In..." “:1

v—v—v,

—--—nn

w,vu

 

. . ‘ -‘Iml I
J I I
I: K‘ .‘ H" "
,I,‘I.(""I-nﬂl'.liv{%ﬁ I, " I
dJIlm.

'I
'I‘

I.
n."

.\ MI
J .."I[
3&4 I

. '1‘

,E‘A‘cv, 5,}:
$ng 1 1'21“
LIL '-~”

..
..I.. i"

‘1
1.2»?

“ ,,~»&§~ﬂ‘
J J

\Hvl.‘
.IW‘

“7-H: 19a; "‘U‘4
I la:

'thI 53:31:; '1‘»:
"I 1\ ﬁg

:5
I-"'I.II
.IHII

‘ II'H.‘ . 4
l!‘ll‘.'w

0‘9"”. ' .z‘lr"-‘,T;._-.:, M
‘IIIL ﬂ ﬁ_f' : “~va

I r'I'I‘

"I‘IEI'I I: WW] ' l
*ﬂmI W‘“ I8
'I III

Jﬂ‘g

%v—v--

. 1"“
W I”?
I‘.

' (”‘1
.ulw
Jquﬁ

" u

I ribs:
391'

1.1%?

 

THESIS

TATE UNIVERSITY LIBR

Iiiii‘ii’iil llll Nil 1 Imiiiii

31293 01716 3472

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

dissertation entitled
"Electron Dynamics in Thin Epitaxial Cu( 100) Films
Studied by Adsorbate Induced Broadband Reflectance

and Resistivity Change Measurements"
presented by

EVSMINWKRAS'EV

has been accepted towards fulﬁllment
of the requirements for

Ph.D. degree in _Eh¥sics_

Qmjﬁ;

 

professor

Date November 14. 1997

MSU is an Afﬁrmative Action/Equal Opportunity Institution 0-12771

 

LIBRARY

Michigan State
Unlversity

 

 

 

PLACE IN RETURN BOX
to remove this checkout from your record.
TO AVOID FINE-S return on or before date due.

 

MTE DUE MTE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1m Wu

 

 

 

ELECTRON DYNAMICS IN THIN EPITAXIAL Cu(100) FILMS STUDIED BY
ADSORBATE INDUCED BROADBAND REFLECT AN CE AND RESISTIVITY
CHANGE MEASUREMENTS

By

Evstatin Todorov Krastev

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
Center for Sensor Materials

1997

ABSTRACT

ELECTRON DYNAMICS IN THIN EPITAXIAL Cu(100) FILMS STUDIED BY
ADSORBATE INDUCED BROADBAND REFLECT AN CE AND RESISTIVITY
CHANGE MEASUREMENTS

By

Evstatin Todorov Krastev

Gas adsorption on clean metal surfaces has been actively investigated in recent
years. It was found that for a number of systems the adsorbed gas causes a decrease in the
broadband infrared reﬂectance of the metal surface. The ﬁ'actional reﬂectance change
AR/R is too big and of the wrong sign to be accounted for by the dielectric properties of
the adsorbates. This effect has been explained by an increase in the diﬁ‘use conduction
electron scattering from the surface due to the adsorbates. The model also relates AR/R to
an increase in the resistivity Ap of the metal.

I present an experimental test of the linear relationship between AR/R and Ap
predicted by the scattering model. The samples used were epitaxial Cu(lOO) thin ﬁlms
thermally evaporated on Si(100) substrates. A great deal of effort was put into developing
the deposition techniques and characterizing the crystal structure, surface quality and
electrical properties of the ﬁlms. The epitaxy was veriﬁed by X-ray crystallography and
low energy electron diffraction. The ﬁlms exhibit 3-D growth with an nns surface

roughness of 1 to 2 nm on a 1 pm2 scale, found by atomic force microscopy in air. Films

of thickness 50 nm exhibit 2 to 3 times higher resistivity than pure bulk Cu. Auger
electron spectroscopy detected carbon and oxygen levels of less than 5 atomic percent,
with no other impurities above ~2%.

Simultaneous in situ measurements of changes in the dc resistivity and broadband
infrared reﬂectance induced by oxygen and forrnate (HCOO) adsorption on the Cu ﬁlms
demonstrate that the mechanism of the resistance change is chemically speciﬁc. For
oxygen, a linear relationship between the resistivity and reﬂectance changes conﬁrms that
conduction electron scattering is dominant, as previously deduced from infrared
measurements on single crystals. The experimental values of the proportionality constant
show a wide spread and in most cases deviate from the theoretical value based on the
electron density in bulk copper.

On identically prepared Cu ﬁlms, adsorbed formate induces a comparable
resistivity change, but no detectable reﬂectance change. This difference is inconsistent
with the scattering mechanism, for which the ratio of the two effects should depend only
on the substrate and not on the adsorbate. The resistivity change induced by formate may
arise ﬁ'om a reduction of the conduction electron density.

This work is the ﬁrst study based on in situ measurements of well characterized
samples demonstrating striking deviation from the scattering model. It is a strong
motivation for ﬁirther experimental and theoretical work in order to clarify the

mechanisms through which adsorbates affect the electron dynamics of metal thin ﬁlms.

To My Family

ACKNOWLEDGMENTS

First I want to thank Prof. Roger Tobin, who guided me through the long process
of working towards completion of this dissertation. He taught me so many things for
which I am deeply grateful. He helped me through times difﬁcult for me, not just being
my advisor but a good ﬁiend.

I want to thank my beautiful wife Karen for her love, encouragement, and the joy
and happiness she brought into my life.

I am grateﬁil to my mother Ivanka, father Todor and sister Vessela. Although they
are far away in Bulgaria, 1 have always felt their love, support and concern about me.

Many people in the Department of Physics and Astronomy at MSU have
contributed to this work by providing valuable help and assistance. I enjoyed working
with Tom Palazzolo, Jim Muns and Tom Hudson at the machine shop. Their expertise
and advise has always been of great benefit to my design work. I strongly appreciate the
assistance from Stephanie Holland, Ann Kirchmeier, Sandy Teague and Mary Curtis.

They made my life much easier. I am thankful to Prof. Jerry Cowen for providing me with
an ofﬁce during the last year and being always very nice. My coworkers, which I shared

my time in the lab with, deserve special regard -- Dennis Kuhl, Ching-Ling Hsu, Larry

vi
Voice, Kathy Lin and Hong Wang. I learned a lot through our common work and being
with them was always fun.

I thank the Center for Electron Optics of the Pesticide Research Center, Michigan
State University, for the use of its AFM facility and Stanley Flegler for his help and
valuable advise. I thank Thomas R. Bieler and Cheong Soon-Wuk, Department of
Material Science and Mechanics, Michigan State University for their help in obtaining and
analyzing pole ﬁgures of our thin ﬁlms.

I am grateful to the people at the Center for Sensor Materials at MSU, as most of
my funding came from there. This work was supported by the National Science
Foundation under Grant Nos. DMR-9201077, DMR-9510267, DMR-9696233 and DMR-
9400417 (MRSEC).

Last, but not least I want to thank Branislav (Bane) Blagojevic, Gyorgy Filep and
all the numerous ﬁiends of mine for the nice times spent together and the exciting

memories, which I will carry with me wherever I go.

TABLE OF CONTENTS

List of Tables
List of Figures
Chapter 1 Introduction
l-l. Motivation for the Study
1-2. Electrical Resistivity of Clean and Gas Covered Thin Metal Films

1-3. Broadband Infrared Reﬂectance Change of Metal Surfaces Due to Gas
Adsorption

1-4. Relation Between the Adsorbate-Induced Broadband IR Reﬂectance
and DC Resistivity Change of Thin Metal Films

References
Chapter 2. Apparatus and Experimental Technique
2-1. F ar-infrared surface spectroscopy system
2-1-1. Optical setup
2-1-2. Main UHV chamber
2-2. Thin ﬁlm sample preparation

2-3. Combined Resistivity and Broadband Reﬂection Change Measurement
of Thin Cu(lOO) Films Due to Gas Adsorption

2-3-1. Resistivity Change Measurement
2-3-2. Broadband Reﬂectance Change Measurement

References

vii

ix

24

32

35

35

37

40

42

46

46

49

52

viii

Chapter 3. Development of a Novel UHV Sample-Transfer and Sample-Handling 54
System

3-1. Introduction 54
3-2. General Operation of Our UHV Transfer System 57
3-3. Carrier Disk 62
3-4. UHV Sample Mount 64
3-5. Concluding Remarks 67
References 69

Chapter 4. Surface Morphology and Electrical Conductivity of Epitaxial Cu(lOO) 71
Films Grown on H-Terrninated Si(100)

4-1. Introduction 71
4-2. Film structure and surface morphology 73
4-3. Electrical conductivity 84
4-4. Conclusions 90
References 92
Chapter 5. Multiple mechanisms for adsorbate-induced resistivity: Oxygen and 94
forrnate on Cu(100)
5-1. Introduction 94
5-2. Experimental 97
5-3. Results and Discussion 99
References 1 07
Chapter 6. Conclusion 110

References l 15

LIST OF TABLES

Table 1-1. Adsorbate induced ﬁ'actional reﬂectance change AR/R for adsorbate
coverage na measured experimentally and calculated using equation 1-28 and
published values for tAp. The table is reproduced from reference [40].

Table 1-2. Comparison of the theoretical and experimental values for the slope of
the linear dependence between reﬂectance change and DC resistivity change for
different adsorbates on Cu and Ag polycrystalline ﬁlms. The table is reproduced
from the study by Hein and Schumacher [45].

Table 4-1. Characteristics of 100 nm copper ﬁlms annealed in vacuum at different
temperatures. All of the ﬁlms were deposited at rates of 0.1 - 0.2 nm/s, with no
intentional heating of the substrate during deposition.

Table 5-1. Values of the coefﬁcient nezlmc deduced from the slope of AR/R vs.
tAp.

29

30

82

102

LIST OF FIGURES

Figure 1-1. Conductirnetric sensing. a) clean surface, b) adsorbates on the
surface - the resistance increases, c) combined broadband reﬂectance and
resistivity change measurement.

Figure 1-2. Resistivity increase due to gas adsorption on thin metal ﬁlms for
different gas-substrate systems. The ﬁgure is reproduced from reference
[17].

Figure 1-3. Broad-band reﬂectance change for 0.25 ML oxygen adsorption on
Cu(100) single crystal for the frequency range 200 - 2000 cm". The gaps
in the spectrum and the noise at high frequencies are due to the poor
transmittance of the polyethylene windows. The solid line is a ﬁt to the
Persson-Volokitin scattering model. The graph is reproduced from
reference [41].

Figure 1-4. Antiabsorption resonance observed by Hirschmugl et a]. [3 7] using
RAIRS. The absorption of the IR light decreases at the resonance
frequency. The graph is reproduced ﬁom reference [28].

Figure 1-5. Specular and nonspecular scattering of conducting electrons at the
surface.

Figure 2-1. Layout of the far infrared surface spectroscopy system. Reproduced
from reference [1] by C. Chung.

Figure 2-2. Comparison between AES scans for Cu ﬁlms grown on not annealed
(a) and annealed at 750K (b) Si substrate. The surface of the not annealed
ﬁlm is contaminated with about 23 atomic percent (a.p.) C and 3 a.p. O.
The surface of the ﬁlm grown on annealed Si substrate is much cleaner
with C and 0 peaks comparable to the noise level.

12

15

17

19

36

44

Figure 2-3. Resistivity change of Cu(lOO) ﬁlm due to 50L formic acid exposure -
a) row data - voltage drop across the sample, registered by the LIA, b) the
linear baseline is removed and the fractional resistivity change Ap/p
calculated. The maximum zip/p is about 1.1%.

Figure 2-4. Broadband reﬂectance change AR/R of 50 nm thick Cu(100) ﬁlm
exposed to 50 Langrnuirs (5x10-7 Torr for 100 seconds) of oxygen. The

measurement is performed at room temperature. Total AR/R is about
0.5%.

Figure 3-1. Main components of the UHV sample transfer system.

Figure 3-2. Starting at 105 K a copper sample is heated up to 740 K, driving the
sample heater at 9.8 Amp. After that the sample heater is turned OE and
the sample is cooled down to 115K.

Figure 3-3. Sample carrier disk.
Figure 3-4. UHV sample mount.

Figure 4-1. 6-29 X-ray diffraction pattern of 100 nm thick Cu ﬁlm deposited on
H-terrninated Si(100) at 0.1-0.2 nm/s evaporation rate, without intentional
heating. The strong Cu(200) peak and absence of a Cu(l 11) peak indicate
highly oriented growth.

Figure 4-2. Cu(] 1 1) X-ray pole ﬁgure of a 100 nm copper ﬁlm grown on H-
terrninated Si(100). Equal areas in the ﬁgure represent equal solid angles.
The four sharp poles at O = 55° demonstrate that the Cu ﬁlm is epitaxial.
The inset shows a scan through one of the poles in the azimuthal (‘1’)
direction keeping the radial angle constant (®=55°). The width of the pole
is instrument-limited.

Figure 4-3. AFM topographical images of area 1 x l umz for two different ﬁlm
thicknesses -- a) 7.5 nm and b) 100 nm. The z-range for both images is 10

nm and the RMS roughness is 0.71 nm for (a) and 0.95 nm for (b). Both
ﬁlms were grown near room temperature and were not annealed.

Figure 4-4. RMS roughness vs. deposition rate for 100 nm thick ﬁlms grown
without intentional heating of the substrate. Each data point represents a

1 X 1 pm2 area of a different sample.

48

50

58

6O

61

65

74

76

77

79

xii

Figure 4-5. AFM topographic image of area 1 X 1 um2 of a 100 nm Cu ﬁlm
grown under conditions nominally identical to those for the ﬁlm shown in
Fig. 3b. The z-range is 20 nm, and the RMS roughness is 1.8 nm. Note
the much larger grains and deeper holes.

Figure 4-6. AFM topographical image of a 100 nm Cu ﬁlm annealed in vacuum at
175°C after deposition near room temperature. The z-range is 61 nm. The
channels penetrate deep into the ﬁlm, nearly to the Si surface.

Figure 4-7. Apparent resistivity of a Cu ﬁlm of thickness 1 = 120 nm grown near
room temperature at approximately 0.1 nm/s. The dashed line is the best
ﬁt to the Fuchs-Sondheimer theory, Eq. 2. The solid line assumes an initial
insulating layer of thickness to, Eq. 5. The ﬁtting parameters are given in

the text.

Figure 4-8. Effective conductivity oefﬁt) (deﬁned by Eq. 1) of the same Cu ﬁlm
as in Fig. 7. The dashed line is the best ﬁt to the F uchs-Sondheimer

theory, Eq. 4. The solid line assumes an initial insulating layer of thickness
to, Eq. 5. The ﬁtting parameters are given in the text. The inset shows the

0 - 30 nm region on an expanded scale. The arrow indicates the
conductivity of pure Cu at the deposition temperature.

Figure 5-1. a) Fractional reﬂectance change at 2650 cm‘1 for oxygen on epitaxial
Cu(100). The ﬁlm was 49 nm thick. At 110 s the sample was exposed to
5x 10'7 Torr 02 for 100 s. b) Fractional dc resistivity change measured
simultaneously with (a).

Figure 5-2. Fractional reﬂectance change AR/R vs. tAp for oxygen on epitaxial
Cu(100), ﬁ'om Fig. 1. The error bar shown is typical of all the points.

Figure 5-3. a) Fractional reﬂectance change at 2650 cm'1 for formate on epitaxial
Cu(100). The ﬁlm was 51 nm thick. At 130 s the sample was exposed to
5 x 10‘7 Torr formic acid for 110 s. No reﬂectance change is observable.
b) Fractional dc resistivity change measured simultaneously with (a).

80

83

85

86

100

101

103

Chapter 1

Introduction

l-l. Motivation for the Study

Surface science is the study of atomic arrangements and chemical complexes on
solid surfaces and interfaces, their electronic, mechanical and chemical properties. Due to
the interdisciplinary nature of the observed phenomena it is in the spheres of vital interest
for both physicists and chemists. From a technological point of view there are many
processes which beneﬁt from a better understanding of the processes occurring at surfaces
and interfaces. These include the industrial ﬁelds of crystal growth, chemical sensing,
corrosion, catalysis, semiconductor interfaces to name a few.

There are several landmarks in the development of surface science as a separate
discipline, which should be noted. The ﬁrst one is the concept of monolayer adsorption.
It was introduced by Langmuir in 1916 [l] and is considered to be of revolutionary
importance [2]. Most of the earlier studies had been concerned with adsorption on porous
materials with the adsorbed layers being thought to be thousands of molecular diameters
thick due to highly underestimating the area of the surface. Langmuir studied the
chemisorption of different gases on a tungsten ﬁlament which is completed at one

monolayer coverage. His understanding of the chemical bonding, arising from

2

nonsaturated or dangling bonds at the surface and the localization of the adsorbed species
at certain locations (sites) are among the basics in the contemporary theory.

The second landmark is considered to be the work by Lennard-Jones in 1932 [3].
All the earlier theoretical developments had been semi-empirical with very little
involvement of quantum mechanical concepts. The experimental studies had been mostly
occupied with thermodynamical quantities - determination of isotherms and adsorption
heats using calorimetry. Lennard-J ones developed a quantum mechanical representation
of the potential energy of an adatom at the surface and clariﬁed the distinction between
physisorption in terms of van der Waals’ forces and chemisorption due to chemical
bonding.

The third, rather technical than scientiﬁc development of crucial importance for
contemporary surface science was the progress of technology in obtaining a commercially
available clean environment for conveying the surface experiments - a good enough
vacuum and the means of measuring it. In the 193 Os attaining pressures of 1045 Torr was
routine using rotary oil and mercury-diﬂirsion pumps and elastomer seals. At that
pressure, each second about 5x10” molecules from the ambient residual atmosphere arrive
on each square centimeter of the surface. Considering that an atomic monolayer
corresponds to about 10” atoms per cmz, a monolayer of “junk” forms in a couple of
seconds on the “clean” surface of the studied material, which makes it impossible to carry
cut surface experiments. Nowadays pressures in the range of 10'10 Torr and lower are
considered adequate for the needs of the surface scientist. This is the so called ultrahigh

vacuum (UHV), which became possible by development of metal seals, new types of

3
pumps ( ion, titanium sublimation and sorption) and gauges (Bayard-Alpert, quadrupole
and crossed ﬁeld spectrometer). At UHV pressure about a monolayer of molecules on the
surface is formed in about 8 hours giving “enough” time to perform an experiment. By my
experience the time is never enough due to the very delicate nature of the surface work
and obtaining a “good” experiment often takes many attempts and long hours.

The last 30 years mark a period of large development in the ﬁeld of surface
science. Sufﬁciently good theories for the band structure and chemical bonding of bulk
solids became available. This inspired extension of the theories and their experimental
examination to the more complicated case of surfaces. Progress in computing capabilities
facilitated the development of theoretical models and the commercial availability of UHV
provided the test environment. The last but not the least factor was the strong
technological need for better fundamental understanding of the physical and chemical
processes occurring at surfaces and interfaces.

Due to the immense signiﬁcance of thin metal ﬁlms in such areas as chemical
sensing and microelectronic devices, the understanding of their structural and electronic
properties is extremely important from a technological point of view. My work is a study
of the crystal structure, surface morphology and electronic transport properties of thin
copper ﬁlms on silicon substrates using the tools of the surface scientist. The simple
concept of "conductimetric sensing" - the increase of the resistivity of different materials
upon gas adsorption on the surface - is the basis for wide variety of commercial gas
sensors (Figure l-la,b). It is well known that the adsorbed gas molecules also affect the

reﬂectance of metals. In an attempt to better understand the fundamental mechanisms

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 o
L O/L' IR

Source Detector

 

Figure 1-1. Conductimetric sensing. a) clean surface, b) adsorbates on the surface - the
resistance changes, c) combined broadband reﬂectance and resistivity change
measurement.

5
underlying these eﬂ‘ects, my work was concentrated on performing in situ broadband
infrared (IR) reﬂectance measurements together with resistivity measurements of well
characterized samples (Figure l-lc). The ultimate goal was to check the validity of the
earlier proposed mechanism [4-7], which attributes the resistivity and reﬂectance changes
observed experimentally to diffuse scattering of the conduction electrons in the ﬁlm by the
potential of the adsorbates, randomly distributed on the surface. Later in this chapter I
will discuss the scattering theory in detail.

To perform the above experiments there was a need for a suitable substrate, which
had to conform to variety of requirements. The material needed to be a well characterized
metal, thin enough for the change in the resistivity to be easily detected, and thick enough
to act as bulk for the infrared light. Our choice fell on copper ﬁlms of thickness 50 to 100
nm because they meet the above requirements (the skin depth 6 for bulk copper is 27 nm
[8]) and there was evidence in the literature [9-17] that it is not difﬁcult to grow well
ordered thin Cu ﬁlms on H-terrninated Si(100) and Si(111). The plan of my work was as
follows. First, develop the techniques for growing epitaxial Cu(100) ﬁlms on Si(100)
substrates in a high vacuum evaporator and characterize their crystal order, surface quality
and electrical properties. Second, design and build an ultrahigh vacuum (UHV) transfer
and evaporation system for the UHV chamber that is part of our infrared spectroscopy
system. We considered designing our own transfer system because there was no
commercial one suitable for our requirements -- move large samples 1 l/ ” x 1/ ” through
relatively small ports (1 3/8” diameter), available on our UHV chamber, position precisely

inside the chamber and make 6 electrical contacts with the sample holder. All of the above

6
also needed to be at a reasonable price. Once the equipment was tested and functional I
had to perform in situ combined broadband reﬂectance and resistivity change
measurements of our thin copper ﬁlms using oxygen and formate as adsorbates, analyze

the data, and look at the results from the point of view of the existing theory.

1-2. Electrical Resistivity of Clean and Gas Covered Thin Metal Films

Because of their large technological importance, thin metal ﬁlms have been an
object of a lot of experimental and theoretical work [18-20] in recent years, aimed at
better understanding their electrical and structural characteristics. Experimentally there is
a signiﬁcant inconsistency in the data published by different authors. The reason is that
the ﬁlms differ considerably in their properties because of variation in the deposition
conditions. Therefore for a reliable study it is very important that the ﬁlms are grown in
an UHV environment and that their structural and electrical properties are well
characterized.

Usually experiments show that the resistivity of clean thin ﬁlms is higher (up to 10
times) than the resistivity of a bulk single crystal material. Most of the theories explain
this effect with a decrease in the mean free path of the electrons due to surface scattering
or bulk scattering of point defects, dislocations, grain boundaries and impurities.

A size effect treatment by Fuchs [21] and Sondheimer [22] assumes that a fraction
of the conducting electrons scatter diffusely at the surfaces of the ﬁlm this way increasing
the resistivity. The parameter p, called the Fuchs specularity parameter, is the fraction of

the conducting electrons specularly reﬂected by the ﬁlm surfaces. The experimental data

7
for this parameter show extremely high variations which are not explained well in the
literature so far. The above theory makes several simplifying assumptions, some of them
not well realized in real ﬁlms:

1.) the ﬁlm possesses ideal plane parallel interfaces.

2.) the disorder is considered independent of the ﬁlm thickness

3.) single parabolic conduction band

4.) the scattering is isotropic, charaCterized by scattering parameter p and single
mean free path I.
The result from solving the Boltzmann transport equation for the resistivity p of the ﬁlm

under assumptions 1-4 is [18]:

l—e"

__.ds, (1-1)
l-pe

 

 

p 3 °° 1 x2
0 =l-—x1— ———
p 2 ( P)‘[(S3 s5)

where p0 is the resistivity of the bulk with the same concentration of lattice defects as the

ﬁlm, x = t / lo, t is the ﬁlm thickness and 10 is the bulk mean free path.

For t >> 10 (1-1) simpliﬁes to:
3 Io
p=p. MEG-p)? . (1-2)

For x > 0.2, which usually applies for thin ﬁlms, equations (1-1) and (1-2) agree to better

than 10 %.

8
A theory by Mayadas and Shatzkes employs a one dimensional grain boundary
model, with crystallite size D, eliminating assumption 2. The idea is similar to the Fuchs-
Sondheimer, but employing different limits for integration of the Boltzmann equation. The

result for resistivity is [18]:

— 14.31..
p p. 2

l-R ]’ (1'3)

e Io"

where R is the part of the conducting electrons specularly reﬂected at grain boundaries, p0

and 10 have the same meaning as in equation 1-2.

The scattering hypothesis [18] uses the concept of extra resistivity induced by

lattice disorder and diffuse surface scattering and the resistivity is given by:
— 1+Z‘A’ ’0 +ZAI" (14)
p - pa D t '

Here Z and Z. are the number of scattering centers per cm2 of the ﬁlm surface and inner
crystallite surface respectively, A and A‘ are the mean scattering cross section at the ﬁlm
surfaces and grain boundaries respectively and p0 is the resistivity of defectless ﬁlm.
Equation (1-4) is pretty much a combination of equations (1-2) and (1-3) via the

Mathiessen’s Rule.

9

Calculations in reference [18] show that all of the above theories give similar
results for the thickness dependence of the ﬁlm resistivity. This is feasible, as all of them
assume that diffuse scattering of conducting electrons is the main reason for higher than
the bulk ﬁlm resistivity, usually found by the experiment. The concept of diffuse electron
scattering was ﬁrst introduced by Fuchs and Sondheimer.

An approach, which takes into account ﬁlm roughness is introduced by Vancea
and Homnann [23,24]. They start with the expression for the ﬁlm conductivity 0(1) ( t is

the ﬁlm thickness) from the Fuchs theory ( ideal ﬁlm with ﬂat parallel surfaces):

 

Elicia-p)? s

(-
a 2 I t
°° ' l—pexp -7-

tis. (1-5)

 

 

The surface roughness exhibited by real ﬁlms is included in the model by a concept

introduced by Namba [25] :

 

(W(t>>>= —»[l< > f,g;,,drl <1—6)

Here <a(<t>)> is the measured average conductivity, <t> is the average ﬁlm thickness,

t(x) is the local ﬁlm thickness which varies across the ﬁlm, L is the length of the ﬁlm, a... is

10
the ﬁnal ﬁlm conductivity, and a(t(x)) is the local conductivity given by equation 1-5. t(x)

is approximated by a one dimensional sinusoidal ﬂuctuation:

t(x) = (t) + hsin(27r-:), (1-7)

where h is the surface roughness amplitude.

The main characteristic of the Namba model is that a —) 0 fort = h , because the ﬁlm
becomes discontinuous. Since many real ﬁlms are discontinuous at small but non zero
<t> the model gives a smooth interpolation between the conductivity cutoff and the Fuchs
- Sondheimer behavior at larger thickness. Using the Drude conductivity relation for

spherical Fermi surfaces, no. is expressed as:

nw=(.:)"2(:—:J”l-‘:-l” ,

where nx is the conducting electron density in the ﬁlm and hp is Planck’s constant.
Equations 1-5,l-6 and 1-7 were computer ﬁtted to the experimentally measured
<a{<t>)> curves for copper, aluminum, silver, gold, nickel and platinum polycrystalline
ﬁlms, evaporated at UHV and high vacuum conditions on glass substrates. The authors
emphasize that the Fuchs-Namba model is the only one which could be ﬁtted to their

experimental data. From the computer ﬁt the parameters 0..., h, l... - mean free path, and p

1 l
- the specularity parameter were extracted. However it is very difﬁcult and model
dependent to separate 1... and p.

A result of the above treatment is that the decreased conductivity 0.. of the thin
ﬁlms (even for a very thick ﬁlm a... < abulk), was correlated to a reduction in the
conducting electron density n.., rather than due to decreased mean ﬁ'ee path I... The
determined free electron density for different ﬁlms varies between 7% and 90% of the
expected bulk values. This was explained by quantum mechanical trapping of the
conducting electrons inside the separate grains, with the grain boundaries serving as
potential barriers, at which the localized electrons scatter elastically. Only the electrons
which can tunnel through all the barriers contribute to the conductivity.

So far I have discussed metal ﬁlms with surfaces as clean as possible. The change
of the electronic properties of thin metal ﬁlms due to gas adsorption is a very important
eﬁ‘ect technologically because of its application in gas sensing. The experiment usually
ﬁnds that the ﬁlm resistivity increases when gas molecules stick on the surface. Figure 1-2
presents resistivity measurements for different gas - substrate systems found in the
literature [18].

Typically the resistivity change is linear with coverage during the ﬁrst stages of
adsorption. At higher exposures the curve deviates from linearity, reaching saturation or
going through a maximum. It is very important that the surface is not contaminated
beforehand. If the latter is the case, a smaller total resistivity increase, dips in the

resistivity curve or complete absence of resistivity change could be observed. There are

l2

 

 

d)

——. 49 ﬁlm!!!)

 

 

 

 

——-> n [”5010“ng

Figure 1-2. Resistivity increase due to gas adsorption on thin metal ﬁlms for different gas-
substrate systems. The ﬁgure is reproduced from reference [18].

several competing models attempting to explain the resistivity changes due to gas
adsorption, which I will discuss next.

The Suhrrnann model [26] attributes the change in resistivity to a direct charge
transfer to (or from) the adsorbed molecules, which changes the density of conduction
electrons in the near-surface region of the metal. The electron mean free path and the ﬁlm
thickness are assumed not to be affected and the starting point for the model is the Drude

free electron gas approximation. The fractional resistivity change zip/p predicted is:

Ae=K."—°. (1-9)
p t

13

where K 1 is a proportionality coefficient, no is the concentration of adsorbed molecules on
the surface and t is the ﬁlm thickness.

Sachtler and coworkers proposed that the binding of conduction electrons by the
adsorbate is strong enough to produce an insulating layer at the surface, this way reducing
the effective thickness of the metal ﬁlm [27]. The electron mean free path and the

conducting electron density are assumed unchanged. The result is :

ﬂ: K,h(t,AT)-nti, (1-10)
p

where K; is a proportionality coefﬁcient, 11,, and t have the same meaning as before and
the function h(t,A 7) can be determined by differentiating the experimental dependence of
the resistivity on the deposition parameters, ﬁlm thickness and annealing temperature
(AT).

The third and most widely accepted model - the Fuchs - Sondheimer scattering
model [21,22] - explains the change in resistivity by an increase in the diffuse scattering of
conduction electrons due to the localized potential of the adsorbate, this way affecting the
mean free path 10. If the surface is clean the conducting electrons scatter specularly to
some extent even for polycrystalline ﬁlms. The presence of adsorbates on the surface
increases the fraction of the diffusely scattered electrons by creating additional scattering
centers. The fractional resistivity change predicted by this model is:

A

_=K310”_a, (1-11)
p t

14
The above three competing models were compared by Wissmann [18] in a thorough study
of resistivity change due to gas adsorption on copper and nickel thin ﬁlms. Both
materials arefcc metals having close values for the lattice constant, but considerably
different electronic properties. The inﬂuence of the ﬁlm thickness, annealing temperature,
measuring temperature and gas coverage were considered. The experimental data showed
deviation from the models by Suhrmann and Sachtler , but was very well ﬁtted to the

predictions of the scattering model establishing it as the most accepted one nowadays.

1-3. Broadband Infrared Reﬂectance Change of Metal Surfaces Due to Gas
Adsorption

As the investigation of the interaction of gas molecules with clean surfaces is one
of the busiest areas of contemporary surface science, different techniques which study the
vibrational resonances within the adsorbed molecules or between the adsorbed species and
the surface were developed under the collective term Surface Vibrational Spectroscopy.
Infrared absorption spectroscopy (IRAS), electron energy loss spectroscopy (EELS), and
surface enhanced Raman spectroscopy (SERS) are experimental tools revealing the
vibrating modes, coupling and energy transfer between the adsorbate and electronic states
of the surface. The adsorbates however can also affect the optical response of metal
surfaces far from the sharp resonance features - the so called broadband response. With
the development of high quality surface infrared spectroscopy it became possible to
measure these broadband reﬂectance changes. This type of measurement is more

demanding experimentally because it requires the absolute stability of the infrared signal to

15
be 1% or better during the measurement, in order to reliably register the broadband
reﬂectance change, which is frequency dependent and usually does not exceed 1.5%.

Figure 1-3 shows a measurement of the IR broadband reﬂectance change of a

Cu(100) single crystal due to oxygen at room temperature. The experiment was

 

0.000

-0.003

-0.006

AR/R

-0.009

 

'
e a
L

-0.012 . .
e . 0.25 ML : i .

U

 

 

-0.015

I I

0 400 800 1200 l 600 2000
Frequency (cm-l )

 

Figure 1-3. Broad-band reﬂectance change for 0.25 ML oxygen adsorption on Cu(100)
single crystal for the ﬁ'equency range 200 - 2000 cm". The gaps in the spectrum and the
noise at high frequencies are due to the poor transmittance of the polyethylene windows.
The solid line is a ﬁt to the Persson-Volokitin scattering model. The graph is reproduced
from reference [28].

performed at the IR synchrotron source at Brookhaven National Laboratory by Lin, Tobin
and Dumas[28]. The frequency range between 100 cm’1 and 2000 cm'1 was covered using
two detectors: Si:B bolometer up to 650 cm‘1 and Ge:Cu photoconductor for ﬁ'equencies

above 400 cm]. Two sets of measurements were performed - before and after the oxygen

16
exposure. After averaging, the fractional reﬂectance change AR/R was calculated and
plotted. The gaps in the spectra and the noisier data at higher ﬁ'equencies are due to the
lower transmittance of the polyethylene windows at that ﬁ'equency range. This
measurement is a nice example of the frequency dependence of the broadband reﬂectance
change -- AR/R gradually decreases with increasing the IR frequency, reaching saturation
at about 1800 cm]. This reﬂectance decrease is too big and of the wrong sign to be
explained with the dielectric properties of the adsorbate. The electronic polarizability of
the adsorbate would have a very small effect ( about +0.01%) towards increasing the
reﬂectance [29].

The theory which is largely believed to account for the adsorbate induced
reﬂectance change was proposed by Persson and Volokitin [4-6] several years ago. It is
based on a concept of “surface resistivity” and attributes the broadband decrease in the
reﬂectance of metals observed upon gas adsorption to the increase of the difﬁise scattering
of conducting electrons from the adsorbates on the surface. The model predicts the
frequency, coverage and temperature dependence of the reﬂectance change and the
relation between MR and the resistivity change Ap of thin ﬁlms. It also explains
antiabsorption resonances observed by reﬂection absorption IR spectroscopy (RAIRS)

[3 0-33] and effects of atomic scale friction [34]. Lin et al. have pointed out that the
relationship between the reﬂectance and resistivity change is pretty general, originating
from the fact that both of them are connected to diffuse scattering of conducting electrons
ﬁ'om the adsorbate, and does not involve many of the more subtle details of the Persson-

Volokitin model [8].

17
The theory by Persson-Volokitin was inspired by the observation of the remarkable
antiabsorption resonances by Reutt et al. [30] and Hirschmugl et al. [31-33], using
RAIRS. In contrast to the dipole-allowed vibrational resonances (adsorbate vibrations
normal to the surface), the reﬂectance is actually increasing at the resonance frequency,

relative to the nearby ﬁ'equencies (Figure 1-4). They were attributed to molecular

 

    

 

 

 

i
:E .
3 “““““
.0.
'5 IMIRﬂaW’
Q
220 zoo ‘ 360
u(cm"1)

Figure 1-4. Antiabsorption resonance observed by Hirschmugl et al. [31] using RAIRS.
The absorption of the IR light decreases at the resonance frequency. The graph is

reproduced from reference [7].

translations and rotations, parallel to the surface. These are formally dipole-forbidden
vibrational modes as the component of the incident light parallel to the surface is orders of
magnitude smaller than the normal component and thus cannot couple directly to the
adsorbate oscillations.

The originally proposed model uses a local optics approach and is valid for light

frequency a)

18

VP
a)>>—, 1-12

where v; is the Fermi velocity of the substrate and 5 = c/cop is the classical skin depth, c is
the speed of light and a), is the plasma frequency. In the local optics approach the
conducting electron moves only small distances (on the scale of the decay length 6 of the
electric ﬁeld due to the incident light) during one period T =27dar of the light. Practically
the electron does not see the spatial variation of the electric ﬁeld and Drude conductivity
and local optics apply. The ﬁustrated rotation for CO/Cu(100) discovered by Hirschmugl
et a1. is at frequency (0., z 285 cm'1 and vp/(S z 500 cm'1 for copper (Figure 1-4). For this
case the condition (1-12) does not hold. For the low light frequencies 6 < va the
conducting electron experiences a spatially varying electric ﬁeld during one period T,
requiring non-local treatment. The theory was ﬁrrther expanded [7] to include the non-
local effects at lower frequencies using a Boltzmann transport equation approach
originally developed to study the anomalous skin effect . This treatment is valid for all IR
frequencies which satisfy the condition 0) << wp. For the light frequency range used in my
experiments, the condition (1-12) is well satisﬁed and both local and non-local treatment
give the same results.

Next I will introduce the details of the Persson-Volokitin model:

1. The metal is treated in the semi-inﬁnite jellium approximation - the conducting
electrons form a ﬁ'ee electron (Drude) gas, characterized with bulk relaxation time T3 and

a mean free path I ( l = vprg).

l9
2. At the surface there is a random distribution of adsorbates with surface density n... The
conducting electrons experience two types of scattering events at the surface: First, with
probability p, not losing momentum parallel to the surface (p is the Fuchs specularity
parameter), and second, probability l-p = m)? , having their parallel momentum
randomized due to the adsorbates. Z is the adsorbate scattering cross section. The ﬁrst
type corresponds to a specular scattering and the second to a diffuse scattering at the
surface (Figure 1-5). In the second case all of the momentum change goes to the

adsorbate, but not to the lattice or surface defects.

Specular Scattering:

clean surface

        

no contribution to the resistance

Nonspecular Scattering:
adsorbate

09 900 A.

 

 

 

6" film

resistance rises

 

Figure 1-5. Specular and nonspecular scattering of conducting electrons at the surface.

20

3. Only the p-component of the incident light is considered because the resultant of the
incident and reﬂected electric ﬁelds for the s-polarization is practically zero at metal
surfaces due to the close to 7: phase change. The perpendicular electric ﬁeld component
of the p-polarization is orders of magnitude higher than the parallel component at the
surface, so only the molecular vibrations with a dynamic dipole moment normal to the
surface can couple to the incident light - the dipole-allowed vibrational modes. The
electric ﬁeld component parallel to the surface can couple only to the ﬁ'ee electrons in the
metal, but not to the adsorbate, as is much smaller in magnitude. The broadband
absorption occurs inside the metal and is caused by diffuse scattering of conduction
electrons from the adsorbates. Inside the metal it is the ﬁeld parallel to the surface that
dominates. It comes from the p-polarized component of the light, which is strongly
reﬂected at the interface. The s-polarized component is much less effective in giving a

parallel electric ﬁeld inside the metal [3 5]. When the IR ﬁ'equency a) coincides with the

frequency are of the parallel adsorbate vibrations, the adsorbed molecules move in
resonance with the driﬁ motion of the conducting electrons, which results in vanishing of
the additional surface resistivity due to diffuse electron scattering from the adsorbates and
the reﬂectance goes back to its value for a clean surface - the dipole-forbidden
translational and rotational resonances.

The expression for the ﬁactional reﬂectance change AR/R for a p-polarized light

incident at angle 0, which follows ﬁom the non-local treatment is pretty complicated [7]:

21

 

 

 

 

 

AR; - — 3(1—p) w2 RC[G(w/wl,l,6)l . (1-13)
R 7:2 cosawlarpl 4 a) 1m 3') dq .
ncosblarp we(q,ar/a),,l,6)
The complex-valued ﬁtnctions G and a are given by:
°° 1 1 2 4i17ar
G=]dy(———.—)/(y)+ , jdy(—-—.-)f(y) (144)
I y y (co-a) -iarn
where f(y)=qu 1 1 , (1-15)
40 e(q,ar,w,,l,b‘)l+i,8y
and
3. . . . 2 1+51
szqz—J—w[£) (2ﬂ)+[(ﬂj —l]ln-—,ﬂ , (1-16)
413w. 13 ﬂ ,1;
13
where ,Bzé—I-Q. (1-17)
I a),

The rolloﬂ‘ frequency a), = vp/b' speciﬁes the frequency range when the non-local effects
become important as discussed above (equation 1-12). The damping coefficient r]

characterizes the ﬁiction force between the adsorbate and free electrons and is related to

22
the specularity parameter p. Whether or not this coefﬁcient is important for accurately
predicting the broadband reﬂectance change has been discussed in the literature and is still
controversial [3 6,37].

Far ﬁ'om resonances the fractional reﬂectance change due to surface adsorption is

expressed as [38]:
slip—”41‘1” ($1). 048)
R 4ccost9 a), 6

The fractional broadband reﬂectance change is negative, meaning that always the clean
surface has a higher reﬂectance. The frequency dependence is incorporated in the function
f, which involves properties only of the metal, and approaches unity for frequencies

a) >> a), . The coverage dependence comes from the quantity (l-p). Considering that l-p
= n“): , AR/R is linearly related to the concentration of the adsorbates na, if 2' is
independent of no. The dependence on the temperature is indirect and comes through the
expression for the mean free path I = vprg. Both R and AR involve I so the temperature
dependence is complicated and condition speciﬁc.

The earliest reports on adsorbate induced reﬂectance change are by Riffe, Hanssen
and Sievers [39-41] and Reutt, Chabal and Christman [30] who studied a variety of
adsorbates on W(100) and Mo(100) crystals. A minimum in the broadband reﬂectance,
observed by Reutt et al., led to the interpretation of an adsorbate-induced surface

electronic state.

23

Different substrates and adsorbates have been considered since then. The
frequency dependence of the broadband reﬂectance change and the antiabsorption
resonances observed by Hirschmugl et al. [31-33] for CO / Cu(lOO) and (111) are in
excellent agreement with the mechanism proposed by Persson. Lin et al. reported very
good agreement with the scattering theory for the ﬁ'equency dependence of the broadband
reﬂectance change for 0 adsorption on Cu(100) [8,28]. The solid line in Figure 1-3,
reproduced from their paper, represents a ﬁt to the Persson-Volokitin scattering theory.
Obviously the agreement is pretty convincing.

Measurements of the coverage, temperature, and frequency dependence of the
broadband reﬂectance change induced by CO on Pt(111), by Kuhl et al. supported the
validity of the scattering model [38]. Lamont et al. detected the antiabsorption peaks due
to dipole-forbidden parallel vibrations of H and D on Cu(] 1 1) [42]. A discrepancy with
the theory was found for the frequency dependence of the broadband reﬂectance change
of H/Cu(111) for ﬁ'equencies higher than 1000 cm'l, but was attributed to electronic
excitations involving adsorbate induced surface states, similar to what was proposed for
the case of H on W(100) and Mo(100) [30]. Borguet et al. reported a linear increase of
the reﬂectance change with coverage, in agreement with the model, both for near-infrared
and visible light [43,44].

To summarize the above experimental review I want to emphasize that a good deal
of evidence in support of the Persson-Volokitin scattering model has been accumulated in

the recent years.

24

1-4. Relation Between the Adsorbate-Induced Broadband IR Reflectance and DC
Resistivity Change of Thin Metal Films
The main goal of my study is the experimental examination of the linear

relationship, predicted by the scattering model, between the fractional reﬂectance change
AR/R and resistivity change Ap for well characterized epitaxial Cu(100) thin ﬁlms. In this
chapter I will develop in detail the above relationship closely following the derivation by
Lin et al. [8]. This derivation is based on the Persson-Volokitin model but does not
involve its more subtle and speculative details like damping of the adsorbate translations
and rotations parallel to the surface. It demonstrates that the relationship between
reﬂectance and resistivity is pretty general and the connection to the antiabsorption
resonances and ﬁictional damping comes from the fact that all of the above effects involve
electron scattering. The ﬁnal result for the relationship between broadband reﬂectance
and resistivity change, far from resonances, is the same with or without considering atomic
ﬁiction and vibrational damping. Even if antiabsorption resonances come from another
mechanism the relationship between AR/R and zip is still valid.

To account for surface effects on reﬂectance, non-local corrections to the
classical Fresnel formulas are added using Feibelman’s d-parameter formalism [45]. The
fractional reﬂectance change of p-polarized light due to change in the surface conditions is

expressed as[8]:

 

:—-—— , (1-19)

25

where w is the light frequency, c is the velocity of light, 6 is the incident angle. Equation

(1-19) is valid provided that Is] >> (1 / cos2 0) >> 1, e is the complex local dielectric

function of the metal. The parameter d. is given by:

 

d - 12(51de

6: 120
l— [wIsz’ (‘)
la—

where z is the direction normal to the surface and a. is the parallel component of the
complex conductivity for a uniform electric ﬁeld. Im(dt) is negligible compared to
Im(d|), because of the strong refraction experienced by the incident light in the metal.
The Persson-Volokitin theory proposes a slab model: Deep in the metal the
conductivity has its bulk value 03 , while near the surface it is affected by the randomly

distributed adsorbates and can be expressed as:
a,(a))=aB(a))+Ao(a)) (1-21)

The near-surface region where (1-21) is valid is of the order of the electron elastic mean

free path IB=vprB. Using the above model to calculate d. the expression for the fractional

reﬂectance change for p-polarized light becomes:

26

ﬂ = 4015. /cIm Acr(ar) (1_22)
R —— .

 

cosH 03 (w)

As previously discussed, provided that equation (1-12) is valid for the particular material
and light frequency range, a local description of the metal’s dielectric response can be
used. For copper lip/6:310 cm'l , the light frequency used in my study is well above 2000
cm", so the condition (1-12) is satisﬁed. The bulk conductivity, using the Drude free

electron gas approximation, is given by:

2
ne 1,,

m(1—iarr3) ’

03(0)) = (1'23)

Here n is the conducting electron density, m is the electron effective mass, and e is the
electron electric charge. As the adsorbates can affect both the conducting electron density
n and the scattering time rthe change in the conductivity A0 in the near surface region

can be expanded to ﬁrst order in changes An and Ar.

1 A0160) _ An_+ AT
03(0)) ”8 78(l-inB).

 

(1-24)

In the above expression only the term in Ar has a imaginary part, which can result in
reﬂectance change (formula 1-22). The term in An would produce only a resistivity

change. The scattering hypothesis assumes that the main reason for the experimentally

27
observed resistivity and reﬂectance change of metals due to gas adsorption is the diffuse
scattering of conducting electrons from the adsorbates. No change in the conducting

electron density n is presumed. Therefore the term Arr/n3 in (1-24) can be neglected. For

a ﬁlm of thickness 13 , and w>> 1/13, Im(Ao(a))) can be expressed using the dc resistivity

p = l/afa) = 0) as:

m[AS’—(—”’l] 1 “’8’ (1-25)

03((9) _ (073 PB .

As it is well established experimentally that tAp does not depend on the ﬁlm thickness t

[18,20], the reﬂectance change can be related to the dc resistivity change for arbitrary ﬁlm

 

thickness:
33 z _ 413 AMI.) = ___—4 [tAp(t)]. (1-26)
R 01,, cost? p3 chrB c039
Using:
m
[2373 =—"-2", (1-27)

equation (1-26) becomes:

28

95:13: 4 [tAp(t)] (1-28)

R mc c050

 

This is the main equation to be tested in my investigation. Several important points

should be emphasized:

1. For the frequency region a) >> vp/b' , a) >>1/rB , where equation (1-28) is

valid the ﬁactional reﬂectance change does not depend on the frequency;

2. The reﬂectance change is linearly proportional to the resistivity change and the
proportionality coefﬁcient does not depend on the adsorbate, but only on bulk parameters

of the metal;

3. The electron density n is assumed to be constant;

4. AR/R does not depend on the ﬁlm thickness t, for t>>6.

The relationship (1-28) was ﬁrst tested experimentally by Lin et al. [8]. They
measured the ﬁactional reﬂectance change for O and CO on a Cu(100) single crystal and
for CO on a Ni(100) single crystal and compared their results to published resistivity
change measurements for the corresponding thin-ﬁlm systems. Their results are

summarized in Table 1-1.

29
For CO and 0 adsorption on Cu the experimental and calculated values for AR/R
are within factor of 2 and were considered consistent with the scattering model within
the experimental error due to uncertainty in the published resistivity data. However for
CO on Ni the experimental reﬂectance change was a factor of 30 smaller than the
predicted one using the values for tAp found in the literature. Lin et al. point out that the

above experiments cannot serve as a deﬁnite quantitative test of the scattering model as

Table 1-1. Adsorbate induced fractional reﬂectance change AR/R for adsorbate coverage
na measured experimentally and calculated using equation 1-28 and published values for
tAp. The table is reproduced from reference [8].

 

 

System "a (141,0) (AR/R)calc (AR/ R) exp
(lOl‘cm'z) (10'12 Qcmz) (%) (°/°)
CO/Cu(100) 6.1 0.60 -2.2 -1.li0.2
O/Cu(100) 3.8 0.42 -1.5 -1.1i0.2
0.70 -2.6
CO/Ni(100) 6.4 1.47 -5.8 -O.2:t0.2

 

the resistivity and reﬂectance measurements were taken on different samples under
different conditions, and a rigorous test would involve simultaneous resistivity and
reﬂectance measurements of well characterized epitaxial ﬁlms.

Recently Hein and Schumacher reported simultaneous reﬂectance and resistivity
change measurements of in situ grown thin polycrystalline ﬁlms on glass substrates, using
unpolarized IR light [46]. The systems studied were 0, CO, Cu on Cu thin ﬁlms and Ag

on Ag thin ﬁlms. The linearity between resistivity and reﬂectance change predicted by the

30
scattering model was conﬁrmed, but quantitative discrepancies for the actual slope were
found (Table 1-2).

The observed quantitative discrepancy was not explained but the authors
emphasized that it cannot be attributed to uncertainties in the experimental parameters. It
was also pointed out that the theoretical calculations deal with a semi-inﬁnite jellium
model, while the ﬁlms were polycrystalline with grain size of the order of the ﬁlm

thickness.

Table 1-2. Comparison of the theoretical and experimental values for the slope of the
linear dependence between reﬂectance change and DC resistivity change for different
adsorbates on Cu and Ag polycrystalline ﬁlms. The table is reproduced from the study by
Hein and Schumacher [46].

 

 

System (AR/Alphas (AR/zip)“,
(ll-Q cm)’l (#9 cm)’l
O/Cu -0.0085 -0.030
CO/Cu -0.0077 -0.014
Cu/Cu -0.0085 -0.010

AgAg -0.0082 -0.025

 

My work presents a test of the relationship (1-28) for oxygen and formate
adsorption on epitaxial Cu(100) thin ﬁlms, grown in situ in UHV, and carefully
characterized. In Chapter 4 I will discuss the preparation and characterization of the

samples. In Chapter 5 the actual combined reﬂectance and resistivity change

31

measurements are presented and analyzed. The predicted linearity between reﬂectance and
resistivity change was found for oxygen, but for the formate no detectable reﬂectance
change together with a signiﬁcant resistivity change was observed in striking contrast to
the predictions of the scattering model. Thus the present study is the ﬁrst test of the
Persson-Volokitin scattering theory on well characterized epitaxial thin ﬁlms and the ﬁrst
work clearly demonstrating deviation from the model, for the particular system formate on

Cu(100).

32

References

[l] I. J. Langmuir, Am. Chem. Soc., 38, 2221 (1916).

[2] Ralf Vanselow, Chemistry and Physics of Solid Surfaces, CRC Press, Boca Raton,
Florida (1979).

[3] J. E Lenard-Jones, Trans. Faraday Soc, 28, 223 (1932).

[4] B.N.J. Persson, Phys. Rev. B 44 , 3277 (1991).

[5] B.N.J. Persson, Chem. Phys. Letters 185, 292 (1991).

[6] B.N.J. Persson, Chem. Phys. Letters 197, 7 (1992).

[7] B.N.J. Persson and AI. Volokitin, Surface Sci. 310, 314 (1994).

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

[9] C.-A. Chang , Phys. Rev. B 42, 11 946 (1990).

[10] C.-A. Chang, Surf. Sci. 237, L421-L423 (1990).

[l 1] C.-A. Chang, J. Appl. Phys. 68, 5893 (1990).

[12] C.-A. Chang, J. Vac. Sci. Technol. A 8, 3779 (1990).
[13] C.- A. Chang, J. Vac. Sci. Technol. A 9, 98 (1991).

[14] Y.-T. Cheng, Y.-L. Chen, MM. Karmarkar and W.-J. Meng, Appl. Phys. Lett. 59,
953 (1991).

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

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

[18] P. Wissmann in: Surface physics, ed. G. Hohler, Springer Tracts in Modern Physics,
77, Springer, New York, (1975).

33

[19] D. Dayal, H.-U. Finzel and P. Wissmann in: Thin metal ﬁlms and gas chemisorption,
ed. P. Wissman Elsevier, Amsterdam, (1987).

[20] D. Schumacher, Surface scattering experiments with conduction electrons, ed. G.
Hohler, Springer Tracts in Modern Physics, 128, Springer, New York, (1993).

[21] K. Fuchs, Proc. Camb. Phil. Soc. 34, 100 (1938).

[22] EB. Sondheimer, Adv. Physics 1, l (1952).

[23] J. Vancea, H. Hoffmann, Thin Solid Films, 92, 219 (1982).

[24] J. Vancea, H. Hoﬂ‘mann, K. Kastner, Thin Solid Films, 121, 201 (1984).

[25] Y. Namba, Jpn. J. Appl. Phys, 9, 1326, (1970).

[26] R. Suhrmann, K. Schulz, Z. Phys. Chem. (Frankfurt) l, 69 (1954).

[27] W.M.H. Sachtler, G.J.H. Dorgelo, Z. Phys. Chem. (Frankfurt) 25, 69 (1960).

[28] KC. Lin, R.G. Tobin and P. Dumas, Phys. Rev. B 49 (1994) 17273 (1994); 50,
17760 (1994).

[29] R.G. Tobin, Phys. Rev. B 45, 12110 (1992).
[30] IE. Reutt, Y.J. Chabal and SB. Christman, Phys. Rev. B 38, 3112 (1988).

[31] C]. Hirschmugl, GP. Williams, FM. Hoffrnann and Y.J. Chabal, Phys. Rev. Letters
65, 480 (1990).

[32] CI. Hirschmugl, Y.J. Chabal, FM. Hoffmann and GP. Williams, J. Vac. Sci.
Technol. A 12, 2229 (1994).

[33] CI. Hirschmugl, G.P. Williams, B.N.J. Persson and AI. Volokitin, Surface Sci. 317,
L1141 (1994).

[34] J. Krim, D.H. Solina and R. Chiarello, Phys. Rev. Letters 66, 181 (1991).
[3 5] Y.J. Chabal, Surface Science Reports, 8, 211 (1988).
[36] R.G. Tobin, Phys. Rev. B 48 15468 (1993).

[37] B.N.J. Persson, Phys. Rev. B 48 15471 (1993).

34

[38] DE. Kuhl, K.C. Lin, Chilhee Chung, J .S. Luo, Hong Wang and R.G. Tobin,
Chemical Physics 205,1 (1996).

[39] D.M. Riffe, L.M. Hanssen and A]. Sievers, Phys. Rev. B 34, 692 (1986).

[40] D.M. Riﬂ‘e, L.M. Hanssen and A]. Sievers, Surface Sci. 176, 679 (1986)..

[41] D.M. Riffe and A.J.Sievers, Surface Sci. 210, L215 (1989).

[42] C.L.A Lamont, B.N.J. Person and GP. Williams, Chem. Phys. Lett. 243, 429 (1995).
[43] E. Borguet, J. Dvorak, and H.-L. Dai, Laser Techniques for Surface Science,
Proceedings of the Society of Photo-Optical Instrumentation Engineers, Volume 2125 , 12
(1994)

[44] J. Dvorak, E. Borguet, and H.-L. Dai, Surface Science Letters, 369, L122 (1996)
[45] RI Feibelman, Prog. Surf. Sci. 12, 287 (1982).

[46] M. Hein and D. Schumacher, J. Phys. D 28, 1937 (1995).

Chapter 2

Apparatus and Experimental Technique

In order to perform my combined resistivity and broad-band reﬂectance
measurements on Cu(100) epitaxial ﬁlms, I had to modify the existing infrared
spectroscopy system to allow in situ growth of thin ﬁlms in ultrahigh vacuum (UHV)
conditions. The UHV surface analysis chamber had to be furnished with thin ﬁlm
evaporation equipment, which I will discuss in this chapter, and a fast sample-entry load-
lock system together with a new sample mount which is described in detail in chapter 3.

The design, construction and testing of these new facilities took a considerable
amount of time and work but it is a very important addition to our system allowing us to
study freshly grown thin ﬁlms on different substrates. In this chapter I will also describe
our far-infrared surface spectroscopy system and the techniques for measuring broad-band

reﬂectance and ﬁhn resistivity changes used in my investigation.

2-1. Far-infrared surface spectroscopy system

The layout of our far-infrared surface spectroscopy system is shown in Figure 2-1,

which is reproduced ﬁ'om reference [1]. The infrared (IR) light is generated and

35

36

 

 

 

fAR-IlfRNED SlRfACE S‘ECIROSCU’Y SYSIDI ﬂorid-lit! M

Spectral rum: no - no 0-1- Grating Spec rater
Insulation 1- S to»!
Sgt. in: 20 - m1

 
 
  
    
 
    
  

 

 

I
- __~

4

Illtrdridt Vanna Swim halysis
Mt villi Single Crystal Sale
an Ub-aeld mipulotu

Lite-Cooled Si B
Infrared Detector

  
 

Otter aria m:
I! 50“! mm L:- (rug (infirm Oillrclim , , 60 a m
min til-cooled 9110163. W Etaclron ﬁrearm '
Polamr and Omar Ml negation Spectrum

Figure 2-1. Layout of the far infrared surface spectroscopy system. Reproduced from
reference [1] by C. Chung.

modulated in a source cryostat and via transfer optics is directed at a grazing angle
towards the sample. The sample is located in the UHV surface analysis chamber, which is
the only component of the system operating under ultra high vacuum conditions. The rest
of the components are kept under high vacuum (10" -— 10'8 Torr) Alter reﬂecting from
the sample the IR light passes through another set of transfer optics and enters the home-
built Czemy-Tumer grating spectrometer. Part of the IR light with the particular
frequency and polarization of interest exits the spectrometer and via a third set of transfer
optics is directed towards the IR detector. There the optical signal is transformed to a
voltage using a photoconductor, ampliﬁed and after that registered by a lock-in ampliﬁer

(LIA), connected to a computer. Parallel with monitoring of the reﬂectance, the sample

37
resistance is registered via a four-wire resistivity measurement, and the sample
temperature is determined using a thermocouple. The process of data acquisition is
automated using home made software. Having described the general plan and operation

of our system, next I will discuss in more detail its main components.

2-1-1. Optical setup

The IR light is generated by a conventional silicon-carbide globar, resistively
heated to 1300 K. It is situated in one of the focal points of a gold plated adjustable
ellipsoidal mirror. Gold plated mirrors are used through the system because of their
superior reﬂectance performance in the IR range compared to other types of mirrors. A
tuning fork chopper [2] is located in the second focal point of the ellipsoidal mirror. It
modulates the intensity of the 1R light at 800 Hz. This frequency was chosen to avoid
mechanical resonances with other system components. The above elements, except the
globar, are kept at liquid nitrogen (LN) temperature as they are mounted on a copper
plate, in contact with a LN reservoir. This is in order to minimize the background photon
ﬂux modulated by the chopper and sent to the successive elements of the system. The
whole assembly is kept at high vacuum (HV) to eliminate atmospheric absorption of the
IR light and help keeping the optical elements cold.

After the chopper the IR light is directed towards transfer optics (TO), which
shares the same vacuum space with the source elements, but is kept at room temperature.
There is no need of cooling, because the T0 is after the chopper on the optical path. The

T0 consists of two off-axis paraboloidal mirrors, one of which is adjustable from outside

38

of the vacuum chamber. The purpose of the transfer optics is to produce a 1.8 times
magniﬁed image of the chopper at the sample in order to illuminate the whole sample area
and the reﬂected light to be collected by a spectrometer with a focal number f/4. Through
a differentially pumped Csl [3] window the IR light hits the sample, located in the main
UHV chamber. The angle of incidence is 86°. The reﬂected portion enters the second
transfer optics via another Csl window. Csl is chosen because of its high transmission in
the infrared, but it must be handled very carefully, because it is easily damaged if exposed
to high humidity conditions. The second transfer optics is a mirror image of the ﬁrst. It
demagniﬁes the image of the chopper, reﬂected by the sample, and places it on the
entrance slit of the spectrometer.

The home-built grating spectrometer is the most complicated optical subsystem of
the apparatus. It operates under high vacuum and is cooled to LN temperature. All the
elements are mounted on an aluminum plate, kinematically supported via three invar rods.
The aluminum plate is clamped to the outer shell of the spectrometer for mechanical
stability and connected to a LN reservoir, carefully designed to provide adequate cooling.
The IR radiation reﬂected from the sample ﬁrst encounters the entrance slit. There is a
choice of several slits, mounted on a rotatable wheel, allowing change in the resolution of
the spectrometer from outside. Via a folding mirror the light is directed towards the
collimating mirror (off-axis paraboloid) and the collimated beam is diffracted by a gold
plated grating [4]. Three different gratings are used to cover the spectral range of 300 to
3000 cm". Only one grating can be installed at any given time and the spectrometer must

be opened in order to replace a grating. The diffracted light is next collected by a camera

39
mirror (also off-axis paraboloid) and via a second folding mirror sent to a polarizer (a gold
wire grid on a silver bromide substrate [5]) Via the exit slit the p-polarized component of
the light is directed towards a third set of transfer optics. Similar to the entrance slit
assembly, a variety of exit slits could be chosen externally and different low-pass IR ﬁlters
are provided to cut higher diffraction orders.

The third set of transfer optics shares the same vacuum space with the
spectrometer and is LN cooled by its own LN reservoir. The main element is an
adjustable ellipsoidal mirror, having as its two focal points the exit slit of the spectrometer
and the detector.

The IR detector is a boron doped silicon photoconductor (Si:B), 1x1x0.4 mm in
size and operated at liquid helium temperature. It is mounted at the exit aperture of a gold
plated Winston cone - a paraboloidal conical device permitting all the light entering its
larger entrance aperture (10 mm in our case), via successive wall reﬂections to exit
through the smaller exit aperture (1mm in diameter). The whole detector assembly is
installed in a commercial liquid helium dewar [6].

The electrical signal from the detector is boosted by a two stage transimpedance
ampliﬁer. The ﬁrst stage consists of two JFETs with two switchable feedback resistors
(1M!) and 60 M9), mounted on the cold plate of the detector dewar. The two transistors
are heated to 77K to allow normal operation. The second stage is an AC coupled

operational ampliﬁer with a switchable gain (2 and 10), operating at room temperature.

40

2-1-2. Main UHV chamber

The main UHV chamber, which houses the sample, is located in between the two
transfer optics. As discussed in the introduction ultrahigh vacuum (UHV) is the only
acceptable environment for a modern surface science experiment, granting several hours
before the specimen surface is contaminated by the ambient gases. The UHV chamber,
12” (30.5cm) in diameter and 25” (98.4 cm) in height, is custom-made out of stainless
steel. A large number of ﬂanged ports is provided for mounting a variety of equipment. It
is divided into three levels: IR/sample transfer and evaporation (IR/STE) on the bottom,
surface analysis in the middle and precision X-Y-Z-O manipulator [7] on the top. The
sample mount is attached to a LN reservoir (cold ﬁnger), which is mounted on the
manipulator. The middle surface analysis level houses equipment for low energy electron
diffraction (LEED), Auger electron spectroscopy (AES), an ion sputtering gun and a
residual gas analyzer with a temperature programmed desorption (TPD) capability.

The bottom IRISTE level is used to perform the reﬂectance measurement, sample
introduction and thin ﬁlm growth. It is equipped with a fast-entry load-lock system
(FELLS) and thin ﬁlm evaporation equipment. The design and construction of these two
systems was a considerable part of my project. The FELLS is described in detail in
chapter 3. The thin ﬁlm evaporation system consists of a resistive evaporator and a crystal
thickness monitor. The evaporator is built on a standard 4.5” (11.4 cm) ﬂange. There are
two ﬁlaments shielded from each other, providing the capability of evaporating two
different metals. A rotating shutter is available to block one of the ﬁlaments at a time.

The whole assembly is enclosed in an oxygen free high conductivity (OFHC) copper

41

sleeve, tightly ﬁt over a stainless steel reservoir for a cooling liquid. The copper sleeve
consists of two shells for easy removal and installation. A ﬂow of liquid nitrogen or
chilled water can be used to provide the cooling. Cooling during the evaporation is
essential, as warming of the chamber surfaces would result in a pressure increase, which is
highly undesirable. A chimney-shaped shield is attached to the copper sleeve in order to
direct evaporation only towards the sample and the crystal monitor, thus protecting the
rest of the chamber. A stainless steel shield on the sample mount also helps prevent
evaporation on unwanted areas. The evaporator ﬂange is mounted on a bellows-sealed
linear translator so the evaporator assembly (EA) can be retracted and separated from the
main UHV chamber, using a gate valve. This option permits us to change ﬁlaments
without breaking the UHV in the main chamber. Before reintroducing the EA to the
chamber the space of the linear translator is pumped to high vacuum using a turbo-
molecular pump. Baking of the BA for a couple of hours before reintroducing helps to
reduce the pressure spike in the main UHV chamber.

Achieving of UHV after the chamber has been vented to atmospheric pressure is a
lengthy procedure taking about 72 hours. Once sealed, the chamber is rough pumped
using molecular sorption and turbo-molecular pumps. When the pressure gets down in the
10" Torr range the main ion pump [8], supplemented with a titanium sublimation pump, is
opened to the chamber. The next step is a 36— to 48— hour bakeout at about 100 to 110
°C. It is accomplished by heating tapes, placed all around the chamber, together with an
internal IR heater. During the bakeout the chamber is wrapped in aluminum foil, helping

to keep more even heat distribution. The temperature is monitored by numerous

42

thermocouples positioned at various locations. After cooling down to a room temperature
a base pressure of about 5—7x10’ll Torr is achieved.

The sample could be exposed to variety of gases via two leak valves. Backﬁlling
of the UHV chamber or a specially designed multihole effusive doser array [9] is used.
The doser provides 16 times ﬂux enhancement over the backﬁlling. The doser is mounted
on a bellows-sealed linear translator and is usually retracted far from the center of the
UHV chamber when it is not in use. This permits unobstructed maneuvering of the sample

holder around the UHV chamber.

2-2. Thin film sample preparation

The samples used in my study are epitaxial in situ grown Cu(100) thin ﬁlms, on
Si(100) substrates. The technique for growing of the ﬁlms was developed using a separate
HV evaporator. A detailed study of the surface morphology and electrical characteristics
of these ﬁlms is presented in Chapter 4. Once our UHV chamber was equipped with
sample transfer and evaporation equipment we were able to grow the ﬁlms in UHV
conditions. The ﬁlms grown at UHV are epitaxial and show equivalent or lower surface
roughness than the ﬁlms grown in HV.

Before being introduced into the UHV chamber the silicon sample (2.5 x 1 cm) is
ﬁrst cut out of standard 3” Si wafer (B-doped, 20-50 S) cm resistivity). It is degreased in
a 15 minute ultrasonic bath of acetone followed by methanol. In a separate HV
evaporator 150 nm thick silver on top of 10 nm chromium is evaporated on the two ends

of the Si chip, using a mask made out of aluminum foil. This way the two contact stripes

43
are formed. A batch of samples (20 or more) could be processed at the same time, thus
increasing the productivity. The Si sample is etched in a 10% aqueous solution of
hydroﬂuoric acid (HF) for about 60 seconds. This treatment removes the SiOz and
passivates the surface by saturating the Si dangling bonds with hydrogen [10-13]. The
sample is then rapidly mounted to a sample disk and introduced into the UHV chamber
using the fast sample entry system, discussed in detail in Chapter 3.

As our UHV chamber is equipped with Auger spectroscopy (AES), I was able to
investigate the ﬁlm-growing conditions that provide the cleanest ﬁlm surfaces. As
mentioned before a surface as clean as possible is of crucial importance for a surface
science experiment. This kind of studies were not performed for the ﬁlms grown at high
vacuum. It proved that annealing the Si chip for about 15 minutes at 720 to 750 K, alter
introducing to the UHV chamber, dramatically improves the surface purity of the ﬁlm
grown afterwards. Figure 2-2 presents a comparison between the ABS scans for two Cu
ﬁlms grown without and with preannealing of the Si substrate. The top scan (Figure 2-
2a), taken from a Cu ﬁlm grown on a Si chip that was not annealed, reveals substantial
amounts of carbon and oxygen present on the surface - about 23 atomic percent C and 3
atomic percent 0. The bottom graph (Figure 2-2b) is for a Cu ﬁlm grown on a Si
substrate preannealed at 750 K. The C and 0 peaks are much smaller and almost
indistinguishable from the noise - the contamination of the ﬁlm surface is reduced to 3
atomic percent C and less than 1.5 atomic percent 0.

The positive eﬂ‘ect of substrate annealing on the surface purity of our ﬁlms is

attributed to the following. The hydrogen-passivated Si surface is contaminated with

 

 

 

 

 

 

 

 

 

0.5 I I ' I ' I ' I ' I ' I ' I ' I ‘ I

0-4 _' Si substrate not annealed T
A 0.3 " “
.3 ’ 4
g 0.2 -
:1 0.1 - -
a 1 1‘ 1
H , '-
:a _ l
g .04 .. l O l T -
V . Cu 1 Cu (:1. .
”-1 -0.2 - C -
re _ .
2 L 1
c -o.3 . l ,

'°'4 .' 1 Cu 1

.0.5 Q“! 1 l 1 l 1 l 1 l 1 I 1 A 1 l 1 l 1

0 100 200 300 400 500 600 700 800 900 1000

8) Electron Energy (eV)

0.5 I I ' I ' I I I f I ' I ' I ' I ' I

0.4 ' Si substrate annealed at 750 K ‘
A 0.3
3 .
g 0.2
a. 0.1
:3
.3 0.0
5 or ,
{-3 -o.2
2
'o -0.3

-0.4

.0 5 P Qul 1 L 1 1 1 J 1 l 1 l 1 l 1 l 1 1C“.

. 0 100 200 300 400 500 600 700 800 900 1000

b) Electron Energy (eV)

Figure 2-2. Comparison between AES scans for Cu ﬁlms grown on not annealed (a) and
annealed at 750K (b) Si substrate. The surface of the not annealed ﬁlm is contaminated
with about 23 atomic percent (a.p.) C and 3 a.p. O. The surface of the ﬁlm grown on
annealed Si substrate is much cleaner with C and 0 peaks comparable to the noise level.

45

carbons from the atmosphere during loading of the sample into the UHV chamber. During
and after the ﬁlm grth the carbons penetrate through the ﬁlm and show on the surface.
This assumption is supported ﬁ'om the fact that for a not annealed substrate the C peak
grows signiﬁcantly with time, which was not observed for ﬁlms grown on annealed
substrates. The annealing of the Si chip causes contamination (water and carbons) to
leave the substrate resulting in a much cleaner ﬁlm surfaces. It also removes the hydrogen
passivating the surface, so the ﬁlm evaporation should be done immediately after the
annealing.

Before annealing of the Si chip, the cold ﬁnger is ﬁlled with liquid nitrogen. This
keeps cool the engaging stainless steel springs of the sample transfer mechanism during
the heating, thus preventing them from collapsing. It also facilitates cooling down the Si
substrate after the annealing as the best ﬁlms are grown at room temperature (see Chapter
4). In about 30 to 45 minutes the temperature of the Si chip drops to about 300 K. The
temperature is kept steady by running the sample heater, which compensates for the heat
lost to the cold ﬁnger. The ﬁlms are grown at evaporation rates of 0.1 to 0.2 nm/s to
thickness of about 50 nm. During the deposition the pressure in the chamber usually rises
and stays in the range 3x10'lo to 9x10’l0 Torr. No post-deposition annealing is performed

as it does not improve the surface quality (Chapter 4).

46

2-3. Combined Resistivity and Broadband Reﬂection Change Measurement of Thin
Cu(100) Films Due to Gas Adsorption

Once the Cu ﬁlm is grown, the desired temperature has stabilized and the pressure
has gone down to 31x10“lo Torr, the sample holder is moved to the infrared position. A
so-called “time scan” is started by simultaneously monitoring the ﬁlm’s resistivity and
reﬂectance at ﬁxed time intervals. It is important to measure the clean ﬁlm for a long
enough time in order to establish a stable baseline before a particular gas is introdirced into
the chamber. The gas is introduced by setting one of the leak valves at a ﬁxed ﬂow, while
pumping is kept on. A constant pressure is held in the chamber, as monitored by a nude
ionization gauge. After a certain period of time the leak valve is closed and the gas is
pumped out. Following the dosing the measurement is continued long enough to establish
a new stable baseline. The exposure is determined by taking into account the pressure and
the time interval the gas was present in the chamber and is measured in Langmuirs (1L=
1x10'6 Torr s). The change or absence of a change in the sample resistivity and reﬂectance
is registered. Usually the duration of a time scan is between 300 and 800 seconds.
Further in this chapter I will discuss in more detail the resistivity and broadband

reﬂectance change measurement.

2-3-1. Resistivity Change Measurement
The ﬁlm resistance is measured by a four-wire ac technique using a 1.0 mA current
provided by the internal reference of a lock-in ampliﬁer (LIA) [14], and modulated at 3

kHz. Four breakable electrical contacts to the sample and a thermocouple are provided by

47
the custom designed sample holder, discussed in detail in Chapter 3. The voltage drop
across the sample is registered by the LIA, using two coax cables, with their shields
grounded. The LIA measures the difference between the two leads, thus avoiding noise
common to both wires. The capacitive and inductive pickup in the leads is less than 8x10'9
VNHz [15]. The data is collected by a computer connected to the LIA , using home
made software. The measurement is very low-noise and fractional resistance changes
AV/V<10‘4 are easily measurable.

A typical resistivity change measurement for a 50 nm thick Cu(lOO) ﬁlm at room
temperature is presented in ﬁgure 2-3. The raw data (ﬁgure 2-3a), is the voltage drop
across the sample, registered by the LIA. At about the 230th second formic acid gas is
introduced in the chamber. The exposure is 50 Langmuirs - 5x10'7 Torr for 100 seconds.
The linear drift in the baseline is caused by a slight change in the sample temperature
during the measurement. To remove the systematic drift from the measurement a line is
ﬁtted to the raw data between the 0th to 225th second. This line is subtracted from the
raw data and the resulting data is plotted in ﬁgure 2-3b. The fractional resistivity change
Ap/p caused by the gas exposure is about 1.1% in this case. The absolute resistivity
change Ap is calculated taking into account the ﬁlm dimensions ( thickness, length and
width) and the actual resistance of the sample, measured separately. After the end of the
dose a slight decrease in the resistance is observed, which is most likely caused by some of

the gas leaving the ﬁlm surface.

0.00270 ﬁ 1 . u 1 1
Film Thickness 50 nm
' 300K
0.00269 '-
A
Z.
& 0.00268 -
§
g 1
0.00267 1
‘——"
Dosing 50 L Formic Acid
0.00266 ‘ ‘ ‘ ‘ ‘ ‘ ‘
0 l 00 200 300 400 500
a) Time (s)
0.01 2 r - u . 1 u -
’ Film Thickness 50 nm
0.010 - 300K -
0.008 - ‘
0.006 -
e _ ‘.
2 0.004 - ‘
0.002 - i
0.000 1 v v ‘: A * “l
* <——> ‘
-0.002 - Dosing 50 L Formic Acid ‘
0 l 00 200 300 400
b) Time (s)

Figure 2-3. Resistivity change of Cu(lOO) ﬁlm due to 50L formic acid exposure - a) raw

48

 

 

 

 

 

 

 

 

 

 

 

 

 

500

data - voltage drop across the sample, registered by the LIA, b) the linear baseline is
removed and the fractional resistivity change Ap/p calculated. The maximum Ap/p is

about 1.1%.

49
2-3-2. Broadband Reﬂectance Change Measurement

The broadband reﬂectance measurement is performed with the spectrometer
cooled to LN temperature in order to reduce the background IR radiation from walls and
other surfaces in the spectrometer. The detector is set to a high feedback resistor. The
ﬁrst step is to maximize the IR reﬂection from the ﬁlm surface by ﬁne adjusting the sample
position and the mirrors of the transfer optics. The spectrometer is set to the desired ﬁxed
IR ﬁ'equency (far from resonances caused by molecular vibrations). The signal modulated
by the chopper is registered by a lock-in ampliﬁer using the chopper reference. The data is
recorded by a computer.

A typical broad-band reﬂectance change measurement for a 50 nm thick Cu(lOO)
ﬁlm due to oxygen adsorption is shown in ﬁgure 2-4. A linear baseline has been removed.
Begining at about the 210th second the ﬁlm is exposed to 50 Langmuirs of oxygen (5x107
Torr for 100 seconds). The reﬂectance of the sample drops sharply and reaches a
saturation before the end of the dosing. The total broadband fractional reﬂectance change
is about 0.5%.

The reﬂectance measurement usually exhibits a baseline drift of up to 2% over the
period of the measurement. Usually this drift is linear and by subtracting the baseline from
the measurement the reﬂectance change can be detemrined with accuracy better than
0.02% [1,15]. It is interesting to compare the sensitivity of our apparatus to the sensitivity
of the U4IR surface science equipment at the National Synchrotron Light Source (N SLS)
- Brookhaven National Lab, where many broadband reﬂectance measurements have been

performed [16-21]. Although the synchrotron provides radiation orders of magnitude

50

 

 

 

 

 

 

 

0.001 * 1 - 1
' -1
0.000 nght Frequency 2650 cm 1
. Film Thickness 50 nm
-0.001 - 300 K ‘
-0.002 ~ —
g -0003 ~
-0.004 —
-0.005 -
i
-0.006 f I I
41007 _ Dosing 50 L Oxygen _
0 l 00 200 300 400 500
Time (s)

Figure 2-4. Broadband reﬂectance change AR/R of 50 nm thick Cu(100) ﬁlm exposed to
50 Langmuirs (5x10-7 Torr for 100 seconds) of oxygen. The measurement is performed
at room temperature. Total AR/R is about 0.5%.

brighter than the globar, instabilities in the electron beam limit the sensitivity to broadband
reﬂectance changes to 0.2% [18] - an order of magnitude worse than ours. A
disadvantage of our system is that it can perform a measurement only at one particular IR
frequency at a time, while the system at the U4IR beam line covers the whole frequency

range 300 - 3000 cm'1 at a time by using 'a Fourier Transform spectrometer (F TIR).

51

Combined resistivity and reﬂectance change measurements are further discussed in

chapter 5.

52

References

[l] C. Chung, Ph.D. thesis, Michigan State University, unpublished, (1993).

[2] Multi-Scanning Systems Corporation, type RC-2.

[3] R. G. Tobin, C. Chung and J.S. Luo, J. Vac. Sci. Technol. A 12 264 (1994).
[4] Milton Roy Co.

[5] Cambridge Physical Sciences, IGP225.

[6] Infrared Laboratories, Inc.

[7] Therrnionics.

[8] Perkin-Elmer corp.

[9] D. E. Kuhl and R. G. Tobin, Rev. Sci. Instr. 66, 3016 ( 1995).

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

[1 l] T. Takahagi, I. Ishiytani, H. Kuroda, Y. NagasawaH . Ito and S. Wakao, J. Appl.
Phys. 68, 2187 (1990).

[12]Y. J. Chabal, G. S. Higashi, K. Raghavachari and V. A. Burrows, J. Vac. Sci.
Technol. A 7, 2104 (1989).

[13] P. Dumas, Y.J. Chabal and G. S. Higashi, Phys. Rev. Lett. 65, 1124 (1990).
[14] Princeton Applied research Corp.
[15] D. E. Kuhl, Ph.D. Thesis, unpublished, (1996).

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

[17] C. J. Hirschmugl, G. P. Williams, F. M. Hoﬂinann and Y. J. Chabal, J. Vac. Sci. and
Technol. A 12, (1994).

53

[18] K. C. Lin, R. G. Tobin, P. Dumas, C. J. Hirschmugl and G. P. Williams, Phys. rev. B
48, 2791 (1993).

[19] K. C. Lin, R. G. Tobin, P. Dumas, Phys. Rev. B 49 17273 (1994).

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

[21] C. L. A. Lamont, B. N. J. Persson and G. P. Williams, Chem. Phys. Lett. 243, 429
(1995)

Chapter 3

Development of a Novel UHV Sample-Transfer and Sample-Handling
System

3-1. Introduction

The fast sample introduction ﬁ'om air to ultra-high-vacuum (UHV) and sample
transfer between different UHV chambers (evaporation, reaction, analysis), without
breaking the vacuum, is vitally important for material and surface studies. The obtaining
of a UHV usually takes about 72 hours or more, so breaking the vacuum for introducing a
new specimen is very unproductive approach. Different groups and commercial
manufacturers have developed a variety of fast entry load-lock systems complying with
different requirements [1-16]. The usual method is to ﬁrst introduce the sample into a
small loading chamber, separated ﬁ'om the main UHV chamber by a gate valve. This
loading chamber is easily and quickly (in about few minutes) pumped to a high vacuum
(HV: low 10" — 10’8 Torr) pressure using pumps with high pumping speed. When the
right pressure is achieved the gate valve to the main UHV chamber is opened and the
sample is loaded in. The last step must be accomplished as fast as possible, so the
pressure increase or spike in the main UHV chamber is as small as possible.

Different types of UHV transporters are used to move the sample in and out of the

chamber and between different chambers: bellows sealed push-pull or air actuated,

54

55
bellows sealed rack and pinion, wobble stick manipulators and magnetically coupled rods.
The last type [19], used in our design, provides adequate linear travel (up to 36 inches, for
a 2.75” mounting ﬂange), 360° rotational motion, and is relatively inexpensive. The
operation is smooth and reproducible and the positioning is achieved by sliding an external
sleeve, which is magnetically coupled to the transporter rod, on which a sample holder is
attached. Applying too big an axial or rotational force leads to decoupling between the
sleeve and the rod, thus preventing damage to the components of the sample securing and
positioning system.

The general requirements for a sample transfer system include easy and
reproducible transfer, heating and cooling range and method, electrical and thermocouple
contacts to the sample and means of positioning inside the UHV chamber. Usually there is
a variety of speciﬁc needs dictated by the particular experiment and conﬁguration. The
small diameter tubes of the UHV chamber, usually dedicated to the transfer system, and
the limited techniques and materials complying with the UHV requirements of cleanliness
(soldering, taping, gluing and materials like plastics and brass for instance are not allowed
due to their high outgassing rate) make the design of a UHV transfer system a pretty
challenging venture.

Our goal was to make simultaneous measurements of the dc electrical resistance
and inﬁared reﬂectance of epitaxial copper ﬁlms grown in situ on silicon substrates
[17,18]. This application, together with the design of our preexisting chamber, imposed

stringent requirements:

56
1) Optical access to the sample surface at angles of 83 - 89° from the surface

normal, so that there could be almost no projections above the surface plane that would

shadow the surface.

2.) Total of six electrical contacts to be closed in the process of engaging to the

sample holder inside the UHV chamber -- two for type K thermocouple and four for the

resistivity measurement, thus permitting simultaneous and independent monitoring of the

sample resistance and temperature.

3.) A sample size approximately 1 x 3 cm for maximum inﬁ'ared signal.

4.) Fast, smooth and reliable transfer in and out of the UHV chamber. Precise and

reproducible sample positioning for optical stability.

5.) Heating and cooling between 100 and 1000 K.

6.) Access through a 38 mm OD tube perpendicular to the manipulator axis.

7.) Sample electrically isolated from the sample mount.

8.) Good thermal contact between sample and sample mount to ensure effective

cooling.

57

9.) Thin ﬁlm evaporation capability inside the UHV chamber.

Although different groups and commercial manufacturers have developed UHV
transfer systems [1-16], none of them were able to successﬁrlly fulﬁll all of our
requirements. The most critical one for our purposes, the need of six electrical contacts
was not found in any of the existing designs. Similar geometries were available but
providing no more than four electrical contacts. Thus in order to perform our combined
resistivity-reﬂectance measurements I designed a novel UHV transfer system to comply

with all of the nine requirements listed above.

3-2. General Operation of Our UHV Transfer System

The main components of our UHV transfer system are shown in Figure 3-1.
Detailed sketches are shown in Figures 3-3 and 3-4. After preparation, the substrate is
mounted on a disk-shaped carrier and introduced in the load-lock chamber using a special
handling device. The load-lock chamber is a simple “six-cross” equipped with 2
viewports, ionization gauge tube for pressure monitoring, quick-access door and
magnetically coupled transporter with a 3 prong fork. It is separated ﬁom the main UHV
chamber by a metal sealed gate valve.

The carrier disk is engaged to the fork by ﬁrst positioning the spring-loaded prongs
inside 3 of the 6 key-shaped holes ( 2 on the top and 1 on the bottom of the carrier disk)

and after that applying a clockwise rotation. After the quick-access door is closed, the

58

 

   
 

attach to UHV manipulator —--

 

 

spring-loaded prong

 

   
   

electrical contact strip

transfer fork

 

 

 

 

 

sample ‘

magnetically coupled carrier ﬁxed sam le
transporter disk mount p

Figure 3-1. Main components of the UHV sample transfer system.

59

load-lock is pumped by a 90 VS turbo pump mounted just below the chamber and backed
by a mechanical pump. Pressure in the low 10'6 to mid-10", Torr is achieved in a few
minutes and the system is ready for the next step -- transfer of the carrier disk inside the
main UHV chamber.

In order for the pressure spike in the main UHV chamber to be as small as possible
the time interval when the gate valve separating it from the load-lock chamber is open
should be minimized. This is achieved by careﬁrl alignment of the sample mount
beforehand using the precision X-Y-Z manipulator mounted on a 360° rotatable mounting
platform. When the sample mount is properly aligned, the gate valve is open and the
carrier disk is advanced towards the sample mount using the magnetically coupled
transporter, till the 3 spring-loaded prongs of the sample mount enter the 3 empty key-
holes of the carrier disk. The disengagement from the fork and engagement to the sample
mount is achieved simultaneously by counterclockwise rotation of the transfer rod. In the
process of engaging the carrier disk to the sample mount 6 electrical contacts are closed --
2 for the thermocouple and 4 for the resistivity measurement. Transferring of the carrier
disk back to the manipulator fork follows the same procedure, just using opposite rotation
of the transfer rod.

The concept utilizing 6 key-holes and spring-loaded prongs is based on a
commercially available sample-transfer system by Therrnionics NW. In their design the
key shaped holes are evenly distributed around the perimeter of the carrier disk. In our
design the key-shaped holes are clustered in two groups -- on the top and bottom of the

carrier disk, this way allowing longer sample to be mounted on, which is strongly

 

800

700 /A\
600

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o

u 500

3

1'!

o 400

at .
200
100 L L 1 A M 1

 

0 10 20 30 4O 50
Time (minutes)

Figure 3-2. Starting at 105 K a copper sample is resistively heated up to 740 K, then
cooled down to 115K.

beneﬁcial for the reﬂectance measurement. Also the design by Therrnionics did not
provide 6 breakable electrical contacts to the sample, which is of crucial importance for
our experiments.

The sample cooling is accomplished by attaching the sample mount to a liquid
nitrogen (LNz) reservoir or “cold ﬁnger”. The lowest temperature achieved is about 100
K. The sample is radiant heated by a nude tungsten ﬁlament. The temperatures of up
1000K can be readily achieved. Figure 3-2 shows a heating/cooling curve for a copper

sample. The sample was initially at temperature of 105 K. A ﬁlament current of 9.8 A

61

disk, beryllium copper

thermocouple with isolation
two-hole ceramic tube

copper contact strip

 

 

 

 

sample

 

6 key holes

ceramic tube for thermocouple

electrical contact
opening for heating

thermocouple contacts

b) Rear View sapphire Plate

 

Figure 3-3. Sample carrier disk.

62
heated it to 740 in about 5 minutes. The ﬁlament was then turned off. The sample
reached 200 K in about 15 minutes, and 115 K in about 40 minutes. The cooling time
constant is dominated by the thermal contact between the sample mounting disk and the
tabs of the ﬁxed sample mount; it could probably be ﬁtrther improved by coating these

contact surfaces with silver or gold.

3-3. Carrier Disk

The carrier disk (Figure 3-3) is a very delicate device due to the variety of
requirements it has to comply to and its tiny dimensions (32 mm OD and 1.6 mm
thickness). It is made out of beryllium copper in order to provide good thermal contact
with the sample mount for satisfactory sample cooling. At the same time the beryllium
copper is a much harder material than the regular OFHC (oxygen ﬁ'ee high conductivity)
copper thus much less susceptible to wearing which is very important for reliable long-
terrn operation. Besides the 6 key-shaped holes, clustered in 2 groups at the top and
bottom of the disk, there is a large opening in the middle (23 x 10 mm) needed for the
radiant sample heating by a nude tungsten ﬁlament on the sample mount. Also there are 8
holes for socket head screws and a hole for an isolation ceramic tube for the
thermocouple.

Front view of the carrier disk, with sample attached, is shown in Figure 3-3a. The
sample is held to the ﬂoat of the carrier disk by a thin (0.13 mm) copper contact strip at
each end. The strips make electrical contact to the sample, but are electrically isolated

from the disk. These mounting strips are the only part of the assembly that projects

63

beyond the sample surface, and they are thin enough to produce negligible optical
shadowing, even at near-grazing incidence. They are held in place by stainless steel 0-80
screws, which pass through alumina tubes to independent beryllium copper contact plates
on the back side of the disk (Figure 3-3 b). Four additional screws, also electrically
isolated from the disk, hold the contacts in place during assembly and prevent them from
moving during installation and removal of the carrier.

The sample is electrically isolated from the disk by two crystalline sapphire plates
(16 x 3.2 x 0.8 mm, with holes for the 0-80 screws), one at each end. In addition to
providing electrical isolation, the sapphire plates have near-metallic thermal conductivity at
liquid nitrogen temperature to facilitate cooling, but are good thermal insulators above
room temperature, for efficient heating [20-24]. Two more sapphire plates separate the
contact plates from the disk as shown in Figure 3-3b, and their center regions are wound
with chromel and alumel thermocouple wire, respectively. The two thermocouple wires
pass to the front of the disk through a short two-hole alumina tube and are joined to form
a thermocouple junction. The thermocouple is pressed mechanically against the sample
(for metal samples, it could be spot-welded). The four beryllium copper plates and the
two sapphire plates wound with wire form the canier side of the six electrical contacts.
When the carrier is latched to the ﬁxed mount, they press against the six spring contacts of
the mount.

Four-wire resistance measurements generally require four independent electrical
contacts to the sample. In our design the two contacts at each end of the sample are both

made through the thin copper strip that holds the sample to the carrier disk, so any contact

64
resistance between the strips and the sample is included in the measured resistance. We
have found this contribution to be <0.01 O, which is negligible for our purposes (typical
sample resistance is 2 to 3 Q ). The resistance between the ﬁxed spring contacts and the
carrier can be as large as 1 (2, but does not contribute to the measured resistance. A true
four-wire carrier could be constructed with only a slight increase in complexity,
eliminating from the measurement the small resistance between the contact strips and the
sample.

Once assembled the carrier disk is sturdy and tolerates several feet drops without
damage. Another very useful feature is that the sample mounting procedure is fast -- in
about 2 - 3 minutes the sample is secured in place and the disk is ready for transfer. This
is very important for hydroﬂuoric acid (HF) etched silicon samples, when the exposure to

the ambient atmosphere needs to be as short as possible.

3-4. UHV Sample Mount

The UHV sample mount (sample holder) hosts the carrier disk inside the main
UHV chamber. It is attached to a liquid nitrogen reservoir (cold ﬁnger) using an OFHC
copper clamp. The cold ﬁnger is mounted on an X-Y-Z-B manipulator, allowing precise
positioning. Feedthroughs are provided for 2 thermocouples, 4 electrical wires for
resistivity measurement and 2 high current heater leads.

The main body of the sample mount consists of 2 parts, head (SMH) and plate
(SMP), which are held together by 6 socket-head screws (Figure 3-4). The head is made

out of beryllium copper and accommodates the electrical contacts and prong-locking

65

 

 

 

O O
,_1 M,
O O
O
O O

isolation ceramics
7/ heater
/
/

thermocouple contact

. V/ radiation shield

 

 

electrical contact for
resrstrvrty measurement

 

 

 

spring—loaded prong

 

 

   

_. spring
”\positioning nuts and
contact washers

s1de VIEW prongassembly
srde Vlew

Figure 3-4. UHV sample mount.

66

mechanism. The plate, made out of OFHC copper, provides connection to the cold ﬁnger,
holds the sample heater and has a number of threaded holes for attaching shields, isolation
ceramics and a thermocouple. It also plays a role in the operation of the spring-loaded
prong-locking mechanism.

Six F-shaped electrical contacts are distributed around the sample mount head.
They are attached to the SMH using 0-80 stainless steel socket head screws and
electrically isolated via alumina tubes and washers (see contact side view in Figure 3-4).
As it was very difﬁcult to obtain chromel and alumel sheet material for the thermocouple
contacts the following approach was used. Commercially available 0.8 mm (0.03 2”)
chromel and alumel wire was ﬂatten out by hammering to form a sheet about 3 mm wide
and 0.2 mm thick. The contacts made this way proved to have excellent springiness and
made very reliable and reproducible electrical connection with the corresponding contacts
on the sample ring (less than 1 (2 contact resistance). Another advantageous property was
that they completely preserved their springiness while exposed to very wide temperature
range - cooling down to about 100 K and heating at least up to 1000K as they are very
closely positioned to the sample heater. A wire of the corresponding thermocouple
material was spot-welded to each of the contacts and via an alumina tube conveyed out of
the SMH. Using a thermocouple feedthrough guaranteed that each half of the
thermocouple ﬁom the sample to the outside of the UHV chamber is composed of alumel
or chromel, thus avoiding unwanted junctions with other materials. Due to the good
performance of the “home-made” sheet material the 4 electrical contacts for the resistivity

measurement were made also out of chromel.

67

The mechanism for securing the carrier disk into place on the sample mount
incorporates 3 spring-loaded prongs (Figure 3-4). They are positioned on the 3 ﬁngers of
the SMH, designed to provide maximum thermal contact with the carrier disk. As the
carrier disk thickness is 1.6 mm the prong’s head should be elevated above the plane of
the SMH ﬁngers. This is achieved by using positioning nuts and washers, pressing on the
sample mount plate (see prong assembly side view in Figure 3-4).

The sample heating is radiant, provided by a nude tungsten ﬁlament attached to the
sample mount plate, and electrically isolated using ceramics. The heater is embraced by a
copper shield concentrating the heat to the sample and protecting the stainless steel
springs of the prongs from losing their springiness and collapsing. Since both the ﬁlament
and the sample are electrically isolated from ground, it would be straightforward to

implement electron-beam heating.

3-5. Concluding Remarks

In this chapter I presented the development of a novel UHV sample transfer
system, providing 6 breakable contacts between the sample and the UHV sample holder,
optical access to the sample at grazing angles of incidence, capability of transferring large
samples, wide temperature range (110 - 900 K) and easy and reliable operation. None of
the commercially available transfer systems were able to comply to the requirements for
number of contacts and sample size together with the 2.75” ﬂange size limitation imposed
by our conﬁguration. For about one and a half years of reliable operation our transfer

system proved to be perfectly adequate for our needs. It was also successfully used for

68
some experiments performed in the surface science UHV chamber at the U4IR far-infrared
beam line at the National Light Source at Brookhaven National Laboratory showing its
portability and adaptability to different conﬁgurations. We submitted a paper describing

our sample-transfer and sample-handling system to the Journal of Vacuum Science and

Technology.

69

References

[1] CA. Crider, G. Cisneros, P. Mark and JD. Levine, J. Vac. Sci. Technol. 13, 1202
(1976)

[2] JP. Hobson and E.V. Komelson, J. Vac. Sci. Technol. 16, 701 (1979).
[3] Bl Mulder, J. Phys. E 12, 908 (1979).

[4] NJ. DiNardo, J .E. Demuth, W.A. Thompson and PG. Ledermann, Rev. Sci. Instr. 55,
1492 (1984).

[5] J. Klebanoff, V.H. Ritz and RE. Thomas, J. Vac. Sci. Technol. A 2, 1396 (1984).

[6] KW. Nebesny and NR. Armstrong, J. Vac. Sci. Technol. A 3, 1763 (1985).

[7] J.M. Lindquist and J .C. Hemminger, J. Vac. Sci. Technol. A 5, 116 (1987).

[8] GS. Chottiner, W.D. Jennings and KL Pandya, J. Vac. Sci. Technol. A 5, 2970
l9]9.81i7.)Z.Moshfegh and A. Ignatiev, Rev. Sci. Instrum. 59, 2202 (1988).

[10] H.-J. Drouhin, M. Picard and D. Paget, Rev. Sci. Instr. 60, 1167 (1989).

[11] RE. Clausing, L. Heatherly and LC. Emerson, J. Vac. Sci. Technol. 16, 708 (1979).
[12] RS. Polizzotti and IA. Schwarz, J. Vac. Sci. Technol. 17, 655 (1980).

[13] AL. Helms, Jr., W.A. Schiedt and S.L. Bernasek, Rev. Sci. Instr. 59, 1223 (1988).

[14] R. Raval, M.A. Harrison, D.A. King and G. Caine, J. Vac. Sci. Technol. A 9, 345
(1991)

[15] X.-S. Wang, C. Huang, V. Bressler-Hill, R. Maboudian and W.H. Weinberg, J. Vac.
Sci. Technol. A 11, 2860 (1993).

[16] S. Thevuthasan, D.R. Baer, M.H. Engelhard, Y. Liang, J.N. Worthington, T.R.
Howard, J.R. Munn and KS. Rounds, J. Vac. Sci. Technol. B 13, 1900 (1995).

[17] ET. Krastev, L.D. Voice and R.G. Tobin, J. Appl. Phys. 79, 6865 (1996).

[18] ET. Krastev, D.E. Kuhl and R.G. Tobin, Surf. Sci. Lett, in press.

70

[19] Model MT-16, MDC Vacuum Products, 23842 Cabot Blvd., Hayward, CA 94545.

[20] R.W. Powell, C.Y. Ho and PE. Liley, ed., Yhermal Conductivity of Selected
Materials (National Bureau of Standards, 1966).

[21] WE. O’Grady, A.V. Melo, R.W. Hoffman and GS. Chottiner, J. Vac. Sci. Technol.
A 5, 281 (1987).

[22] JD. Beckerle, Q.Y. Yang, A.D. Johnson and ST. Ceyer, Surf. Sci. 195, 77 (1988).
[23] PG. Strupp and PC. Stair, J. Vac. Sci. Technol. A 9, 2410 (1991).

[24] C. Rusu and IT. Yates, Jr., J. Vac. Sci. Technol. A 15, 436 (1997).

Chapter 4

Surface Morphology and Electrical Conductivity of Epitaxial
Cu(100) Films Grown on H-Terminated Si(100)

4-1. Introduction

This chapter is based on our paper published in Journal of Applied Physics [1].

Hydrogen-terminated silicon surfaces prepared by wet chemical etching have been
shown to be excellent substrates for growing epitaxial ﬁlms of a variety of fee and bcc
metals [2-9]. In most cases a thin layer of epitaxial copper is used as a seed layer for
subsequent growth of other metals. For this reason, and because the low electrical
resistivity and large electromigration resistance of copper make it an important
technological material, [10] the properties of epitaxial copper ﬁlms on Si are of particular
interest. We report here an investigation of the crystal structure, surface morphology and
electrical conductivity of such ﬁlms as a function of deposition rate, deposition
temperature and post-deposition annealing. Epitaxial growth is observed over a wide
range of deposition conditions. The smoothest ﬁlms were obtained by near-room-
temperature growth at rates of 0.1-1.0 nm/s; higher temperatures and higher deposition
rates produced noticeably inferior ﬁhns. Post-deposition annealing at temperatures up to

125°C had little effect, while higher temperatures degraded the ﬁlms.

71

72

There have been a number of studies of epitaxial Cu ﬁlms grown on H-terminated
Si, [10-14] but important aspects, including their surface roughness and the nature of the
Si-Cu interface, are not yet understood. It has been reported that the ﬁlms, although
epitaxial, are not atomically smooth [14]. There has been no previous systematic
investigation of the surface morphology, however, or of the inﬂuence of deposition
conditions and post-deposition treatment on the morphology. The present work presents
such an investigation, together with new information on aspects of ﬁlm composition and
structure that affect electrical transport.

An earlier study [14] revealed the existence of a mixed "buffer layer" about 10 nm
thick between the Si and the Cu, which is presumably important in facilitating epitaxy.
Reﬂection high-energy electron diﬁ'raction (RI-IEED) [l4] and grazing incidence X-ray
diffraction [13] studies have shown that the Cu lattice is rotated by 45° relative to the Si
lattice, with Cu(010) parallel to Si(011). The rotation reduces the lattice mismatch
between the two materials from 40% to 6%. Even a 6% mismatch, however, is unusually
large for epitaxial growth. The buffer layer presumably helps to relieve the strain, but its
nature — for example, whether it is compositionally disordered or a stoichiometric silicide
— has not been established. Our work conﬁrms many of these conclusions, and
demonstrates that the buffer layer, though it may contain Si as an impurity, is not a
stoichiometric silicide, but is structurally indistinguishable from pure Cu.

Measurements of electrical resistance as a function of thickness during deposition
provide additional clues to the quality and composition of the ﬁlms. The ﬁrst few nm of

the ﬁlm consist of discontinuous Cu grains. The remainder of the ﬁlm is relatively pure Cu

73
(defect density of a few tenths percent) but with a high level of surface, interface or grain
boundary scattering that strongly affects ﬁlm conductance at thicknesses below about 50

nm.

4-2. Film structure and surface morphology

The copper ﬁlms were deposited in a resistive evaporator with base pressure 1 x
10.3 Torr. The (100) Si substrates (B-doped, 20-50 0 cm resistivity) were prepared by
ultrasonic degreasing with acetone and methanol, etching for 30 - 60 seconds in a 10%
aqueous solution of HF, and pull-drying (slowly and smoothly removing the substrate
from the solution with the surface vertical, so that the liquid sheets off smoothly with no
droplets). It has been shown that such a procedure leads to an extremely ﬂat and
chemically inert surface, with virtually all the Si dangling bonds terminated with H [15-
18]. After etching, the samples were loaded into the evaporator and pumping was begun
as quickly as possible —— within less than ﬁve minutes. A pressure of the order of 5x10'8
Torr was achieved in about 15 minutes. The substrates were clamped to a variable-
temperature stage so that effects of elevated substrate temperature and post-deposition
annealing could be studied. The temperature was measured by a type K thermocouple
mounted on the heating stage, close to the sample. The substrate was not intentionally
heated, but the thermocouple on the evaporation stage typically indicated a temperature of
25-30°C during deposition. The evaporation rate and ﬁlm thickness were measured using
a quartz crystal monitor calibrated with a diamond stylus proﬁlometer. Most of the ﬁlms

had thicknesses of about 100 nm. During deposition the pressure rose to between S x 10'8

74
and 3 x 10'7 Torr. This is an unusually high pressure for epitaxial growth, and results in a
signiﬁcant impurity concentration in the ﬁlm, as we discuss below. It is a remarkable

feature of the H-terrninated Si surface that oriented Cu ﬁlms can be grown even at

pressures of 10" Torr [19].

 

I l I l l l l I '
Cu(200)
'E
:3 _ ..
.6
H
3 - _
>1
30:1; _ Cu(l 1 l) _
l:
o .
E r- Sl(400) q
H

 

 

 

 

 

 

30 40 50 60 701 so
29 (deg)
Figure 4-1. 0-20 X-ray diffraction pattern of 100 nm thick Cu ﬁlm deposited on H-

terminated Si(100) at 0.1-0.2 nm/s evaporation rate, without intentional heating. The
strong Cu(200) peak and absence of a Cu(l 11) peak indicate highly oriented growth.

75

The crystal structure of the ﬁlms was determined using X-ray diffraction (XRD).
The orientation in the direction perpendicular to the surface was characterized by standard
0-20 scans using Cu Kct radiation. Completely disordered (powdered) Cu exhibits a (111)
peak 2.17 times more intense than the (200) peak [20]. Polycrystalline Cu ﬁlms grown on
glass, alumina, or oxidized Si substrates typically show a larger (111):(200) ratio,
indicating preferential (111) orientation, [13,20] but both peaks are much weaker than the
Si(400) peak since only a small fraction of the ﬁlm volume contributes to each Cu peak.
A typical scan of a ﬁlm grown on an etched (H-terrninated) Si(100) substrate is shown in
Figure 4-1. Highly-oriented growth in the direction perpendicular to the surface, with the
Cu[100] direction aligned with Si[100], is signaled by a Cu(200) peak at least as strong as
the Si(400) peak, and no detectable Cu(l 1 1) peak.

The in-plane orientation of our ﬁlms was determined by means of X-ray pole
ﬁgures [21,22]. The source-detector angle was set to the Bragg angle for diffraction from
Cu(lll) planes, and the sample angle was varied over the full range of radial (O) and
azimuthal (‘1’) angles using a four-circle goniometer. Figure 4-2 shows a Cu(l 1 1) pole
ﬁgure [21] of a 100 nm thick copper ﬁlm. The radial angle 0 is the angle between the ﬁlm
normal and the plane of incidence; the azimuthal angle ‘1’ represents rotation about the
surface normal. A completely disordered ﬁlm would produce a uniform intensity,
independent of either angle. A (100) ﬁlm that was disordered in the plane would exhibit a
ring at (9 = 55°. The four pronounced Cu(] 11) poles in Figure 4-2 demonstrate that our
samples are fully epitaxial. By comparing the integrated intensity in the poles with the

background intensity we estimate that >95% of the sample volume is epitaxial. The width

76

 

 

 

 

 

     
       

2.78
1.00
0.36
0.13
0.05

9 (deg)

‘—.

0

180 =

 

30

  

60

 

Intensity
(arb. units)

270

 

 

 

Figure 4-2. Cu(l 1 1) X-ray pole ﬁgure of a 100 nm copper ﬁlm grown on H-temrinated
Si(100). Equal areas in the ﬁgure represent equal solid angles. The four sharp poles at G
= 55° demonstrate that the Cu ﬁlm is epitaxial. The inset shows a scan through one of the
poles in the azimuthal (‘1’) direction keeping the radial angle constant (O=55°). The width
of the pole is instrument-limited.

of the poles in the azimuthal direction, shown in the inset, is limited by the instrumental
resolution. A Si(lll) pole ﬁgure of the Si substrate was very similar in appearance,
except that the four poles were rotated by 45° in ‘1’, in agreement with earlier reports that
the copper lattice is rotated 45° with respect to the silicon lattice [13,14].

The surface morphology was studied in air using a commercial atomic force
microscope [23] (AFM) with nominal 0.1 nm vertical resolution and 2-5 nm lateral

resolution. Most of our scans represent a 1 x 1 pm2 area of the surface. The raw

77

10.0 mt

 

 

10.0 III

   

0
1.00
Ill

Figure 4-3. AFM topographical images of area 1 X l um2 for two different ﬁlm
thicknesses - a) 7.5 nm and b) 100 nm. The z-range for both images is 10 nm and the
RMS roughness is 0.71 nm for (a) and 0.95 nm for (b). Both ﬁlms were grown near room
temperature and were not annealed.

78
topographical images were software processed to eliminate the image bow in the y
direction, spurious horizontal stripes, and tilt in the x axis. The primary measure of
surface roughness used was the root-mean-square (RMS) deviation of the height from the
mean. Scans of bare Si substrates yielded RMS roughness values of 0.2 - 0.3 nm.

Figure 4-3 shows typical AFM images for samples of 7.5 nm and 100 nm
thickness. The deposition rate for both samples was 0.1-0.2 nm/s and the substrate was
not intentionally heated. The brightness represents surface height, with brighter areas
higher than darker ones, and both images use the same vertical and horizontal scales.
Both ﬁlms exhibit a granular structure, with individual grains 5-10 nm in diameter —
comparable to the ﬁlm thickness — for the 7.5 nm thick ﬁlm and 2-3 times larger for the
100 nm ﬁlm, with a few very large (50 nm) grains. The RMS roughness of the thinner ﬁlm
is 0.71 nm versus 0.95 nm for the 100 nm ﬁlm. Clearly the ﬁlms, although epitaxial, are
rough on the atomic scale. These measurements conﬁrm and quantify the conclusion of
Demczyk et al., [14] based on RHEED patterns, that the ﬁlm growth is three-dimensional.
On a larger length scale, however, the 100 nm ﬁlm is relatively smooth, varying in height
by a few nm over a lateral distance of tens or hundreds of nm.

X-ray diffraction from the 7 .5 nm thick ﬁlm revealed only a distinct Cu(200) peak,
as in Figure 4-1; there was no evidence of copper silicide peaks. At this thickness the ﬁlm
consists of grains of crystalline copper which are still not ﬁilly connected. If there is

intemiixing of the Cu and Si as reported by Demczyk et al., [14] it does not aﬁ‘ect the

crystalline stmcture.

79

We investigated the eﬂ‘ect of varying the deposition rate ﬁ'om 0.1 to 3.5 nm/s,
without intentional heating of the substrate. All of the ﬁlms showed good adhesion to the
substrate and appeared mirror-like with the characteristic red color of Cu. X-ray analysis
showed strongly oriented grth for deposition rates up to 2 nm/s, but ﬁlms deposited at
3.0 - 3.5 nm/s exhibited a Cu(] 1 1) peak comparable in intensity to the (200) peak,
indicating polycrystalline structure. All of the epitaxial ﬁlms showed surface morphologies
similar to that of Figure 4-3b, and the roughness depended only very weakly on deposition
rate. Figure 4-4 shows RMS roughness as a function of deposition rate. For rates up to 1
nm/s the RMS roughness is 0.5 - 2.0 nm; at higher deposition rates the roughness

increases to as much as 4 nm.

 

 

 

IIIVTYVTﬁIIIII'IIII

4b 'f' ..
’a‘ . .
a
3 3.. ..
E
50 P q
3 n—I—a
°‘ 2' i-I-r ‘
0
3 3+ +
g 6" r—I—t
U) 1% -
m 1i"
E _ "11$.

0 nunllnnnnlnnlnl

 

o 1 2 3 I ' I ' 4
Evaporation Rate (nm/s)

Figure 4-4. RMS roughness vs. deposition rate for 100 nm thick ﬁlms grown without
intentional heating of the substrate. Each data point represents a 1 X l umz area of a
different sample.

 

80
It is apparent from Figure 4-4 that ﬁlms grown under nominally identical
conditions can exhibit signiﬁcantly diﬁ‘erent surface roughness. Figure 4-5 shows an AFM
micrograph of a ﬁlm grown under conditions nominally identical to those used for Figure
4%. The lateral scale is the same as in Figure 4-3, but the vertical range is twice as great,
since the RMS roughness of the ﬁlm in Figure 4-5 is 1.82 nm, nearly double that of Figure
4-3b. Much greater coalescence of the crystallites has taken place, leading to larger grains

and deeper holes.

 

Figure 4-5. AFM topographic image of area 1 X l umz of a 100 nm Cu ﬁlm grown under
conditions nominally identical to those for the ﬁlm shown in Fig. 3b. The z-range is 20
nm, and the RMS roughness is 1.8 nm. Note the much larger grains and deeper holes.

8l

These variations are perhaps not surprising in view of the high background
pressure during deposition. We believe that they arise ﬁ'om uncontrolled variations in the
experimental procedure, such as pull-drying technique, time in air before pumpdown,
background pressure during deposition, atmospheric humidity, or variations in deposition
rate. In the case of Figures 4-5 and 4-3b, we suspect that the differences are related to
atmospheric humidity, which was much higher for the ﬁlm shown in Figure 4-5 than for
those shown in Figure 4-3. The bulk crystal structure, as revealed by XRD, was
insensitive to such variations.

Intentionally heating the substrate during deposition, to temperatures between
100°C and 250°C, resulted in extremely poor ﬁlms. The ﬁlms had a dark brown color, and
XRD scans showed no peaks attributable to Cu, but a variety of weak peaks similar to
those observed by Chang, [12] and attributed to a mixture of Cu3Si and Cu4Si. These
results agree with the observation of Demczyk et al. that substrate temperatures above (or
below) room temperature produce inferior ﬁlms [14].

We also investigated the effect of post-deposition annealing in vacuum, for ﬁlms
grown near room temperature at deposition rates of 0.1 - 0.2 nm/s. The results are
summarized in Table 4-1. Following deposition the stage was heated to an annealing
temperature (125, 150, 175 or 200°C) over a period of about 15 minutes. The stage
would then be held at the annealing temperature for 15 minutes and allowed to cool in
vacuum for about eight hours before the sample was removed from the evaporator for

characterization.

82

Table 4-1. Characteristics of 100 nm copper ﬁlms annealed in vacuum at different
temperatures. All of the ﬁlms were deposited at rates of 0.1 - 0.2 nm/s, with no
intentional heating of the substrate during deposition.

Annealing
Temperature

    

 

Visual Appearance

Crystal
Structure

      

Surface
Morphology

RMS
Roughness

   

 

 

 

 

 

 

 

 

not annealed reddish metal shine, epitaxial Cu grains 1-2 nm
mirror like
125°C reddish metal epitaxial Cu grains 1-2 nm
shine,mirror like
150°C reddish, cloudy epitaxial Cu clusters and 2 nm
holes
175°C reddish, cloudy epitaxial Cu channels 8 nm
200°C dull gray-brown Cu3 Si not studied not studied
L—L—————i———-———l—-—-i

 

 

Epitaxial structure was preserved for annealing temperatures up to 175°C. Films

annealed at 125°C were essentially indistinguishable ﬁom unannealed ﬁlms in surface

morphology and roughness. A ﬁlm annealed at 150°C exhibited the same RMS roughness

as an unannealed ﬁlm, but the AFM images showed the formation of larger copper grains

and the appearance of deep holes in the surface. Visually the surface appeared cloudy, as

though covered by a whitish haze. These features became more pronounced for ﬁlms

annealed at 175°C, as shown in Figure 4-6. The holes became larger and deeper,

extending almost to the surface of the Si (50-60 nm) and forming channel-like structures

around the enlarged copper clusters.

annealing temperature of 175°C.

The RMS roughness increased to 8 nm for an

83

4 ;:;.;.-; .

   

5 0.50 0.75 1.00
”It

0 0.2

Figure 4-6. AFM topographical image of a 100 nm Cu ﬁlm annealed in vacuum at 175°C
after deposition near room temperature. The z-range is 61 nm. The channels penetrate
deep into the ﬁlm, nearly to the Si surface.

The picture changed dramatically for the ﬁlms annealed at 200°C. Immediately
after removal from the evaporator they appeared shiny but gray, and XRD performed
within a few hours after removal from vacuum showed only a distinct peak attributable to
the (320) planes of Cu38i, indicating that the ﬁlm had reacted completely with the Si. This
is in agreement with the Chang’s observation that the (100) Cu ﬁlms react rapidly with Si
at 200°C [12] After the samples annealed at 200°C were kept in air for one day the color
changed to dull gray-brown and instead of the pronounced (320) Cu;Si peak a large
number of small peaks was detected by XRD. This behavior is consistent with rapid

oxidation of the copper silicide at room temperature.

4-3. Electrical conductivity

A signiﬁcant amount of work on the electrical properties of our ﬁlms was done by
Larry Voice. F our-wire DC resistance measurements were made of selected ﬁlms during
deposition. For these samples contact strips (150 nm Ag on top of 10 nm Cr) were
deposited on each end of the substrate. The substrate was then removed from the
evaporator and etched as described in section 11. Some pitting of the silver was
observable, but the contacts remained intact. The sample was mounted on the evaporation
stage using four phosphor-bronze clips pressing on the contacts, with separate electrical
connections for the current and voltage leads. Cu was then deposited at a rate of 0.1 - 0.4
nm/s to a thickness of 100-400 nm using the procedures described above. The resistance
of the ﬁlm was continuously monitored as a ﬁmction of ﬁlm thickness 1 during the
deposition.

Figure 4-7 shows the apparent ﬁlm resistivity p(t) as a function of t for a typical
sample. The resistivity drops rapidly with increasing thickness. The very large range of p
values, strong variation with t, uncertainty in the absolute thickness, and non-negligible
parallel resistance due to the substrate make it difﬁcult to analyze the data in this form. A
more tractable quantity is the conductance derivative dG(t)/dt, where G(t) is the
conductance of the ﬁlm at thickness t. This quantity is plotted in Figure 4-8 for the same
ﬁlm. It is apparent in Figure 4-8, for example, but not in Figure 4-7, that little change in
conductance occurs for about the ﬁrst 5 nm. Our use of the conductance derivative is
similar to the approach of Fischer, Hoffmarm and Vancea in their study of rough platinum

ﬁlms [24].

85

 

10‘"

 

10‘0_

p (0m)

10—7 ',

 

 

 

I“ L A I . l a L A I . l .
o 20 4o 60 80 100 120
Thickness (nm)

Figure 4—7. Apparent resistivity of a Cu ﬁlm of thickness t = 120 nm grown near room
temperature at approximately 0.1 nm/s. The dashed line is the best ﬁt to the Fuchs-
Sondheimer theory, Eq. 4-2. The solid line assumes an initial insulating layer of thickness
to, Eq. 4-5. The ﬁtting parameters are given in the text.

 

 

 

 

 

 

 

 

 

 

30
ﬂ
r—-r T
T A
A E
E 20 I
I c:
g 3.
t:
% 3
:3 1° o
1:
. o
o 20 40 60 so 100 120

Thi ckness(nm)

Figure 4-8. Effective conductivity 09190) (deﬁned by Eq. 1) of the same Cu ﬁlm as in
Fig. 7. The dashed line is the best ﬁt to the Fuchs-Sondheimer theory, Eq. 4-4. The solid
line assumes an initial insulating layer of thickness to, Eq. 4-5. The ﬁtting parameters are
given in the text. The inset shows the 0 - 30 nm region on an expanded scale. The arrow
indicates the conductivity of pure Cu at the deposition temperature.

To remove geometrical factors we deﬁne an effective conductivity:

mm)
06.3““) —W dt (4-1)

where W and L are the width and length of the ﬁlm, respectively. In the limit that the
thickness is much greater than the electronic mean free path, 0.17 can be interpreted as the
bulk conductivity of the layer of material at thickness 1. For uniform material Ceﬂ' should

approach the bulk conductivity at large thicknesses, as it does in Figure 4-8.

 

87
At smaller thicknesses, diffuse scattering of conduction electrons from surface
defects and grain boundaries reduces 0.1;. The effects of surface scattering are often
described within the Fuchs-Sondheimer model of the classical size effect, [24—32] which

gives an expression for the effective resistivity of a thin ﬁlm:

PU) = Po[1+‘§']

(4-2)
where p0 is the bulk resistivity of the material, and
'2 =_83_(1__ pv +psjlo
2 (4—3)

is an effective mean-ﬁee—path, with p. and p, the phenomenological specular scattering
probabilities at the ﬁlm-vacuum and ﬁlm-substrate interfaces respectively and lo the bulk
electron mean ﬁ'ee path. Although strictly valid only for I >> [0, Equation (4-2) has been
shown to be a good approximation for t/Eo > 0.1 [29-32]. For copper at room
temperature [0 z 39 nm. Ifthe intrinsic conductivity of the deposited material is uniform,
the effective conductivity deﬁned by Equation 4-1 will have the form:

22

ceﬂ(t):00 1-—~—
(t+ ()2 . (44)

where so is the bulk conductivity. Equations 4-2 to 4-4 assume plane parallel interfaces,
uniform ﬁlm composition and structure, and thickness-invariant surface scattering.
Signiﬁcant deviations from Equation 4-4 would imply that one or more of these

assumptions is violated.

 

88

The dashed curve in Figure 4-8 shows the best ﬁt to Equation 4-4, and clearly fails
to reproduce the data. Instead of increasing sharply at low I, the data are nearly ﬂat up to
about 6 nm. This indicates that the ﬁlm’s resistance does not initially decrease as rapidly
as we would expect for a uniform continuous ﬁlm of copper. Our XRD data for ﬁlms less
than 10 nm thick clearly show that the ﬁlm consists of crystalline copper and not, for
example, a high-resistance silicide. The most likely explanation, consistent with the AFM
images, is that at thicknesses less than about 6 nm the ﬁlm is discontinuous. Similar
results were observed for Pt ﬁlms by Fischer et al. [24]. In our measurements some
decrease in resistance occurs even at low thicknesses because the presence of the Cu
grains decreases the apparent resistance of the Si substrate.

Our goal is to determine whether the basic form of the 047(1) curve can be
explained by surface scattering, and to estimate the specularity parameter p. Since the
detailed form of the curve is rather complex and varied from run to run, we have not
attempted a detailed analysis in terms of a roughness parameter, such as was used in Ref.
24. As a ﬁrst approximation, we model the ﬁrst few nm of the ﬁlm as an insulating layer
of thickness to. Subsequent grth of uniform material will then give an effective

conductance curve of the form:

22
0,170): 00 1-——7—2— t2to
t—to + 3)
(4-5)

 

89

The solid line in Figure 4-8 shows the best ﬁt of Equation 4-5 to the data for t > 8 nm.
The best-ﬁt parameters for this ﬁlm were:

00 = 55 :3 (tram)-1

2’ = 15 :3 nm

to = 7.0 fl): nm . (4-6)
Although the model is clearly oversimpliﬁed it accounts for the major features of our
results. For comparison, the dashed and solid lines in Figure 4-7 also show the results of
the basic Fuchs-Sondheimer model (Equation 4-2) and the model with an insulating layer
of thickness to (Equations 4-5, 4-6).

Several conclusions can be drawn from these results. First, the value of do is close
to that of pure Cu at the deposition temperature, 54 (it!) m)-1, and the data are consistent
with this value being uniform for t > 8 nm. Above this thickness, then, the ﬁlm consists of
reasonably pure Cu (see below). Second, the value of 2 indicates a high level of diffuse
scattering. In fact, if Equation 4-3 is taken literally we ﬁnd that the average specularity
parameter pm = (p. + p,)/2 is very small or even negative: —O.37 < pm < 0.11. Negative
p values, while apparently unphysical, have been observed previously [30,33] and indicate
a breakdown in the approximations of the Fuchs-Sondheimer theory, notably the
assumption of plane parallel surfaces. The AFM images shown here clearly show the
inadequacy of this assumption. Nevertheless if surface scattering is dominant the low
value of 2 shows that the Cu conduction electrons see very rough boundaries at both

interfaces. The roughness of the Si/Cu interface may arise from the interdifﬁrsion of Cu

and Si as suggested by Demczyk et al. [ l4]

 

90
The low value of 2 could also, however, indicate a high level of internal grain-
boundary scattering rather than, or in addition to, scattering from rough interfaces.
Scattering from internal defects is normally assumed to be independent of thickness, but in
thin ﬁlm growth the grain size often increases with t, as our AFM images suggest. In that
case grain-boundary scattering can give a thickness-dependent resistivity that closely
approximates Equation (4-2) [28,29,34]. Our data cannot distinguish between these

mechanisms.

4-4. Conclusions

Cu ﬁlms grown on H-terminated Si( 100) represent an unusual and useful system.
Epitaxial growth occurs readily with little sensitivity to background pressure, deposition
rate or substrate temperature, and in spite of a 6% lattice mismatch and a rough and
atomically mixed interface. These remarkable features are worthy of study in their own
right, and also give the ﬁlms great utility as a seed layer, [1-9] a model system for training
students, [19] and a substrate for surface studies. We have determined the ﬁlms’ surface
morphology, which is smooth on the nm scale though not on the atomic scale (RMS
roughness 1-2 nm) and by quantifying the effects of deposition rate, substrate temperature
and post-deposition annealing — the smoothest ﬁlms are obtained near room temperature,
without annealing, at deposition rates below 3 nm/s.

For thicknesses below 5-10 nm the ﬁlms consist of disconnected grains, but those
grains are structurally indistinguishable from pure Cu. There is no silicide formation. The

high level of surface scattering observed in electrical measurements, however, suggests

 

91
that the Cu-Si interface, as well as the Cu-vacuum interface, is quite rough on the atomic
scale. It remains a puzzle how such excellent and robust epitaxy is achieved in a system
that departs so radically from the ideal of layer-by-layer growth. More detailed studies of
ﬁlm structure and composition during the initial stages of growth may help to resolve the

issue.

 

92

References

[1] E. T. Krastev, L. D. Voice, and R. G. Tobin, J .Appl.Phys. 79 (9), 6865 (1996).
[2] C.-A. Chang, Surf. Sci. 237, L421-L423 (1990).

[3] C.-A. Chang, J. Appl. Phys. 68, 5893 (1990).

[4] C.-A. Chang, J. Vac. Sci. Technol. A 8, 3779 (1990).

[5] C.- A. Chang, J. Vac. Sci. Technol. A 9, 98 (1991).

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

 

[7] Y.-L. Chen and Y.-T. Cheng, Mat. Lett. 15, 192 (1992).
[8] Y.-T. Cheng and Y.-L. Chen, Appl. Phys. Lett. 60, 1951 (1992).

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

[10] T.Ohrni, T.Saito, T.Shibata, and T.Nitta, Appl. Phys. Lett. 52, 2236 (1988).

[11] M.Sonowski, H. Usui and I. Yamada, J. Vac. Sci. Technol. A 8, 1470 (1990).

[12] C.-A. Chang, J. Appl. Phys. 67, 566 (1990).

[13] C.-A. Chang, J.C. Liu, and J. Angilello, Appl. Phys. Lett. 57, 2239 (1990).

[14] B.G.Demczyk, R.Naik, G. Auner, C.Kota and U.Rao, J. Appl. Phys 75, 1956 (1994).

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

[16] T. Takahagi, I. Ishiytani, H. Kuroda, Y. Nagasawa, H.Ito and S.Wakao, J .Appl.
Phys. 68, 2187 (1990).

[17] Y.J. Chabal, G.S. Higashi, K. Raghavachari and VA. Burrows, J. Vac. Sci. Technol.
A 7, 2104 (1989).

93

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

[19] Minsu Longiaru, E. T .Krastev, and R. G. Tobin, J. Vac. Sci. Technol., A, 14(5),
(1996)

[20] J. Yang, C. Wang, K. Tao and Y. Fan, J. Vac. Sci .Technol. A 13, 481 (1995).

[21]J.S.Kallend, U.F.Kocks, A.D.Rollett, H.-R.Wenk, Materials Science and Engineering,
A 132, l ( 1992).

[22] H.-R.Wenk, Preferred Orientation in Deformed Metals and Racks: An Introduction
to Modern Texture Analysis, (Academic Press, New York, 1985).

[23] Digital Instruments Nanoscope III Scanning Probe Microscope.

[24] G. Fischer, H. Hoffmann and J. Vancea, Phys. Rev. B 22, 6065 (1980).
[25] AA Baski and H. Fuchs, Surf. Sci. 313, 275 (1994).

[26] K. Fuchs, Proc. Cambridge Phi. Soc. 34, 100 (193 8).

[27] E.H. Sondheimer, Adv. Phys. 1, 1, (1952).

[28] P. Wissmann, in Surface Physics, edited by G. H6hler, Springer Tracts in Modern
Physics 77 (Springer, New York, 1975).

[29] D. Schumacher, Surface Scattering Experiments with Conduction Electrons, edited
by G. Hohler, Springer Tracts in Modern Physics 128 (Springer, New York, 1993).

[30] H. Sugawara, T. Nagano and A. Kinbara, Thin Solid Films 21, 33 (1974).

[31] H. Sugawara, T. Nagano, K. Uozumi and A. Kinbara, Thin Solid Films 14, 349
(1972)

[3 2] J. Bass, in Landolt Banistein Numerical Data and Functional Relationships in
Science and Technology, vol. 15, edited by K.-H. Hellwege (Springer-Verlag, Berlin,
1982)

[33] D. Dayal and P. Wissmann, Thin Solid Films 44, 185 (1977), and references therein.
[34] AF. Mayadas and M. Shatzkes, Phys. Rev. B 1, 1382 (1970).

[3 5] L.J. van der Pauw, Phillips Research Reports 13, 1 (1958).

 

Chapter 5

Multiple mechanisms for adsorbate-induced resistivity: Oxygen and
formate on Cu(100)

5-1. Introduction

In Chapter 4 I discussed the techniques of growing of epitaxial Cu(lOO) thin ﬁlms
on Si(100) substrates in high vacuum (HV, base pressure 10" Torr) and their crystal
structure, surface morphology and electrical conductivity. In Chapter 3 I presented the
equipment we build in order to grow the Cu( 1 00) ﬁlms in ultra-high vacuum (UHV; base
pressure < 2x10'lo Torr). This chapter is dedicated to simultaneous in situ adsorbate-
induced reﬂectance and resistivity change measurements of our thin copper ﬁlms,
performed in UHV.

Recent years have seen a resurgence of interest in adsorbate-induced changes in
the electrical transport properties of metals. It has been known for many years that
adsorbates change (usually increase) the resistance of thin metal ﬁlms [1-6]. Models
proposed to explain these changes include modiﬁcation of the conduction electron density
in the near-surface region caused by direct charge transfer to (or from) the adsorbed
molecules [7,8]; a reduction in the effective thickness of the ﬁlm [9], and the scattering of
conduction electrons by the localized potential of the adsorbate [1,3,10,11]. The

scattering model has generally been the most successful and widely accepted.

94

 

95

With the development of high-quality surface infrared spectroscopy, adsorbate-
induced decreases in the broadband reﬂectance (i. e. far from any vibrational resonance) of
bulk metal crystals have been measured [12-24]. The changes are too large and of the
wrong sign to be explained by the dielectric properties of the adsorbed layer itself [25], so
they are attributed to changes in the substrate’s electronic properties. Early results on W
and Mo were attributed to a surface electronic state [15] and such a state has also been
invoked as a partial explanation for H on Cu(] 11) [19]. A series of careful spectroscopic
experiments on Cu [16-22] and Pt [24] single crystals have strongly supported the
scattering model. Theoretical work by Persson and Volokitin has extended the model to
explain resistivity changes, reﬂectance changes, and antiabsorption resonances due to
hindered rotations and translations of the adsorbates [18,19,26-30].

The scattering model predicts that the fractional reﬂectance change AR/R of a bulk
sample, for p-polarized light at near-grazing incidence, is proportional to the dc resistivity

change Ap of a thin ﬁlm of the same material under the same surface conditions [20,26]:

A); = {Mi—jam.) . <5-1)

mccosB

Here t is the ﬁlm thickness, n is the electron density, assumed to be unaffected by the
adsorbates; m is the effective mass, and 6 is the angle of incidence of the light. The
frequency must be sufﬁciently high, (0 >> 1/t, vp/B, where r is the bulk electron scattering
time, v; is the Fermi velocity and 6 is the classical skin depth. The quantity tAp is
independent of ﬁlm thickness [1,3]. The proportionality between AR/R and tAp is quite

general. It requires the Dmde approximation for the metal’s dielectric function, that the

 

96
adsorbates affect only 'c and not n, and a slab model for the depth dependence of the
conductivity [20]. Most modiﬁcations of the model, such as relaxing the slab model, using
separate scattering times for dc and optical ﬁequencies, incorporating ﬁlm anisotropy,
etc., will affect the value of the proportionality constant, but not the structure of Eq. 5-1,
in which AR/R and tAp are related through a constant that depends only on the substrate,
not on the adsorbate.

There have been only a few attempts to test Eq. 5-1. Lin et al. compared
reﬂectance measurements on bulk single crystals with published resistivity data for
polycrystalline ﬁlms, and found rough agreement for CO and 0 on Cu, but a large
discrepancy for CO on Ni [20]. Because the reﬂectance and resistance measurements
were made on different samples under different conditions, the comparison was
unavoidably rough. Hein and Schumacher made simultaneous near-IR reﬂectance and
resistance measurements for oxygen, C0 and Cu on Cu ﬁlms and for Ag on a Ag ﬁlm
[31]. They conﬁrmed the proportionality predicted by Eq. 5-1, but their coefﬁcients were
larger than predicted by factors of 1.2 - 3.5 and varied considerably ﬁom sample to
sample. Their ﬁlms were deposited by thermal evaporation onto glass slides, giving
polycrystalline ﬁlms with probably a predominant (111) texture [3].

We present the results of simultaneous measurements of the changes in do
resistivity and nonresonant infrared reﬂectance induced by oxygen and formate adsorption
on epitaxial Cu(100) thin ﬁlms. For oxygen we conﬁrm the linear relationship predicted
by the scattering model. Fonnate adsorption on identically prepared samples, however,

caused no measurable reﬂectance change, despite a resistivity change comparable to that

 

97
induced by oxygen. This surprising behavior strongly contradicts the scattering model. It
implies, ﬁrst, that formate is a very weak scatterer and, second, that formate induces a
resistivity change through a different mechanism, probably a reduction in conduction
electron density.
5-2. Experimental

The samples were epitaxial Cu(lOO) ﬁlms grown in situ in ultrahigh vacuum on
Si(100) substrates. Contact strips (150 nm Ag on 10 nm Cr) were deposited on 1 x 2 cm
pieces of Si in a separate evaporator. The samples were then etched in a 10% aqueous HF
solution as described elsewhere [3 2,33], removing the native oxide and leaving a
hydrogen-passivated surface. The sample was mounted on a custom-designed sample
holder that incorporates four electrical contacts for the resistivity measurements and two
thermocouple contacts and provides a temperature range of 100 to 900 K [34] and
immediately (within a few minutes) transferred into UHV via a high-vacuum load-lock.
The sample was baked in UHV at 725 K for 15 minutes to remove impurities before
deposition of the Cu.

The Cu ﬁlms were grown by resistive thermal evaporation. A liquid-nitrogen-
cooled shield maintained the pressure during deposition in the range of 5 - 10 x 10'10 Torr.
The sample temperature was 300 - 310 K, the deposition rate was 0.1 - 0.5 nm/s and the
ﬁlm thickness was typically 50 nm, measured with a quartz-crystal thickness monitor
calibrated against a diamond-stylus proﬁlometer. We have shown that ﬁlms grown in this
way in HV are fully epitaxial, but have rrns surface roughness of 1-2 nm [32]. X-ray

diffraction and AFM studies of the UHV-grown ﬁlms showed that they are epitaxial and

 

98

with a comparable surface roughness to the HV-grown ﬁlms. The ﬁlms exhibited hazy
low energy electron diﬂ‘raction spots characteristic of their (100) orientation, conﬁrming
epitaxy and indicating surface ordering on a length scale ~10 nm. Auger electron
spectroscopy detected carbon and oxygen at levels of less than 5 atomic percent, with no
other impurities above ~2%. The samples had room temperature resistivities of 3 - 6 HQ
cnr, two to four times higher than pure Cu, independent of ﬁlm thickness (for t > 50 nm).
Copper ﬁlms of comparable thickness (50 nm) grown in high-vacuum conditions show
room temperature resistivity of 3 110 cm (see Figure 4-7) and at higher thickness their
resistivity approaches the value for the pure copper (1.7 it!) cm). Currently we are
investigating the reason(s) for the higher resistivity of the UHV-grown ﬁlms.

The inﬁ'ared spectroscopic system has been described elsewhere [24]; it permits
measurements of nonresonant reﬂectance changes at a ﬁxed frequency (350 - 3000 cm")
with a sensitivity of 0.02%. All measurements reported here were made at a frequency of
2650 cm'l, well away ﬁ'om any vibrational resonances and in a frequency range where the
approximations appropriate to Eq. 5-1 are well satisﬁed [21]. The ﬁlm resistance was
measured by a quasi-four-wire ac lock-in technique (sensitive to contact resistance but not
to lead resistances), using a 1.0 mA current modulated at 3 kHz. Fractional resistance
changes AV/ V < 10" were easily measurable.

All measurements were made at room temperature. Gas exposures were
accomplished by backﬁlling the chamber and were calculated ﬁ'om the uncorrected
pressure measured at the ionization gauge. The formic acid was puriﬁed by repeated

freeze-pump-thaw cycles. F onnic acid is known to dissociate on Cu, leaving adsorbed

99
formate (HCOO) and hydrogen [22,3 5,36]. At room temperature we expect the hydrogen
to leave the surface. Auger spectroscopy after formic acid dosing revealed increased C

and O signals consistent with formate adsorption.

5-3. Results and Discussion

A typical combined resistivity and reﬂectance measurement for oxygen on Cu(100)
is shown in Figure 5-1. Figure 5-la shows the ﬁ'actional reﬂectance change of a Cu(lOO)
ﬁlm (thickness 49 nm) as a function of time. Figure 5-lb shows the fractional resistivity
change. A linear baseline has been subtracted from both curves. At 100 s, 5 x 10'7 Torr
02 was introduced into the chamber for 100 s. The reﬂected IR intensity decreased
sharply, accompanied by a sharp increase in the resistivity. The maximum fractional
reﬂectance change AR/R was —0.0070i0.0005 and the maximum fractional resistivity
change Ap/p was 0.032. When the oxygen was turned off, the reﬂectance increased (and
resistance decreased) slightly, probably due to oxygen leaving the surface.

In the period before and after dosing, the rrns ﬂuctuation in AR/R is 0.0004; this
represents the uncertainty in each point. Since we have 40-50 independent measurements
both before and after the dose, the statistical uncertainty in the overall reﬂectance change
is reduced to 0.0001. The uncertainty in AR/R, however, is generally dominated by

systematic uncertainty in the baseline subtraction.

100

 

.001 ‘ '
0 Infrared Light, 2650 cm’1

0.000
-o.oor
p.002

g p.003
-o.004
-o.oos
-0.006

-0.007
-0.008

 

1' V
I

 

 

Reflectance

‘ I

 

 

l a l

 

Dose 50 L Oxygen

 

0.04 ' f V I V v ' I r I

Resistivity

F
l

0.03

i
L

0.02

0.01

I
I

 

 

0.00 ..

 

 

L

 

-0.01 L ‘ ‘ ‘ ‘ ‘
0 50 1 00 150
time (s)

200 250

Figure 5-1. a) Fractional reﬂectance change at 2650 cm'1 for oxygen on epitaxial
Cu(lOO). The ﬁlm was 49 nm thick. At 110 s the sample was exposed to 5x10” Torr O;
for 100 s. b) Fractional dc resistivity change measured simultaneously with (a).

101
Figure 5-2 shows a plot of AR/R vs. tAp for the dosing period, ﬁ'om Figure 5-1.
The error bar, which is the same for all the points, represents the single-point rms noise of
i0.0004. Although neither the resistivity nor the reﬂectance varies linearly with exposure,

the relationship between them is linear, as predicted by the scattering model (Eq. 5-1).

 

I 1 I ' I ' l r I

Dose 50 L Oxygen

    

0.000 °

-0.002

g .0.004

-0.006

I

I

I n I n I n l J

o 2 l 4 6 8 10 12
tAp (10'13 Qcmz)

 

 

-0.008 ‘

 

Figure 5-2. Fractional reﬂectance change AR/R vs. tAp for oxygen on epitaxial Cu(lOO),
from Fig. 1. The error bar shown is typical of all the points.

The value of nezlmc deduced from the slope is 1.2 x 108 Q" cm‘z. The failure of the line
to pass through the origin is attributable to a small discrepancy in the time scales of the

two measurements; it does not affect the value of the slope. We performed ﬁve similar

102
experiments on different samples, always obtaining a linear relationship between AR/R and
tAp, with values of nezlmc varying from 0.6 to 1.2 x 108 Q" cm’z. Table 5-1 compares
our value of nezlmc for oxygen on Cu with the data of Hein and Schumacher [31]. The

experimental values show a wide spread and in most cases deviate from the theoretical

Table 5-1. Values of the coefficient nez/mc deduced from the slope of AR/R vs. tAp.

 

 

ne’lmc Source
[10° a" cm'2]
Theoretical for pure Cu 0.61
O/Cu(100) 0.06 - 0.12 This work
O/poly Cu 2.13 Ref. 31
CO/poly Cu 1.06 Ref. 31
Cu/poly Cu 0.71 Ref. 31
Formate/Cu(100) < 0.01 This work

 

value based on the electron density in bulk copper. We return to these numerical
discrepancies below; at this point we emphasize that the linear relationship predicted by
the scattering model is well conﬁrmed for oxygen on Cu.

A more stringent test of the scattering model can be made by comparing different
adsorbates on identically prepared substrates. Figure 5-3 presents a combined resistivity-
reﬂectance measurement for formic acid dosing on a Cu(lOO) thin ﬁlm (thickness 51 nm)

grown under conditions identical to those used for Figure 5-1. The resistances of the two

103

 

0.000 ‘

 

 

 

-0.002 ‘ Reflectance ‘

, l

E; 41004~ *

-0.006 ' -

Infrared Light, 2650 cm“1 '

-0.008 ' , . . '
,t .

 

Dose 55 L Formic Acid

 

' V I f

0.010 . Resistivity

Ap/p
O
O
O
U!

 

 

 

 

0 i 100 i 200 l 300
Time (s)

Figure 5-3. a) Fractional reﬂectance change at 2650 cm'1 for formate on epitaxial
Cu(lOO). The ﬁlm was 51 nm thick. At 130 s the sample was exposed to 5 x 10'7 Torr
formic acid for 110 s. No reﬂectance change is observable. b) Fractional dc resistivity
change measured simultaneously with (a).

104
ﬁlms diﬂ‘ered by only 15%. Similar results were obtained in three other experiments on
different samples. The formic acid dose was 5 x 10'7 Torr for 110 s. The maximum Ap/p
was 0.011. In sharp contrast to the oxygen case, however, no reﬂectance change is
detectable. Based on the noise level and baseline stability, we can set an upper limit of
0.0002 on AR/R, leading to a nominal value of nezlmc less than 0.1 x 108 Q" cm'z, at least
a factor of ten smaller than for oxygen on an identically prepared sample.

This striking deviation from the scattering model demonstrates, ﬁrst, that the
scattering cross section for formate is anomalously small, and second, that another process
must be responsible for the resistance change. Assuming a formate coverage of 0.5
monolayer (ML: 1 ML represents one formate per surface copper atom), we set an upper
limit on the scattering cross section for formate of 0.04 A2, as compared with 0.8 A2 for
CO on Pt(111), itself a weak scatterer [24]. The lack of a reﬂectance change for formate
on Cu(100) was previously reported by Lin et al. [21,22] from single crystal experiments,
and tentatively explained in terms of the absence of an adsorbate-derived electronic state
close to the Fermi level [22].

The resistance change for formate is most readily explained by a change in the
conduction electron density n. Within a Drude model, which is a good approximation for
Cu at infrared frequencies [37], the resistivity and reﬂectance changes of a ﬁlm of

thickness t are given to ﬁrst order by [15,20]:

_ n t

angels]
p t

(5-2)

 

105

where the changes in n and r are assumed to affect a slab of thickness [3 = VF‘L', the bulk

electron mean ﬁ'ee path, and t is larger than the skin depth. (Eq. 5-1 is recovered if An =
0.) A change in n will affect the resistivity but not the reﬂectance, as we observe for
formate [3 8]. If we assume a bulk electron density n equal to that of pure Cu and a
maximum formate coverage of 0. 5 ML, the observed resistivity changes require that each
formate molecule bind approximately nine conduction electrons — an implausibly large
number.

We suggest, however, that n in our ﬁlms is considerably smaller than in pure Cu.
If the high room temperature resistivity of our ﬁlms compared to pure Cu (see section 2)
is due to n, then each formate would need to bind only 2-3 electrons to account for our
results, a far more plausible number. A reduced value of it would also partially explain the
low values of nezlmc deduced ﬁ'om our reﬂectance vs. resistance curves (Table 5-1). We
observe a weak but positive correlation between ne2/mc and the ﬁlm conductivity, which
tends to support this hypothesis. Reduced n (by as much as a factor of 50) has been
observed previously in ﬁlms of Cu and other metals and explained by localization of
electrons within microcrystallites or at grain boundaries [11,39-41].

In summary, we have found that the mechanism for adsorbate-induced resistance
changes is chemically speciﬁc. While the scattering model successfully explains results for
a number of systems, including oxygen on Cu(lOO), it fails in the case of formate on
Cu(lOO). On identically prepared ﬁlms, the ratio of the reﬂectance change to the
resistivity change is at least a factor of ten smaller for formate than for oxygen. It is likely

that the resistance increase induced by formate is caused by a reduction in the conduction

106

electron density. Chapter 5 is based on our paper, published in Surface Science Letters

[42].

107

References

[1] P. Wissmann in Surface Physics, ed. G. Hohler, Springer Tracts in Modern Physics
Vol. 77 (Springer, New York, 1975).

[2] D. Dayal, H.-U. F inzel and P. Wissmann in Thin metal ﬁlms and gas chemisorption,
edited by P. Wissmann (Elsevier, Amsterdam, 1987).

[3] D. Schumacher, Surface Scattering Experiments with Conduction Electrons, edited by
G. thler, Springer Tracts in Modern Physics Vol. 128 (Springer, New York, 1993).

[4] T.J. Coutts, Electrical Conduction in Thin Metal Films (Elsevier, Amsterdam, 1974).
[5] MR. Shanabarger, J. Vac. Sci. Technol. A 4, 623 (1986).

[6] AF. Hebard, R.R. Ruel and CB. Eorn, Phys. Rev. B 54, 14 052 (1996).

[7] R. Suhrmann and K. Schulz, Z. Phys. Chem. (Frankfurt) 1, 69 (1954).

[8] G. Wedler and M. Fouad, Z. Phys. Chemie Neue Folge, 40, 12 (1964).

[9] W.M.H. Sachtler and G]. H. Dorgelo, Z. Phys. Chem. (Frankfurt) 25, 69 (1960).
[10] K. Fuchs, Proc. Cambridge Phil Soc. 34, 100 (1938).

[11] E.H Sondheimer, Adv. Phys. 1, l (1952).

[12] D.M. Riffe, L.M Hanssen and A.J. Sievers, Phys. Rev. B 34, 692 (1986).

[13] D.M. Riffe, L.M Hanssen and A.J. Sievers, Surf. Sci. 176, 679 ( 1986).

[14] VA. Burrows, S. Sundaresan, Y.J. Chabal and SB. Christman, Surf. Sci. 180, 110
(1987)

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

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

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

108

[18] C]. Hirschmugl, G.P. Williams, B.N.J. Persson and AI. Volokitin, Surf. Sci. 317,
L1141 (1994).

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

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

[21] KC. Lin, R.G. Tobin and P. Dumas, Phys. Rev. B 49 (1994) 17 273; ibid 50, 17 760
(1994)

[22] KC. Lin, R.G. Tobin and P. Dumas, J. Vac. Sci. Technol. A13, 1579 (1995).
[23] J. Dvorak, E. Borguet and H.-L. Dai, Surf. Sci. 369, L122 (1996).

[24] DE. Kuhl, K.C. Lin, C. Chung, J .S. Luo, H. Wang and R.G. Tobin, Chem. Phys.
205, l (1996).

[25] R.G. Tobin, Phys. Rev. B45, 12 110 (1992).

[26] B.N.J. Persson, Phys. Rev. B 44, 3277 (1991).

[27] B.N.J. Persson, Chem. Phys. Lett. 197, 7 (1992).

[28] B.N.J. Persson and AI. Volokitin, Surf. Sci. 310, 314 (1994).
[29] AI. Volokitin and B.N.J. Persson, Phys. Rev. B 52, 2899 (1995).

[30] P. Dumas, M. Suhren, Y.J. Chabal, C.J. Hirschmugl and GP. Williams, Surf. Sci.
371, 200 (1997).

[31] M. Hein and D. Schumacher, J. Phys. D 28, 1937 (1995). Their Eq. 6 differs from
my Eq. 5-1 because of their use of unpolarized light and a smaller angle of incidence. We
have used their Eq. 6 in analyzing their data.

[32] ET. Krastev, L.D. Voice and R.G. Tobin, J. Appl. Phys. 79, 6865 (1996).

[33] M. Longiaru, E.T. Krastev and R.G. Tobin, J. Vac. Sci. Technol. A 14, 2875 (1996).
[34] ET. Krastev and R.G. Tobin, submitted to J. Vac. Sci Technol.

[35] BA. Sexton, Surf. Sci. 88, 319 (1979).

109

[36] PA. Taylor, P.B. Rasmussen, C .V. Ovessen, P. Stoltze and I. Chorkendorff, Surf.
Sci. 261, 191 (1992).

[37] MA. Ordal, R.J. Bell, R.W. Alexander, Jr., L.L Long and MR. Querry, Appl. Opt.
24, 4493 (1985).

[3 8] We are assuming that 1 remains constant; this diﬂ‘ers slightly ﬁ'om the assumption of
constant (’3 considered in Refs. 1,6-8.

[39] H. Hoffrnann and J. Vancea, Thin Solid Films 85, 147 (1981).
[40] J. Vancea and H. Hoffrnann, Thin Solid Films 92, 219 (1982).
[41] J. Vancea, H. Hoffrnann and K. Kastner, Thin Solid Films 121, 201 ( 1984).

[42] E. T. Krastev, D. E. Kuhl and R. G. Tobin, Surf. Sci. Lett., 387, L105], (1997)

 

Chapter 6

Conclusion

The way adsorbed molecules affect the electron dynamics in the near surface
region of metals has been actively investigated both experimentally and theoretically
during the last years. A theory which correlates the presence of alien molecules on the
clean metal surface with an increased diffuse scattering of the conduction electrons [1-4]
has been established as dominant. The theory predicts a linear relationship between
reﬂectance change AR/R and resistivity change Ap, with a proportionality constant that
does not depend on the type of adsorbate. The research presented in this dissertation was
concentrated on testing the above model by experimental examining the relationship
between AR/R and Ap.

My work comprised the following three stages:

1. Design of UHV sample transfer, sample handling and thin ﬁlm evaporation
equipment. The developed system was optimized for infrared reﬂectance and resistivity
measurements and incorporates several features that are not commercially available --

grazing optical access and transfer of large samples with six breakable electrical contacts.

110

1 l 1
2. Development of methods for growing epitaxial Cu(100) thin ﬁlms on H-
terrninated silicon substrates. Thorough characterization of their crystal structure, surface
morphology and electronic properties. Despite the large lattice mismatch between copper
and silicon, etching of the substrates in aqueous solution of hydroﬂuoric acid (HF) allows
growing of highly epitaxial ﬁlms by simple thermal evaporation and with minimal in situ
preparation. Although epitaxial, the ﬁlms are microscopically rough, showing RMS

surface roughness of 1-2 nm on a 1pm) scale.

3. Investigation of the changes in the resistivity and the broadband reﬂectance of
the ﬁlms caused by adsorbates. Simultaneous in situ resistivity-reﬂectance measurements
were performed using oxygen and formate as adsorbates and the results were analyzed

ﬁ'om the point of view of the scattering theory.

Although the heart of my study is the resistivity-reﬂectance change experiments, all
three stages should be considered as a whole as the ﬁrst two were crucial for performing
accurate measurements and obtaining reliable and respectable results. Most of my work
was presented in four articles, published in reputable scientiﬁc journals [5-8].

For the case of oxygen adsorption the linearity between AR/R and Ap was
observed in agreement with the scattering mechanism. However a discrepancy was found
between the experimental and theoretical values of the proportionality constant. The
lower value found in our experiments could be attributed in part to a lower, compared to

bulk, concentration of conducting electrons in our ﬁlms. Lower conduction electron

112
concentration has been experimentally observed before [9,10] and is consistent with the
fact that our ﬁlms have higher resistivity than single crystal copper.

For the case of formate adsorption on identically prepared ﬁlms a considerable
resistivity change was found with no measurable reﬂectance change. This result is in
striking contradiction with the statement of the scattering theory, that the relationship
between AR/R and Ap is adsorbate independent. Therefore it can be concluded that the
mechanism of the adsorbate induced resistivity change is chemically speciﬁc and a
mechanism other than scattering must be responsible for the resistivity change observed
for formate. One possible reason could be a decrease in the conduction electron density
due to binding of electrons by the formate species on the surface.

Obviously there is an imminent need for more experimental work exploring other
adsorbate - thin ﬁlm systems. A good potential candidate is the system O/Ni(100), which
could also show a deviation from the scattering model, as suggested by a previous study
[14]. The investigation of the above system would be facilitated by the fact that epitaxial
Ni(100) ﬁlms can be easily grown on silicon, using copper as a seed layer [20].

The scientiﬁc signiﬁcance of my work can be seen in the following aspects:

1. This is the ﬁrst study experimentally probing the relationship between the
adsorbate induced resistivity and reﬂectance change, performed in situ and on well

characterized epitaxial ﬁlms.

113

2. My work rigorously demonstrates a deviation from the scattering mechanism.
Previous studies performed on single crystals [1 1-18] and thin ﬁlms [19] generally show
support to the scattering theory to greater or lesser extent. This work indicates that other
mechanisms or a combination of several should also be considered while trying to clarify

the effects of adsorbates on the electron dynamics in the close vicinity of the metal surface.

3. My research has been directly involved with the issue of energy transfer between
moving adsorbates and substrate degrees of freedom, which is of large importance in the
fast developing ﬁeld of nanoscale tribology -- the study of atomic scale friction. It also
rises several intriguing questions for the surface scientist. What are the mechanisms
underling the formate induced resistivity change? Is the relation between resistivity and
reﬂectance change adsorbate independent even for systems where the scattering

mechanism is dominant and the linearity holds?

4. If the resistivity and reﬂectance changes are not so simply related by adsorbate
independent constant, the combination of the two measurements could provide a measure

of chemical selectivity in potential gas sensing methods.

Based on the above I consider my study as a strong motivation for further
experimental and theoretical investigation aimed at better understanding of the
complicated phenomena occurring at the gas-metal interface and their possible

implementation in gas sensing applications. Apart ﬁ'om studying other adsorbate -

l 14
substrate systems, further experimental work concerning coverage, temperature and
frequency dependence of the adsorbate induced resistivity-reﬂectance change would be
strongly beneﬁcial. From a theoretical point of view further development is needed in
order to explain the formate/Cu(lOO) behavior and build a more integral picture of the
acting mechanisms, through which different adsorbates affect the electron dynamics in the

near-surface region of metals.

115

References

[1] B.N.J. Persson, Phys. Rev. B 44, 3277 (1991).

[2] B.N.J. Persson, Chem. Phys. Letters 185, 292 (1991) .

[3] B.N.J. Persson, Chem. Phys. Letters 197, 7 (1992).

[4] B.N.J. Persson and AI. Volokitin, Surface Sci. 310, 314 (1994).

[5] E. T. Krastev, L. D. Voice, and R. G. Tobin, J .Appl.Phys. 79 (9), 6865 (1996).

[6] Minsu Longiaru, E. T .Krastev, and R. G. Tobin, J. Vac. Sci. Technol. A, 14(5), 2875
(1996)

[7] E. T. Krastev, D. E. Kuhl and R. G. Tobin, “Multiple mechanisms for adsorbate-
induced resistivity - Oxygen and Forrnate on Cu(lOO)”, in press Surf. Sci. Lett.

[8] E. T. Krastev and R. G. Tobin, “Ultrahigh vacuum sample transfer system with
multiple electrical connections and grazing optical access”, submitted to J. Vac. Sci.
Technol.

[9] J. Vancea, H. Hoffrnann, Thin Solid Films, 92, 219 (1982).

[10] J. Vancea, H. Hoffrnann, K. Kastner, Thin Solid Films, 121, 201 (1984).

[11] CI. Hirschmugl, G.P. Williams, FM. Hoffrnann and Y.J. Chabal, Phys. Rev. Letters
65, 480 (1990).

[12] C]. Hirschmugl, Y.J. Chabal, FM. Hoffmann and GP. Williams, J. Vac. Sci.
Technol. A 12,2229 (1994).

[13] C]. Hirschmugl, G.P. Williams, B.N.J. Persson and AI. Volokitin, Surface Sci. 317,
L1141 (1994).

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

[15] KC. Lin, R.G. Tobin and P. Dumas, Phys. Rev. B 49 (1994) 17273 (1994); 50,
17760 (1994).

 

116

[16] BA Sexton, Surf. Sci. 88, 319 (1979); PA. Taylor, P.B. Rasmussen, C.V. Ovessen,
P.Stoltze, and I. Chorkendorff, Surf. Sci. 261, 19] (1992).

[17] DE. Kuhl, K.C. Lin, Chilhee Chung, J .S. Luo, Hong Wang and R.G. Tobin,
Chemical Physics 205,1 (1996).

[18] C.L.A Lamont, B.N.J. Person and GP. Williams, Chem. Phys. Lett. 243, 429 (1995).
[19] M. Hein and D. Schumacher, J. Phys. D 28, 1937 (1995).

[20] Chin-An Chang, J. Vac. Sci. Technol. A 8, 3779 (1990).