scatter conduction electrons.
This discrepency, along with the apparent lack of a linear relationship between AR/ R and
tAp , could indicate that the scattering model does not apply to formate adsorption on

Cu(l 00).

109

References

‘K. Fuchs, Proc. Cambridge Phil. Soc. 34 (1938) 100.

2T. Holstein, Phys. Rev. 88 (1952) 1427.

3G.E.H. Reuter and EH. Sondheimer, Proc. Roy. Soc. London, Ser. A 195 (1948) 336.
411.13. Dingle, Physica 19 (1953) 311.

511.13. Dingle, Physica 19 (1953) 729.

6P. Wissmann in: Surface Physics, ed. G. Htihler, Springer Tracts in Modern Physics Vol.
77 (Springer, New York, 1975).

7D. Dayal, H.-U. Finzel and P. Wissmann in: Thin Metal Films and Gas Chemisorption,
ed. P. Wissmann (Elsevier, Amsterdam, 1987).

8D. Schumacher, Surface Scattering Experiments with Conduction Electrons, ed. G.
Htihler, Springer Tracts in Modern Physics Vol 128 (Springer, New York, 1993).

9M. Watanabe and A. Hiratuka, Surf. Sci. 86 (1979) 398.

“’11. Merkt and P. Wissmann, z. Phys. Chem. Neue Folge 135 (1983) 227.
”M. Watanabe and P. Wissmann, Surf. Sci. 138 (1984) 95.

12J.E. Reutt, Y.J. Chabal and SB. Christman, Phys. Rev. B 38 (1988) 3112.

13OJ. Hirschmugl, G.P. Williams, F.M. Hoffmann and Y. J. Chabal, Phys. Rev. Lett. 65
(1990) 480.

”B.N.J. Persson, Phys. Rev. B 44 (1991) 3277.

15KC. Lin, R.G. Tobin, P. Dumas, C.J. Hirschmugl and GP. Williams, Phys. Rev. B 48
(1993) 2791.

”Charles Kittel, Introduction to Solid State Physics (John Wiley and Sons, Inc., New
York, 1986) 144.

110

”RC. Lin, R.G. Tobin, and P.Dumas, Phys. Rev. B 49 (1994) 17273.

”A. Spitzer and H. Luth, Surf. Sci. 118 (1981) 121.

'9M.H. Mohamed and LL. Kesmodel, Surf. Sci. Lett. 185 (1987) L467.

20M. Wuttig, R. Franchy, and H. Ibach, Surf. Sci 213 (1989) 103.

2|H.C. Zeng, R.A. McFarlane, and K.A.R. Mitchell, Surf. Sci. Lett. 208 (1989) L7.
221K. Robinson, E. Vlieg, and s. Ferrer, Phys. Rev. B 42 (1990) 6954.

23MC. Asensio, M.J. Ashwin, A.L.D. Kilcoyne, D.P. Woodruff, A.W. Robinson, Th.
Lindner, J .S. Somers, D.E. Ricken, and A.M. Bradshaw, Surf. Sci. 236 (1990) 1.

24Ch. Woll, R.J. Wilson, 3. Chiang, H.C. Zeng, and K.A.R. Mitchell, Phys. Rev B 42
(1990)]1926.

25R. Mayer, C.-S. Zhang, and K.G. Lynn, Phys. Rev B 33 (1986) 8899.

26NW. Ashcroft and ND. Merrnin, Solid State Physics (Holt, Rinehart, and Winston,
New York, 1976) 5.

”WA. Reed and E. Fawcett, J. Appl. Phys. 35 (1964) 754.

28M. Hein and D. Schumacher, J. Phys. D: Appl. Phys. 28 (1995) 1937.

29M. Bowker and R.J. Madix, Surf. Sci. 102 (1981) 542.

3"RC. Lin, R.G. Tobin, P. Dumas, J. Vac. Sci. Technol. A 13 (1995) 1579.

Chapter 6

Conclusions

I have employed a novel form of reflection-absorption infrared spectroscopy,
utilizing a unique liquid nitrogen-cooled grating spectrometer, to probe the effects of gas
adsorption on the electron dynamics in the near-surface region of metals. Information
gained from this study contributes to our fundamental understanding of surface
properties. The results of this research provide support for a model due to Persson and
Volokitinl'4 that relates a number of different physical phenomena, including a
broadband absorption of infrared light and an increase in resistivity, to the scattering of
conduction electrons by adsorbates. This work also provides impetus for further research
regarding the scattering model.

Due to the small size of the effect, the difficulty of maintaining a sufficiently
stable baseline, and the low frequency range where nonlocal effects are significant, few
measurements of the broadband reflectance change AR/ R have been reported?‘4 Here I
present measurements of the adsorbate-induced change in the broadband IR reflectance
AR/ R of Pt(1 l l) as a function of frequency, temperature, and CO coverage. The results
are compared with the scattering model."4

The frequency dependence of AR/ R for CO on Pt(1 11) is at least qualitatively

consistent with the theoretical prediction, but experimental nonreproducibility and

uncertainty in the material parameters prevent a rigorous test.

111

112

The magnitude of AR/ R at 2500-2800 cm'1 is smaller at 90 K than at room
temperature (~315 K) by about 30%. This result appears surprising, since the scattering
model predicts that AR should be larger at the lower temperature.4 The clean surface
reflectance R, however, is also larger at low temperature, with the result that a decrease in
AR/ R at these frequencies is consistent with the theory. This demonstrates the previously
unrealized fact that it is important to consider the temperature dependence of the clean
surface reflectance R in addition to the temperature dependence of the absolute change in
reflectance AR when dealing with the fractional change AR/ R.

The magnitude of the reflectance change increases with increasing coverage for
low coverages, peaks at a coverage of about 0.33 ML, and decreases toward saturation.
The presence of a peak in the coverage dependence cannot be attributed to changes in
adsorption site, but could be attributable either to partial ordering in the overlayer or to
changes in CO's electronic structure. The slope of the curve at low coverage corresponds
to a scattering cross section per CO on the order of 1 A2. The determination of this
quantity is significant because it shows that broadband reflectance measurements can be
used as a new probe to determine physically interesting quantities for systems where the
scattering model is valid. Calculating and predicting the values and variations of these
quantities, such as the cross section, presents a challenging and fertile field for theoretical
investigation.

It is significant that the Persson-Volokitin scattering model is successfill in
treating the temperature- and coverage-dependence of CO on Pt(1 1 l), a transition metal

surface. The scattering model assumes that the electrical conductivity is described by the

113

Drude free electron model. Since the Drude model is not expected to apply to transition
metals, the success of the scattering model with regard to the CO/Pt(1 l 1) system provides

valuable information regarding the extant of its applicability.

Another prediction of the scattering model is that there exists a linear relationshipl
between AR/ R and tAp , the adsorbate-induced change in resistivity (times the sample’s
thickness). Previous research into this relationship was done either on dissimilar samples
at different times” or simultaneously on polycrystalline films.16 I present here the first
study of the predicted linear relationship using simultaneous reflectance and resistivity
measurements on a single, well characterized, epitaxial thin film sample.

The adsorption of O and formate on a Cu(100) thin film grown epitaxially by
UHV evaporation onto an etched Si(100) substrate was studied. Strong evidence in
support of the linear relationship predicted by the Persson scattering model is seen for 0
adsorption on Cu(100). There is, however, a large discrepancy between the
experimentally measured slope and that predicted by the model. Linearity was not
confirmed for formate adsorption on an O predosed Cu(100) film, but cannot be ruled
out. As the formate replaces the O on the Cu(100) surface, AR/ R appears to return very
close to zero suggesting that the formate does not scatter conduction electrons. tAp ,
however, does not return nearly as close to zero suggesting that formate does scatter
conduction electrons. This discrepancy, along with the possible lack of a linear
relationship between AR/ R and tAp for formate, could indicate that the scattering model

does not apply to formate adsorption on Cu(100).

114

Taken together, these measurements provide support for the validity of the
scattering model in explaining the reflectance changes induced by adsorption on metals,
and indications of the limits of the model’s validity. In cases where scattering is the
dominant effect of adsorbates on the electron dynamics of a metal, this work
demonstrates that infrared reflectance measurements can offer a new probe of surface
dynamics leading to accurate quantitative information.

Interesting future work could include simultaneous reflectance and resistivity
measurements of CO adsorption on Ni(100). It has been shown that an epitaxial Ni(100)

”1

film can be grown using a Cu(100) film as a “seed layer, 7 thus it is possible to produce
a well characterized thin Ni(100) film appropriate for this type of study. The CO on
Ni(100) system is interesting because an earlier test of the scattering model’s predicted
quantitative relationship between AR/ R and tAp using AR/ R measured on a bulk single
crystal and tAp measured on a thin film showed a discrepancy much larger than
predicted8 — far larger than seen here for O on Cu(100). A second possibility is to
further investigate the issue of coverage-dependent AR/ R. The adsorption of O on
Ni(100) is a strong candidate system with which to test the possibility of an ordering
effect. The O/Ni(100) system is known to have a single adsorption site, to adsorb in a
disordered state at low coverage, and to undergo two ordering transitions observable by
LEED at room temperature as the coverage is increased.18 Through these and other

experiments, it should be possible to extend our understanding of the affect of adsorbates

on the electron dynamics of metals.

115

References

1B.N.J. Persson, Phys. Rev. B 44 (1991) 3277.
2B.N.J. Persson, Chem. Phys. Lett. 197 (1992) 7.

3B.N.J. Persson and A.I. Volokitin, J. Electron Spectrosc. Relat. Phenom. 64/65
(1993) 23.

4B.N.J. Persson and A.I. Volokitin, Surf. Sci. 310 (1994) 314.

5D. E. Kuhl, K. C. Lin, C. Chung, J. s. Luo, H. Wang, and R. G. Tobin, Chem.
Phys. 205 (1996) I, invited paper in special issue on “Surface Reaction
Dynamics.”

61B. Reutt, Y.J. Chabal and SB. Christrnan, Phys. Rev. B 38 (1988) 3112.

7C.J. Hirschmugl, G.P. Williams, F.M. Hoffmann and Y. J. Chabal, Phys. Rev. Lett. 65
(1990) 480.

8RC. Lin, R.G. Tobin, P. Dumas, C.J. Hirschmugl and GP. Williams, Phys. Rev. B 48
(l993)2791.

9C.J. Hirschmugl, Y.J. Chabal, F.M. Hoffmann and GP. Williams, J. Vac. Sci. Technol.
A 12 (1994) 2229.

loK.C. Lin, R.G. Tobin and P. Dumas, Phys. Rev. B 49 (1994) 17 273 (1994); ibid. 50
(1994) 17760.

”CJ. Hirschmugl, G.P. Williams, B.N.J. Persson and A. I. Volokitin, Surf. Sci. 317
(1994) L1141.

12E. Borguet, J. Dvorak and H.L. Dai, SPIE Proceedings 2125 (1994) 12.
'3 M. Hein and D. Schumacher, J. Phys. D: Appl. Phys. 28 (1995) 1937.
l“CLA. Lamont, B.N.J. Persson and GP. Williams, Chem. Phys. Lett. 243 (1995) 429.

lsK.C. Lin, R.G. Tobin, P. Dumas, C.J. Hirschmugl and GP. Williams, Phys. Rev. B 48
(1993) 2791.

116

16M. Hein and D. Schumacher, J. Phys. D: Appl. Phys. 28 (1995) 1937.
l7Chin-An Chang, J. Vac. Sci. Technol. A 8 (1990) 3779.

18U. Starke, P.L. de Andres, D.K. Saldin, K. Heinz, and J.B. Pendry, Phys. Rev. B 38
(1988) 12277.