PLACE II RETURN BOX To roman this chockwtflom your "cord.
TO AVOID FINESMumonorbdon ddocluo.

I DATE DUE DATE DUE DATE DUE
\ ” ‘ W i #J i ’—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

fiflfl

MSU I. An mm W“ Oppomnny lm

 

 

 

 

ADSORBATE-INDUCED BROADBAND
INFRARED REFLECTANCE CHANGES
ON METAL SURFACES

By
Keng-Ching Lin

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

1995

Hi

br

UR

Sui

re?

11“

ABSTRACT

ADSORBATE-INDUCED INFRARED REFLECTANCE
CHANGES ON METAL SURFACES

By

Keng-Ching Lin

This research focuses on the broadband infrared reflectance changes of
metal surfaces upon adsorption of foreign species. The 1-2% observed
adsorbate-induced reflectance decrease is much larger than the 0.001%
reflectance increase expected from the polarizability of the adsorbate itself.
Therefore, this reflectance drop must result from changes in the dielectric
properties of the substrate. This effect has received little attention since the
signal is difficult to measure reliably, and most surface IR spectroscopy studies
emphasize the vibrational modes of the adsorbate, which give rise to relative
narrow absorption bands. By distinguishing the physical origin of these
broadband changes, we can characterize the modification of electronic
transport in the substrate by the adsorbate. This knowledge is important in
understanding the fundamental physics of a wide range of technological

subjects, such as mechanisms of chemical sensing and transport in thin films.

I have done a series of experiments to investigate the broadband

reflectance changes for different adsorption systems using two reflection-

absorption infrared spectroscopy (RAIRS) systems, a synchrotron-based
system at Brookhaven National Laboratory and a novel cryogenic
spectrometer at Michigan State University. The experimental results provide
a test of a proposed model of conduction-electron scattering from disordered
adsorbates that incorporates nonlocal electrodynamics for the interactions of

probing electric fields with conduction electrons.

A direct relation between surface resistivity changes and reflectance
changes is derived and discussed by comparing the measured reflectance
changes and published resistivity changes on thin metallic films for CO on
Cu, O on Cu, and CO on Ni. The results are consistent for CO and O on Cu. A
discrepancy found for CO on Ni stimulates further discussion of the theory

and experimental tests.

For O on Cu(100) the reflectance change at different coverages were
measured as a function of infrared frequency and fitted with a theoretical
function. At coverages below 0.25 ML, the reflectance changes are well fitted
by a universal form with a characteristic rolloff frequency that determines the
transition between local and nonlocal optics. The coverage-dependence of the
reflectance change is different from theoretical predictions. Possible
explanations are investigated with reference to the adsorption systems of O

on Cu(100), CO on Pt(111) and formate on oxygen-predosed Cu(100).

To my Parents

iv

 

em

ACKNOWLEDGMENTS

The completion of this work is under numerous supports. I would
like to express my deepest appreciation to my advisor Dr. Roger G. Tobin.
Roger’ s intellect and insight of experimental physics provide me the best
training I ever expected. His understanding of all the difficulties I

encountered during these years encourages me to accomplish my best.

Dr. Gwyn P. Williams designed the infrared beamline and Dr. Carol J.
Hirsucmugl built up the surface science system at the National Synchrotron
Light Source. Gwyn and Carol made this research possible by their primary
works at the beamline. The collaborations with Dr. Fritz Hoffmann and Dr.
Paul Dumas have been invaluable. Fritz and Paul’s scientific goals and

personalities taught me a lot about being a researcher.

Iwould like to thank all the members in Roger’ 3 group, Dr. Chilhee
Chang, Dr. ].S. Luo, Dr. Hong Wang, Dennis Kuhl, E.T. Krastev, and Larry
Voice. Chilhee handed me the rein of the system. Sam has always been a
good friend. The discussions with Sam and Wang are very useful. Dennis
came at the time when Ineeded a second hand to run the experiments and
succeeds in expanding the project. The discussion with Ati and Larry of

undergoing experiments is beneficial. I am also grateful to all my friends who

V

share my life in Michigan. They provide me joy, laughs and different

prospects of life.

My parents are my anchor for all these years. Being the youngest one
of three children in my family, I indulge the love from my two sisters. The
final thanks go to my husband, Woei-Wu Pai. His encouragement helps me

to finish this work and head for a more challenge stage of life.

vi

List of '
list of i
Chaple

Chapte

Chapte.

FJ
. ,._)

Ix) I J

[\J [\J .

I\)

Table of Contents

List of Tables
List of Figures
Chapters
Chapter 1 Introduction
1.1 Electron transport in a metal
1.2 Thin film resistivity studies
1.3 Surface RAIRS studies and surface effect on reflectance
1.4 Scattering theory of conduction electrons in adsorbed layer
1.5 Experimental designs
References
Chapter 2. Instrumental Introduction
2.1 Surface signal and light sources
2.2 Noise, detectors and relevant optics
2.3 Spectrometers
2.4 General surface science tools
2.5 Surface science branch at U4IR beamline, NSLS, BNL
2.6 RAIRS system at Physics Department, MSU
2.7 Summary and discussion

vii

vii

viii

13
19
26
28

31

31

#88

53
56
60

Chapter 3.

3.1
3.2
3.3
3.4

Chapter 4.

4.1
4.2
4.3
4.4
4.5
4.6
4.7

Chapter 5.

5.1
5.2
5.3
5.3.1
5.3.2

References

Adsorbate-Induced Changes in the Infrared Reflectance
and Resistivity of Metals

Experimental technique

Surface effect on reflectance

Comparison between reflectance and resistivity
Summary

References

Adsorbate—Induced Changes in the Broadband Re-
flectance of a Metal: Oxygen on Cu(100)

Experimental procedures

Data reduction

Experimental results

Frequency dependence

Coverage dependence

Dynamic dipole moment of oxygen on Cu(100)
Summary

References

Infrared Spectroscopy of Formate on Oxygen-predosed
Cu(100): Broadband Reflectance and Low-Frequency
Vibrations

Introduction

Experiments

Results and discussions
Broadband reflectance change

Copper-formate vibrations

viii

62

65

72

78

80

80

86
89
95
100
102
103

106

106
108

109
112

5.4 Summary 115

References 117

Chapter 6. Adsorbate-Induced Reflectance Change of a Metal: 120
CO on Pt

6.1 Introduction 120

6.2 Experimental 122

6.3 Results and discussion: Coverage-dependence 128

6.4 Summary 137

References 138

Chapter 7 Conclusion . 141

Appendix 144

References 145

ix

T.

Table 2.1

Table 3.1

Table 3.2

Table 4.1

Table 6.1

List of Tables

The performance numbers of two RAIRS systems at
NSLS and MSU.

Bulk material parameters for Cu and Ni.

Adsorbate-induced changes AR,,/Rp in the reflectance of
p-polarized light at the adsorbate coverage 11‘ as indicated,
measured experimentally and calculated from Eq. 3.8

using the indicated values of tAp for thin films of
thickness t, the electron density 1: given in Table 3.1, and

an angle of incidence 0=85°.

Best-fit parameter values and uncertainties obtained by
fitting Eq. 4.2 to the spectra in Figure 4.2, and
transforming the fitting parameters according to Eq. 4.3.

The weighted average of the col values is 3383234 cm'l.
Also listed are the values of a and a)1=co,,up/c found for

CO on the same Cu(100) crystal, and the value of col
predicted from a free electron model.

Bulk material parameters for Pt.

61

73

74

92

128

‘U
. THC
Wu .. .
Wsfiv Fur

Figure 1.1

Figure 1.2
Figure 1.3

Figure 1.4

Figure 1.5

Figure 1.6

Figure 1.7

Figure 2.1

Figure 2.2

List of Figures

CO on Cu(100) surface. Distinguishable background
absorption observed together with the dipole-selection
forbidden hindered rotation mode of the CO molecule at

288 cm'1 and the Cu-CO stretch at 345 cm’1 over the
frequency range 200-500 cm’l.

A schematic of resistivity measurement on a metal film.
A schematic of a deposited metal film on a substrate.

CO induced resistivity change on a 100 A thick copper
film as a function of coverage.

Reflection of polarized light on a conductor surface. EP
denotes the field parallel to the plane of incidence; Es
denotes the field perpendicular to the plane of incidence.

0 is the angle of incidence and 8(a)) is the dielectric
function of the substrate.

A schematic of penetrating fields in the near surface

region of a conductor. E 1 notes the field perpendicular to
the surface; Ell notes the field parallel to the surface. The
field inside the metal is attenuated.

The near surface region of an adsorbate-covered surface.

(a) The reflectivity of 2000 cm“ p-polarized light on Cu
surface. (b) The reflectivity of 2000 cm”1 p-polarized light
on Pt surface.

The integrated vibrational signal on Pt surface with
uncoupled oscillators model. The dashed line is for an
oscillator with effective dipole moment and vibration
frequency 2000 cm". The solid line is for an oscillator
with effective dipole moment and vibration frequency
500 cm".

xi

1O

13

14

15

22

33

33

figu.

1. ,
\hh.b
. u .

J.
WU
Pr.

3.. m.
. WAN...» mLNw
t . DU.

5313.

U

Figure 2.3

Figure 2.4

Figure 2.5

Figure 2.6

Figure 2.7

Figure 2.8

Figure 2.9

Figure 2.10

Figure 3.1

Figure 4.1

Figure 4.2

The radiation power from a blackbody source with (a) 0.5
mmzsr and (b) 0.05 mmzsr throughput, temperature 1000

K, 1 % efficiency and 1 cm“ bandwidth using i4 deg of
the incident angle 84° and i3 deg of the incident angle 87°
with a 25mmx5mm sample.

Illustration of synchrotron radiation emission.

Synchrotron radiation output curve with electron
current 1 ampere, bending radius p 1.91 meters, 2 cm'1
bandwidth, OH 100 milliradians and summing over the
vertical angle (iv.

The bandpasses of selected optical components used in
RAIRS experiments.

Optical schematic of the Michelson interferometer.

Czerny-Turner mounting of a reflecting grating system
showing light at oblique incidence onto a grating and the

definition of the angles 4) and 0.

The schematic of the infrared beam extraction at
synchrotron light source and the floor plan of surface
science branch.

Optical layout of the MSU RAIRS system.

ARP/ Rp for clean Cu surface, CO/Cu(100), O/ Cu(100) and
CO/Ni(100) over the frequency range from 300 to 2000
-1

C111.

Ratio of the oxygen KLL to copper LMM Auger
intensities as a function of oxygen dosage at 300 K. The
coverage scale on the right axis is obtained by assuming a
saturation coverage of 0.50 ML.

Oxygen-induced fractional reflectance change AR/ R for
four oxygen coverages: (a) 0.08 ML; (b) 0.17 ML; (c) 0.25
ML; (d) 0.35 ML. The gaps in the spectra and the noise at
high frequencies are due to the low transmittance of the
polyethylene windows. The solid lines are fits to Eq. 4.2
discussed in section 4.4; for spectrum ((1) no fit was
possible.

xii

37

39

39

43

47

55

57

67

83

88

L1 I
(1‘;

Pin

CIT-4 -
£3

:17
,.
(r?

r.
I 1:27;:

5v

Scare

Figure 4.3

Figure 4.4

Figure 4.5

Figure 5.1

Figure 5.2

Figure 6.1

Figure 6.2

Oxygen —induced reflectance change for (#025, with the
instrumental offset C’-C subtracted. The solid line shows
the best fit to Eq. 4.2 and the dashed line shows the
results of Eq. 4.1 using the parameters given by Eq. 4.3.
Over the frequency range studied, the two curves are
virtually indistinguishable.

Asymptotic (high frequency) limit of the reflectance
change as a function of coverage (filled square). The
values and uncertainties are taken from Table 4.1 for the
three lowest coverages, and estimated from Figure 4.2 (d)

for 0:0.35 ML. For comparison, the open circles show
the resistivity change of a 40 nm-thick copper film.

The low frequency spectrum from Figure 4.2 (c) for 0.25
ML oxygen on Cu(100) surface after removing the
electron-scattering induced background reflectance
change. The O-Cu stretch is observed around 340 cm".

Adsorbate-induced fractional reflectance changes AR/ R
due to (a) 0.25 ML oxygen on Cu(100); (b) formate on
Cu(100) obtained by exposing the oxygen-predosed
surface to 50 L formic acid. The data reduction
procedures are discussed in the text. The solid line in (a)
is the best fit to the electron scattering model.

High resolution (6 cm'l) RAIR spectrum of formate on
Cu(100), showing the bands assigned to the Cu-formate
vibration. The lines show the best fit to a sum of two
Lorentzians plus a linear baseline.

Fractional change in reflected IR intensity from Pt(111) as
a function of time at a frequency of 2800 cm‘1 and room
temperature. The Pt(111) sample was exposed to CO for
100 seconds at about the 200th second. Background
pressure during dosing was 8x10‘10 torr, while the
pressure at the sample was increased by a factor of 16 due
to the effusive array doser. A linear drift has been
subtracted.

(a) Fractional reflectance change AR/ R as a function of
coverage for CO on Pt(111) at 315 K. Data shown were
taken at two IR frequencies, 2500 and 2800 cm‘l.

(b) Work function change for CO on Pt(111).

xiii

93

97

101

111

114

125

130

Figure 6.3.

The reflectance change lAR/R I vs. US, the ratio between
the bulk electron mean free path and the skin depth, at a

frequency w/ (01:4 where the reflectance change
magnitude approach to an asymptotic value for the
universal frequency-dependent function.

xiv

136

mes
tech:
SON:
inter;
tells:
infon
the

Spect
LL18 b
bells}
of the

hOW ;

Slhch
lTLil‘a f
3‘19er

‘3

mom

Chapter 1

Introduction

The interactions of adsorbed gases on metal surfaces are crucial to the
fundamental understanding of heterogeneous catalysis and electrochemical
reactions. The interplay between experimental studies and theoretical
investigations along with the prosperous improvements of experimental
techniques has led to a deeper understanding of the underlying physics of
solids. Among the numerous experimental techniques that probe the
interactions between substrates and adsorbates, light-based probes, such as
reflection absorption infrared spectroscopy (RAIRS), have yielded fruitful
information on the nature of bonding. The most common interest is to study
the vibrational resonance of adsorbates on surfaces. However, the
spectroscopic response of the substrate to light with various wavelengths, e.g.
the broadband reflectivity, can also provide a clear picture of the interactions
between adsorbate and substrate. Studying the changes in collective response
of the substrate to the probing electric field upon gas adsorption can reveal

how adsorbates modify the dielectric properties of metal substrates.

In this work, I present experimental results obtained with a
synchrotron-based RAIRS system, for the adsorbate-induced changes in
infrared reflectance on metal surfaces and compare them to a scattering
theory that incorporates nonlocal electrodynamics. The immediate

motivation of this work comes from the measurement of the broadband

reflec
1939 1

the d

curl.

Perssc
the m.
resuhs
to V9:
tramp.-

region

2

reflectance change due to CO on Cu(100) by Hirschmugl and coworkers in
1989 [1]. They observed distinguishable background absorption together with

the dipole-selection forbidden hindered rotation mode of the CO molecule

and the Cu-CO stretch within the frequency range 200-500 cm'1 as shown in
Fig. 1.1. The magnitude of reflectivity change increases monotonically with

frequency and reaches an asymptotic value at high frequencies around 2000

cm'l. The significance of the broadband reflectance change is noted by

Persson’s theory [2] which attributes the change of the dielectric properties of
the metal substrate to the increase of diffuse scattering of the electrons that
results from the damping of the parallel motion of adsorbates. It is important
to verify the theory, which characterizes the effects of adsorbates on the
transport properties of conduction electrons of metals in the near surface

region.

Electron transport in a metal will be reviewed in section 1.1 followed by
a review of earlier studies of surface-scattering effects in thin film resistivity
in section 1.2. A brief introduction about surface IR studies will be in section
1.3. Persson’s [2-7] theory of electron scattering and broadband reflectance
change is introduced in section 1.4. A brief outline of conducted experiments

is in section 1.5.

 

  
   

 

 

 

 

a r - u - I - t
_ c(2 x 2) CO/Cu(100]
- -‘ 2 cm" _.
E . l
5
2
O
(I
\ - '4
O
O
m
3‘.
’n
- A J . n . on - 1 -
200 250 300 350 400 450

Frequency (cm")

Figure 1.1 CO on Cu(100) surface. Distinguishable background absorption
observed together with the dipole-selection forbidden hindered rotation

mode of the CO molecule at 288 cm'1 and the Cu-CO stretch at 345 cm'1 over
the frequency range 200-500 cm'l. ( reproduced from Ref. [27])

1.1

the ma
:pom
matter's
dielectr
configu
are treat
level are
field an
Coulom
and the
”collisic
879mm
(0111310,
be1/t

metal Ca

1.1 Electron transport in a metal

When electromagnetic waves are incident on a solid, the charges inside
the matter rearrange themselves as a response to the external fields. The
response of the matter to the electric field, e.g. the polarizability, defines the
matter's macroscopic dielectric properties. A detailed understanding of the
dielectric properties nevertheless calls for a microscopic picture of electronic
configuration and electron transport inside the object. For metals, the ions
are treated as fixed at lattice sites. The conducting electrons above the Fermi
level are treated as a free electron gas which will be driven by the incoming B-
field and cause the large polarizability of a metal. In the Drude model, the
Coulomb interactions of a electron with other electrons and ions is neglected
and the interaction which the electron experiences are only instantaneous
"collision" events [8]. These collisions are assumed to maintain local

thermodynamic equilibrium. If the mean time interval between two

collisions is 1', the probability of collisions for a electron per unit time would

be 1/ 1. Within the free electron model, the DC electrical conductivity of a

metal can be expressed as:

 

j= O'E , with 0' = p“1 = ne T Eq.1.1
m
j : current density E : electric field strength

0': conductivity p : resistivity

The me
defined
electron

The Pet
hmmt
resStiri‘
model :
r00m te
electron
mmdmd
IQSlSfiVit

the 0rde

equafiOn

mamfle

This fem

be...”

my

lOrm

5

n : electron density 1 : relaxation time

711 : effective electron mass e : elementary charge

The mean distance an electron travels between two collision events can be

defined as the electron mean free path : 1: DOT where no is the average

electronic speed. For conducting electron, it could be the Fermi velocity vF .

The Fermi velocity of the electron can be calculated with the known electron
density: v; =(h/m)-(37tzn)2. The product of the mean free path and the
resistivity depends only on the electron density within the free electron

model : p-l=e‘2(37:2)” 311’“ 3. The order of the electron mean free path at
room temperature is about hundreds of angstroms. For tungsten, the
electron mean free path is about 144 A and 1140 A for copper which are
hundred times of the interatomic spacing [9,10]. Room temperature

resistivities are typically of the order of one microhm-centimeter which gives

the order of the relaxation time about 10'14 sec for metals.

The AC electrical conductivity of a metal can be given by solving the
equation of motion for the momentum per electron and the current induced

in a metal by a time-dependent electric field:

0'0 nezt
_ . do =
1-zan' m

 

 

Kw) = 0(w)E(w) ;0(w) = Eq.1.2

This form correctly reduces to the DC result at zero frequency. One point to
bear in mind is that the equation of motion is derived by assuming a spatially

uniform force. If the external field is from the propagating electromagnetic

 

 

radiati

of the

compa:
electric
should
in a m.
Wavelei

dielectr

Where (

fiEldS (

frequm

t“? of

Path m,

,4
h

I
l

6
radiation in a metal, the validity of the Eq. 1.2 will be questioned. The effect

of the magnetic field on the motion of electrons can always be neglected since
the magnetic force is much smaller than the electric force. However, the
current density at one certain point is entirely determined by the summation
of the effect of the E-field on each electron at that position since its last
collision. If the electric field does not vary appreciably over a distance
comparable to the electron mean free path, the relation of current density and
electric field in j(a),r)=0'(co,r)E(a),r) is justified. The wavelength of the field
should be large compared to the electron mean free path about hundreds of A

in a metal. This is satisfied by wavelength larger than visible light which

wavelength is of the order of 103 to 104 A. Therefore the macroscopic complex

dielectric function at infrared frequency region can be given by:

2
2 _ 47me2

p
I a) =
(02(1-1/ian) ” m

 

 

s=1+i%g,s(m)=1+ Eq. 1.3

where (up is known as the plasma frequency. At low frequencies a)<<a)p, the E

-fields decay in the medium. When the frequencies are higher than plasma
frequency, the metal should become transparent and the E-field can propagate

through the material.

The conductivity of a metal is intimately correlated to the relaxation
time of the conduction electrons. The relaxation time depends in turn on the
microscopic electron scattering processes. In the Drude model, the mean free
path and the relaxation time were introduced without any reference to

different types of collisions or scattering mechanisms. To study the relation

betweer
xahefi:
Huh?
dmmg
each sc.

flame

Mums

asasen‘e

 

0f metal

In
S‘Jrface IE
importan.
EDIEraCn-O:
kmumg

Sm ~
L €1.ng

 

7
between scattering and resistivity, one first treats the different types of

scattering mechanism as independent events. The generalized form of
Matthiessen's rule in which the resistivity in the presence of several
distinguishable scattering processes is simply the sum of the resistivities of
each scattering mechanism turns out to be a good approximation. The

effective relaxation time is [9]:
l = 2i Eq. 1.4
1' i 1,.

i : different scattering processes

1. electron-electron scattering rate

2. electron-phonon scattering rate

3. an effective scattering rate due to the collision with

surface or impurities
Ifwe disregard the simplicity of this approximation, the assumption that the
common influence of the different types of scattering processes can be treated
as a series connection is an explicit or implicit precondition of many theories

of metal conductivity.

The electron-surface scattering affects only the electrons in the near
surface region. Its contribution to the bulk resistivity is very small. But the
importance comes in when one studies surface phenomena and the
interaction of substrate, adsorbates and the probe. The electron-surface
scattering can be divided into two categories: specular scattering and diffuse

scattering. Specular scattering is the collision process which conserves the

 

 

paralle

scatter:
the sun
may als
complic
harsitit
spectros

hroadha

surface

paramet.
Change .
Rid-om]
0’3““ t3?
1 Which
(”Omept .
electmnE

Elle‘vec

£6.me

8
parallel momentum of electrons. The diffuse scattering, on the other hand,

causes momentum loss and heat dissipation and is generated by disordered
scatterers such as the missing or extra atoms in the top surface layer, steps on
the surface, or foreign atoms adsorbed at the surface. N onspecular scattering
may also arise from the surface umklapp scattering when the Fermi surface is
complicated. The umklapp process involves the electronic interband
transitions by surface phonon or IR photon. From reflection-absorption IR
spectroscopy, one can probe the interband absorption and also the background

broadband absorption by electron-surface scattering.

The overall scattering behavior of conduction electrons from the
surface is often phenomenologically described by the Fuchs specularity
parameter p [11,12]. If all electrons are reflected with no tangential velocity
change then p=1. The specularity p=0 indicates that all the electrons are
randomly scattered into all possible scattering angles due to diffuse scattering.
Other types of scattering behaviors are characterized by a value between 0 and
1 which indicates the effective specularity of the outgoing electrons. This
concept is widely used in the discussion of the scattering of the conduction
electrons in metallic fihn. The study of the resistivities of metal films is

reviewed in next section.
1.2 Thin film resistivity studies

A thin continuous smooth metal film with a well defined structure

can be mapped to the near surface region of a thick metal crystal. The electric

 

 

“sent
scattem
Ol in lht
hsuhth
empm
thnat

apphedl

Fig»k
08p

For

9
resistivities measured on these metal thin films reveal, to a large extent, the

scattering properties of conducting electrons at the near surface region instead
of in the bulk. The metal film is usually made by depositing atoms or ions on
insulating material substrates such as Si crystal or glasses. The experimental
setup for resistivity measurements is well known as four-wire method. The
schematic is shown in Fig. 1.2. The direction of the electric current and the

applied field are specified.

 

 

 
    

I 1: film thickness

 

 

 

 

Figure 1.2 A schematic of resistivity measurement on a metal film.
(reproduced from Ref. [14])

For clean thin metal films, the surface roughness is the origin of the
diffuse scattering. The specularity parameter can also be used to quantify the
fraction of conduction electrons which are specularly reflected at the
boundaries of the metal films. The calculation is based on a number of rather

restrictive preconditions: a single parabolic conduction band, a homogeneous

and thic

sketched

10
and thickness-independent bulk structure, and plane-parallel boundaries as

sketched in Figure 1.3.

V/

 

 

 

insulattr substrate

 

outer crystallite surface

 

 

the innercrystallite grainboundary

 

1: thickness

Figure 1.3 A schematic of a deposited metal film on a substrate
(reproduced from Ref. [12])

Although this Fuchs-Sondheimer [11,12,13] approach is rather popular
for describing the measured dependence of the resistivity on the thickness of
thin metal films, it might be unsuitable due to the intrinsic preconditions
mentioned above. However, it is worthwhile to adopt this approach and see
how the film resistivity behaves under this framework. The resistivity of the

metal film can be solved from the Boltzmann transport equation as [12] :

pa 3 t .. 1 x2 1—e" .
=1__ _ _ 1.. __— ————, th =t I E .15
p (2)10( P)l,.(s3 $5)1_pe_s WI x /0 q

where p and t are the resistivity and thickness of the film, p0 is the resistivity

of the bulk with the same density of lattice defects as the film. For sufficiently

high thickness(t>>10), the fractional change of the resistivity can be expressed:

r
tip wil

Q‘JAI‘Hllflt

Ir
material
in thin (
and the 2
hard to <
inner g1
EVidence
mdepem
Stages 0:
material,
rate, an
Regimen
forms, t}

SubstTate

11

Ap 3 l0
- 1 . .
—0 ——8( -p)—t Eq 16

As mentioned before the product of the resistivity and the mean free path is a
constant depending only on the electron density 11. Since the electron density

depends only weakly on the temperature and the defect density, the value of
tAp will mainly depend on the effective specularity parameter which

quantifies the properties of the film.

In general, the properties of the metal film depend strongly on the
material and deposition conditions. For example, recrystallization can occur
in thin continuous films or during heat treatment, causing grain boundaries
and the Fuchs specularity p of the same material can vary widely. It is usually
hard to distinguish whether the scattering events of electrons happen at the
inner grain boundaries or at the film surfaces. There is experimental
evidence showing that the assumption of a homogeneous and thickness
independent bulk-like structure is hard to meet in thinner films. The initial
stages of film growth are strongly influenced by several factors such as
material, structure, temperature and the purity of the substrate; evaporation
rate, apparatus geometry, residual gas pressure and composition, sample
treatment and measurement temperature [12]. Before a homogeneous film
forms, the structure of the film is more or less like islands on top of the

substrate.

 

 

For
complicai
change 0:
measurin;
adsorptio:
resistivity
Wissrnanr
ispnmari
adSOI'ptior
100ml thic.
on the 85
FtSlStii'jty

the effect c
Ap :

where 3p i

12
For the study of the electrical resistivity upon gas adsorption, the

complications introduced by film quality can be ignored if the fractional
change of resistivity induced by the adsorbates is measured as long as the
measuring conditions of the substrate remain the same before and after the
adsorption. In the last two decades, the adsorption effect on the electric
resistivity of thin metal films has been extensively investigated by
Wissmann, Schumacher and other scientists [12,14-17]. The diffuse scattering
is primarily responsible for the increase in the resistivity of thin films upon
adsorption. Fig.1.3 shows the CO adsorption induced resistivity change on a
100 A thick copper film [12]. Although the detailed form of the curve depends
on the gas species adsorbed, common features are a linear increase of
resistivity change at low coverage and saturation at high coverage. To explain

the effect on the resistivity of the gas adsorption, we can differentiate Eq.1.6:

 

Ap Eq. 1.7

where Ap is proportional to the adsorbate density

LA") =K'3fi K’a LA”) Eq.1.8
dna "rm f ' dna "‘40

This characterizes the initial slope of the measured curves of the resistivity
change. The linearity between specularity change and the coverage is based

on the assumption of the independence of scattering events.

Adsorption can also affect the conduction properties by modifying the

conducting electron density. Changes in n have been invoked to explain

13
thin-film resistivity changes due to surfaces, defects, and adsorbates, and

oxygen incorporated at grain boundaries in a thin Cu film has been shown to
reduce n [17]. But the influence of conduction electron density on surface

resistivity and the reflectance of a single crystal surface remains

 

 

 

 

undetermined.
50 . . a
40 ~ _ . J a
O ' I
I.
g; 30 ~ 9' ~
E“. I'
E
20 - a
I,“
10 - _ .. -' .
I. - l
u ' .
o - ' i 4 I I .
0.0 0.5 1.0 1.5
n. (l 015 moleculdcmz)

Figure 1.4 CO induced resistivity change on a 100 A thick copper
film as a function of coverage. (reproduced from Ref. [12])

1.3 Surface RAIRS studies and surface effect on reflectance

Reflection absorption infrared spectroscopy has become a standard
technique in surface science. The advantages of RAIRS include the high
energy resolution, strict dipole selection rules, time resolution of dynamic
processes, and convenient analysis of the vibrational lineshape and intensity.

This technique has been successfully used to obtain useful information such

14
as the nature of chemical bonds of adsorbates, adsorption sites, molecular

orientations, adsorbate-adsorbate interactions, and electrostatic effects of
surface defects [18,19]. Among these investigations, the interest is usually

focused on the vibrational absorption bands of the adsorbed species.

Figure 1.5 shows a schematic view of the incident and reflected IR

beams on a metal surface.

 

 

 

 

 

interface-k ( ) W

Figure 1.5 Reflection of polarized light on a conductor surface.
denotes the field parallel to the plane of incidence; Es denotes the

field perpendicular to the plane of incidence. 0 is the angle of
incidence and 8(0)) is the dielectric function of the substrate.

The s- and p- polarized electric fields of the radiation are perpendicular and
parallel to the plane of incidence. On a metal surface, any dipole moment
will induce an image dipole.Any molecular dynamic dipole parallel to the

surface will be effectively screened out. For the dipole perpendicular to the

15
surface, the apparent dipole moment is enhanced. The vibrational modes

with dynamic dipole perpendicular to the surface could be observed by p-

polarized light through the electrodynamic interaction. When p-polarized

light is incident at an angle 0 on a metal surface, the El field will decay more

 
 
 

: electron mean free path

 

 

 

 

8 : classical skin depth

Figure 1.6 A schematic of penetrating fields in the near surface

region of a conductor. Ei notes the field perpendicular to the
surface; Ell notes the field parallel to the surface. The field inside
the metal is attenuated.

rapidly into the surface than the parallel component E. . as sketched in Figure

1.6. Both components propagate exponentially into the metal with a

characteristic skin depth 50 as shown in Fig. 1.6. The difference is that the

value of 15.. at the surface is larger than by a factor ~\/|;|.The conducting

electron in the near surface region will experience the external field and

generate a current density corresponding to the field strength. This accounts

16
for the absorption of incoming light. The reflectivity of a metal will change, if

the dielectric properties in the near surface region are modified. One can
consider that the surface condition changes with a thin adlayer such as gas
molecule adsorption or atom deposition. If the thin adlayer is a weakly-
absorbing film (i.e. for a layer of oscillating dipoles, that condition will be
satisfied far from the resonance frequency), the reflectivity change will be
determined primarily by the changes of the optical properties of the substrate.
Although people have been aware of the importance of this effect for the
quantitative interpretation of RAIRS measurements [20,21], the background
absorption caused by the modification of substrate conduction properties has

usually been ignored.

Standard analysis is based on the local response of the dielectric

functions of metal substrates. The problem arises from the validity of the
definition of the skin depth. If the incident light frequency a) is not too high
i.e. much smaller than the plasma frequency, the classical skin depth is given

by: 1/60 = c/Jeraa) where o is the bulk conductivity of the metal. The skin

depth is around hundreds of A for normal metals [9,10]. The electric field is
localized to a very thin layer at the metal surface. The validity of this picture
is that the field in the metal varies little over the distance one conducting

electron could travel during one cycle of the oscillation of the field. So it

requires the frequency much larger thanvF/B which is defined as the

characteristic frequency and the value is around 400 cm'1 [22]. When the

frequent“)
field ove:
skin ef‘cct
one cycle
on reflect
Nonlocali
that an e
motion 0
dielectric

from a f'm
estimate .
Feibehnan
llHIll of a
meaSUre tl
Parts repre
this analt‘s
POlaered 1

IEI-left a It Ce

FOIE

Scattenng 0]

tr; . .'
.EaQCllllt}. ‘

Bem'lEt [24]

17
frequency drops below vF/fi, the simple picture of an exponentially decaying

field over a distance Sobreaks down. This phenomena is called anomalous

skin efl'ect [8]. If the distance that an electron on Fermi surface travels during
one cycle of field is larger than the skin depth, the calculation of surface effects
on reflectance requires nonlocal corrections to the classical Fresnel equations.

Nonlocality means a contribution to the time dependence of the electric field
that an electron experiences during the time period 1/(1) comes from the

motion of the electron in the spatially varying external electric field. The
dielectric response at one certain point in space would be the summation
from a finite spatial range defined by the travel of the electron. One can also

estimate the dielectric properties in the near surface region using the
Feibelman d-parameter formalism which is valid for C0>>Up/8 [23]. For the

limit of a metal with no bulk absorption, the real parts of d-parameters
measure the effective surface locations for the electric field and the imaginary
parts represent the phase lag of the field in the near surface region. Within
this analysis, the p—polarized light reflectance is far more sensitive than s-
polarized light to the dielectric response. Even on a clean metal surface, the

reflectance is so complicated that it is hard to get an analytic form.

For good conductors such as Ag and Au, it has been shown that the
scattering on the surface is specular if the surface is flat and free of defects. The
reflectivity on these metal surfaces has been measured within local optics by

Bennet [24] using the Drude model. The infrared absorption at a metal

18
surface would be larger if the collisions of the conducting electron with the

surface are diffuse. The effective specularity of the conducting electrons can be
used to characterize the reflectance change. Holstein used the specularity
parameter to estimate the reflectance change mainly associating with the s-
polarized field [25]. In order for surface-electron scattering to have a major
effect on the reflectance of an Optically thick sample, two conditions must

apply. The first is that the mean-free path 10 of the electrons must be at least
comparable to the classical skin depth 80 of the metal. If 10 <<80 then the bulk

scattering (electron-electron scattering and electron-phonon scattering)
dominates the absorptance. The other condition is that the energy of the
impinging radiation should be below the region where the interband
transitions dominate the optical response of the metal. For good conductors
at room temperature the range between 100 and 4000 cm'1 usually satisfies
these two conditions [9]. The broadband absorption yields detailed
information about the interaction between the surface and the probing E-field

as well as how the adsorbate modified the surface.

In the late-1980s, broadband reflectance changes upon gas adsorption
were reported for H on W and M0 by Riffe, Hanssen, and Sievers [9] and
Reutt, Chabal, and Christrnan [26]. Riffe and coworkers [9] observed a

frequency-dependent reflectance change over the frequency range from 886 to

1088 cm'l. Qualitatively they attributed the changes of the attenuation

coefficient of surface electromagnetic wave to the changes in the specularity of

19
scattered free carriers caused by the adsorbate-induced reconstruction of the

W(100) surface atoms for high coverage adsorption. Riffe also studied other
adsorbates and discussed free—carrier absorption. Later, Reutt and coworkers
showed the reflectance change vs. frequency exhibiting a definite minimum
for H on W and Mo which led to an interpretation of an intrinsic adsorbate-
induced surface state around Fermi level. The possibility of the contributions
by the free-carrier scattering was ruled out on the basis of a measured

electronic band corresponding to the IR absorption.

Hirschmugl et a1. [1,27] showed that the magnitude of the reflectivity

change for CO on Cu increases monotonically with frequency and reaches an

asymptotic value at high frequency region around 2000 cm'l. An asymmetric
anti-absorption resonance assigned to the dipole-forbidden rotation of CO and
one symmetric absorption of the Cu-CO vibration mode were also observed
along with the broadband reflectance change. It is intriguing about the
physical origins of the anti-absorption peak and the broadband reflectance
change lineshape. The new theory proposed by Persson explains the
underlying physical mechanism for the experimental results well. We will

discuss the theory in the next section.
1.4. Scattering theory of conducting electrons in adsorbed layer

The scattering theory of conducting electrons in an adsorbed layer,
proposed by Persson and Volokitin [2-7], involves the damping of the parallel

motions of the adsorbates on metal surfaces through electron-hole pair

productit

 

wmwn
theory, th
damping ;
surface is
expects tht
because the
screened or

electronic c

 

 

friction torn
between the
forbidden er
electrons in
line shape f‘
forbidden n
decrease of
affounts fOr
Optics i5 am
will Propaga
“.cle of th e (
Elecmc field

20
production. It gives a microscopic picture of how the adsorbates modify the

surface resistivity and the surface reflectivity. According the description of the
theory, the diffuse scattering of the conducting electrons increases due to the
damping process, hence the surface resistivity also increases. When the metal
surface is probed by p-polarized light at near grazing angle incidence, one
expects the surface parallel motions of adsorbates would not be detected
because the dipole moment is perpendicular to the external electric field and
screened out by the image dipole. If the dipole forbidden mode couples to the
electronic current density induced in the metal by the external field, the
friction force working on an adsorbate is proportional to the relative velocity
between the adsorbate and the collective motion of the electrons. This dipole-
forbidden excitation of adsorbate vibrational modes can be mediated by the
electrons in the metal. In this case, an asymmetric band with anti-absorption
line shape for the parallel vibrations is observed. In addition to the dipole-
forbidden modes, the background broadband reflectance decreases. This
decrease of reflectivity exhibits a frequency-dependent character which

accounts for the local-nonlocal optics transition on metal surfaces. Local

optics is accurate only if m>>vP/8. If 6<<vF/ a), an electron on the Fermi surface

will propagate over a distance of the order of the decay length during one
cycle of the oscillating field. Hence the electron will be influenced by the
electric field over a large spatial region, implying non-local optics. A

contribution to the time dependence of the electric field which an electron

experier

electron

Fig

21

experience during the time period 1/a) comes from the motion of the

electron in the spatially varying external electric field.

randomly distributed adsorbates
.__, :
metal surface

- t
scattered electron

A
V

   
     
    

p

 

t: near surface region or thin film

Figure 1.7 The near surface region of an adsorbate-covered surface.

Let us consider a thin metal film randomly covered with adsorbates as
shown in Figure 1.7. The adsorbates on the surface can vibrate parallel to the
surface in addition to the perpendicular stretch mode. This parallel motion of

adsorbate is a hindered motion localized to the surface potential well. The
lifetime of the parallel motion of the adsorbate is Ta- The relative motion of
the adsorbate and the drift motion of electrons in the metal introduces a
friction force f =—nM(11-v,/) where M is the adsorbate mass and 1),, is the
drift velocity of the conduction electrons and it is the time derivative of the
parallel coordinate of the adsorbate and is the velocity of the hindered

motion. The friction force coefficient 1) is equal to the scattering rate 1/re_h,

22
which indicates the excitation of the electron-hole pairs. This friction force

introduces a damping term in the equation of motion for the adsorbate. The
power dissipated in the conduction electrons is equal to the energy transferred

from the adsorbates. So the conductivity will change if the scattering rate is

changed. The relaxation time of scattered conduction electrons TM
corresponds directly to the life time of the damped vibrations Ta. The effective
scattering time in the near surface region will be modified as 1/ 1'97 =1 / rB+1/ Ta

where 13 is the relaxation time for conducting electrons in the bulk. If the

modification affects the electron transport in an effective thickness t inside
the metal surface and the Joule heating of the induced electron current is
equal to the power loss from the damping of the adsorbate parallel vibration,

the resulting damping rate is related to the resistivity change in the thin layer:

1 _ nzezt 8p
M an

d a n,=0

 

Eq. 1.9

where n is the conduction electron density, 11,; is the adsorbate density. dip/8n“

is the initial slope of the surface resistivity change due to the gas adsorption
from the measurements on a metallic film with same thickness t. Since the
connection between surface resisitivity change and the diffuse electron
scattering is established through the direct interaction of the adsorbates and
the e-h pairs, it is enhanced if there is an adsorbate-induced electronic state

near the Fermi level. The effective cross sections for the diffuse electron

23
scattering of different adsorption systems depend on the overlapping

adsorbate-induced surface state and the conduction bands.

The reflectance change is caused by the absorption of the incident field.

The intensity of the incident beam is determined by the poynting vector and

is given by I0 =cEg/87r. The power dissipated in the conduction electron

driven by the external field, i.e. o/ /~exp(-ia)t), is equal to the work done by

the friction force. The time-average adsorbate-induced power absorption is,
P=N-(f-(u,, —a)) Eq. 1.10

where N is the number of the adsorbate. The adsorbate-induced change in

the IR reflectivity is given by :
AR =__.P_ Eq.1.11
10A

where A is the cross-sectional area of the incident photon beam. Under local

optics, the reflectance change is,

2_ 22
AR=———‘M ’7 (w ‘00) Eq.1.12

 

When Ito—(no I >>n, the frequency is far away from the resonance of the

parallel hindered motion of the adsorbate the reflectance (no, then the

reflectance change is a term depending on the electron scattering rate:

4nM 1 1
cnmcosOrH

 

Eq. 1.13

Then the direct relation between resistivity change and reflectance change is

justified under the assumptions of the theory.

lt
try/5. It
with cor
optics. '
derivatic
electric fi

It
An apprc
relation t
infinite 11
toward t]
OSClllathr
Which in
a-‘Surned

Can be Wr

5.
Where T is
EleClTOn C
fiaCtion p '

Lhe YES] are

llDl‘athn U

24
It is more complicated when the incident field frequency drops below

vF/S. The problem is to figure out the induced current by the incoming field

with conduction electrons colliding to the adsorbates under the nonlocal
optics. The approach for the theory is briefly described below. A detailed
derivation is given in Ref. [7]. For nonlocal optics, the spatially varying
electric field have to be considered in calculating the induced current,
[(x,a)) = [dax - o(x,x’;a))E(x’,a)) Eq. 1.15

An approach based on Boltzmann’s equation is used to obtain the non-local
relation between current density and the electric field. If one assumes a semi-
infinite metal with its surface in the x-y plane, the positive z-axis directed
toward the interior of the metal, the adsorbate can perform harmonic
oscillation in the x-direction, and also neglecting the normal field component
which inside the metal is much weaker than the parallel component

assumed along the x-axis, the electron distribution function f(z,vx,vy,vz)
can be written in the form,

.13: £511 in—f‘fo E 116
at+mau,+”‘az’ r q"

where Tis the relaxation time and f0 is the Fermi distribution function. The

electron distribution function will satisfy the boundary condition that a

fraction p of the electrons arriving at the surface are scattered specularly while
the rest are scattered diffusely with average no (the velocity of the adsorbate

vibration in the x-direction),

hmhi

 

motion 0
the electr
electron r.
M:
The electr
nz;
From Eq
solution,
calculated

fer‘lectivih

Per
Scattering
YEtlectanCE
dellt'ed a

l

refleCtsince

Mi

\

R

the 919C [IO

25
f(0,v,,vy,vz) = pf(0,vx,vy,—vz)+(1-p)f(vx —v0,vy,vz) Eq. 1.17

To take into account that a relative motion between an adsorbate and the drift
motion of the electrons leads to friction clue to transfer of momentum from
the electrons to the adsorbate, the friction force acting on an adsorbate is the
electron momentum flux divided by the number of adsorbates per unit area:
M55+Mm§x+ 11,, /n, = o Eq.1.18
The electron momentum flux in x—direction can be written as
III: = m]d3vv,v,f(0,vx,vy,vz) Eq. 1.19
From Eq. 1.15-1.19 , the distribution function f(x,u) can be found. With the

solution, the drift current induced by the external electric field Ex(z)can be
calculated. Along with Maxwell equations, it is possible to calculate the

reflectivity of the adsorbate-covered metal surface.

Persson and Volokitin elaborate the nonlocal electrodynamics and the
scattering effect of conduction electrons to calculate the adsorbate-induced
reflectance changes on metal surfaces. The reflectivity for p-polarized light is
derived and some numeric results are tabulated in Ref [7]. The fractional

reflectance change can be expressed as follows:

AR____—4a - (m/wl)2Re[g(a)/w1,l/6)]
R 7:2 1_[—4—0)—]Im£; dq

mop cosa 2(w/w1,l/5,q)

 

Eq. 1.20

 

where g(x,z) and e(x,z,q) are complex-valued functions defined in Ref. [7],l is

the electron mean free path, a is the magnitude of the reflectance change at

26

higher frequency and equal to %g—-—’;)vC—F, mp is the plasma frequency of the
cos

metal, 5=c/a)p is the classical skin depth, and (01 is a characteristic roll-off

frequency, near and below which the nonlocal effects become important.
Although the connection of electron scattering to vibraitonal damping and
anti-absoprtion resonances is much emphasized in the theory, there are still

questions in the assumptions [28].

If the frequency arise higher than the characteristic frequency (1),, Eq.
1.20 reverts to local optics and the reflectance change reaches an asymptotic

value. The spectral shape of the broadband absorption depends on when and

1/6. All the quantities are properties of the substrate material. The

interaction between the adsorbate and the substrate mainly determines the
magnitude of the reflectance change, not the shape of the frequency

dependence.

The frequency dependence is the characteristic of broadband reflectance
changes. The roll-off frequency characterizes the local-nonlocal transition
and shows the conduction electron properties. The theory is based on a linear
dependence on adsorbate coverages. It is of great interest to see how the

adsorbate population would affect the reflectance change.

27

1.5 Experiment Designs
My study is motivated by the proposed theory.

As a test of whether the observed reflectance changes are related to the
modification of electron transport, a comparison between resistivity changes
and observed reflectance changes would show whether the scattering
assumption holds in different experimental systems. A simple relation
between measured resistivity changes on thin metallic films and observed
reflectance change on a crystal is derived for the comparison. The
relationship is confirmed within certain satisfaction for CO and oxygen on Cu

surfaces. The details are discussed in Chapter 3.

The frequency-dependent shape of the broadband reflectance changes ,
predicted by the theory is investigated with the system of atomic oxygen on
Cu(100). The measured reflectance changes are fitted to the predicted
functional form in chapter 4. At low coverages, the frequency dependence
agrees well with the universal form. Some discrepancy remains at higher

coverage. The characteristic of coverage dependence is also investigated.

While the evidence suggests the direct interaction between adsorbate
and substrate is through electron diffuse scattering, other adsorbate species
show different results. I also study formate on oxygen predosed Cu(100) with
the synchrotron based FTTR system and no reflectance change is found.

Possible explanation is discussed in Chapter 5.

28
I use CO on Pt(111) as a further approach for the coverage-dependence.

The nonmonotonic behavior of the reflectance change vs. coverage is similar

to O on Cu(111). The reflectance change increases as coverage increases for

coverages below 0.33 ML, reaches a peak at 0.33 ML, then declines toward the

saturation coverage. The overlayer structure and work function change trend

also switch at 0.33 ML. The correlation between reflectance change, overlayer

structure, and surface electronic structure is discussed in Chapter 6.

References:

1.

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

B. N. I. Persson, Phys. Rev. B44, 3277 (1991).

B. N. I. Persson, Surf. Sci. 269/270, 103 (1992).

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

B. N. ]. Persson, Phys. Rev. B 48, 15 (1993); B. N. J. Persson, J. Chem.
Phys. 98(2), 1659 (1993).

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

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

N. W. Ashcroft and N. D. Mermin, Solid State Physics (Saunders

College, Holt, Rinehart and Winston, Philadephia, 1976).

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

29
D. M. Riffe, L. M. Hanssen and A. J. Sievers, Phys. Rev. B 34, 692 (1986);

D. M. Riffe, L. M. Hanssen and A. I. Sievers, Surf. Sci. 176, 679 (1986); D.
M. Riffe and A. J. Sievers, Surf. Sci. 210, L215 (1989).

KC. Lin, R.G. Tobin, P. Dumas, CJ. Hirschmugl and GP. Williams,
Phys. Rev. B 48, 2791 (1992).

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

P. Wissmann, in Surface Physics, edited by G. Hohler, Springer Tracts
in Modern Physics, Vol. (Springer, New York, 1975).

EH. Sondheirner, Advan. Phys. 1, 1 (1952).

D. Schumacher, Surface Scattering Experiments with Conducting
Electrons, Springer Tracts in Modern Physics (Springer, New York,
1975).

D. Dayal, H.-U Finzel, and P. Wissmann inThin Solid Films and Gas
Chemisorption, edited by P. Wissmann (Elsevier, Amsterdam, 1987).
H. Buck, R. Schmidt, and P. Wissmann edited by D. Hirschfeld
(DGM_Informationsgellschaft, Wiesbaden, 1988), p.219.

1. Vancea and H. Hoffmann, Thin Solid Films 92, 219 (1982).

YJ. Chabal, Surf. Sci. Reports 8 (1988); R.G. Tobin, Surf. Sci. 183, 226
(1987).

FM. Hoffmann, Surf. Sci. Reports 3 (1983).

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

R.G. Greenler, J. Chem. Lett. 44, 310 (1966); ].D.E. McIntyre and DE.

Aspnes, Surf. Sci. 24, 417 (1971).

24.

26.

27.

28.

30
KC. Lin, R.G. Tobin and P. Dumas, Phys. Rev. B 49, 7273 (1994).

PJ. Feibelrnann, Prog. Surf. Sci. 12, 287 (1982).

HE. Bennet and ].M. Bennet in Optical Propertiesand Electronic
Structure of Metals and Alloys, edited by F. Abeles (John Wiley and
Sons, New York, 1966).

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

LE. Reutt, Y. J. Chabal and SB. Christrnan, Phys. Rev. B 38, 3112 (1988).
CJ. Hirschmugl, Y. I. Chabal, F. M. Hoffmann and GP. Williams, J.
Vac. Sci. Technol. A 12, 2229 (1994).

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

Chapter 2

Instrumental Introduction

The major part of the experimental work was done at the U4IR
beamline at the National Synchrotron Light Source at Brookhaven National
Laboratory [1-5]. A home-built RAIRS system at Physics department,

Michigan State University, is also used in this work [6]. These two systems

are particularly designed for low energy studies, less than 2000 cm"1 (400
meV). I will discuss the aspects for the design of a RAIRS system in following

sections, then introduce these systems at the end of the chapter.

2.1 Surface signal and light sources

A generic RAIRS system consists of the source, spectrometer, UHV
sample chamber, detector and relevant optics. Proper modulation can be
chosen to separate a signal from superimposed background for sensitive
detection. Primary concerns are related to optimizing the surface signals
either due to the adsorbate vibrations or the changes in the surface optical

properties of the metal.

To optimize the performance of an experimental system, one seeks the
highest signal and the lowest noise. The N/ S ratio value indicates the
sensitivity of an operating system and expressed in term of root Hertz to take
into account the measuring bandwidth [7]. The signal of interest in a RAIRS

system is the fractional change in the intensity of the reflected light before and

31

32

after the gas adsorption on the substrate. This small change of the reflectance
caused by the interaction of the substrate, the adsorbates, and the probing
infrared radiation is obtained by comparing two spectral scans. Non-
reproducibility of the raw spectra can appear as noise or drift in the
comparison spectrum. High stability is therefore required for precise

measurements.

The absorption of incoming light by a solid surface is due to the
polarization of the substrate which is related to the dielectric pr0perties and
the dipole interaction on the surface. Gas adsorption can be modeled as a
surface covered by a thin, homogeneous, and isotropic layer [8,9]. The surface
region of a substrate where the dielectric properties are different from bulk

values can also be treated as a distinct region [8]. If we work out the boundary
conditions for the interfaces, the reflection intensity R (8, s) will be a function

of the incident angle, and the dielectric properties of the layers. For the case of
gas adsorption, the proper polarization can be chosen for the adsorbates
which can link the microscopic parameters to the macroscopic dielectric

functions [8].

The high conductivity of metals makes the reflection properties
dramatically different for various polarizations of light. We are interested in
p-polarized light due to surface sensitivity. Fig. 2.1 shows the incident angle
dependence of the reflectivity on a bare metal surface using a Drude model

for the dielectric functions of Pt and Cu [10] for p-polarized

33

 

 

 

 

 

 

 

 

1.0 . . .

0.9 3......“ ..
>‘ 0.8 - ............. ‘
.2: . '

3.
«6 0.7 - ..
.9. - .
Q-
o 0.6 _ ct
a b
i
0.5 - .
' 4
70 75 80 85 90
1.0
t .1
0.9 - .
“uh-o...
0.00.0... ......
0" _ .......... ..
3‘ ...........
:E ..................... r
3 0.7 - ”m” -l
0
a .
0 i
M 0,5 - ‘
OJ ' c1
70 75 80 85 9O
Incident angle

Figure 2.1 (a) The reflectivity of 2000 cm”1 p-polarized light on Cu
surface. (b) The reflectivity of 2000 cm‘1 p-polarized light on Pt
surface. Dielectric parameters are adapted from Ref. [9].

34

light at the frequency 2000 cm'l. For p-polarized light at near grazing angle,
85-90 degree, less than 80% of incident beam is reflected. If the incident light
is set close to grazing angle, the reflected background intensity is low and the
fractional change due to the absorption of the adsorbates or the change of the
dielectric properties in the near surface region is corresponding large. The

grazing angle incidence is essential for detectability with the dipole selection
rule for metal surfaces and the angle dependence 1/ c058 of the effective

density of adsorbates. If one is interested in gas adsorption signals, the
integrated fractional signals from the GO internal vibration and substrate-
adsorbate vibration calculated using a non-interacting harmonic oscillator
model on Pt are shown in Fig. 2.2. The absorptivity of the internal vibration
is about several percent, while the external vibration is about half of tenth

percent.

So noise to signal ratio levels on the order of 10'2 percent are needed to

get a convicing results for low frequency signals. The achieved sensitivities
are 0.02% peak to peak Hz'”2 for the U4IR surface beamline [11] and 0.02%

rms Hz'“2 for the MSU RAIRS system [12]. Both systems are marginally
capable for adsorbate-substrate vibration measurements. However, the signal
of interest in this study is broadband background reflectance changes which
require the fluctuation of the absolute intensity of the source be less than the

changes from surface effects about 1% to 0.2% over 20 mins.

35

 

 

 

 

 

0.4 ’ t ’ t ' l
A
E
3
_. 0.3 r ‘
a r"\
= ,’ \
.9.“ ,
VJ //
F‘ I] ‘\
e x .
.2 0.2 _ ’1’ 1—4
H ’ t
o ” t
a ,” |
h r’ \.
rt. .
'0 ,o”’ '1
:5 0.1 v" if
DD 1
0 l
H l
C t.
H l/N
l
000 . I i l ' l 1
7O 75 80 85 90

Incident Angle

Figure 2.2 The integrated vibrational signal on Pt surface with
uncoupled oscillators model. The dashed line is for an oscillator
with vibration frequency 2000 cm". The solid line is for an
oscillator with vibration frequency 500 cm". (reproduced form Ref.

[6] )

36

The existing apparatuses offer different advantages for broadband
measurements. We will get back to the comparison later. For now, we just

keep the typical signal of interest in mind.

When one sets the incident angle close to grazing for optimizing the
surface effect to the signal, the amount of radiation reflected off the sample
will drop sharply if the solid angle of reflection at the sample becomes
smaller. For an optical system, the acceptance of light for the apparatus can be

defined as throughput:
Throughput = accepting area xthe solid angle of acceptance 2 AQeffl

If the incident angle approaches grazing incidence, the solid angle accepted by
the sample will be reduced. One should keep a certain level of radiation to

minimize the contribution from the source fluctuation to system noise. The

typical size of single crystal for surface measurements is less than 2 cm2. The
trade-off between maximizing the surface signal and throughput is important
in designing a RAIRS system. In most cases, the sample is the optical
component to limit the thoughput for grazing incidence and practical size.
Fig. 2.3 shows the output spectra expected from a conventional 1000K globar

IR source limited by the different throughput of a typical RAIRS system.

 

1In optics, people usually use the f-number(f/#) to characterize the ”speed" or the angular
acceptance of an optical device. The f/# is equal to the ratio of the effective focal length of an
optical component and the diameter of the circular device. The photon flux density varies
inversely to the square of the f/#.

37

 

l 0'7

IIIIIIP

    
    

' l

l 0-8

ITIIHII["

'l

l 0‘9

I rrrrrr] *1

\
i 1 1 11111

Watts

\
1

10-10

r rrrrrrr'r'r

1 1111111-..

L1

  

ij'

10‘11

\
4111 n 1111111 -1

 

 

10-12’ 11'111 It"'rl . .....
10‘ 102 103

Wave number (cm-1)

 

Figure 2.3 The radiation power from a blackbody source with (a) 0.5
mmzsr and (b) 0.05 mmzsr throughput, temperature 1000 K, 1 %

efficiency and 1 cm'1 bandwidth using :4 deg of the incident angle

84° and :3 deg of the incident angle 87° with a 25mmx5mm
sample.2 [5].

 

2The power from a thermal source with temperature T at frequency v in the unit of wave number ,
equal to 10., in an interval Av for a particular polarization is given by the Planck distribution
law: pvdv = hc2v[exp(hcv / kT) — 1]". The throughput is calculated as : sample size x
cos(incident angle) x sqr(expanding angle)

38

So a brighter source such as synchrotron radiation may be a solution to
this restriction since the radiation is emitted into a cone with high flux

density. The emission from an accelerated electron is shown in Fig. 2.4. The
total emitted power emitted per electron P(o), Q, u) is a fuction of frequency,

the solid angle, and the electron velocity from classical electrodynamics. The
emitted radiation from a synchrotron light source can be expressed in

practical units, per horizontal milliradian per ampere into a 0.1% bandwidth

with electrons in the storage ring, integrated over vertical angle 8v as [13]:

le=2.547x1013xsxcl(y)
where E is the electron energy in GeV, G1(y) is the Bessel function with

y=Ac/h and Ac is the critical wavelength which bisects the power spectrum

and is given by k(A)=5.59xp/E3 where p is the bending radius of the storage

ring in meters. The universal synchrotron radiation output curve is

illustrated in Fig. 2.5. For the NSLS VUV ring at Brookhaven at a
wavelength of 10 um (1000 cm'l), operating at 0.75 GeV, bending radius 1.91
meters, electron current 1 ampere, collecting 100 milliradians horizontally
(about 6°), 1 milliradian verticaly and assuming 2 cm'1 bandpass, that

corresponds to F.(,.=9.9x10'8 Watts [5]. For a 1000K black body soure at 1000 cm'

1 with 1cm'l bandpass, an opening throughput 0.5 mmzsr (typical throughput
for RAIRS measurements), and 10% emittance the power is 1.8x10'8 Watts.

These numbers are close to real experimental situation.

39

electron current cross section

   
    

 

 

radiation area

Figure 2.4 Illustration of synchrotron radiation emission.
(reproduced from Ref. [5], p.23)

 

  
   
 

 

 

 

.- J-I I . I - I .17 ~
I I I 1
10-6 :- I I :4
E I I 2:
I ' ' 1'
I. l a
I
T I I '1
I I I 1
10-7 ;- I I 1
I I 1‘
I ' ' I
.- I I 4
I”, .. I I .
z __ I I ‘
. I . .
3 E Infrared reglon . 1 Ultra Violet region 1
10~8 e . . 1
': I I :
'_’ I I I“
I- I I at
>- I I -I
I I . . o .4
g . .4— VISIble hght ]
. I I
10’9 :' I I 1
: I I .1
: I I 1
.. I I
h- I I
~1Hl 1-1 1 1111111-1 I L'II‘JII ' 1;;1-141 - I -.r,.-I . 1'I'
103 104 105 105 107 108

Wave number (cm-1)

Figure 2.5 Synchrotron radiation output curve with electron current 1
ampere, bending radius p 1.91 meters, 2 cm'1 bandwidth, 8H 100

milliradians and summing over the vertical angle 8v. (reproduced and
rescaled form Ref. [5])

40
A synchrotron source is brighter than a blackbody radiation source.

However, the synchrotron source intensity is proportional to the electron
current in the storage ring and strongly depends on the geometry of the
electron orbit. The source fluctuation turns out to be the major uncertainty of
the experiments. The experiments were done with only limited adjustments
available to offset the synchrotron source fluctuations. A strict data selection

criterion is essential for data taken at NSLS. It will be discussed in chapter 4.
2.2 Noise , detectors and relevant Optics

The noise we are concerned with is related to the detected signal. It is
important to characterize the intrinsic noise from the detector and the noise
contained in the signal arriving at the detector. There are several possible
sources of noise from the system: source intensity fluctuation, ambient
photon noise, mechanical instability, electrical noise, and the detector noise.
Mechanical vibrations usually result in low frequency oscillations which can
be chopped out by high frequency modulation [6]. Photon shot noise either
from the source or from the ambient cannot be eliminated by lock-in
techniques. The high ambient photon flux can saturate sensitive detectors.
A cooled optical system can reduce the background radiation [6,7]. The designs
of a cooling spectrometer and relevant optics take into account of cryo-state
properties, such as thermal contraction which would affect the optical

alignment, lubrication for mechanically driven systems, etc. [6]. One major

41
advantage of brighter sources is that larger ratios of signals to noise can be

obtained with less effort to control the ambient thermal flux.

The noise—equivalent power NEP (or the noise-equivalent photon rate

NE N) is defined as the amount of radiant power (photon rate) that must be

incident on a detector to produce a signal equal to the RMS output noise of
the detector in a one Hz bandwidth. The NEP(NE N) is thus measured in W/
Hz”2 (photons/secHzm). The noise level of the detector indicates the

sensitivity. Typical RAIRS requires NEP better than 10'10 W /Hz“2 to achieve
the desired sensitivity of the system [4] for a black body source, less than 1%.

Modern photoconductive detectors operated at low temperature have noise
levels close to the photon rate 107 per sec, corresponding to the NEP=2X10'17

W/Hz“2 at 1000 cm’1 with conventional amplifiers [7]. The photoconductive
detector generates a current proportional to the incident photon flux and
produces a voltage across a load resistor. Specialized amplifiers are available
for noise levels below ~100 photons/ sec. Photoconductive detectors have a
cut-off frequency due to the impurity binding energy of the material which
limits the probing frequency region. For low frequency broadband
measurements, thermal detectors can be a better choice. The most sensitive

thermal detectors are semiconductor bolometers. In low backgrounds, such
bolometers have NEPs approaching 10'15 W/ Hz”2 when operated at liquid

4He temperature [7]. The resistance of the detector element will change

exponentially with the temperature rise responding to the incident power

42
when a constant current is flowing in. A voltage proportional to the signal

which is generated across a load resistor can be recorded by the data

acquisition system.

The measurable frequency range is an important element for a RAIRS
system. The spectral range is determined by transmission of the optical
components such as windows, the beamsplitter of interferometer, filters for
the grating spectometer, and the operating range of the detector. A sketch for
the band passes of selected windows, the beamsplitters, and detectors is shown

Fig. 2.6.

One should be aware of the Fabry-Perot interferences caused by the
windows which produce fringes in a period of 1/ dn on output spectrum
where d is the thickness in centimeter and n is the refractive index of the

window material. For example, a Csl window with thickness about 7 m m
will result in the fringes ~ 0.4 cm'l, which won’t affect the spectrum with

resolution 1 cm'l.

2.3 Spectrometer

A chosen spectrometer will give the required resolution of the
measured spectrum and should not limit the throughput. Two types of

spectrometers are used in this study.

 

43

 

l ' r ' l

diamond window

 

silicon window/beamsplitter (dashed line transmittance < 10% )

.------------‘----------------‘--------

 

cesium iodide window

 

KRS-S polarizor substrate

 

graded polyethelene (dashed line, major absorption bands)

.........................................

 

Cu doped Ge photoconductive detector

 

Silicon bolometer

 

0 500 1000 1500 2000
Frequency (cm'l)

 

 

Figure 2.6 The bandpasses of selected optical components used in
RAIRS experiments. ( modified from Ref. [5])

44

A. Fourier Transfrom Infrared Spectrometer (FTIR)

A Fourier Transform Infrared Spectrometer (FTIR) consists of a
Michelson interferometer, a fast analog-to—digital converter, and a computer
station with a fast Fourier Transform calculation algorithm. The schematic of
a Michelson interferometer is shown in Fig. 2.7 [5]. The collirnated source
input is divided by the beamsplitter into two components. One is reflected
back by the fixed mirror, the other by the moving mirror. They are
recombined at the beamsplitter and directed into the detector. The signal at
the detector is a function of the optical path difference between the fixed and
moving mirrors that results in an interference pattern. The interferogram

can be Fourier-analyzed to reconstruct the broadband spectrum as a function

 

 

 

 

 

 

of frequency [14].
fixed mirror
It D
‘ n I
beam splitter movmg mirror
4— _>
detector D All
‘ 4 ¥ 1
L/ 2

 

 

light source

Figure 2.7 Optical schematic of the Michelson interferometer.
(from Ref. [5])

45
Three major advantages over dispersive monochrometers for a FTIRS,

the throughput advantage, the multiplex advantage and the rapid-scan mode

providing a modulation of the signal [5,14].

The resolution of an interferometer is determined by the total path

difference of the two interfering rays. The theoretical maximum value the
resolving power is: RaZd Mo, where 110 is the wavelength of the incident

monochromatic radiation and d is the maximum optical path difference. The
resolution doesn’t depend on the aperture size. When one chooses higher
resolutions, the throughput will not be degraded. This is the throughput

advantage; it is not necessary to reduce throughput to get higher resolution.

The multiplex advantage (Fellgett’s advantage) exist if the noise is
independent of the spectral bandwidth over which signal power falls on the
detector, such as the inherent detector noise. The FTTRS can measure the
entire spectrum of N spectral elements in the same time that a grating
spectrometer with similar efficiency and throughput takes to obtain a single

spectral element with the same signal to noise ratio. An improvement of a

factor of N”2 is gained compared in the normal situation.

The signal sampling can be triggered by a He-Ne laser that shares the
optical path. The moving mirror is traveling at a constant velocity and the

signal is recorded by the fast AD converter sampling at fixed path difference

46
intervals. This mode provides a specific modulation frequency, which is

proportional to the mirror velocity, for every incident wavelength.

To study the frequency dependence of the adsorbate-induced broadband
reflectance change, the convenience of getting the whole spectrum range in
one scan is the major reason one chooses a FTIR. However, although FTIR
also offers various advantages, for RAIRS, sample limits throughput and
high resolution isn’t necessary— surface vibration modes are usually wide.
The measurements are never really limited by noise, but by signal fluctuation
or instabilities—multiplex advantage may not exist either. The simplicity and
ability to do monochromatic measurements with a grating spectrmeter can

give better sensitivity to changes in broadband reflectance.

B. Reflection grating spectrometer

The Czerny-Turner mounting shown in Fig. 2.8 of a reflecting grating
is widely chosen for reflection grating spectrometer [6]. The resolution of the
spectrometer is determined by the dispersive power of the grating, the slit
width and the focal length of the focusing mirror. The throughput is
determined by the slit size and the f/# of the collimator. A larger slit width
will give higher throughput, but degrade the resolution. It is crucial to
maintain adequate resolution and maximize the throughput to increase the

sensitivity. The angular resolution of the spectrometer is determined by the

slit width w and the effective focal length f of the focusing mirror: d8=w / 2f.

47

collirnating mirror

\

optical axis

fightout

lightin _ e gran' ; normal

 

 

_—
—
—
_—
_—
—
—
—
—
—
—
_—
__-
.—
—

 

Izgrating anJe

 

 

grating

focal point focal point
entrance slit
exit slit

Figure 2.8 Czerny-Turner mounting of a reflecting grating system
showing light at oblique incidence onto a grating and the definition

of the angles o and 8.

48
The spectral resolution is also related to the dispersion of the grating.

A grating spectrometer has to be calibrated to get accurate frequency
measurements. The calibration is done by using a known monochromatic
source to get the spectra of different order diffraction peaks compared to
calculation. A He-Ne laser can be used. The diffracted angle of the beam line

of n-th order is expressed as the grating equation:

n}. = 2dsin0n cos¢

where A is the wavelength of the laser, 6328A, n the order of diffraction, d
the grove spacing of the grating, 6,, the grating angle at the n—th order peak,

and 4) the half angle between the incident and diffracted beams of light on the
grating face as shown in Fig. 2.8. By measuring the laser spectra of several
orders along with 0“, cost) can be determined from the slope. By
differentiating the equation and dividing them, one can get the spectral

resolution and resolving power:

(12. dv fix 1
A v 2f tan0

In the system in MSU, the grating spectrometer is cooled to reduce the
background radiation [6]. Measurements show different coso values for room

temperature and liquid nitrogen tempertaure; 0.983 and 0.980. It is consistent
with the thermal contraction of aluminum which is the substrate material for

the grooved grating. Filters are used to remove the higher than first order

49
diffraction which would contribute about 10% to the signal if it were not

removed.

2.4 General surface science tools

A contamination-free metallic surface has to be maintained for the
study of the interaction between adsorbate and metal substrate. A clean
surface is prepared under well controlled conditions to ensure the absence of
foreign matter. This requires the development of an ultrahigh vacuum
chamber to keep clean for a period of time, cleaning tools and recipes inside
the UHV, and also reliable experimental techniques for characterizing the

solid surface. A working UHV system should maintain the pressure on the

order of 1x10'10 torr. That leaves the sample free of the gas adsorption about

one monolayer for 10,000 secs [15].

The UHV techniques primarily concern the pumping system designs,
from roughing pumps for low vacuum to ion pumps or turbomolecular
pumps for ultrahigh vacuum. In order to reach ultrahigh vacuum, a bake-
out process of the metal chamber wall is required to reduce the outgassing
rate at room temperature. In general, stainless steel is chosen as the material
of the chamber for the strength to the pressure difference and the low out-
gassing rate at high vacuum. High quality aluminum alloy can be used
sometimes. The titanium sublimation pump is usually equipped for

removing residual gas molecules after extensive gas dosing in the chamber.

50
The main chamber has to be sealed below atomic leaking level. Knife edge

flanges with soft metal gaskets are used for UHV. Technical information is

available with vacuum product manufacture’s catalogs [16].

Proper surface analysis tools can reveal useful informations for a clean
surface while UHV provides the suitable environment. Standard surface
analysis tools are including: Auger Electron Spectroscopy (AES), Low Energy

Electron Diffraction (LEED), and Temperature Programmed Desorption (TPD).

Auger electron spectroscopy is widely used to determine the chemical
composition of a clean surface. A high energy 2-5 eV electron beam is
directed on the sample and the backscattered electrons of a selected energy
N(E) are collected. A typical Auger process is: the incident electron collides
with an atom and ionizes a ls electron, a 25 electron drops into the hole and

the transition energy ejects a second electron, Auger electron from the 2p

level. This Auger process is called KLL3 . To distinguish Auger electrons with
characteristic energy from scattered electrons which lose their energies in the
inelastic collisions, the derivative signal dN(E)/dE is more sensitive to
identify the peaks from Auger transitions. The electrons with kinetic energies
in the range 15-1000eV have a very short mean free path in matter (<10 A). 50

the signals are from the top layers of the solid and the information obtained is

 

3The atomic levels in an Auger transition are labelled in accordance with conventional X-ray
spectroscopic nomenclature, i.e. K, L, M, for the n=1, 2, 3,... principal quantum numbers of the
atimic shells.

51
about surface properties. The binding energy of a core level electron is a

sensitive function of atomic species. Auger electron energies are well known
and tabulated [17]. Every element with Z > 3 exhibits some Auger decay
within the critical range for surface sensitivity. Since every surface atom
leaves its fingerprint in the kinetic energy spectrum, Auger spectroscopy is
suited perfectly for surface elemental analysis. Using AES data for
quantitative analysis requires a well-calibrated reference system. Otherwise
the information will be limited to identify the contamination species on
surfaces. The sensitivity can usually reach about 1%, i.e., the impurity density

higher than 0.01 ML will show up in Auger measurements.

Low energy electron diffraction is used to observe the surface structure.
The LEED pattern is an image of the surface reciprocal net when viewed along
the surface normal at a great distance from the crystal. The existence of a
sharp spot pattern implies the existence of a well-ordered surface and
provides direct information about the symmetry of the substrate.
Translational invariance in two dimensions ensures the diffraction occurs if
the two-dimensional Laue conditions are satisfied:

(k. —k,)-as =2nm; (k. —k,)-bs =27m,

where ki and k; are the wave vectors of the incident and scattered electron, a5
and b3 are the primitive vectors defining a unit mesh, m and n are integers
[15]. If adsorbates form an ordered structure on top of the substrate, the

expected pattern is found by superimposing the diffraction pattern induced by

52
the adsorbate-overlayer and the substrate diffraction. In principal, one can

decide the overlayer coverage by an observed LEED pattern and use the

information for coverage calibration for ABS and TPD.

After adsorption happens on a surface, desorption becomes one of the
elementary surface processes. The desorption rate of adsorbed species on a
substrate depends on the desorption energy and the temperature. If one
ramps the temperature of the substrate in a controlled way, the desorption
will happen and a measurable quantity, e.g. partial pressure of the adsorbed
species, can be related to this process. By examining the temperature profile
of the desorption, one can either get the information of the desorption energy
[15,18] or integrate the desorption curve to obtain relative coverage
informations for adsorbates. However, quantitative measurements also
require well-calibrated reference system and residual gas analyzers which are

usually subjected to change to different vacuum conditions.

Contaminations on the surface can be removed by heating or ion

sputtering. Sputtering process is performed in low inert gas pressure around
1><10'4 torr with an electron gun to ionize the gas. The ions collide on surface

and remove top layer atoms. In general, after sputtering the surface becomes
rough and annealing is required to restore a atomic flat sample. Gases are
leaked into the UHV chamber through metal seal leak valves by backfilling
the chamber or with specilly designed dosers which enhance the dosage on

sample surfaces [20].

53
2.5 Surface sciencebranch at U4IR beamline, BNL

The basic optical layout of the RAIRS system at the BNL, U4IR
beamline is sketched in Fig. 2.9 [2,3,5] The source radiation is extracted from
the electron storage ring by mirror M1, travels through transfer optics, and is
introduced to the surface science branch platform after M5 mirror. The
opening angle of the beamline is 90 mrad x90 mrad to preserve the source
brightness at long wavelength. M1 is a silicon-carbide mirror coated with a
highly conducting surface, also equipped with water-cooling system to
dissipate the heat generated by the radiation. The ”periscope”, M1 and M2,

sends the beam in the upstream direction to the ellipsoidal M3 which deflects

the beam horizontally by 90°. M4 deflects the beam vertically to another
ellipsoidal M5. An image of the source with a magnification 0.9 is produced

between M4 and MS with a slit assembly for alignment. The ring vacuum of

~10'10 torr is separated from the spectrometer vacuum ~10'3 torr by a piece of
type 2A diamond window with thickness 0.8mm which can transmit the

whole IR region and visible light for alignment. But it produces Fabry-Perot

interference fringes. Measurements with resolution higher than 4 cm"1
cannot be achieved. It is later modified with a CsI window for high
resolution measurements. After the diamond window, the diverging
synchrotron beam is collirnated and directed into the spectrometer. The
collimation is achieved by a combination of a spherical mirror and a plane

mirror.

54
A modified Nocolet 20f Michelson vacuum interferometer is used for

both transmission and reflection measurements. After the interferometer,

the collimated light is focused onto the single crystal sample in the UHV

chamber with f/ 10 optics at 85° angle of incidence. The reflected beam is
collected by an off-axis f/3.5 ellipsoid and directed into the detector. The
sample manipulator is facilitated for the translation motion along x-, y-, and
z-directions and the 360° rotation to reach AES, LEED, sputtering and infrared

positions. The material for the windows on the UHV chamber in the optical
path was polyethylene which permits 80% of light in the range of 10-600 cm'l,
but has strong absorption bands in the mid-IR (600-2000 cm'l). Two detectors
were used to cover the frequency range from 80-2000 cm'lz a boron-doped

silicon bolometer at 4.2 K for low frequency measurements 80 to 650 cm'1 and

a copper-doped germanium photoconductive detector also at 4.2 K for higher

frequency measurements 300 to 2000 cm'l. The thermal equilibrium state of
the detector is crucial for the experiments. The bolometer is equipped with
liquid helium cooled filters to reduce the high frequency radiation
background. There is a Winston cone in front of the detector element which

collects the incident beams within a fixed solid angle as a converging lens.

55

UW Chamber

    

NSLS SOURCE Kinematic Mirror —- I
(Roll-cud Upwardl From Below)

\ Nlcale! 20F Michelson

,Ptom Mina lnlulcromglu

 

Otomond Window

 

 

Elliplicol Minor

M5

to experimental branch

M2 /
/ M4
\ bending radius

1\ electron beam

/
I

 

4’ /
\ infrared radiation
33.5 deg
M1 port

Figure 2.9 The schematic of the infrared beam extraction at synchrotron

light source and the floor plan of surface science branch. (modified from
Ref. [3,4])

56
2.6 RAIRS in physics department, MSU

The basic optical layout of the RAIRS system in MSU is shown in Fig.
2.10 [6]. The source is a conventional silicon carbide globar which is
surrounded by a liquid-nitrogen-cooled housing. The globar is heated

resistively to approximately 1300K by an HP model 6264 B power supply. The

source cryostat is pumped out by an Ultek ion pump to about 1x10° torr and
cooled by liquid nitrogen. An automatic liquid nitrogen feeding device is
installed. The LN2 can is supported by three stainless steel rods which are

kinematically mounted on the bottom of the source chamber. The radiation

is collected, reflected, and focused by an 60° off-axis ellipsoidal mirror to an
exit slit where the tuning fork chopper is located. A plane mirror is also
mounted on a turret mirror mount for reflection of a laser beam coming
though a window from outside of the cryostat into the infrared optic axis. It is
usually out of the infrared optical path and is moved in for alignment. A
tuning fork chopper is chosen for the operation in vacuum and at liquid
nitrogen temperature. The tuning fork chopper is a custom made one from
Multi-scanning System Corp. type RC-2 operating at 800 Hz. The assembly
consists of the fork, a pick up coil, a driving coil, a magnet and a mounting
base. The AC voltage generated by the motion of the vanes in the pick-up coil
is amplified and the amplified voltage is applied to the drive coil in the

proper phase to cause the vanes to vibrate at the fork’s resonance frequency.

57

 

 

l! ran-mares mace mecrmm msrcu norm. ~Cooled

! Spectral rmqe: :00 - woo a-l Grating Spec router
Resolution: I- 5 a—l
Saple lap: 20 - 1200 K

          
 
  
  
 
  

._.. . . . . . eon-om coo-_-

 

 

 

 

 

Ultrdridr Vocqu Sirloce Analysis
Outer uith Single Crystal Saple
on Ute-cooled uripulotor

Uic~Cooled Si :8
lnira’ed Detector

 

 
 

Other striooe probes:
ll! Source 0703“! Low [terry Electron Diffraction
IIUI LII-cooled SIIeIds. W [lectrm hum ‘ x 50 I), [ll
Polmzer and Clara-er "ml Desorption Spectroscopy

Figure 2.10 Optical layout of the MSU RAIRS system. The source,
chopper aperture, sample, spectrometer entrance and exit slit are
conjugate points. (from Ref. [6])

58

The signal from the amplifier is synchronous with the vane motion and used
for a reference for modulation. All the optical components are mounted on
a thick copper plate which forms the bottom of the liquid nitrogen can. the
cooling of the optical components can be monitored through Si diodes

connected with electrical feedthroughs.

There are two transfer optics; one in between the source cryostat and
the UHV chamber, and the other in between the UHV chamber and the
spectrometer. Together they enable the infrared radiation to illuminate the
whole sample with an adequate size of the infrared windows and to be

collected by a spectrometer with a relatively low focal number f/4.2.

The UHV chamber is a custom-built stainless steel ultrahigh vacuum
UHV chamber 12" o x 25” h. The chamber is divided into three levels:

infrared level in the lowest, surface analysis level above the IR level and a
manipulator on the top. The travel of the manipulator can be 1” in x- and y-
directions and 10” in z-direction and rotated 360° with a differentially
pumped rotary seal. Surface analysis tools are equipped on the surface
analysis level. A Perkin-Elmer Model 10-155 Cylindrical-Mirror Analyzer is
installed for AES analysis. A reverse view low energy electron diffraction
optics (Princeton Research Instrument Co.) is used to investigate the surface
structures. A Varian Ion Bombardment Gun, Model 981-2043 is installed for

ion sputtering to get rid of impurities on the surface. A Dycor quadrupole

59
mass spectrometer, Model MA100M, is controlled through R5232 interface to

measure Temperature programmed desorption (TPD) measurements and
residual gas analysis. On the infrared level, there are two differentially
pumped ports for CsI windows and two leak valves connected to the gas doser

[19,20].

The spectrometer and the cryostat were home built at MSU. The
system consists of a spectrometer shell, spectrometer top cover, liquid
nitrogen can, spectrometer base plate and radiation shield. All the optical
components, such as diffraction grating, folding mirrors, slit and filter wheels,

the polarizer, and paraboloidal mirrors, are mounted on the base plate. The

pressure of the spectrometer can be pumped down to 5x10"7 torr. The cryostat
design is similar to the source cryostat. The grating drive mechanism is a
combination of a stepping motor, a controller with a 360:1 worm gear, a 100:1
speed reducer. The speed reducer is engaged or disengaged to the driving
shaft for fine or coarse grating movements using a computer controlled
clutch. Precise angular position reading of the grating is possible through an
lnductosyn rotary position transducer. A complete description is given in

Ref. [6]

The infrared beam, which is modulated by the tuning fork chopper, is
collected by the detector and converted to an electrical voltage signal by the
transirnpedence amplifier in the detector. The detector signal is demodulated

by a PAR 5209 lock-in amplifier and read through a GPIB interface in an IBM

60
AT compatible computer. The computer can control the grating drive by

communicating with the stepping motor controller and lnductosyn to read
and change the grating angle with feedback and also permit precise data
manipulation. The interfacing control program was written in Asyst

programming language by Prof. R. G. Tobin.

2.7 Summary and discussion

This chapter presents general ideas of a generic RAIRS systems. Two
specific systems were described. Basic considerations are presented under
idealized situations. Table 2.1 lists the practical numbers for these two
systems. I have used these two systems to study adsorbate-induced broadband
reflectance changes for atomic oxygen and formate on Cu(100) and CO on
Pt(111). The measured signals are also listed in Table 2.1. Although the U4IR
beamline can access longer wavelength and slightly better sensitivity, the
MSU system is far more stable for broadband measurements. The numbers
given are from the average performing results. The instability from source
fluctuation and cryostat thermal drift are still expected sometimes during
measurments. Global orbit feedback system has been installed in the storage
ring at NSLS to stabilize the source fluctuation recently. And wavelength
modulation is chosen for MSU system to promote the overall performance of
the system. Further improvements in the instruments will bring more

abundant results.

61

Table 2.1 The performance numbers of two RAIRS
MSU.

systems at NSLS and

 

 

 

 

 

 

 

 

RAIRS System U4IR NSLS MSU
Frequency range 80cm'1 to mid-IR 350 cm'1 to 2500 cm'1
Sensitivity 0.02% 0.02% rms
Stability 0.2% over 20 mins 0.02% over 20 mins
ReSOIUtion 1 CHI-1* 1 CHI-1
Adsorption Systems CO/ Cu(100) CO/Pt(111)
O/Cu(100)
Signal measured ~1.1% <0.3%

 

 

 

* measured with a CsI window

 

Refer

s»

 

 

62

References

10.

11.

12.

13.

14.

15.

GP. Williams, Nuclear Instruments and Methods, 195, 383 (1982); W.
D. Duncan and G. P. Williams, Appl. Opt. 22, 2914 (1983).

GP. Williams, Int. J. Infrared Millimeter Waves 5, 829(1984).

G.P. Williams, Conceptual Design Report BNL #35985.

C. J. Hirschmugl, G. P. Williams, F. M. Hoffmann and Y. I. Chabal, Rev.
Sci. Instrum. 60, 2176 (1989).

CJ. Hirschmugl, doctor’s dissertation, Yale University, 1994.

C. Chung, doctor’s dissertation, MSU,1993.
P.L. Richards and R.G. Tobin, in Vibrational Spectroscopy of Molecules

on Surfaces, edited by ].T. Yates Jr. and TE Madey, pp. 413-483 (Plenum,

New York, 1987)

Y. J. Chabal, Surf. Sci. Reports 8, 211 (1988).

M. A. Ordal, R. J. Bell, R. W. A. In, L. L. Long and M. R. Querry,
Applied Optics 24, 4493 (1985).

R. G. Greenler, J. Vac. Sci. Technol. 12, 1410 (1975).

K. C. Lin, R. G. Tobin and P. Dumas, Phys. Rev. B 49, 7273 (1994).
Laboratory Notes in 1993 of Dr. R. G. Tobin’s Lab.

K.-]. Kim, Lawrence Berkeley Laboratory Report (1986).

RJ. Bell, Introductory Fourier Transform Spectroscopy, Academic,

New York (1972).

A. Zangwill, Physics at Surface 5 (Cambridge University Press, 1988)

16.

17.

18.

19.

20.

63
Product and Vacuum Technology Reference Book, Leybold Vacuum

Products, Inc. 1994
Perkin Elmer Physical Electronics, Auger Handbook.
P.A. Redhead, Vacuum 12, 203 (1962).

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

D. E. Kuhl and R. G. Tobin, Rev. Sci. Inst. 66, 3016 (1995).

betwe
resisti
and C:
comps;
systerr

chapte

Chapter 3
Adsorbate-Induced Changes in the Infrared Reflectance and

Resistivity of Metals

In this chapter, I present an experimental test of the relationship
between adsorbate-induced changes in the infrared reflectance and dc

resistivity of a metal. The changes in reflectance for O/Cu(100), CO/Cu(100)

and CO/Ni(100) over the frequency range 300-2000 cm’1 were measured and
compared to the published resistivity changes for the corresponding thin film
systems. These results were previously published in Ref. [1], on which this

chapter is based.

It has been widely known that adsorbates increases the dc resistivity p

of thin metal films [2-4] and there have been several earlier attempts [5-7] to
relate both optical (in visible light region) and resistivity measurements to
scattering of conducting electrons, which modifies the dielectric properties of
the substrate. The adlayer-metal interaction is taken into account through the
p-parameter and the modified dielectric constants. The scattering is found to

be one dominant factor in the experimental observation.

We derive a simple formula to present the relationship based on the
ideas of Persson’s model [8]. We intend to test the scattering hypothesis, by
comparing our IR reflectance measurements with published data on thin film

resistivity changes with this ancillary model. We emphasize that the

64

relatic-
refererl
conne.
resona

arein;

but a I;

3.1

infrare
Nation;

to the:

to freq-
bombar
combin
to 11001
Cu 920-

Was det

 

300Kfo'

 

65

relationship between resistivity and reflectance is very general; it makes no
reference to the damping of the adsorbate’s hindered translation and the only
connection to the proposed model of atomic friction and antiabsorption
resonance is that they all invoke electron scattering. The experimental results
are in good agreement to theoretic predictions for CO/Cu(100) and O/Cu(100),

but a large discrepancy is found for CO/Ni(100).
3.1 Experimental technique

The reflectance experiments were performed using the U4IR far-

infrared beamline at National Synchrotron Light Source at Brookhaven

National Laboratory. The light was p-polarized and incident at an angle of 85°

to the surface normal. We used a Ge:Cu photoconductive detector sensitive

to frequencies above 300 cm'l. The Cu crystal was cleaned using N e-ion

bombardment at 500K followed by annealing to 750K, and the Ni crystal by a
combination of ion bombardment at 600K, oxygen treatment, and annealing
to 1100K. No surface impurity showed an Auger peak larger than 0.3% of the
Cu 920-eV peak or 1.0% of the Ni 100-eV peak. For O on Cu(100), the coverage
was determined from the oxygen Auger peak height, assuming a saturation
coverage of 0.50ML at room temperature. For CO/Cu and Cu/ Ni, the
coverage was determined by TPD and LEED. The sample temperatures were

300 K for CO/Ni(100) and O/Cu(100), and 90K for CO/Cu(100).

66
Fig. 3.1 (a) shows the fractional reflectance change ARp/Rp for two

concessive spectra of a clean surface; the deviation from zero indicates the

level of systematic error. Taking the uncertainty of the current and beam drift
in the storage ring into account, we can determine ARp/Rp to an absolute
accuracy of 1:0.2%. The spectra are composed from the measurements of two
detectors. Spectra (b)-(d) of Fig. 3.1 show ARp/Rp for CO/Cu(100), O/Cu(100)
and CO/Ni(100) over the frequency range from 300 to 2000 cm’l. The low
frequency region in (d) is measured by a Si bolometer. The spectral resolution
was 6 cm'1 for O/ Cu and CO/Ni and 16 cm'1 for CO/Cu. The data above 750

cm'1 have been smoothed to a resolution of 23 cm'l. The noise at the high
frequency region and the absence of data in certain regions in spectra (a), (c)
and (d) are due to the low transmittance of the polyethylene windows.
Cesium iodide windows were used for spectrum (b). The data for CO/ Ni are
almost consistent with zero reflectance change. The reflectance change seems
slightly less than zero but shows no frequency dependence. The magnitude of

the change is within the experimental uncertainty and much smaller than

the observed values for CO/ Cu and O/ Cu. For 0.4 ML CO on Cu(100), ARp/Rp
approaches zero at low frequencies and reaches a constant value of -1.1i0.2%
above 1500 cm'l. This is consistent with a value of -1.9:t:0.1°/o measured at an

incident angle of 87° by Borguet and Dai [9]. For 0.25ML O on

Flg’Ur
CO/r
101-m -

67

 

 

 

 

 

0.5 -
5 3 I 1‘
l.’ -; ; .."
00 _...- - M‘_IWME\II§- - -\,~‘;:“?‘:‘*3§1..:i;
(a) clean '.
~05 * - 1 1 i
0.5 . . . 1
0.0 _. --------------------
(b) CO/Ni
-0.5 * e a
0.5 - .
0.0 c..-.- - - --- _,-_..._, — -.- - -...... - .. .

.0... \ .
-1.0 - W WA:

ARp/Rp(%)

 

.0.—
mu
1

.1

 

00 _.-.-.-....-.-.-.-.......-......._....-.-.-.-.l

 

 

 

0.5 e .
.1 g
> 1 = E ‘I ': "
l S (d) O/Cu . " .-
' o 400 800 1200 1600 2000

Frequency ( crn'l )

Figure 3.1 ARp/Rp for clean Cu surface, CO/Cu(100), O/Cu(100) and

CO/Ni(100) over the frequency range from 300 to 2000 cm'l. (modified
form Fig. 1 in Ref. [1] )

(in

3511

exp

can:

32

corr
treq‘
isst
to a

para

llere

electr
rehac

thESL

68
Cu(100), the reflectance change shows the same characteristic as CO/ Cu. The

asymptotic value is also about -1.1.t0.2% at high frequency.

These reflectance changes are too large and of the wrong sign to be
explained by the dielectric properties of the adsorbates [10]. They must be

caused by the changes of dielectric properties of the metal.

3.2 Surface effect on reflectance

The calculation of surface effects on reflectance requires nonlocal
corrections to the classical Fresnel formulas [11,12]. We mainly focus on the
frequency region where the local description of the metal’s dielectric response
is still valid. The fractional change in the reflectance of p-polarized light due
to a change of surface properties can be calculated using Feibelman’s d-

parameter formalism [12]:

 

ARL=_4(” Im(d,,)

RP c c056 '

Eq. 3.1

provided that ls l >>(1/cosG)>>1, with e the complex (local) dielectric function

of the metal, 0 the angle of incident, a) the frequency of the light, and

d. I = [2(80. , / az)dz/] (80'. I / 3Z)dz Eq. 3.2
Here 0.. is the complex conductivity parallel to the surface for a uniform

electric field, and z is the direction normal to the surface. (Because of strong
refraction by the metal, the electric field inside the metal is nearly parallel to

the surface, even for p-polarized light at near grazing angle incidence [13].)

We 2155

Near t}
allECl t '
tree pa

Per

C)

v

(I)
(I)

value

conduc.

the mc

The co

adsor":

reflect:

reSpoy

ElEClTic

 

 

QIECII'C
llEQUe:
Drude

69
We assume [8] that Im(d 1) is negligible compared to Im(d, I)- The problem

of finding the reflectance change therefore reduces to that of calculating o. .(z).

Deep inside the metal, the conductivity simply has its bulk value 03.

Near the surface, it is modified by the presence of adsorbates. These changes

affect the electron current to a depth on the order of the electron elastic mean
free path lgzvprg (1);: is the Fermi velocity and the TB the bulk scattering time).
Persson proposes a slab model [8] in which the conductivity has a constant

value 05(m)=03(0))+AO'((D) from the surface to a depth 13, and the bulk

conductivity on throughout the rest of the solid. When d I . is calculated from

the model, it follows that:

&=_aea_1_hn[em], Eq....

Rp c cos0 03(0))

The conductivity of a film of thickness 13 in this model is simply as, so if the

adsorbate-induced change in conductivity of such a film is known, the

reflectance change for the same adsorption on a bulk crystal can be calculated.

It is appropriate to use a local description of the metal’s dielectric
response, provided that 5>>uF/co, where 8 is the skin depth [8]; that is the

electric field should not vary appreciably over the distance traveled by an
electron during one cycle of the electric field. For both Cu and Ni in the
frequency range studied, the local response is well described [14, 15] by the

Drude model, in which the bulk conductivity is given by :

 

where

bulk r
princip

in the

respect

 

At zerc

thick

they

70
nezrB
m(1—im3) '

 

03(0)) = Eq. 3.4

where n is the conduction-electron density and m is the electron mass. The

bulk material parameters are listed in Table 3.1. The adsorbate can in

principle affect both n and rnear the surface. We can expand A6 to first order

in the changes An and A1 in the electron density and scattering time,
respectively:

Ao(ro)_§£+ Ar
03(0)) n3 tau-ions)

 

 

Eq. 3.5

At zero frequency, both terms enter equally, but only the term in Ar has an
imaginary part that can give rise to a reflectance change through Eq. 3.3 (A
second-order term in AnAt has been used to model absorption by a surface

state [16,17]) More detailed treatments lead to essentially the same conclusion

[5,7,15].
Ifwe assume that scattering dominates (An/n3<<At/rg), then Ao(a)) can

be predicted from the change in the DC resistivity p=1/o((o=0) of a film of

thickness 13. For (D>>1/1.'B, we find that :

1m[éa_(w_)] = _ 1 [AP (13 ):| Eq. 3.6

as ((9) 601' 8 p3

 

In fact, we can use the DC resistivity change for a film of arbitrary thickness t,
since it is well established that Ap(t)oc1/ t [2,4]. The reflectance change can

therefore be written as:

pretax

limit

 

micro

PIOPC

indep
optica
contri’:
scatter

that d.

can Wr

This is
and 49
dampii
Ea cot
the Sca

Chapted

 

71
fl = 413 APUB) ___ ___4___
Rp CTB c050 p8 chrB c036

 

[tAp(t)]. Eq. 3.7

Only the term tAp refers to the thin-film sample; all of the quantities in the
prefactor relate to the bulk crystal. The refelctance change at high frequency

limit is predicted to be independent of frequency and proportional to tAp. A
microscopic scattering calculation by Watanabe and Hiratuka also found ARP

proportional to Ap [5].

For simplicity, we have derived this result asuming a frequency-

independent scattering time 1'3, but the same equation is obtained if separate

optical (top) and dc (tdc) [15] times are used, provided that the adsorbate
contribution to the scattering rate is frequency independent. Then the DC
scattering timet'dc appears in the denominator. But p31'dc=m/ne2 is a constant

that depends only on the density of conduction electrons in the metal, so we

can write Eq. 3.7:

AR, _4ne2 1

 

 

 

[tAp(t)], Eq. 3.8

Rp mc c059

This is similar to Eq. 40 of Ref. [8]. It can also be obtained by combining Eqs. 41
and 49 of Ref. [8], but the present derivation avoids any assumption about the
damping of adsorbate vibrations. We should note that the resistivity change
is a coverage-dependent term. How the population of the adsorbates affects
the scattering is an interesting question, and we will address it in the next

chapter.

 

33

refle
COUI
Inea
equi
best
Inaxi

hon}

avafle
depei

varie:

 

72
3.3 Comaprison between reflectance and resistivity

We attempt to test Eq. 3.8 by comparing our measurements of IR
reflectance with published data on thin film resistivity changes. It would of
course be preferable to make simultaneous reflectance and resistivity
measurements on well-characterized films deposited in situ, but we were not

equipped to attempt this difficult measurement. In Table 3.2, we present the

best available literature values for the resistivity change tAp, our measured

maximum reflectance change ARP/ RP values, and reflectance change predicted

from Eq. 3.8 using the electron density n values listed in Table 3.1.

For 0 on Cu and CO on Ni, explicit plots of tAp versus coverage 11,, are
available in the literature [3,18,19]. For CO on Cu, the shape of the coverage
dependence Ap(na) is quite reproducible [2,6,7], while tAp at saturation value
varies considerably. Wissmann [20] reports an average saturation value of

tAp (for ~50 measurements) of 1.5x10'12 Qcm2 for films annealed at 60 °C,
while a series of films annealed at room temperature gave values of tAp
ranging from 1.1 to 2.6x10'12 (2cm2 [6]. The effect of annealing temperatures

higher that 60 °C is unknown, as films develop cracks. To arrive at the values
of tAp listed in Table 3.2, we have scaled the coverage-dependence curves by

Wissmann’s average saturation value.

lat
elec
con

deg
elet
div

J

l j

 

l

Table 3.1 Bulk material parameters for Cu and Ni. Values for conduction
electron density n and Fermi velocity 1),: are calculated assuming one

conducting electron per atom [23,24]. The optical scattering time IB and skin
depth are taken from Drude parameters given by Ordal et a1. [14]; the optical

electron mean free path is given by IB=UPTB. Frequencies in s’1 have been
divided by 27m to convert to cm'l.

 

 

 

 

 

 

 

 

 

Metal n 73-1 DF/s 1., 5
(1022 CIn’3) (cm'l) (cm'l) (A) (A)
Cu 8.5 73 310 1140 270
Ni 9.2 350 210 240 400

 

 

lat“.
light I
calci.
thick
9:85

 

 

74

Table 3.2 Adsorbate-induced changes ARp/Rp in the reflectance of p-polarized
light at the adsorbate coverage na as indicated, measured experimentally and

calculated from Eq. 3.8 useing the indicated values of tAp for thin films of
thickness t, the electron density n given in table 3.1, and an angle of incidence

6=85°. See the text for discussion of the uncertainties in tAp.

 

 

 

 

 

 

 

 

 

 

 

System n, tAp Angry)“l A(RP/Rp)exp
(10“ cm'zl (10‘12 (2 cm2) PM PM

7267511000) 6.1 0.603 -2.2 -1.1¢0.2

4.1101"
—C-)/Cu(100) 3.8 0.42c -1.5 -1.1:tO.2

0.70“1 -2.6
"To/Niuom 6.4 1.47" -5.8 0210.2
3 See text

b Reference [9], corrected for the difference in incident angles.
° Reference [3].

d Reference [18].

e Reference [19].

 

exper
mode

small

mode

 

 

greate
mode
for su
for C
a‘PPro
0» Tzl
At 1m

COUEC.

75
Variable sample preparation and uncertainties about thickness and

surface sturcture make all the values of tAp reliable only within about a factor

of two, as listed values for O on Cu suggest. Uncertainty also arises from the
crystallographic orientation of surfaces; the films are polycrystalline, but are
believed to be predominantly of [111] orientation [2], while (100) crystals were
used for the reflectance measurements. This difference in orientation could

affect the strength of electron scattering by the adsorbates.

For the two adsorbates on Cu, both tAp and ARp/Rp are the same within

experimental error. These results must be regarded as consistent with the
model, even though the measured reflectance change is roughly a factor 2

smaller than predicted. This discrepancy could be due to the simplicity of the

model, the uncertainty in tAp or the differences between the surfaces

For CD on Ni, however, the predicted ARp/Rp is nearly a factor 30

greater than the observed value. Uncertainty in n, inadequacy of the slab
model, and deviation from the Drude approximation are not likely to account
for such a large discrepancy -- particularly since the agreement is much better
for Cu. The disagreements cannot be assigned to a failure of the

approximations leading to Eq. 3.8, even though the derivation requires
m>>rg1 and vp/S, conditions that are only marginally met in our experiment.
At 1000 cm'l, (ofi/vp~3 for Cu, and m3~3 for Ni (see table 3.1), the fractional

correction to Eq. 3.8 due to these effects, however, is at most only ~10°/o; these

76

errors are much smaller than the uncertainty in tAp, and are roughly the

same for Cu and Ni. For Ni, unlike Cu, lB<5; according to Ref. [11], this would
render the d-parameter formalism invalid. But the formalism in fact requires

only, IlEl<<8, where I: =(w+i/TB)/vp. For Ni above 1000 cm'l, lE~1)F/(D, so

this requirement simply reduces to the locality condition a1>>vF/5 discussed
above. According to the latest paper of the theory [21], the discrepancy for CO
on Ni is explained by the scaling factor 5/1 which affects the characteristic of

the universal function. Based on the free electron model and under the
assumption of the same specularity caused by diffuse scattering from the
adsorbates, the reflectance change for Ni would be a factor two small than for
Cu. From Table 3.2, the reflectance change for CO on Ni is much less than

0.5%. The scaling effect is still too small to account for the discrepancy.

Another possible explanation is that the reported tAp for CO on Ni is

overestimated. Somehow the film may have been of nonuniform thickness;
there are apparently no more recent experiments to compare with. However,
the discrepancy between theory and experiment is so large that it suggests
that for this system either scattering is not the dominant cause of the change
in the resistivity or the theory relating reflectance and resistivity is seriously
incomplete. Changes in n have been invoked to explain thin film resistivity
changes due to surfaces, defects, and adsorbates, and oxygen incorporated at

grain boundaries in thin Cu films has been shown [21] to reduce n ; such a

mec

Hm

this

3.4

 

77
mechanism could in principle account for the small reflectance change.

However, more experimental evidence is required to resolve the origin of

this discrepancy.

3.4 Summary

I have reported measurements of changes in the broadband IR
reflectance for CO and atomic oxygen on Cu(100) and CO on Ni(100) and
compared them with the changes predicted from published thin-film
resistivity measurements, using a simple model based on the scattering of
conduction electron by the adsorbates. For the adsorbates on Cu, the results
are consistent with the model, but a large and unexplained discrepancy was
found for CO on Ni. Uncertainties in the resistivity data , and doubts about
the comparability of the thin-film and single crystal surfaces, however,
prevent a definitive and quantitative test. There is a clear need for more
experimental work. Reflectance and resistivity measurements should be
performed simultaneously on well-characterized samples, such as epitaxial

films.

Refe

7m

78

References

1.

10.

11.

12.

13.

KC. Lin, R.G.Tobin, P. Dumas, CJ. Hirschmugl, GP. Williams, Phys.
Rev. B 48, 2791 (1992).

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

D. Dayal, H.-U. Finzel, and P. Wissmann in Thin Solid Films and Gas
Chemisorption, edited by P.Wissmann (Elsevier, Amsterdam, 1987).

D. Schumacher, Surface Scattering Experiments with Conducting
Electrons, Springer Tracts in Modern Physics Vol. 128 (Springer, New
York, 1975).

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

U. Merkt and P. Wissmann, Z. Phys. Chem. Neue Folge 135, 227 (1983).
M. Watanabe and P. Wissmann, Surf. Sci. 138, 95(1984).

B.N.]. Persson, Phys. Rev. B 44, 3277 (1991); B.N.I. Persson and A. I.
Volokitin, Chem. Phys. Lett., 185, 292 (1991).

E. Borguet, I. Dvorak and H.-L. Dai, Laser Techniques for Surface
Science 2125, 12 (1994).

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

W.L. Schaich and W. Chen, Phys. Rev. B 39, 10714 (1989), and
references therein.

PJ. Feibelman, Prog. Surf. Sci. 12, 287 (1982).

Y. I. Chabal, Surf. Sci. Reports 8, 211 (1988).

13.

17.

18.

19.

 

14.

15.

16.

17.

18.

19.

20.

21.

23.

24.

79
M. A. Ordal, R. J. Bell, R.W.A. In, L. L. Long and M. R. Querry, Applied

Optics 24, 4493 (1985).

F. Abeles in Optical Properties of Solids, edited by F. Abeles (John Wiley
and Sons, New York, 1966).

].E. Reutt, YJ. Chabal, and SB. Christrnan, Phys. Rev. B 38, 3112 (1988).
D.M. Riffe, L.M. Hanssen, and AJ. Sievers, Phys. Rev. B 34, 692 (1986);
D.M. Riffe, L.M. Hanssen, and AJ. Sievers, Surf. Sci. 176, 679 (1986);
D.M. Riffe and AJ. Sievers, Surf. Sci. 210, L215(1989)

H. Buck, R. Schmidt, and P. Wissmann in Nichtrnetalle in Metallen,
edited by D. Hirschfeld (DGM_Informationsgesellschaft, Wiesbaden,
1988), p.219.

G. Wedler, H. Reichenberger, and G. Wenzel, Z. Naturforsch. 26, 1452
(1971).

P. Wissmann, Thin Solid Films 13, 189 (1972); P. Wissmann (private
communication).

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

J. Vancea and H. Hoffmann, Thin Solid Films 92, 219 (1982).

N. W. Ashcroft and N. D. Mermin, Solid State Physics (Saunders
College, Holt, Rinehart and Winston, Philadephia, 1976).

W. A. Reed and E. Fawcett, J. Appl. Phys. 35, 754 (1964).

broac

ofat.
at to

the I

0.25 l
defin

to be

 

At 0.
predi
devia
Surfac

based

4-1 E)

the \

Chapter 4
Adsorbate-Induced Changes in the Broadband Reflectance

of a Metal: Oxygen on Cu(100)

In this chapter I will present an investigation of the adsorbate-induced

broadband reflectance change for atomic oxygen on Cu(100). Four coverages

of atomic oxygen were measured through the frequency range 100 to 2000 cm'1
at room temperature. The observed frequency-dependence provides a test of
the conduction electron scattering model which incorporates nonlocal
electrodynamics. We find good agreement with theory for coverages below
0.25 ML. The value of the characteristic roll-off frequency (opvp/c, which is
defined through conduction electron properties of the substrate and predicted
to be independent of the adsorbate, is consistent for CO and oxygen on copper.
At 0.35 ML coverage, the frequency dependence deviates strongly from the
predicted form and the magnitude of the reflectance change decreases. These
deviations may be related to an incipient ordering and reconstruction of the
surface. These results have been published as Ref. [1], on which this chapter is

based.

4.1 Experimental Procedures

The experiments were performed at the U4IR far-infrared beamline at

the National Synchrotron Light Source at Brookhaven National Laboratory.

The light was p-polarized and incident at an angle of 85° to the surface

80

81

normal; these conditions are nearly optimal for the observation of reflectance
changes due either to adsorbate vibrations or to changes in surface electronic

properties [2]. Ultrahigh-vacuum-compatible polyethylene windows allow
the spectral range to expand down to 50 cm'1 but have low transmittance
above 750 cm'l. We measured the spectra from 100 to 2000 cm"1 using two
detectors: a Si:B bolometer for frequencies below 650 cm'1 and a CuzGe
photoconductor for frequencies above 300 cm'l. The nominal resolution was

6 cm'1 for the bolometer data and 16 cm'1 for the photoconductor data. During
each run the sample temperature was stabilized to avoid thermally induced
signal drift (a change in the temperature of 5 K produced observable intensity
changes). The data presented here represent runs at varied sample
temperatures between 300 and 330 K; the magnitude of the reflectance

changes were found to be insensitive to the temperature within this range.

The Cu(100) crystal was oriented within 0.5° and mechanically

polished. It was cleaned by 500 eV N e+ ion bombardment at 600 grazing
incidence to increase the efficiency and cover the whole surface more
uniformly. The temperature was held at 550K for annealing during
sputtering and annealed to a temperature between 700K and 850 K. Cycles of
sputtering and annealing were applied until the contamination level was less
than 1% of a monolayer. AES was used to monitor the contamination level.

S, O and C are the most common impurities found on copper surface. No

 

COVEI
lh‘ca
Tiiac

Sh0w

Obsex
Overt
annea
We di

1'51“

 

82
impurity revealed an Auger peak intensity larger than 0.3% of the Cu 920-eV

peak (LMM).

Dosing was accomplished by backfilling the chamber with Oz gas. At
room temperatures 02 dissociates on the surface leaving an overlayer of
atomic oxygen [3-10]. Heating and cooling the sample invariably resulted in
severe baseline drift, so it was not annealed after dosing. The coverage of
atomic oxygen on Cu(100) was determined from AES measurements. The
Auger intensity ratio of the O KLL line (512 eV) to the Cu LMM line (920 eV)
is proportional to the oxygen coverage. The characteristic oxygen coverage vs
dosage curve is shown in Fig. 4.1, which presents our data for the ratio of
Auger intensities of O/Cu as a function of dosage at 300 K. The saturation
coverage for atomic oxygen is about one half monolayer at room temperature
by calibrating the AES signal to a well known adsorption system, oxygen on
Ni according to Wuttig, Franchy and Ibach [5]. For comparison, Fig. 4.1 also

shows their results of the corresponding coverage vs. exposure data.

We also conducted qualitative low-energy electron diffraction (LEED)
observations. At the coverages and temperatures discussed here, no
overlayer ordering was observed, which is consistent with other reports that
annealing is required for optimum ordering. At higher coverages (> 0.45ML)
we did observe a c(2x2) pattern, which according to Wuttig, Franchy and Ibach

is in fact characteristic of the reconstructed phase.

Flgm
fimct
obtai
are Q
COVe

0fth

 

83

 

 

 

 

 

 

 

0.20 v r I T I
5?: O ‘0 ‘ o o o 1 05
o .. O I ‘
v 0K4 ‘. A
5 015 ~ 2 h
(:3 .-‘ . « 0.4 S
5 '0“ g
3 ‘ 009 7 Y -. E
.3 010 P ' -'.3A S0.24 - 0.3 :3
1'! i '6 t 8
m 0'06’ "'5 «016 a
b > . I3 . .1 02 0
t 003.” ' i”
3 0.05 ~. ‘ 39 +0.03] 5
S i l 01
L60 >3 0.000 15 30 ‘ 45 0°00 '
g S

0.0

 

0.00 A l A l l l i l A I
0 400 800 1200 1600 2000
Oxygen Dosage (L)

Figure 4.1 Ratio of the oxygen KLL to copper LMM Auger intensities as a
function of oxygen dosage at 300 K. The coverage scale on the right axis is
obtained by assuming a saturation coverage of 0.50 ML. The solid symbols
are our measurements on two separate days; the open circles show
coverage vs. exposure data from Ref. [5]. The inset shows an enlargement

of the low-dosage region. (from Ref. [1])

4.2 Data Reduction

The intensity of synchrotron radiation is proportional to the electron
current in the storage ring [11]. All the spectra were normalized to the beam
current before analysis. The normalization process is: at the beginning of
each scan, the time was recorded by the data acquisition system. The beam
currents were also recorded periodically. The normalizations were performed
after the experiments were completed. The beam current was derived for the
starting time for each spectrum by interpolation from a least-squares linear fit
to the measured currents. Although the beam decays exponentially due to
loss of electrons in the stored beam, a testing result of this normalization
shows that the reproducibility was better than 0.2% over a 20 min. period
with a CuzGe detector. For a Si bolometer detector, the reproducibility is
slightly worse which results from the strong dependence of its responsivity to
temperature. The overall systematic error of the absolute reflectance

measurements is about 0.2%.

The uncertainty of the beam current and the drift of the stored electron
beam position, however, can also cause intensity variations that mimic
reflectance changes. To exclude such spurious effects, we adopted a stringent
and unbiased system of data selection and analysis, as described in the

following.

85
After the sample had been cleaned and its temperature was stable, four

to eight consecutive reference spectra were measured. Each spectrum
comprised 128 scans and took 55 sec. The sample was closed with oxygen and
another set of spectra was measured. Ratios of the individual, current-
normalized reference spectra with each other were examined, as well as ratios
among the individual post-dosing spectra. If any of these ratios of nominally

identical spectra showed baseline deviations greater than 0.2% between 300

and 2000 cm'l, the entire experiment was discarded. Comparisons of the post-
dosing spectra with the reference spectra were not used in selecting the data.

Approximately one-third of the measurements met the stability criterion,
leaving three photoconductor and two bolometer runs for 0=0.25ML and one

run with each detector at 0.08, 0.17 and 0.35 ML coverages. Because of the long
dosing time required, spectra at higher coverages invariably showed excessive

drift.

The reference spectra (R) and post-dosing spectra (R’) from each run
were averaged. Then the fractional reflectance change AR/R (AR=R’-R) was

calculated. Owing to systematic errors, there was in general an offset (typically

~0.05°/o) between the bolometer and photoconductor spectra in the frequency

range (400-650 cm'l) covered by both detectors. To obtain a final composite

spectrum the bolometer data were offset to agree with the photoconductor

signal. For 0 =0.08 and 0.35 ML, baseline deviations were found below 300 cm'

1 and the low-frequency region was discarded.

86
As a final check against instrumental drift, the oxygen dosage was

increased to 50L and the sample was exposed to 50L of formic acid, which

removed the adsorbed oxygen, leaving a layer of formate species on the
surface [12]. The reflectance change AR/R (relative to the clean surface)

returned to zero within 0.2%. It clearly demonstrates that the reflectance
changes shown are in fact oxygen induced. For oxygen coverages >0.25 ML
formic acid does not completely remove the oxygen and this test could not be

used.
4.3 Experimental results

Fig. 4.3 presents the composite reflectance spectra for four coverages of
O on Cu(100). The gaps in the spectra and the noise at high frequencies are
due to the low transmittance of the polyethylene windows. All show the
same qualitative behavior-- the reflectance change is negative, and its

magnitude is small at low frequencies, increases rapidly between 200 and 600

cm'l, and reaches an asymptotic value above 1500 cm'l. Theoretical

considerations require AR/R to approach zero at zero frequency, as the

wavelength and the penetration depth of the field become infinite [13—16].
We therefore attribute the apparent nonzero intercepts to systematic errors.
The solid lines are fittings to the scattering model and will be discussed in

next section.

87

Figure 4.2 Oxygen-induced fractional reflectance change AR/ R for four
oxygen coverages: (a) 0.08 ML; (b) 0.17 ML; (c) 0.25 ML; (d) 0.35 ML. The
gaps in the spectra and the noise at high frequencies are due to the low
transmittance of the polyethylene windows. The solid lines are fits to Eq.
4.2 discussed in section 4.4; for spectrum (d) no fit was possible.

AR/R

AR/R

AR/R

-0.005

-0.010

-0.005

-0.010 r

-0.015

-0.005

-0.010 -

-0015 '

-0.020

0.005 _

-0.010 -

-0.015
0

 

0.000

 

 

 

l _ 0 : t :7 °‘, 4
(a) e ‘- 0.08 . B ‘ e
.- .1

 

0.000

T

 

 

 

 

3 ‘3; '53!

v‘ :P:

®)0=017 1 ;.
o t

 

0.000

 

¢)e=025

 

 

 

0.000

 

(d) 0 = 0.35

I T I I I o
. I
i O ' Eel
t ‘. ‘
. 3 53 '5
0.. e .’ a ‘
3 t‘. "-‘l' ' -r
.00... U h.‘

t}. ’0. 0” k ‘
.1-r. -.°

0.. ‘ o : .0

8' :

 

 

800 1200

Frequency ( cm'I )

400

1600

2000

COI‘

llllt

the

lllE‘t

4.4

m0<

 

 

the
and

diff

the

w}

the

th.

be

89

The Cu-O stretch vibrations are expected near 340 cm'1 at low coverage

and near 290 cm"1 at higher coverage [4,5,13]. We did not observe a
convincing Cu-O stretch signal despite a sensitivity of 0.02%. This
information would place an upper limit to the dynamic dipole moment of
the vibration. I will make a comparison with other experimental and

theoretical results in section 4.6.
4.4 Frequency Dependence

In a series of papers [14-17], Persson and Volokitin have elaborated a
model of adsorbate-induced-reflectance changes, based onthe assumption that
the electron transport in the near surface region is perturbed by foreign atoms
and the optical response to the incoming IR field drops due to the increase of
diffuse conduction electron scattering by disordered adsorbates. As I discussed

in section 1.4, the universal function for adsorbate-induced reflectance

changes, valid for free electron metals at [R frequencies w>>r'1, is:

 

2
A; —4a (co/a2.) Re[g(w/wpl/5] , Eq. 4.1

— 732 40 ~ dq
1—[7rcop cosellm-L e(w/w,,l/6,q)

where g(x,z) and e(x,z,q) are complex-valued functions defined in Ref. [17], l is

 

the electron mean free path, 0),, is the plasma frequency of the metal, 6=c/(0p is

the classical skin depth, and (01:60pvp/C is a characteristic frequency, near and

below which nonlocal effects become important.

 

 

wh

Sin

Nc
bu

inc

C111

90
One of the theoretical predictions tested here is that only the

magnitude a should depend upon the adsorbate or the coverage; the spectral
shape and the parameter (01 should be properties only of the substrate.
Applying Eq. 4.1 to the data of Hirschmugl et al. for CO on Cu(100), Persson

found a good fit with a value of (01:500 cm'l, on the same order as theoretical

estimates of 444 cm‘1 from a jellium model; an analysis of newer data found a

value of 400 cm"1 [18]

Equation 4.1 is computationally very cumbersome, and we did not
attempt to fit it systematically to our data. Over the frequency range studied

here, it is extremely well approximated by a much simpler expression,

AR ~ 2
"R-=C-a‘—£6—, Eq.4.2
(02 +362

where for l/6=4.2, appropriate for copper at room temperature, the parameter

sin Eq. 4.1 and 4.2 are related by:

C = -—0.00064a ,
a = 0.96a , Eq.4.3
a = 1.21ch1 .

Note that Eq. 4.2 is very similar to the approximation proposed by Persson

building on the work of Dingle [19].

I have performed weighted least-squares fits of Eq. 4.2 to my data,
including an additional offset C’ to account for systematic errors in the

current normalization. The solid lines in Fig. 4.2(a)-(c) show the results and

lahl
Eq. -
COVt
exa:

QUE

COX“.

see-

the
Son
thec
nea
zerc
all?
Wh

 

 

 

inn

for

Wit

V0]

91

Table 4.1 gives the values and the uncertainties of the parameters a and (01 in

Eq. 4.1 obtained through Eq. 4.3 and of the instrumental offset (C'-C), for each
coverage at which a fit was possible. The uncertainties were estimated by

examining the contours near the minimum x3 and for normally distributed
errors represent better than 99% confidence limits [20]. The values of XE

indicate an excellent fit for the two lowest coverages. The fit to the 0.25 ML

coverage is still good, but there is a small but significant discrepancy in the

800-1200 cm" region.

Fig. 4.3 compares Eq. 4.2 (solid line) with Eq. 4.1, using parameters
related through Eq. 4.3 for our data for 0:0.25 ML with C’ subtracted. Within

the accuracy of our measurements, the two curves are indistinguishable.
Some disagreement is apparent at very low frequencies, where the general
theory predicts that the reflectance change goes to zero with nonzero slope but
near-zero curvature. The data, on the other hand, seem to be approaching
zero slope with a distinct negative curvature--similar to the behavior of the
approximate expression. A similar trend is visible in the data for CO [15,21].
While the discrepancies are too small to be conclusive, they may hint at an
incompleteness in the theory at very low frequencies, and suggest the need

for further theoretical work.

For the three lowest coverages the excellent agreement of Eq. 4.1 and 4.2
with the data provides strong support for the model of Persson and

Volokitin.

92

Table 4.1 Best-fit parameter values and uncertainties obtained by fitting Eq.
4.2 to the spectra in Fig. 4.2, and transforming the fitting parameters according

to Eq. 4.3. The weighted average of the (01 values is 3384.34 cm'l. Also listed
are the values of a and ml=mpvp/c found for CO on the same Cu(100) crystal,
and the value of a); predicted from a free electron model.

 

 

 

 

 

 

 

 

O/Cu(100) a(%) 601(cm'1) C’-C(%) 15
coverage
0.08 $47333 5553;: 4.15333 1.06
0.17 4.6333; 3:31:57; -0.163;3§ 1.10
0.25 4.04:3;3 339:3; 4.233% 1.34
Wuflm)
0.4“ -1.23 500
0.5b 4.33 400
Predicted 444c
310d

 

 

 

 

 

 

 

aReference [21] and [16]
bReference [18]
cReference [15], based on one electron per Cu atom.

dReference [24], based on an electron density deduced from optical
measurements.

93

 

0.000

 

 

 

 

0004
g -0.008
<1 .,
-0012 ~ ' " . f ';.
0020 L ~ 1 l '
400 800 1200 1600 2000

Frequency (cm-l )

Figure 4.3 Oxygen-induced reflectance change for 0:20.25, with the
instrumental offset C'-C subtracted. The solid line shows the best fit to Eq.
4.2 and the dashed line shows the results of Eq. 4.1 using the parameters
given by Eq. 4.3. Over the frequency range studied, the two curves are
virtually indistinguishable.

94
The values of (01 for the three coverages are also consistent. (The large
uncertainty in an for 0:0.08 ML is caused by the weak variation of the

reflectance and by the absence of data at low frequencies.) Assuming that (01 is
independent of coverage- as predicted by the theory— the weighted average

for the three curves gives (01:3382t34cm'1. It is surprising that this estimate

differs from the values of 500 and 400 cm'1 found for CO on exactly the same
Cu(100) crystal, since wlzwva/c should be independent of the adsorbate and

the temperature.

The high-coverage (0.35ML) data could not be fit by Eq. 4.2 (The
minimum 13 value was 10.2). Since the form of the frequency dependence
depends only on the assumption that the dominant effect of the adsorbate on
conduction is increased scattering, this deviation suggests that other processes
may be important at high coverages. Within the scattering model (see section

1.4, the magnitude a of the asymptotic reflectance change at high frequencies

(a)>>a)1) is related to the change of the electron relaxation time which results

in the change in dc resistivity Ap of a thin film of the same metal according

to:

_ 4ne2 l

[tAp(t)], Eq. 4.4

 

 

mc c089

where n is the conduction electron density, 111 is the effective mass, t is the

thickness of the metal film, and 6 is the angle of incidence. The quantity tAp

95
is independent of t for a given adsorption system [22,23]. In the previous

chapter, we compared the reflectance change at high frequency (>1500 cm'l)
for single coverages of oxygen and CO on Cu(100) and of CO on Ni with the
values predicted from Eq. 4.4, using published data for the resistivity change
of thin films [25]. Given the large variations and uncertainties in the
resistivity data, and in the comparison between measurements on dissimilar
samples, the agreement was reasonably good for the adsorbates on Cu. For CO
on Ni(100) [24], however, the predicted reflectance change was large (~-6%)
while essentially no reflectance change (<0.4%) was found experimentally. A
qualitative explanation has been offered [16,17] that can account for roughly a
factor of 2 difference between the predicted and measured values, which is
similar to what is observed for CO on Pt and will be discussed in chapter 6, but
the factor of 30 difference is too large to explain. Testing the scattering model
by focusing on the frequency dependence of reflectance change is more
convincing, since it avoids the large uncertainties in the resistivity data. In
the next section, I also examine the dependence of the reflectance change on

coverage.
4.5 Coverage dependence

An important consequence of the scattering model is a qualitative
prediction of the dependence of the reflectance or resistivity change on
adsorbate coverage. At low coverage the adsorbates are randomly distributed,

in the absence of long-range interactions, and the resistivity or reflectance

96
change should vary linearly with coverage. At higher coverage, the

adsorbates may form an ordered overlayer. The effect of ordering on the
resistivity and reflectance is an unsettled issue both theoretically and
experimentally. No diffuse scattering is possible from an ordered surface, but
Bragg scattering will contribute to the resistivity if the unit cell of the
overlayer is larger than that of the clean surface. We would expect the Bragg
scattering contribution to be less than that from diffuse scattering from the
same adsorbate coverage, so that ordering should produce a decrease in the

resistivity and in the magnitude of the reflectance change.

Resistivity measurements on the films sometimes exhibit a maximum
as a function of adsorbate coverage [25], but the resistance does not approach
the clean-surface value and there is no direct evidence associating the
decrease in resistance at high coverage with ordering. The roughness of the
film is difficult to control and therefore the overlayer structure is not well
defined. An IR reflectance study of H on W(100) by Riffe and Sievers [26]
found a decrease in the magnitude of the reflectance change upon ordering, as
expected. For CO on Cu(100), however, Hirschmugl et a1. observed a linear
increase with no detectable change in slope as the ordered c(2x2) structure

formed [27].

Figure 4.4 presents our data for the asymptotic (high-frequency)
magnitude of the fractional reflectance change as a function of coverage. For

the three lowest coverages, the values and uncertainties are those listed in

97

 

 

 

 

1.5 . .
. < 0.16
33: 1.2 e
1
\ L O
3;“ 4* - 0.12
O

30 0.9 l
1 o 1
u "“ - 0.03
3 0.6 ~ 0 '21:." l
C
s +
_23 o
‘15 0.3 ~ ‘ 0'04
05

0.0 I . 0.00

0.00 0.15 0.30 0.45 0.60

Oxygen Coverage (ML)

Resistivity Change An (1 (Milan)

Figure 4.4 Asymptotic (high frequency) limit of the reflectance change as
a function of coverage (filled square). The values and uncertainties are
taken from Table 4.1 for the three lowest coverages, and estimated from

Fig. 4.2(d) for 020.35 ML. For comparison, the open circles show the
resistivity change of a 40 nm-thick copper film (from Ref. [25] ).

98
Table 4.1 for the parameter a, which is the magnitude of the reflectance
change. For 0:0.35 ML, we have made an estimate from the high-frequency

portion of Fig. 4.2 (d). The uncertainty in the coverage includes both errors in

dosing and uncertainty in the coverages vs. dosage curve in Fig.4.1.

A qualitative change in behavior at the highest coverage is apparent
both from Fig. 4.4 and from Fig. 4.2; the magnitude of the reflectance, rather
than continuing to increase, drops about 20% from its value at 0.25 ML
coverage. At the same time, the shape of the frequency dependence
undergoes a marked change, reflected in the inability to fit the 0.35 ML
coverage data, even approximately. We suggest that this change of behavior
is related to the tendency of oxygen on Cu(100) to form an ordered,

reconstructed surface at a coverage close to 0.35 ML. A number of

experiments have established the existence of a (2J2 x «[2- )R45° structure
involving a ”missing row” reconstruction [5-10]: every fourth row of Cu
atoms is vacant. According to the LEED study of Wuttig, Franchy, and Ibach

[5], there is a first-order phase transition from a disordered layer into the
ordered phase at critical coverage 0c=0.34 ML, close to the highest coverage we

were able to study.

Our own LEED observation did not reveal the (242 x J2 )R 45° pattern
at the coverage studied, probably because we were not able to anneal the

overlayer (heating and cooling the sample invariably resulted in baseline

99
drift). Equilibration of the overlayer and formation of the ordered phase are

known to require annealing to at least 400K. It is possible that some partial or
local ordering was occurring at room temperature, which could not be
observed with LEED but had a noticeable effect on the reflectance change.
Qualitatively, we would expect some reduction in diffuse scattering for
ordering on a length scale greater than the electrons, Fermi wavelength,
approximately 5 A for copper; it is therefore plausible that ordering on a
length scale too short to be observed with LEED could have an observable
effect on the reflectance. We present this possibility as a speculation; clearly

additional experiments are needed for a definitive conclusion.

In Fig. 4.4, the coverage dependence of the resistivity change observed
for oxygen on a thin copper film is also displayed. The three lowest coverage
points behave similarly to the resistivity, with the reflectance change
increasing monotonically, but with decreasing slope. The drop in the
magnitude of the reflectance change at 0.35 ML coverage is not mirrored in
the resistivity curve. The resistivity curve does exhibit a smaller (~10%) drop
at higher coverage (> 0.6 ML, not shown in figure). If the drops are associated
with surface ordering, it is not surprising that two curves are different since
the surfaces are dissimilar (the films are heterogeneous, but expected to be
dominantly of [111] orientation [2]). Alternatively, it is possible that the
actual adsorbate coverages in the resistivity experiments were substantially

lower than estimated.

100
4.6 Dynamic dipole moment of oxygen on Cu(100)

We did not observe a convincing Cu-O stretch signal, despite a
sensitivity of 0.02%. Figure 4.5 shows the spectreum for 0.25 ML oxygen on
Cu(100) with the frequency range from 100 and 600 cm’1 after removing a

baseline disscussed in previous sections. Using an estimated linewidth of 50

cm'l, we find the dynamic dipole moment e"<0.12 e. This compares with

values from 0.12 e to 0.18 e estimated from electron energy loss spectroscopy

intensities [28,29], assuming typical values of 1-1.5° for the spectrometer
acceptance half-angle [29,30]. (Details of the calculation formalism are in
Appendix A.) Possibly the lines were broadened in our experiment because
the sample was not annealed, or part of the EELS intensity may result from

impact scattering.

The effective charge value for the Cu-O mode is rather small compared
to other adsorbate-substrate vibrations [29]. This may be understood
qualitatively in terms of screening of the incident electric field by the copper
conducting electrons. Structural studies have found that the oxygen is less
than 0.25 A above the center of the top Cu layer, a position which according to
jellium calculation [31] places it well inside the screening electron density.
However, a theoretical calculation of e" that include the effects of screening
gives much larger values, in the vicinity of e*=1.0 8 [32,33]. This implies

stronger polarization of O—Metal bonding. The large discrepancy between

101

 

 

0.0005 ‘ l ‘ T I I I l . u . v
0.0003 ~ , ; . " ..
r' . . 1
0.0001 ~° 1...“... . . ; ’1'. ".;.;.' -
g -.,' -.--j.j'..:; 15's:
-0.0001 -; " if}? : xyfif 2.
'24:. 5' 2° ° ’ l
-000003 '- .".:: _.
" 0 = 0.25 ML

 

 

0000? 00 ‘ 200 300 400 500 600

Frequency ( cm-l )

Figure 4.5 The low frequency spectrum from Fig. 4.2 (c) for 0.25 ML
oxygen on Cu(100) surface after removing the electron-scattering induced
background reflectance change. The O—Cu stretch is observed around 340
cm".

 

102

experimental results and theoretical predictions merits further theoretical

investigation.
4.7 Conclusion

Ihave measured the adsorbate-induced change in infrared reflectance

over a wide frequency range (100-2000 cm'l) for oxygen on Cu(100) near room
temperature. The data offer a test of a proposed model of surface scattering
that incorporates nonlocal electrodynamics. For coverages below 0.25 ML, the
data strongly support the model. A qualitative change in behavior is
observed at the highest coverage studied (0.35 ML), where the magnitude of
the reflectance change drops and the shape of the frequency dependence
changes markedly. These changes may be associated with an incipient
ordered phase, although annealing the surface to obtain long range ordering
could not be obtained for experimental reasons. Further experimental and
theoretical work is called for to clarify the effects of ordering, the generality of
the scattering model, and the very-low-frequency behavior of the reflectance

change.

References
1. K. C. Lin, R. G. Tobin and P. Dumas, Phys. Rev. B 49, 17 273 (1994).
2. Y.]. Chabal, Surf. Sci. Rep. 8, 211, (1988)
3. A. Apitzer and H. Liith, Surf. Sci. 118, 121 (1981).
4. M.H. Mohamed and LL. Kesmodel, Surf. Sci. 185, L467 (1987) F‘
5. M. Wuttig, R. Franchy, and H. Ibach, Surf. Sci. 213, 103 (1989). i
6. HO Zeng, R.A. McFarlane, and K.A.R. Mitchell, Surf. Sci. 208, L7
(1989).
7. I.K. Robinson, E. Vlieg, and S. Ferrer, Phys. Rev. B 42, 6954 (1990). in:
8. MC. Asensio, M.]. Ashwin, A.L.D. Kilcoyne, D.P. Woodruff, A.W.
Robinson, Th. Linder, ].S. Somers, D.E. Ricken, and A.M. Brawshaw,
Surf. Sci. 236, 1(1990).
9. Ch. W611, R. J. Wilson, S. Chiang, H.C. Zeng, and K.A.R. Mitchell, Phys.
Rev. B 42, 11 926 (1990).
10. R. Mayer, C.-S. Zhang, and K.G. Lynn, Phys. Rev. B 33, 8899 (1986).
11. W.D. Duncan and G.P.Williams, Appl. Opt. 22, 2914 (1983).
12. I. Stohr, D.A. Outka, R.]. Madix, and U. Dfibler, Phys. Rev. Lett. 54, 1256
(1985).
13. BA. Sexton, Surf. Sci. 66, 56 (1977).
14. B.N.]. Persson, Phys. Rev. B 44, 3277 (1991).
15- B.N.]. Persson, Chem. Phys. Lett. 197, 7 (1992).
16. B.N.]. Persson and A. I. Volokitin, J. Electron Spectrosc. Relat. Phenom.

103

 

64/65, 23 (1993)

17.

18.

19.

20.

21.

23.

24.

26.

104
B. N. I. Persson and A. I. Volokitin, Surf. Sci. 310, 314 (1994).

C. I. Hirschmugl, G. P. Williams, B. N. I. Persson and A. I. Volokitin,
Surf. Sci. Lett. 317, L 1141 (1994).

RB Dingle, Physica 19, 729 (1953).

RR. Bevington and D.K. Robinson, Data Reduction and Error Analysis
for the Physical Sciences, 2nd ed. (McGraw-Hill, New York, 1992), pp-
69-70, 258 and 259.

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

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

D. Schumacher, Surface Scattering Experiments with Conducting
Electrons, Springer Tracts in Modern Physics Vol. 128 (Springer, New
York, 1975).

K. C. Lin, R. G. Tobin, P. Dumas, C. I. Hirschmugl and G. P. Williams,
Phys. Rev. B 48, 2791 (1992).

D. Dayal, H.-U. Finzel, and P. Wissmann in Thin Solid Films and Gas
Chemisorption, edited by P.Wissmann (Elsevier, Amsterdam, 1987); H.
Buck, R. Schmidt, and P. Wissmann in Nichtmetalle in Metallen,
edited by D. Hirschfeld (DGM_Informationsgesellschaft, Wiesbaden,
1988), p.21; G. Wedler, H. Reichenberger, and G. Wenzel, Z.
Naturforsch. 26, 1452 (1971).

D.M. Riffe and AJ. Sievers, Surf. Sci. 210, L215(1989).

 

27.

28.

29.

30.

31.

32.

33.

105
C. I. Hirschmugl, Y. J. Chabal, F. M. Hoffmann and G. P. Williams, J.

Vac. Sci. Technol. A 12, 2229 (1994).

S. Anderson and ].W. Davenport, Solid State Commun. 28, 677 (1978).
H. Ibach, Surf. Sci. 66, 56 (1977); A.M. Barc’), H. Ibach, and H.D.
Bruchman, Surf. Sci. 88, 384 (1979).

M. Persson, S. Andersson, and PA. Karlsson, Chem. Phys. Lett. 111, 597
(1984).

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

PS. Bagus and F. Illas, Phys. Rev. B 42, 10 852 (1990).

EA. Calhoun and ].E. Inglesfield, Phys. Rev. Lett. 66, 2006 (1991).

Chapter 5
Infrared Spectroscopy of Formate on Oxygen-Predosed Cu(100):

Broadband Reflectance and Low-Frequency Vibrations

In this chapter I will discuss the adsorbate-induced broadband infrared
reflectance change and low frequency vibrational modes of formate on an
oxygen-predosed Cu(100) surface at room temperature. In contrast to CO, O,
and H on the same copper crystal surface, no broadband reflectance change is
detected over the frequency range 200-2000 cm'l. The difference may be
associated with the lack of a formate-derived. electronic state near the Fermi
level. Two vibrational bands are observed, at 365 and 380 cm‘l, attributed to
oxygen-metal stretch vibrations; previous electron energy loss spectroscopy at
lower resolution detected only a single band. We suggest that the doublet
may indicate the coexistence of two formate species with different binding
sites. These results have been published in Ref [1’], on which this chapter is

based.
5.1 Introduction

Interest in the chemisorption of formate (HCOO) on copper single
crystals arises from its importance in heterogeneous catalysis: Formate is
well-established as a surface intermediate in methanol synthesis and the
catalytic decomposition of formic acid over copper surfaces. This system has

been studied by a variety of surface sensitive techniques such as XPS,

106

107

NEXAFS, etc. and also investigated theoretically [2-17]. However, such basic

issues as the binding site and adsorption geometry are still controversial.

We present a reflection-absorption infrared spectroscopy (RAIRS) study
of formate on oxygen-predosed Cu(100), using synchrotron radiation to span
the frequency range from 200 to 2000 cm'l. There have been previous
vibrational studies of formate on Cu(100) using electron energy loss
spectroscopy (EELS), with lower resolution [1], and a RAIRS study of formate
on Cu(110) by Hayden et al. [5], restricted to frequencies above 1300 cm‘l. To
our knowledge this is the first reported RAIRS study of formate on Cu(100)
and the first far-IR study of adsorbed formate on any surface. We find that the
low-frequency metal-formate stretch vibration previously detected with EELS

is actually a doublet, suggesting that formate may occupy two binding sites.

We are also able to investigate the formate-induced change in the
broadband IR reflectance of the surface. It has been shown that CO [18,19],
oxygen [20,21], and hydrogen [2] all reduce the broadband reflectance by ~1%
changes that are too large and have the wrong sign to be explained by the
dielectric properties of the adsorbate layer [23]. Persson and Volokitin [24,25]
have attributed the change, together with an antiabsorption feature observed
for CO [18], the damping of adsorbate vibrations parallel to the surface and
adsorbate-induced changes in the resistance of thin films, to diffuse scattering
of conduction electrons by the adsorbates. Their formalism incorporates

nonlocal electrodynamics to predict the frequency dependence of the

108
broadband reflectance and the line shape of the antiabsorption feature [25] and

is in good agreement with experiment [19,20], at least for low coverages,
where the adsorbates are randomly distributed. These comparisons between
experimental results and theoretical predictions have been discussed in
previous chapters. We find that formate, unlike these other adsorbates,

produces a negligible change in the broadband reflectance.
5.2 Experiments

The experiments were performed at the U4IR beamline at the National
Synchrotron Light Source at Brookhaven National Laboratory. Details of the
optics have been described elsewhere [26,27]. The light was p-polarized and
incident at an angle of 85° to the surface normal. Polyethylene windows on
the UHV chamber permitted the use of frequencies down to 50 cm'l, but had
low transmittance above 750 cm'l, which resulted in noise and gaps in the
spectrum. We used two detectors: a Si:B bolometer for frequencies below
550 cm'1 and a Ge:Cu photoconductor for frequencies above 300 cm‘l. The
internal vibrational modes of formate, near 760 cm"1 and above 1000 cm‘1 lie
in regions of low window transmittance. As a result, only the Cu-formate
stretch mode was studied. The nominal resolution was 6 cm'1 for the
bolometer data and 16 cm'1 for the photoconductor data. All spectra were
normalized to the electron beam current in the synchrotron ring. During
each run the sample temperature was stabilized to avoid thermally induced

signal drift (a change in temperature of 5 K produced an observable intensity

109
change). The data presented here were taken near room temperature. A

monolayer of formate is reported to be stable up to 400 K [2].

The sample was cleaned by Ne ion bombardment at 550 K, annealed to
about 800 K, then cooled to the experimental temperature. No impurity
showed an Auger signal larger than 0.3% of the Cu 920 eV peak. The formic
acid gas was purified by repeated freeze-pump-thaw cycles. Although the
formate species can also be obtained through direct decomposition of formic
acid on Cu(100), predosing the surface with oxygen increases the sticking
coefficient of formate. In these experiments the sample was first dosed with
50 L (1 L = 10‘6 torr sec) oxygen, giving an oxygen coverage of 0.25 ML, then
exposed to 50 L of formic acid, giving a saturation coverage of formate. The
absolute formate coverage was not determined. Temperature-programmed
desorption (TPD) was used as an additional check on the formate layer

formation.
5.3 Results and discussions
5.3.1 Broadband reflectance change.

Great care is required in the analysis of broadband reflectance data to
avoid systematic errors due to drift of the stored electron beam, uncertainty in
the beam current, instrumental instabilities, and temperature fluctuations.

Our procedures are fully described in chapter 4. Approximately one third of

110
the experimental runs met our unbiased stability criteria and are included in

the analysis.

Figure 5.1 compares the broadband reflectance change for 0.25 ML of
oxygen with the spectrum obtained after subsequent exposure to 50 L of
formic acid. The solid line through the oxygen spectrum is a fit to the
scattering theory of Persson and Volokitin [25], as detailed in chapter 4. The
formate spectrum shown in Fig. 5.1(b) combines the average of four separate
photoconductor spectra with the average of four bolometer spectra; the
averaged bolometer spectrum was offset by 0.07% to match the
photoconductor spectrum between 340 and 400 cm'l. There is more
systematic uncertainty in the formate data than in the oxygen or CO
measurements because of the longer time required for the UHV system
pressure to recover after dosing. Nevertheless it is very clear that replacing
oxygen with formate dramatically reduces the magnitude of the broadband
reflectance change. In fact the formate-induced reflectance change (relative to
the clean surface) is smaller than 0.3%, and consistent with zero. Apparently
formate is a much weaker scatterer of conduction electrons than oxygen, CO
or hydrogen.

It is initially surprising that formate should be such a weak scatterer. It
is stable on Cu(100) to higher temperatures than either CO or H, indicating
relatively strong bonding; theoretical estimates of the binding energy bear out

this conclusion [14,17]. In addition, the formate moiety is expected to carry a

111

0.005 a

 

E O. . .l‘ tfiht‘sw
0.000 ""' - --- "1M 3% _________ L-h!:1-:1

-0.005

AR/R

 

-0.010

 

-0.015 . -. .3.

 

 

-0.020 .1. 1214'!-
0 400 800 1200 1600 2000

Frequency( cm-l )

 

Figure 5.1 Adsorbate-induced fractional reflectance changes (AR/. R) due
to (a) 0.25 ML oxygen on Cu(100); (b) formate on Cu(100) obtamed by
exposing the oxygen-predosed surface to 50 L formic acrda The. data
reduction procedures are discussed in the text. The SOlld line m (a) IS the
best fit to the electron scattering model of Ref. [25]. The gaps m the spectra
and the noise at high frequencies are due to the low transmittance of the
polyethylene windows.

112

large negative charge [14], which might be expected to lead to strong scattering.
A possible explanation is found in the electronic structure. Persson postulates
[24] that scattering is strongly enhanced by the presence of an adsorbate-
induced electronic state overlapping the Fermi level. For oxygen on Cu(111)
and Cu(110), an inverse photoemission study by Jacob et al. [28] shows an
adsorbate-induced state about 3 eV above the Fermi level which could extend
into the Fermi energy region. Several experiments that probe the electronic
structure of adsorbed CO indicate the presence of states within 2-3 eV of E;
[29]. For formate on copper surfaces, however, the highest occupied state
detected with ultraviolet photoelectron spectroscopy is about 5 eV below E;
[11,12]. There appears to be neither experimental nor theoretical information
regarding the unoccupied states. The absence of states near E; may accounts

for formate's weak effect on the surface reflectance.
5.3.2 Copper-formate vibrations

It is established that formate bonds to Cu in a bidentate configuration
through its two oxygen atoms, but the exact binding site and geometry are still
debated [3,4,7,8,9,14,16,17]. No ordered low energy electron diffraction pattern
has been reported for this system, indicating that no long-range ordered
structure is formed. It has been assumed that only one binding site is
occupied, based in large part on EELS measurements showing single bands
corresponding to the Cu-formate stretch, symmetric COO stretch, and OCO

deformation modes [1]. In addition, XPS data showed a single sharp O ls line

 

 

113
[2] and a RAIRS study [5] of formate on the Cu(110) surface showed only a

single COO stretch mode — although close inspection of the spectra suggests
the possibility of a low-frequency shoulder. An apparent doublet in the CH

stretch region was attributed to a combination band [5].

Figure 5.2 shows the average of four high-resolution spectra measured
with the bolometer in the Cu-formate stretch region. There is clear evidence
for a doublet. (Each of the individual bolometer spectra also showed evidence
for a doublet; the lower resolution photoconductor scans, however, were less
consistent.) The lines show the best fit to a sum of two Lorentzians after
removing a linear baseline. The best fit values for the resonance frequencies
of the two peaks are 365 and 380 cm'l, with linewidths of 18 and 13 cm‘l,
respectively, and effective charges of 0.24e and 0.30e, arbitrarily assuming
equal concentrations of 0.5 ML for each species. Both of these frequencies are
considerably higher than observed in the EELS studies, which found the peak
frequency to be 320-340 cm'l. The EELS experiments, however, did not use
oxygen predosing, so the formate coverage was probably lower than in this
experiment and some of the structural studies [3,4,10]. This coverage
difference — or other effects arising from the oxygen predose — may account

for the difference in frequency.

The presence of a doublet suggests the possibility that multiple binding
sites are occupied. The two sites most consistent with experimental and

theoretical results are the "short-bridge" and "cross-bridge" sites [13]. We are

l. ' 4'." ’"J

 

114

 

 

0.03 r .“
0.00 £:;=u-..\._‘ '-‘__._.:Q.‘::;2?‘=’J_:_f_:.m
" .07.‘\\\o . I” ”Z / a.“ ' .
gs -0.03 - .-\‘\.\ 1/ - r
V ' \ ’1
E -0.06 - ‘\° 0.] ..
-009 — \3 g} _. E
, . ‘1 . .. .
‘0.1 2 " J .4

 

 

 

310 330 350 370 390 410 430 450
Frequency ( cm:1 )
Figure 5.2 High resolution (6 cm'l) RAIR spectrum of formate on

Cu(100), showing the bands assigned to the Cu-formate vibration. The
lines show the best fit to a sum of two Lorentzians plus a linear baseline.

115
not aware of any calculation of the vibrational frequencies of formate for

these sites on Cu(100); however Ushio et al. [30] have calculated the
frequencies for two sites on Ni(110). They find a frequency of 368 cm'1 for the
short-bridge site and 252 cm‘1 for the atop site. To the extent that these results
can be carried over to Cu(100), they tend to confirm the identification of the
binding site(s) as bridging, and to suggest that the Cu-formate stretch
vibration is rather sensitive to the binding site. The small frequency
difference (15 cm'l) between our observed peaks therefore suggests sites that
are chemically, and perhaps geometrically, very similar. This similarity could

account for the observation of only a single O ls line in XPS.

Without additional experimental or theoretical information, we
cannot determine the nature of the two sites. One possibility is a mixture of
short-bridge and cross-bridge sites. Another is that all the sites have
essentially the same geometry, but the long-range structure of the overlayer
generates chemical differences due, for example, to antiphase boundaries

between locally ordered regions.
5.4 Summary

We have investigated the chemisorption of formate on an oxygen-
predosed Cu(100) surface with RAIRS, using synchrotron radiation. No
change in the broadband IR reflectance of the surface is detected upon
adsorption of formate species. This suggests that formate scatters conduction

electrons much more weakly than oxygen, CO, or hydrogen; the difference

 

116
may be due to formate's lack of electronic states near the Fermi level. Two

low-frequency vibrational modes are observed, attributable to the Cu-formate

stretch vibration. This suggests a mixture of two different binding sites.

There is a clear need for additional RAIRS experiments on this system,
to investigate the higher frequency internal vibrations, the effects of varying
the coverage and the temperature, and the influence of the oxygen predose.
Theoretical efforts to calculate the vibrational frequencies associated with

different binding sites would also make an important contribution.

117

References

1.’

10.

11.

KC. Lin, P. Dumas, and R.G. Tobin, J. Vac. Sci. Technol. A 13, 1579
(1995).

BA. Sexton, Surf. Sci. 88, 319 (1979); RA. Taylor, P.B. Rasmussen, C.V.
Ovesen, P. Stoltze and I. Chorkendorff, Surf. Sci. 261, 191 (1992).

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

D.A. Outka, RJ. Madix and]. Stéhr, Surf. Sci. 164, 235 (1985).

I. Stéhr, RJ. Madix, D.A. Outka and U. Débler, Phys. Rev. Lett. 54, 1256
(1985).

BE. Hayden, K. Prince, D.P. Woodruff and A.M. Bradshaw, Surf. Sci.
133, 589 (1983); Phys. Rev. Lett. 51, 475 (1983).

A. Puschmann, J. Haase, M.D. Crapper, C.E. Riley and D. P. Woodruff,
Phys. Rev. Lett. 54, 2250 (1985).

MD. Crapper, C.E. Riley and DP. Woodruff, Surf. Sci. 184, 121 (1987).
MD. Crapper, C.E. Riley and DP. Woodruff, Phys. Rev. Lett. 57, 2598
(1987).

MD. Crapper, C.E. Riley, D.P. Woodruff, A. Puschmann and I. Haase,
Surf. Sci. 171, 1 (1986).

DR Woodruff, C.F. McConville, A.L.D. Kilcoyne, Th. Lindner, I.
Somers, M. Surman, G. Paolucci and A.M. Bradshaw, Surf. Sci. 201, 228
(1988).

Th. Linder, I. Somers, A.M. Bradshaw and GP. Williams, Surf. Sci. 185,

75 (1987).

 

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.
23.
24.

25.

26.

118
P. Hofmann and D. Menzel, Surf. Sci. 191, 353 (1986).

LS. Caputi, G. Chiarello, M.G. Lancelotti, G.A. Rizzi, M. Sambi and G.
Granozzi, Surf. Sci. 291, L756 (1993).

SP. Mehandru and AB. Anderson, Surf. Sci. 219, 68 (1989).

].A. Rodriguez and GT. Campbell, Surf. Sci. 183, 449 (1987).

A. Wander and B.W. Holland, Surf. Sci. 199, L403 (1988).

M. Casarin, G. Granozzi, M. Sambi, E. Tondello and A. Vittadini, Surf.
Sci. 307-309, 95 (1994).

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

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

KC. Lin, R.G. Tobin and P. Dumas, Phys. Rev. B 49, 7273 (1994).

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

GP. Williams, private communication.

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

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

B.N.]. Persson and Al. Volokitin, J. Electron Spectrosc. Relat. Phenom.
64/65, 23 (1993); Surf. Sci. 310, 314 (1994).

GP. Williams, Int. 1. Infrared Millimeter Waves 5, 529 (1984); Nucl.

Instrum. Methods A 291, 8 (1990).

 

27.

28.

29.

30.

31.

32.

119
GP. Williams, C]. Hirschmugl, E.M. Kneedler, E.A. Sullivan, and DP.

Siddons, Y.J. Chabal, F. Hoffmann and K.D. Moeller, Rev. Sci, Instrum.
60, 2176 (1989).

W. Jacob, V. Dose, and A. Goldmann, Appl. Phys. A 41, 145 (1986).

J. Rogozik, V. Dose, K.C. Prince, A.M. Bradshaw, P.S. Bagus, K.
Hermann, and Ph. Avouris, Phys. Rev. B 32,4296 (1985).

K.-D. Tsuei and PD. Johnson, Phys. Rev. B 45, 13 827 (1992).

R. Courths, B. Cord, H. Wern, H. Salfeld, and S. Hiifner, Sold State
Commun. 63, 619 (1987).

J. Ushio, I. Papai, A. St-Amant and DR. Salahub, Surf. Sci. Lett. 262,

L134 (1992).

 

Chapter 6
Adsorbate-Induced Reflectance Change of a Metal:

CO on Pt

Ipresent an investigation of the coverage dependence of the infrared
reflectance changes induced by CO adsorption on Pt(111) she near-surface
conductivity of a metal by the adsorbates causes a decrease in the

reflectanceurface at room temperature. The modification of t. The

magnitude of the reflectance change AR/R, at room temperature and high

frequencies, increases monotonically at low CO coverages, peaks at 0:033 ML,
and then decreases toward saturation coverage. The monotonic dependence
is explained by a model of conduction electron scattering from the randomly
distributed adsorbates. Possible explanations for the nonmonotonic behavior
include coherent scattering from a partially ordered overlayer or a coverage-
dependent scattering cross section. The maximum reflectance change is
0.28%:t0.03%, a factor of three smaller than 1.1% observed for oxygen or CO

on Cu(100). The discrepancy can be explained by the scaling factor of 1/6, the

ratio of skin depth and electron mean free path. This chapter is largely based

on a paper to be published in Chemical Physics [1’].
6.1. Introduction

CO on Pt(111) has been extensively studied due to the practical

application to the catalytic oxidation of carbon monoxide. On Pt(111), CO is

120

121

adsorbed non-dissociatively with the carbon atom bonded to the surface in a
linear (on top) or a bridged configuration depending on the coverage. The
internal C=O and external Pt-C stretching modes for CO on Pt(111) are well
studied by infrared spectroscopy [1-3] for the complexity of the coupling and
energy transfer between vibrational and electronic states of the adsorbates and
substrate. Even without strong coupling, adsorbates can affect the electron
dynamics in the near surface region as impurities from which conduction
electrons will scatter without conserving the parallel momentum. These
scattering events cause an increase in the dc resistance since the adsorbate
perturbs the electrical transport to a depth on the order of the electron mean-
free-path I, typically ~100 A, even though the scattering potential is screened
within a few A [4]. It was also recognized [5—8] that the diffuse scattering on
the surface would reduce the infrared reflectance. The observation is difficult
to achieve because the broadband reflectance measurements demand that the
absolute intensity be held constant to an accuracy considerably better than 1%
between the measurement on the clean surface and that on the adsorbate-
covered surface. Vibrational measurements on the other hand, are relatively
insensitive to change in signal level as long as the background remains

relatively smooth.

Recent experiments [9-21] and theoretical developments [22-26] have
revived interest in these scattering effects as they relate to adsorbate-substrate

dynamics. A number of careful infrared studies on single crystals in

 

 

122
ultrahigh vacuum (UHV) have revealed adsorbate-induced decreases (~1%)

in the broadband reflectance of metal surfaces that I have discussed in
previous chapters. In this chapter, I mainly report the coverage-dependent
adsorbate-induced reflectance changes for CO on Pt(111) and discuss some

possible explanations for the observations.
6.2 Experimental

The single crystal platinum sample was purchased from Aremco, Inc.
The original sample was rectangular, with dimensions 3 cm wide x1 cm high

x0.1cm thick. We cut off the sample edge at one corner by 1.5 cm x 0.3 cm in

a triangular shape due to the presence of a domain misoriented by 9° from the
[111] direction. The sample was repolished mechanically with 0.5 micron

diamond paste. The normal to the sample surface of the remaining crystal is

oriented at 27° from the [111] direction, as determined by the back reflection

Laue X-ray technique.

The crystal was also examined with low energy electron diffraction
(LEED). The LEED pattern shows a doublet that is from the ordered step
structure. The intra-doublet separation is inversely proportional to the
distance between steps, so the terrace width can be obtained by comparing the
doublet separation with a major spot separation. The terraces are 21-atoms
wide as calculated from the doublet separation of the LEED pattern. The
intensity distribution within the doublets can be obtained as a function of step

height from kinematic analysis [27]. The calculated step height is 2.12 A.

 

123
From the terrace width and step height, the normal to the sample surface is at

273° from the [111] direction, which is in good agreement with Laue X-ray
diffraction. We don't expect the 5 % density of step sites resulting from the
miscut to affect our results; at the coverages studied the step sites should be
saturated with CO, and the change in scattering rate due to the terrace should
be independent of the presence of scattering centers on the step edges

(Mathiessen's rule).

The sample was mounted at the end of a stainless steel tube which can

serve as a liquid nitrogen reservoir for cold fingers, in UHV at a base pressure

S 1x10'10 torr. The sample was heated radiatively by a tungsten filament and
the temperature was measured by two chromel-alumel thermocouples spot-
welded to the edge of the sample. The sample could be closed with various
gases by means of a leak valve and a multihole effusive array [28], which

provides a flux enhancement of about 16 over background dosing.

The primary contaminants in the crystal were carbon and calcium.
Cleaning involved cycles of argon ion sputtering, annealing to 1170 K,
exposure to oxygen at various elevated temperatures, and heating to 1360 K
(to remove the atomic oxygen). After numerous cycles of sputtering and
oxygen treatments, residual concentrations of C, Ca and O could be reduced to
less than 3% of one monolayer (the noise level of ABS, i.e. 4% of the Pt 237 eV
peak) and the integrated intensity of the oxygen TPD peak reached a

maximum.

 

124
The spectroscopic system was the MSU instruments descirbed in

chapter 2 [3]. A conventional silicon-carbide globar was used as the IR source,
which was intensity modulated with a tuning fork chopper. The source

housing, chopper, and source optics were cooled to liquid nitrogen

temperature in high vacuum. The IR light was incident on the sample at 860
through a Csl window [29]. The reflected IR light passed through a second CsI
window and focused on a home-built Czerny-Turner type infrared grating
spectrometer housed in high vacuum and cooled to liquid nitrogen
temperature to reduce the background photon flux. After the
monochromator, the p-polarized component of the light is selected and

detected by a Si-B photoconductor operated at liquid He temperature.

The data taking procedure went as follows. First the sample was
flashed to 975 K to clear the surface of a passivating layer of CO. The
spectrometer was then set to a single frequency position and the IR reflected
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 6.1 shows the change in the reflected IR intensity versus time
for a single measurement, normalized to the initial value, with a systematic
linear baseline drift subtracted. The uncertainty was determined form the

slope of the baseline drift prior to closing. This particular run shown exhibits

 

125

 

0.0005 "' .4

0.0000 '1 '1‘ '1‘ ..'.1H."Hl'lll‘..'."l -

 

 

 

 

 

 

 

 

 

 

~0.0005 - a
% -0.0010 - a
- 0.42:t0.03 ML CO/Pt(100) .
-0.0015 - 2800 cm'l -
315 K n
1- . 1 ll 1 l I n.
-0.0020 [ll “I”. “l ‘
.
-0.0025 - -—>1 dosing 4_ ..
0 100 200 300 400 500

Time (s)

Figure 6.1 Fractional change in reflected IR intensity from Pt(111) as a
function of time at a frequency of 2800 cm'1 and room temperature. The
Pt(111) sample was exposed to CO for 100 seconds at about the 200th
second. Background pressure during dosing was 8x10’lo torr, while the
pressure at the sample was increased by a factor of 16 due to the effusive
array doser. A linear drift has been subtracted.

126
an extremely stable baseline. In most cases, the baseline drifts ~ 0.2% over the

course of a run and the accuracy of the reflectance change measurement is
limited. Usually, the drift was quite linear and AR / R can be measured to an

accuracy of 0.02% or better. During this run the sample was closed with CO for
100 seconds beginning at about the 200th second at a background pressure of
8.0x10‘10 torr. The fractional change in the reflected IR intensity for this run

was 0.00200i0.00002.

It is of interest to compare the sensitivity of this home-built system for
broadband reflectance measurements to that of the U4IR beamline at
National Synchrotron Light Source of Brookhaven National Laboratory,
where most of the previous studies have been carried out [12—16, 20].
Although the IR flux from the synchrotron radiation is greater than the
traditional IR source in our system, electron beam movement and
uncertainties in beam current limit the sensitivity to 0.2% [14, 20], about an
order of magnitude above our level. It would be extremely difficult to
investigate the reflectance change of the adsorption system such as CO on
Pt(111), which have a maximum magnitude less than 0.3%. A significant
drawback of our system is that we only measure a single frequency in each
run, whereas the FTIR spectrometer at NSLS obtains the entire spectrum at

once.

The reflectance changes are measured for various coverages at 2500 and

-1 .
2800 cm where the reflectance changes are predicted to reach an asymptotic

127

value within the scattering model [24,25], i.e., AR/ R becomes frequency-
independent for a)>> 1'1, (01. As shown in Table 6.1, 1'1 corresponds to at most

970 cm'1 while (01 is considerately lower. Also the absoprtion due to the

vibrational resonances is less than 0.003% from the Lorentzian oscillator

approximation. The closest resonance peaks are the internal stretch of CO on

Pt at ~1850 ch for bridge-bonded CO and at ~2100 cm.1 for atop-bonded CO

[1.3]-

The CO coverage was determined with reference to the C=O stretching

frequency of atop CO. When measuring AR / R at a certain frequency, we first

measured the clean-surface reflectance from 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 CO stretching vibration. The
saturation coverage at room temperature was assumed to be 0.5 ML [2] and
the rate of change of frequency with coverage was taken from Fig. 3. of Ref. [1],
which is a compilation of published data. Uncertainty in the actual room
temperature saturation coverage of our sample could introduce an overall

shift in the coverage scale.

128

Table 6.1. Bulk material parameters for Pt.

 

 

 

 

w,(cm'1) 011cm") 1 1cm") 1 (A) 5030
Opticala 41,500 133 558 91 380
[)(:b 77,000 372 968 79 207

 

 

 

 

 

 

aDrude parameters calculated from optical data [30].

bConduction electron density based on one electron per atom; scattering time
calculated from DC resistivity measurements [31].

6.3 Results and Discussion: Coverage Dependence

In Fig. 6.2 (a), the fractional reflectance change AR/ R is shown as a

function of CO coverage for CO on Pt(111) at 315 K at both 2500 and 2800 cm”1

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. The overall coverage-
dependence is similar for O on Cu(100) (see chapter 4). The magnitude of the
reflectance in that case increased monotonically up to 0.25 ML, but decreased

by about 20% when the coverage increase dto 0.35ML.

 

129

Figure 6.2. (a) Fractional reflectance change AR/R as a function of
coverage for CO on Pt(111) at 315 K. Data shown were taken at two IR
frequencies, 2500 and 2800 cm'l. (b) Work function change for CO on
Pt(111) from Refs. [34] (triangles) and [37] (circles).

AR/R

A<I>(eV)

130

0.0000 . , . T . j ,

 

_l

 

-0.0005 — _._ 2800 cm"
+ 2500 cm"

 

 

 

-0.0010 '-

i—L—i
-0.001 5 '- +

t—l——-1
5—1
-00020 - HEL-
” l—iF—Fj a 1
I 1 I

 

 

 

- 1— H
0.0025 Egg:
-00030 — (a) i

1 I 1 I 1 I 1

 

 

‘ I ' I ' T T I ' I

0.00 ~ ‘

-0.04 - ‘

-0.08 r A

-0.12 - O

-0.16 - .

 

_0.20 1 I 1 I 1 I 1 I 1 I

 

 

0.0 0.1 0.2 0.3 0.4 0.5
CO coverage (ML)

131

For a dilute, disordered overlayer, the scatterers are independent and
the scattering of electrons will be incoherent-- the scattering rates from
different adsorbates will simply add. If the scattering cross section per
adsorbate is constant, the resulting reflectance change will vary linearly with
adsorbate coverage. A linear dependence is observed at low coverages [10,
12,13,15, 16, 18, 19]. For CO on copper surfaces, the linear dependence persists

up to saturation coverage [13, 16].

For other systems, however, nonmonotonic dependence at higher
coverages have been observed. Riffe et al. [9, 10] measured the attenuation of
surface electromagnetic waves (SEW) for N2, 0;, CO, H2, and D; on W(100).
The magnitude of the SEW attenuation coefficient a increased monotonically
with coverage for N2, while for both O2 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 study of H on W, Riffe and
Sievers correlated the coverage-dependence of the reflectance change with
ordering of the overalyer, at a frequency where the reflectance change was

dominated by scattering [17] rather than electronic state effects [11].

At least two explanations can be offered for a nonmonotonic coverage-
dependence. The scattering cross section per adsorbate could vary with
coverage, either because different binding sites are being occupied or due to

coverage-dependent changes in the electronic structure of the adsorbates.

132
Alternatively, ordering of the overlayer could reduce 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 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 the scattering amplitudes are in any case different, the reflectance
need not return to its clean surface value [16]. On the other hand, it can be
expected to be different from the reflectance from a disordered overlayer of
the same surface density. Indeed, Riffe et al. interpreted their observations in
terms of ordering, and found correlations between the observed coverage-
dependence and adlayer patterns observed in separate low energy electron
diffraction (LEED) experiments [32]. For O on Cu(100) surface, although long
range order observable with LED could not achieved under our
experimental conditions, the reflectance drops at a coverage where formation
of an ordered overlayer structure is suggested by detailed studies of the

overlayer structures [15].

It is interesting to note that 0.33ML is the coverage at which the CO

overlayer forms a (43 Xw/3)R30° ordered structure on Pt(111) at 100 K, as

observed by LEED [32]. At 0.5 ML the overlayer forms a c(4 x2) ordered
structure [32]. It is tempting to attribute the decrease in AR/R to ordering.

Since long-range order cannot be achieved at room temperature, however,

133
the effect would have to be assigned to partial, short-range ordering, for

which there is no direct experimental evidence.

Another possible explanation for the nonmonotonic coverage-
dependence is a coverage-dependent scattering cross section caused by changes
in the chemical bonding of the CO. It is well know that the CO occupies both
atop and bridging sites on Pt(111) and that the relative populations of the two
sites are coverage-dependent [34,35]. These two species have distinctly
different electronic state and dipole moments [35]. So it is plausible that they

might have different scattering cross sections for conduction electrons. Site
exchange alone, however, can not account for the drop in AR/R seen in Fig.

3.2 (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:

AR
——R—(0.5ML) — 0.25

%(0.33ML) 0-33

 

= 0.76 Eq. 6.1

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 0.5 ML [34, 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 measurements of the work function

 

134
change Ad) [34, 36, 37], which show a coverage dependence strikingly similar

to that of AR / R, as shown in fig. 6.2 (b) where the work function data of Ref.
[34,37] are reproduced. (It should be noted, however, that the coverage scale
for Ref. [34] was set by assuming that the minimum value of Ad) occurred at
0.33 ML; in Ref. [37] the coverages were determined directly from TPD and
LEED). The rapid increase in Ad) between 0.33 and 0.4 or 0.5 ML is attributed

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 Fig. 6.2 (a), but the ordering hypothesis

cannot be ruled out.

The magnitude of the fractional reflectance change AR/ R for p-

polarized light on the surface is strongly dependent on the substrate

properties and can be written [24]:

£ _ 301:0" __P_)
(R )_— __f(w/w1'l/6)'4ccose Eq.6.2

where p is the specularity parameter, which is a phenomenological

parameter to describe the specularity of the scattering events of the
conduction electrons on the surface ,0 is the angle of incident of the light and

Up is the Fermi velocity in the metal, which within the Drude model depends

only on n. The universal function for the frequency dependence fan/(1)1 ,1/5)

depends on the frequency a) relative to 01:13:60,, / c and the ratio of the electron

135

mean-free—path I to the classical skin depth fizc/a)P. Near the characteristic
frequency (01 an electron travels a distance comparable to the skin depth

during one cycle of the magnetic field; (1); therefore specifies the frequency
range below which nonlocal electrodynamic effects are important [22-24]. As
I have pointed out the parameters that enter into f do not involve properties
of the adsorbate; the frequency dependence depends only on the metal, while
the specularity depends on the adsorbate-induced surface scattering. Under

the same influence of the adsorbates, i.e. same p values, different metals
exhibit different 1/5 values which scale the function f and affect the

magnitude of the reflectance changes at high frequency limit. The calculated

curves for these characteristic function are plotted in Fig. 6.3. from Ref. [24].

For copper, 1/8 is about 4.0 and the maximum reflectance change observed is

~1.1%. For Pt, 1/5 is about 0.2 and the maximum reflectance change observed

is ~0.3%. That explains the large discrepancy of the reflectance changes shown
between copper and Pt. This is also a characteristic due to the anormalous

skin effect for metals.

136

 

 

 

 

E 2.5 1 1 r

5 0

e ./

.1: 2.0 — 0/ -1

“é /

«3 Pt

: 0 Cu

E 1.5 " “

g}, 1.0 — ° -

:1

C6

.5:

0

g 0.5 - -

S:

«3 O

8

<1)

31:, 0.0 . I .

9‘ 10'2 10'1 100 101 102
Scaling factor 1/5

at a frequency of 60/61), = 4

Figure 6.3 The reflectance change AR/R vs. US, the ratio between the

bulk electron mean free path and the skin depth, at a frequency 60/(11)l = 4
where the reflectance change magnitude approach to an asymptotic value
for the universal frequency-dependent function. (reproduced from Ref.

[24])

137
6.4 Summary
We have measured the adsorbate-induced change in the broadband IR
reflectance change of Pt(111) as a function of CO coverage. The IR reflectance
decreases when CO adsorbs on the Pt(111) surface. After increasing with

increasing coverage for low coverages, the magnitude of the reflectance
change peaks at a coverage of about 0 = 0.33 ML, and decreases toward

saturation. The presence of a peak in the coverage dependence could be
attributable to ordering in the overlayer or changes in the electronic structure
of CO-Pt bonding. The observed reflectance change for CO/Pt is a factor of

three smaller than that measured for O/Cu. The discrepancy can be

qualitatively explained by the scaling factor 1/5.

138

References

1.’

10.

11.

12.

13.

14.

DE. Kuhl, K.C. Lin, Chilhee Chung, J.S. Luo, Hong Wang, and R.G.
Tobin, Chemical Physics, in press.

J.S. Luo, R.G. Tobin, and D.K. Lambert, Chem. Phys. Lett. 204, 445 (1993).
1.]. Malik and M. Trenary, Surf. Sci. 214, L237 (1989); LR Sutcu, J.L.
Wragg and H.W. White, Phys. Rev. B 41, 8164 (1990).

C. Chung, Dr. thesis, MSU, 1993, references there in.

 

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

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

G. E. H. Reuter and E. H. Sondheimer, Proc. R. Soc. London, Ser. A 195,
336 (1948).

R. B. Dingle, Physica 19, 311 (1953)

R. B. Dingle, Physica 19, 729 (1953).

D. M. Riffe, L. M. Hanssen, and A. J. Sievers, Phys. Rev. B 34, 692 (1986).
D. M. Riffe, L. M. Hanssen, and A. J. Sievers, Surf. Sci. 176, 679 (1986).

J. E. Reutt, Y. J. Chabal, and S. B. Christman, Phys. Rev. B 38, 3112 (1988.
C. J. Hirschmugl, G. P. Williams, F. M. Hoffmarm, Y. J. Chabal, Phys.
Rev. Let. 65, 480 (1990).

C. J. Hirschmugl, Y. J. Chabal, F. M. Hoffmann, G. P. Williams, J. Vac.
Sci. Technol. A 12(4), 2229, (1994).

K. C. Lin, R. G. Tobin, P. Dumas, C. J. Hirschmugl, and G. P. Williams,

Phys. Rev. B 48, 2791 (1993).

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28

29.

139
K. C. Lin, R. G. Tobin, and P. Dumas, Phys. Rev. B 49, 17273 (1994).

C. J. Hirschmugl, G. P. Williams, B. N. J. Persson, A. I. Volokitin, Surf.
Sci. Let. 317, L1141 (1994).22.

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

E. Borguet, J. Dvorak, and H. L. Dai, Laser Techniques for Surface
Science, 2125, 12 (1994).

M.Hein and D. Schumacher, J. Physics D Applied Physics, 29, 1937
(1995).

C.L.A. Lamont, B.N.J. Persson and GP. Williams,Chem. Phys. Lett., 24,
429 (1995).

J. Krirn, D.H. Solina and R. Chiarello, Phys. Rev. Lett. 66, 181 (1991).
B.N.J. Persson, Chem. Phys. Lett. 197, 7 (1992).

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

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

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

J. B. Sokoloff, Pys. Rev. B, to be pulished.

M.B. Webb and MG. Lagally, in Vol. 28 of Solid State Physics, edited by
H. Ehrenreich, F. Seitz and D. Turnbull, p.301 ( Academic, New York,
1973).

D. E. Kuhl and R. G. Tobin, Rev. Sci. Inst. 66, 3016 (1995).

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

(1994).

30.

31.

32.

33.

35.

36.

37.

140
M. A. Ordal, R. J. Bell, R. W. A. Jr., L. L. Long and M. R. Querry, Appl.

Opt. 24, 4492 (1985).

J. Bass and K. H. Fischer, in Landolt-Bornstein Numerical Data and
Functional Relationships in Science and Technology, 15a, edited by K.
H. Hellwege (New York, 1982), p. 63.

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

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

P. R. Norton, J. W. Goodale, and E. B. Selkirk, Surf. Sci. 83, 189 (1979).
W. D. Mieher, L. J. Whitman, and W. Ho, J. Chem. Phys. 91, 3328 (1989).
K. Horn and J. Pritchard, J. Phys. (Paris), Colloq. C4, Suppl. No. 10, 38,
164 (1977).

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

Chapter 7

Conclusion

In this work, Ihave studied the adsorbate-induced reflectance changes
on metal surfaces. Experimental results support the scattering model that
suggests the reduction in reflectance and increases in surface resistivity are
caused by conduction electron scattering. While the frequency-dependence of
reflectance change reveals the conduction electron properties, the magnitude
of the reflectance change provides information on the adsorbate-substrate

interactions.

The direct relation between reflectance change and resistivity change
for O and CO on Cu(111) surfaces is proven within satisfaction by comparing
the measured reflectance change and published resistivity change on thin
metallic films under an assumption that the change in electron relaxation
time is much larger than the change in conduction electron density. For CD
on Ni, the reflectance change is much smaller than expected from resisitivity
measurements, which can not explained by the scaling factor that implies the
interaction between probing light and conduction electron and is the ratio of
the classical skin depth to the electron mean free path. The discrepancy may
result from different experimental condition and is subject of further
investigation: that is the simultaneous optical and resistance measurements

on well-defined surfaces.

141

142

The adsorbate-induced reflectance change is also characteristic of the
interaction between conduction electrons and the incident electric field. The
frequency dependence of the reflectance change shows the local-nonlocal
transition, which is predicted by calculating the induced current due to the
probing light from the distribution function of conduction electrons under
the influence of the diffuse scattering from adsorbates. The characteristic
frequency that determines the roll-off of the reflectance change depends only
on the conduction electron properties. The magnitude of the reflectance
change indicates the strength the interaction, i.e., scattering, between
adsorbate and substrate and can be expressed through the specularity
parameter. For oxygen on Cu(100) at coverages below 0.25 ML, the reflectance
changes are well fitted by the universal form. At higher coverage, the
frequency dependence can no longer be fitted by the theory and the
magnitude of the reflectance changes and coverages are no longer in a
monotonic relation. Possible explanations are: the formation of an ordered

overlayer formation and the decrease of the effective scattering cross section.

Upon exposure of an oxygen-predosed copper surface to formate, the
reflectance returns to the clean surface value; there is no reflectance change
detected, unlike H, O, CO on Cu surface. The explanation of the absence of
the reflectance change is lack of an adsorbate-derived surface state since the
direct interaction between adsorbate and conduction electron requires

overlapping of surface induced electronic states and the Fermi level.

143
A nonmonotonic coverage dependence of the reflectance change for

CO on Pt(111) is also observed. The magnitude of the reflectance change

AR/R, at room temperature and high frequencies, increases monotonically at

low CO coverages, peaks at 020.33 ML, and then decreases toward saturation

coverage. The reduction of the reflectance change occurs when the ordered
overlayer starts to form and the work function starts to decrease. However,
we cannot really distinguish which effect dominates the reduction of the

diffuse scattering.

My study shows the scattering theory well explains the origin of the
broadband reflectance changes and this information can provide a new
prospect of surface IR study. We learn more about the conduction electron
properties in the near surface region and how the adsorbate interacts with
substrate that is exactly what surface scientists are trying to do in the last

thirty years.

Appendix:

The intensity of an energy loss peak divided by the elastic intensity is

(for the case of specular reflection):

 

Iv 47We 2 2 N s

— = ———e (E , a)

10 11250 W cosaf 0

IV: measured vibrational mode peak intensity

IO: elastic peak intensity
me: electron mass

h: Planck’s constant

N 5: number of dipoles per cm2
E0: primary electron ( incident electron energy)
0:: angle of incident

11V: effective dipole moment

h
#v = e *2 2———
mama
m #2 effective mass of the oscillator
a)”: vibration frequency of the oscillator in wave numbers
e*: effective charge of the oscillator

f(E0,a) = (sinz— 2cos2 a)Y +(sin2— 2cos2 a)lnX

144

 

 

145
__ 6% . -eiwé
012+90’ 03

 

61: half angle of the aperture of the detector

90: beam width 3‘1
2130

For uncoupled. oscillators:

 

 

 

2
AR 7: e* 2 1 sin261+rp
I—dv= —-7 — —NS 2
R c e m” c036 Ir I
P
0: incident angle of incoming light
rpt

reflectivity constant given by the Fresnel equations for p-polarized light

The correlation between measured IR absorption intensity and the intensity

of EEL signal is linked through dynamic electron charge.

References

1. S. Anderson and J.W. Davenport, Solid State Commun. 28, 677 (1978);

IS. Luo, Reading Notes, 1990.

2. H. Ibach, Surf. Sci. 66, 56 (1977); A.M. Baro, H. Ibach, and H.D.

Bruchman, Surf. Sci. 88, 384 (1979).

3. C. Chung, Dr. thesis, MSU, 1993.

nICHIan STATE UNIV. LIBRQRIES
lllllllllllllllllllllllllllllllllllllllllllllll
31293013908383