. ...
2: ..
23.1..»
ft? pl
.2941.
.:

r: .10
i. r 7“.

3
1W -
dark
iguana

 

.0 ’31:
‘1': «I

.. L
‘1‘; Us.
. {.1714
< \ l .

‘
V .

 

. $§
I.
fl‘v.txt
A...- v. 30; o
.3. wVvI-dl.

 

H‘|.|.rj.V...L-.. a; It'bl'. |

 

THESIS

ERSIIBTYL BRARIES

IHII IIHI l

 

II

 

 

 

 

 

 

 

 

 

 

                         

HHIWHIWlllltilHiU

3 129300

This is to certify that the

dissertation entitled

DEVELOPMENT OF A NEW FAR-INFRARED REFLECTION-
ABSORPTION SPECTROMETER AND MEASUREMENTS OF
DIOXYGEN ON THE Pt(111) SURFACE

presented by

Chilhee Chung

has been accepted towards fulfillment
of the requirements for

Ph.D. , Physics
degree in

 

 

Q7 774

flMajor professor

Date July 22, 1993

 

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

 

 

LIBRARY
Michigan State
University

 

I

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

‘ DATE DUE DATE DUE DATE DUE

r j: ‘J‘

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

7 T f TA
_1

 

 

 

 

zfi

 

 

 

 

 

 

 

 

MSU Is An Affirmative Action/Equal Opportunity Institution
cWUnS-o.|

 

 

 

 

 

 

 

 

 

 

DEVELOPMENT OF A NEW FAR-INFRARED REFLECTION-
ABSORPTION SPECTROMETER AND MEASUREMENTS OF
DIOXY GEN ON THE Pt(111) SURFACE

By

Chilhee Chung

A DISSERTATION

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

DOCTOR OF PHILOSOPHY .

Department of Physics and Astronomy
and Center for Fundamental Materials Research

1993

ABSTRACT

DEVELOPMENT OF A NEW FAR-INFRARED REFLECTION -
ABSORPTION SPECTROMETER AND MEASUREMENTS OF
DIOXYGEN ON THE Pt(111) SURFACE

By

Chilhee Chung

Infrared spectroscopy has long been widely used for examining
adsorbates on solid surfaces, but until recently its use was largely
confined to high frequency modes (above ~100013Ed), mostly internal modes
of molecular adsorbates. In recent.years several groups have successfully
measured the adsorbate-substrate vibrations of molecular adsorbates, and
in ‘very limited cases atomic adsorbates, which lare usually at low
frequencies below 1000 cmfl. Vibrational spectroscopic measurements of
such low frequency modes with high resolution, high sensitivity and
excellent baseline substraction will likely play an important role in the
understanding of adsorption, surface dynamics, and other surface
phenomena. There are only a limited number of surface’ infrared
spectroscopic instruments that can be operated in the low frequency region
with those capabilities.

We have built a new far-infrared reflection-absorption spectroscopy
system designed specifically for the frequency region 350 - 1000 cm"1
(usable up to 3500 cm“) . It incorporates a liquid-nitrogen—cooled grating
spectrometer and other cooled optics to minimize noise due to ambient
background radiation, a highly sensitive SizB photoconductive detector and

a UHV sample chamber equipped with LEED, Auger and TPD. The instrument

currently achieves a level of sensitivity necessary for the study of the

low frequency modes, ~2 x 10“ l/fl'E at ~850 cm'l. We have measured the
0-0 stretch mode of molecularly adsorbed 02 on Pt(111) at ~875 cm'1 and the
molecule-substrate vibration of CO on Pt(111) at ~466 curl. The dominant
noise source is non-reproducibility from one scan to the next. An
improvement of the N/S ratio by a factor of 4 is expected by improving the
scan-to-scan reproducibility.

As a first experiment, the isotope effect on the linewidth of the 0-0
stretch mode will be measured. The shifts in the linewidth are expected
to be 1.2 cm"1 if inhomogeneous broadening is dominant, or 12.5 cm"1 if
electron-hole pair broadening is dominant. With the resolution, N/S ratio
and baseline subtraction performance of our new system, those shifts can
be measured. We have measured the linewidth for 1‘02 on Pt(111) , 21.6 cm-1
and will measure the linewidth for 1802 to sort out the line-broadening

mechani sm .

To My Wife and Son

iv

ACKNOWLEDGMENTS

I would like to thank my thesis adviser, Professor Roger G. Tobin,
for his support, insightful and patient guidance, and encouragement
throughout this work. He has taught me how to design, construct and
evaluate the apparatus, and how to do surface experiments. Without his
patient advising, this dissertation could not have been finished.

I am grateful to many other people at MSU, in particular, the physics
department machine shop staffs, Tom Palazzolo, Jim Muns and Tom Hudson,
for their help in building the instrument, Peggy Andrews for her patient
handling of many administrational work, and other members of our research
group, Kathy Lin, HongflWang and.J.S. Luo for their help and collaboration.

I would like to give special thanks to my wife and son, my mother,
and parents-in—law for their support and.encouragement. I also would like
to thank many friends at Sansung for their support and encouragement.

This work is financially supported in part by the Center for
Fundamental Materials Research, by the Petroleum Research Fund,
administered by the American Chemical Society, and by the National Science

Foundation under Grant No. DER-8815616.

TABLE OF CONTENTS

List of Tables .................................................
List of Figures ................................................
Introduction ....................................................

. Scientific Background and Objectives ..............................
1.1 Line broadening mechanism .................................
1.1.1 Lifetime broadening .................................
1.1.2 Dephasing ...........................................
1.1.3 Inhomogeneous broadening ............................
1.2 Experimental determination of the line broadening mechanism
1.2.1 Experimental methods ................................
1.2.2 Summary of previous experiments .....................
1.3 0—0 stretch.mode of 02 on Pt(111) ..........................
1.4 Summary ....................................................

References ....................................................

. Feasibility of the IRAS Experiment ................................
2.1 Comparison of IRAS with other vibrational spectroscopies ..
2.2 Surface selection rule on metal surfaces ..................
2.3 Incidence angle dependence of the surface signal ..........
2.4 Requirements for an IRAS system ...........................

2.5 Calculation of photon noise ...............................

vi

xi

10

10

ll

12

15

18

22

24

28

28

30

32

36

37

2.6 General considerations of the IRAS instrument .............

2.6.1 Radiation source ....................................
2.6.2 Spectrometer ........................................
2.6.3 Detector ............................................
2.6.4 Modulation ..........................................

2.7 Design of the IRAS system .................................
2.7.1 Angle of incidence and throughput ...................
2.7.2 Spectrometer ........................................
2.7.3 Minimizing ambient radiation ........................
2.7.4 Detector ............................................

2.8 Summary ...................................................
References ....................................................
. Description of the Apparatus ....................... _ ...............
3.1 Radiation source in a liquid-nitrogen cryostat ............
3.1.1 Cryostat ............................................
3.1.2 Globar source in a copper housing ...................
3.1.3 Ellipsoidal and plane mirrors .......................
3.1.4 Tuning fork chopper..................................

3.2 Transfer optics ...........................................
3.3 UHV system ................................................
3.3.1 Surface analysis tools ..............................
3.3.2 Sample mount ........................................
3.3.3 Infrared window .....................................
3.3.4 Gas doser ...........................................

3.4 Grating Spectrometer ......................................
3.4.1 Cryostat and structure ..............................

vii

41

41

43

45

46

48

49

50

52

54

SS

56

68

68

69

70

70

71

72

73

75

76

77

77

79

79

3.4.2 Optical

3.4.

3.4.

3.4.

3.4.3 Grating

2.1

2.2

2.6

components ..................................
Entrance slit ...............................
Folding mirrors .............................
Paraboloidal mirrors ........................
Diffraction gratings ........................

Exit slit and filters .......................

3.5 Cooled Transfer Optics ....................................

3.6

Detector

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

3.7 Data Acquisition ..........................................

3.8 Optical Alignment ........................................

References

3.8.1 Optical alignment of the spectrometer ................

3.8.2 Optical alignment of the whole system ................

. System Performance ............................................

4.1 Calibration of spectral frequency .........................

4.2 Spectrometer resolution ...................................

4.3

4.4

4.5

4.6

Detector calibration ......................................

Optical efficiency of the system ..........................

Calculated noises .........................................

4.5.1 Noise of the detector circuit .......................

4.5.2 Photon noise ........................................

Measured noises ...........................................

4.6.1 Noise from the detector circuit .....................

4.6.2 Photon noise ........................................

viii

83
83
83
84
85
86
86
86
89
89
91
92
92
94
95

114

114
116
117
124
126
127
130
132
132

132

5.

4.6.3 Behavior of system noise ............................

4.7 Improving noise-to—signal ratio ............................

4.8 Comparison of our IRAS system and others ..................

4.9 Summary ...................................................

References

Measurements on 02/Pt(1 1 1) and CO/Pt(111) .........................

5.1 Sample preparation and characterization ...................

5.2 Measurements of vibrational spectra for CO on Pt(111) .....

5.3 Measurements of 0-0 stretching band of Ozton Pt(111) .......

References

Conclusion

Reference

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

ix

134
136
139
141
143
156
156
158
160
167

178

180

2.4.1

3.4.1

4.3.1

4.5.1

4.8.1

5.1

LIST OF TABLES

Peak IR intensities and frequencies for the adsorbate—
substrate and internal vibrations for a number of molecular
and atomic adsorbates on metal surfaces. External modes have
weaker signals and lower frequencies, in general.
Specifications of the diffraction gratings.

Detector Calibration. The responsiVity of the detector has
been measured for three different temperatures, using the
calibration setup shown in Figure 4.3.1. The efficiency of
the detector optics is included in the responsivity and the
amplifier gain A. used is 10.

Summary of calculated and.measured noises. See sections 4.5
and 4.6 for details.

Comparison of IRAS systems. No groups, BNL group and
Bermudez group, are using brighter sources to improve the
signal intensity; while others are using<conventional thermal
sources and room—temperature spectrometer, mostly operated in
vacuum to reduce the atmospheric absorption. MSU is using a
liquid-nitrogen-cooled spectrometer and optics to reduce
background noise.

Voltages for singlets of maximum intensity and doublets of

equal intensity and step height in A from equation 5.1.

58

96

144

145

146

169

2.

2.

2.

2.

2.

2.1

2.2

2.3

3.1

3.2

LIST OF FIGURES

Polarization of infrared light. The electric field
components of the polarizations parallel and perpendicular to
the plane of incidence are denoted by EF and 8,, respectively.
The electric field E/Eo and (E/E°)2/coso for platinum at (a)
500 cm"1 and (b) 2100 cm’1 as a function of the angle of
incidence 0 [Ref.8, p.420]

A schematic of dipoles on a metal surface, showing the
screening of a dipole oriented parallel to the surface and
the enhancement of a perpendicular dipole [Ref.9, p.318]
P-polarized light is incident at an angle 0 on a surface from
vacuum side. 6(a)) is the dielectric function of the
substrate.

Reflectance of a platinum surface for p-polarized light at
500 cm‘1 (solid line) and at 2000 cm’1 (dotted line) as a
function of the angle of incidence, calculated using a Drude
model for the dielectric function with Drude parameters of
conduction electron density N - 5.77 x 1021 cut“3 and mean free

time rc - 7.83 x 10'15 sec [23].

xi

59

60

61

62

63

2.3.

2.5.

2.6.

2.7.

3.1

3

1

l

1

Integrated fractional IRAS signal from the adsorbate-
substrate stretch vibration (solid line) and the internal
vibration (dotted line) of CO on Pt(111). calculated using a
non-interacting harmonic oscillator model with the same e' but
different u's. The difference of the two signals comes from
mostly from the difference in u's. Note that the signal has
a peak at near grazing angle and at 500 cm'1 the peak occurs
closer to grazing.

Photon-noise-limited sensitivity of IRAS experiments on a
platinum surface as a function of frequency for three cases;
source photon-noise-limited case. 100 K and 300 K ambient
photon—noise-limited cases.

Schematic of a generic IRAS instrument. Source could be
thermal or non-thermal. Chopper can go anywhere. For FTIR.
modulation usually comes from the moving mirror.
Spectrometer could be grating, FTIR or CVF. but it has to
have particular An and Au. Spectrometer can, and usually
does. precede sample [1.6.8.9]. For diode laser source. no
spectrometer may be needed [24].

Photon flux per unit frequency for a single polarization from
a blackbody at 100 K. 300 K and 1300 K. as a function of
frequency. The vertical line indicates the detector cut-off
frequency.

Optical layout of the MSU far-infrared spectroscopy system.
The source. chopper aperture. sample. spectrometer entrance
and exit slits. and detector entrance aperture are conjugate

points.

xii

64

65

66

97

Schematic of the source cryostat.

Cooling curves for the source cryostat.

Schematic of the source housing.

Layout of the source optics.

Schematic of the lower part of the sample mount.
Differentially pumped CsI window. ‘Vacuum seals are provided
by three viton.0—rings. The CsI window is pressed on only by
the O-rings.

The calculated uniformity of the gas doser.

Vertical cross section of the spectrometer.

Detailed diagram of the clamp mechanism. Three screws are
used to clamp the spectrometer base plate to enhance the
mechanical stability.

Cooling curves for the spectrometer.

Optical layout of the spectrometer.

Grating table assembly with Inductosyn rotary position
transducer.

Schematic of the grating drive.

Schematic of the liquid-nitrogen-cooled transfer optics.
The circuit diagram of the SizB detector amplifier.

Block diagram of data acquisition system.

Light at oblique incidence onto a grating. showing the
definition of the angles d and 0.

Calibration of spectral frequency for cold spectrometer. The
calibration curve for room—temperature spectrometer is not
shown since it is indistinguishable. A nominal value of

d is used for 2dsin0.

xiii

98

99

100

101

102

103

104

105

106

107
108

109

110

111

112

113

147

148

4.2.1. Measurement of the angular resolution of the spectrometer. 149
Spectra of a He-Ne laser are measured for different slits.

Peaks are not centered at zero degrees since grating angle
are not calibrated when the measurements were made. The
intensities are arbitrarily sealed for convenience.

4.2.2 Measurement of angular resolution of the spectrometer. Both 150
calculated and measured values are shown. The measured
resolution behaves as if the effective focal length were
74 cm instead of 65 cm.

4.3.1 Calibration setup for the detectorx The distance between the 151
source aperture and the detector is 10.9 cm. The temperature
of the blackbody is measured with a R-type thermocouple
attached to the copper cone.

4.5.1 A noise equivalent circuit of detector and preamplifier. 152
4.6.1 Measured noise as a function of input test voltage. The line 153
has a slope of 1/2 indicating the photon noise limited

operation of the detector with n - 0.1.

.4.6.2 A.noise spectrum.measured at the input test with source light 154
through but with the chopper off. The chopping frequency is
chosen as 800 Hz. where 1/f noise is fading out. Note that
there are no sharp features due to mechanical vibrations.

4.6.3 Noise from a ”time scan” and "real scan”. The frequency 155
label of the x-axis applies to the real scan. It takes
approximately 3 minutes to measure the time scan. Deviation
from the mean value is shown for the time scan and a
comparison of two subsequent spectral scans is shown for

the real scan.

xiv

LEED pattern of the crystal with a doublet that is from the
ordered step structure. The step is monatomic and the
terrace is 21 atoms wide. The normal to the surface is at
2.73° from the [111] direction.

An Auger spectrum for clean Pt(lll) surface. Note that a few
percent of carbon (273 eV) appears and oxygen (510 eV) is
hardly seen.

A TPD spectrum measured after dosing the Pt(111) sample with
0.5 L of CO gas using the doser.

IRAS spectra of the C-0 stretch.mode of on—top CO on Pt(111).
CO gas exposures are in Langmuir and are referred to ion
gauge readings. The effective exposures at the surface were
larger by a factor ~20 because of the doser. A linear
baseline is subtracted for each spectrum.

IRAS spectrum of the CO—Pt stretching mode. The spectrum is
smoothed and baseline is subtracted.

A TPD spectrum measured after dosing the sample with 0.8 L of
02 gas. The sharp strong peak at 160 K is from molecularly
adsorbed oxygen and the rather broad peak at about 800 R is
from atomic oxygen on the surface.

IRAS spectra of the 0-0 stretching band for different 02
exposures. Spikes are caused mainly by the
nonreproducibility of the grating angles (see also section
4.7).

An IRAS spectrum of the 0—0 stretch vibration for 1‘02
molecules adsorbed on Pt(111) measured at the saturation

coverage. and fit to the Fano lineshape.

170

171

172

173

174

175

176

177

Introduction

The surfaces of solids have been studied by scientists for many
years. motivated by a desire for a basic understanding as well as a
technological interest in heterogeneous catalysis, crystal growth.
oxidation. corrosion. and semiconductor surface devices [1.2]. The laws
of the interactions on surfaces are basically the same as those
[appropriate to the bulk. but the boundary condition and.the environment of
the atoms on the surfaces are very different from those in the bulk. The
lack of symmetry in the direction perpendicular to the surface greatly
complicates theoretical calculations. Many new phenomena have been
observed.on.the surfaces: surface structure relaxation.and.reconstruction.
steps and other defects unique to surfaces. two-dimensional phase
transitions. adsorption and desorption. and others.

Advances in ultrahigh vacuum (UHV) techniques have played a major
role in advancing surface science. ‘UHV is indispensable for preparing and
maintaining atomically clean surfaces. usually single crystals, for most
surface science experiments. An exposure of one Langmuir (1 L - l x 10'6
Torr for one second) will build up one monolayer of molecules on a clean
surface at room temperature, if each impinging molecule sticks to the
surface. assuming that the surface atomic density of the substrate is
3 x 101‘ cm’z. In order to have a reasonable amount of time to perform an

experiment on a clean surface, UHV is thus a necessity. For an example.

2
suppose that a clean surface is placed in a UHV chamber of a pressure
p - l x lO'H’Torr. and it takes just little more than two and a half hours
before a monolayer of molecules can accumulate on the surface.

Preparing clean surfaces is not an easy task. and there is no general
method to obtain clean surfaces, either. With an appropriately oriented
crystal. clean surfaces can be obtained by a few different ways; (1)
cleaving inside a UHV chamber. (2) ion sputtering and annealing in a UHV
chamber, after mechanical polishing and chemical cleaning outside the UHV
chamber. together with chemical reactions at elevated temperature using
hydrogen and/or oxygen gas. and (3) bringing carefully into a UHV chamber
with special chemical treatments outside the UHV chamber (e.g. Si(111)
[3]). To measure the impurity content and crystalline quality of a
surface. a combination.of.Auger electron spectroscopy (ABS) and low energy
electron diffraction (LEED) are used. AES can only detect as low as 1 Z
of a monolayer, at best [1]. The clean surface is fairly dirty by bulk
standards. but it is not easy to achieve even that level of cleanness.

The interaction of gases with surfaces is a widely studied field in
surface science. which includes simple static adsorption on well
characterized clean surfaces as well as complicated catalytic processes.
As a gas atom or molecule approaches a surface. it feels an attractive
potential and will be adsorbed on the surface provided that it has
sufficient energy to overcome any activation barrier. and is able to
transfer energy to the surface rapidly enough to become trapped.
Depending on the strength of the potential the adsorption results in
chemisorption or physisorption.and.the adsorbed atom or molecule is called

an adsorbate. The strength of the potential can be expressed in terms of

3

the heat of adsorption Alia“, which is defined by the average binding
energy per molecule between the gas atoms and the surface atoms [1].
Physisorption lacks a true chemical bond between the adsorbate and surface
and is formed by the weak van der Waals interaction. with All“, < 10
kcal/mole (1 kcal/mole z 0.04 eV/molecule). Chemisorption is
characterized by a chemical bond that gives larger binding energies. All“,
> 15 kcal/mole. Many surface scientists investigate the structural.
dynamical. and electrical properties of the chemisorbed layer to study
gas-surface interactions and many surface chemical reactions.

Among many experimental techniques used to investigate surface
properties, surface vibrational spectroscopy (SVS) has been known as an
important tool because it has the capability to determine the chemical
nature of adsorbates and surfaces that are not readily available by other
techniques such as Auger electron spectroscopy. and ultraviolet or X-ray
photoelectron spectroscopies (UPS or XPS). While electron and
photoelectron spectroscopy are used to study electronic properties of the
.core and valence states of adsorbates. surface vibrational spectroscopy
can be used to identify the types of adsorbed atoms or molecules on a
surface, and to determine bonding sites and adsorbate geometry. Bond
strengths and distances of adsorbates can be estimated from the
vibrational frequencies. Furthermore. information about surface dynamics
can be obtained directly from time-resolved laser measurements or
indirectly from line shape and linewidth analysis of SVS.

There are many SVS techniques available; surface, infrared
spectroscopy. electron energy loss spectroscopy (EELS). surface enhanced

Raman spectroscopy. inelastic tunneling spectroscopy. and inelastic atom

4

or neutron beam scattering. Surface infrared spectroscopy can be further
classified; infrared reflection-absorption spectroscopy (IRAS). infrared
emission spectroscopy (IRES), surface electromagnetic wave spectroscopy
(SEWS). and. electroreflectance ‘vibrational spectroscopy (EVS). among
others. Since each SVS technique has its own advantages and limitations.
the investigation of a particular surface vibrational phenomenon usually
requires a specific SVS technique. In certain cases. it is also necessary
to use more than one technique to perform an experiment; the combination
of IRAS and EVS is used for a direct measurement of the Stark tuning rate
for adsorbed molecules on surfaces. IRAS and EELS are among the most
popular and successful techniques (see section 2.1). partly due to high
sensitivity and versatility.

The energy transfer between the adsorbed molecule and the substrate
plays an important role in such surface dynamical phenomena as adsorption,
desorption. surface diffusion, and surface chemical reactions. Therefore.
the mechanisms and rates of the energy transfer have been among the most
intensively studied subjects. both experimentally and theoretically, in
surface dynamics [4-15]. Nevertheless. the understanding is far from
clear for most adsorbate modes. particularly at low vibrational
frequencies (below 1000 cm'l), partly due to the limited experimental data
available. (There is no clear threshold for the term low frequency --
there are some experiments below 1000 cm’l. though not a lot. and only a
handful below 800 cm".)

One of the experimental approaches is. direct time-resolved
measurements of the vibrational lifetime. A direct measurement would

unambiguously identify the energy transfer channel and separate

5

homogeneous and inhomogeneous contributions to vibrational linewidths.
The experiments are very difficult, and were at first limited to high-
surface-area insulators. and long-lived modes (2 100 ps) [10]. Recent
advances in the time-resolved laser technique have made it possible to
directly measure the vibrational lifetimes for some adsorbates on well-
characterized single crystal surfaces [ll-15]. An intense resonant
picosecond infrared (pump) pulse creates an excited state population of an
adsorbate vibrational mode. The subsequent relaxation of the excited
vibrational mode is monitored by an infrared spectroscopy based on a
second infrared (probe) pulse. For the second probe spectroscopy, both
IRAS and surface IR-visible sum frequency generation (SFG) spectroscopy
have been successfully used. The IRAS probe method has the advantage that
the observed spectral line shapes may be compared to the vast available
IRAS data. In the SFG spectroscopy method. the signal is generated
through the nonlinear response of the surface at the sum frequency of the
visible and infrared beam. and therefore it can be detected using the
highly sensitive photomultiplier tube. The time-resolved measurements
were performed for CO on Pt(111) [11.12] . cnas (methyl thiolate) on Ag(lll)
[13], CO on Cu(100) [l4]. and H on Si(111) [15]. but they are still
limited to high-frequency modes. mainly due to the lack of tunable far-
infrared picosecond lasers.

The other experimental approach is to examine the vibrational spectra
of adsorbates on the surface. The frequencies. linewidths. and line
shapes of the vibrational modes. and their dependences on coverage and
temperature, contain information about the adsorbate-adsorbate and

adsorbate-substrate interactions [5.8.9]. The linewidth. for example, can

6
be a measure of the finite lifetime of the vibrationally excited state,
and therefore of the rate of energy transfer between the adsorbed molecule
and the substrate. It is remarkable that information about a process on
a picosecond time scale can be obtained through the static measurement of
the vibrational linewidth. The linewidth. in general, can also be
affected by other processes, such as dephasing and inhomogeneous
broadening. In order to sort out these effects. it is essential to
perfonm a precise line shape and linewidth measurement, which requires
high sensitivity and resolution. and.the capability to vary such dynamical
variables as temperature and coverage over a wide range. Electron energy
loss spectroscopy (EELS) has good sensitivity, but does not have the
resolution required for linewith studies. Surface infrared spectroscopy
has good resolution and sensitivity, and it is regarded as nearly the only
experimental technique adequate for studies of the linewidths and line
shapes of adsorbate modes. The spectral region of conventional surface
infrared spectroscopy has historically been limited to frequencies above
1000 cm'l. because of the low intensity of thermal IR sources and bright
background radiation. Only a limited number of surface infrared
spectroscopy instruments can be routinely operated in the low frequency
region [17-20]. Most adsorbate-substrate vibrations have frequencies in
the far-infrared region, below 1000 cm'l. and they play very important
roles in dynamical processes. That is why we have built a new far-
infrared spectroscopy system. which is specifically designed for the far-
infrared region. below 1000 cm'l. This dissertation describes the design.

construction. characterization. and preliminary results of this system.

1. Scientific background and objectives

Surface infrared spectroscopy has proved to be one of the most useful
experimental techniques for investigating the dynamics and interactions of
adsorbed atoms and molecules on clean surfaces [4-7] . Due to such
features as high resolution, strict selection rules, and quantitatively
interpretable line shapes and intensities, it has been successfully used
to obtain information about the chemical nature of adsorbates. adsorption
sites. molecular orientation, adsorbate-adsorbate interactions]. adsorbate-
substrate energy transfer. and electrostatic effects of surface defects.
For the most part, however these investigations have been confined to high
frequency modes, above 1000 cm'l. which are mostly the internal vibrational
modes of molecular adsorbates.

The low frequency adsorbate-substrate modes, usually below 1000 cm'l,
have been historically inaccessible by conventional infrared spectroscopy.
in part because the signals are usually weak and are interfered with by
room temperature radiation [16]. They are. however, even more important
than the high frequency modes to such surface phenomena as adsorption.
desorption. surface diffusion. heterogeneous catalytic reactions. and
other surface phenomena. In recent years. some adsorbate-substrate modes
have been measured. but still they are not readily accessible. Here at
MSU we have built a unique IRAS system specifically designed for studying
such adsorbate-substrate modes.

In this chapter. I discuss one of the subjects that will be
investigated by the new instrument: the study of line-broadening

mechanisms of molecular and atomic adsorbates on single crystal metal

8
surfaces. Of course. the instrument can and will be used for other
studies. such as adsorbate-substrate vibrations of atomic and molecular
adsorbates, alkali metal adsorption on metals. adsorption on
semiconductors and broadband IR absorption of metal surfaces. which
require high resolution, very high signal-to-noise ratio and excellent

baseline subtraction.

1.1 Line broadening mechanism

One of the most remarkable features of the infrared vibrational
spectrum of a chemisorbed molecule is the large linewidth. The linewidth
of the C-0 stretch vibration of chemisorbed CO. for example. is typically
~ 10 cm'l. some 1010 times greater than the natural linewidth of the same
vibration in the gas phase [5]. The broadening processes can be largely
classified as lifetime broadening. dephasing. and inhomogeneous
broadening.

1.1.1 Lifetime broadening

The vibrational energy of the adsorbed molecule could decay and
'result in lifetime broadening by exciting or creating such excitations as
photons. phonons. adsorbate vibrational modes other than the one under
study. and electron-hole pairs (for metal substrates). First, the decay
process which generates a photon is entirely negligible. since the
radiative lifetime of the C-0 stretch vibration is estimated to be on the
order of 10"1 second, which is some 1010 times longer than the lifetime
implied by the adsorbate linewidth [5]. Second. lifetime broadening by
phonon emission has received a lot of attention at least for modes with

vibrational frequencies less than two or three times the maximum substrate

9
phonon frequency [21-24.33,38]. Since typical maximum phonon frequencies
are 200 - 500 cm*l[2]. this process is hardly significant for such high-
frequency modes as C-O at ~2000 cm'1 on metal surfaces. but it has long
been thought be at least potentially important for low frequency modes.
Recently. however, Persson and Ryberg [25] have pointed out that the
direct phonon contribution is negligible on metal surfaces, although it
can still dominate on semiconductors [26] and insulators [27]. Third. the
decay of a vibrational mode could occur through the excitation of other
adsorbate vibrational modes, accompanied by the emission or absorption of
one or more substrate phonons. This process is difficult to explore
experimentally. since some vibrational modes cannot be measured with
infrared spectroscopy due to the surface selection rule (see section 3.2)
and others are even unknown. There is, however. a pretty clear example.
Harris et a1. [13] have directly measured the vibrational relaxation time
of the C-H stretch mode of CHQS on Ag(lll). using SFG spectroscopy. They
have observed two component relaxation times. and have assigned them to
intramolecular vibrational relaxation. proposing the C-H asymmetric
stretch and/or bending modes as candidates for the intermediate
vibrational mode. Lastly. on metallic substrates, electron-hole pair
generation could be the dominant channel for vibrational decay [8,28-
32.42.54]. Since the number of accessible electron states increases
linearly with the energy of the vibration. the decay rate for this process
increases with vibrational frequency. It is therefore most likely to
dominate the linewidth for high frequency modes that are well above the
maximum substrate phonon frequency. Langreth [30] has argued. based on

the idea of an adsorbate-induced resonance near the Fermi level. that the

10

electron-hole pair mechanism necessarily produces a characteristic,
asymmetric line shape with a extended tail toward either side of the peak
frequency depending on certain relevant parameters. Volokitin [32] has
further predicted that a similar asymmetric line shape should be produced
for a vibration parallel to the surface. as well as for a vibration
perpendicular to the surface. if the line is broadened by the electron-
hole pair mechanism. This argument is discussed more fully in
section 5.3.

1.1.2 Dephasing

In addition to lifetime broadening. dephasing can provide a
homogeneous contribution to the linewidth [8]. The process of dephasing
has been well known in other fields. such as nuclear magnetic resonance
and optical spectra of impurity ions in crystals. Contrary to lifetime
broadening, pure dephasing results in broadening the vibrational linewidth
with no decay in amplitude. The dephasing process can be caused by an
impulsive elastic collision of the vibrating molecule with a phonon
[8.33], by a coupling to another adsorbate vibrational mode at lower
frequency [32,34]. or by the interaction of the adsorbate vibration with
electron-hole pairs [32,35].

1.1.3 Inhomogeneous broadening

The inhomogeneous contribution to the linewidth is believed to
dominate in partial overlayers of adsorbates on metals since the narrowest
linewidth is observed for the well ordered overlayer. In reality. some
degree of inhomogeneous broadening is expected even for well ordered
overlayers. since all real adsorbed layers have inhomogeneities due to

finite substrate. finite domain size and domain boundaries. and a certain

11
amount of ‘point, defects and. contaminants that always exist on the
substrate. These inhomogeneities disturb the local chemical environment
to break the ideal periodicity of the well ordered overlayer, resulting in
shifts of the vibrational frequencies of nearby adsorbates [5]. If long
range dipole coupling is important. these effects could extend over a
appreciable region of the surface.

The extent of inhomogeneous broadening is not. however. completely
known for the well ordered overlayer [8]. Since the inhomogeneity of a
_real surface is often too complicated and is caused by many different
factors. it is extremely difficult to analyze either the line shape or the
linewidth even for a complete layer. It is often assumed that the line
shape will be asymmetric. Recently Jacob et a1. [36] have made
measurements for H on Si(lll) prepared by an etch in a buffered HF
solution. a surface that closely approximates an ideal complete layer.
They have measured an asymmetric line shape with a low frequency tail and
a linewidth of 0.05 cm'l. which is apparently due to inhomogeneous
broadening. and have shown that two major sources are responsible for the
asymmetric inhomogeneous broadening: normal modes other than the in-phase
normal modes and the domain size distribution.

It is generally expected that the extent of inhomogeneous broadening
will be sensitive to the number. nature. and distribution of defects and
contaminants. Inhomogeneous broadening is therefore likely to depend on

the preparation of the surface.

1.2 Experimental determination of the line broadening mechanism

As discussed above the line broadening contains the combined effects

of non-radiative vibrational decay (T1 process). dephasing (T2 process).

12
and inhomogeneous broadening. In general. it is difficult to determine
experimentally the dominant broadening mechanism for a given vibrational
mode. The extent of inhomogeneous broadening should be sorted out to
estimate the homogeneous line broadening. This task is. however. very
difficult to realize. It is essential to minimize the inhomogeneous
effects and to work.on.well characterized systems for the careful study of
linewidth and line shapes. The surface should be smooth. as free as
possible of crystal defects and contaminants. as determined by X-ray
diffraction and Auger electron spectroscopy (AES). and adsorbate
overlayers should be well ordered and complete. as determined by low
energy electron diffraction (LEED). Nevertheless. the possibility of
inhomogeneous broadening always exists even for the well characterized
system [5]. In this section. experimental methods to distinguish the line
broadening mechanism and previous results are presented.
1.2.1 Experimental methods

The homogeneous broadening could be isolated from the inhomogeneous
contribution experimentally by careful.measurements of the line shape. and
of the mass and temperature dependences of the linewidth. Each line
broadening mechanism is expected. or predicted, to give characteristic
features for the measurements. For studies of linewidth and line shapes.
the measurements require‘very“high signal-to-noise ratio. high resolution,
and a very flat or smooth baseline over an extended region. since the
differences are very small and often subtle.

First. homogeneously broadened lines are expected to be Lorentzian
[8], if the line is symmetric. and are also predicted to be asymmetric if

they are caused by the generation of electron-hole pairs in the substrate

13
[30]. Inhomogeneous lines are often highly asymmetric, and it is often
assumed (though with little justification) that they will be Gaussian.[8],
if they are symmetric.

Second. the mass of the adsorbate can be' changed by isotopic
substitution without affecting the chemical environment. The
characteristic mass dependence for various line broadening mechanisms.
however. is not clear and is controversial in some cases [5]. Let us
define the quantity a - (a 1n 7/6 1n mg) to represent the mass dependence
of the linewidth. where 1 is the linewidth. and m.Jr is the reduced mass of
the adsorbate.

The isotope dependence of electron-hole pair line-broadening has been
predicted theoretically. Persson. et a1. [37] have predicted that
electron-hole pair decay should have a - -1. Volokitin [32] has also
predicted the same mass dependence of the linewidth. for vibrations both
parallel and perpendicular to the surface. and has further predicted that
the intensities at peak frequency have different mass dependences: mgqu
_for perpendicular vibration and no mass-dependence for the parallel
vibration. The isotope dependence of dephasing due to phonons has been
predicted to have a - -l [33]. but it will be much more complex when other
vibrational modes of the adsorbate are involved in the dephasing.

Inhomogeneous broadening due to defects and impurities should be
characterized by a - -0.5. since the chemical effect on the force constant
Should be mass-independent, while the frequency is proportional to mgqu.
Tobin [5] argued that the same dependence should occur for domain boundary

inhomogeneity. if the lateral interactions are predominantly chemical in

nature or dipolar coupling. Therefore the isotopic substitution

14
experiment can be very effective.to.sort out the inhomogeneous broadening.
in particular.

Third. measurements of the temperature dependence of the linewidth
are most widely used and fruitful for determining the line broadening
mechanism [5.8.24]. Lifetime broadening by electron-hole pair generation
should be essentially independent of temperature. since the characteristic
electronic energies for an adsorbate on a metal surface are the Fermi
energy of the substrate metal and the width of the adsorbate resonance;
the adsorbate resonance level is a few eV wide and is located at a few eV
from the Fermi level. The adsorbate resonance is an electronic level
induced by the adsorbate near the Fermi level. As the adsorbate vibrates.
the resonance is perturbed by the change in the atomic positions of the
adsorbate. and its electronic occupation oscillates at the vibrational
frequency. resulting in a dynamic transfer of charge between the adsorbate
and the metal substrate. In the case of dephasing by interaction with
electron-hole pairs. the linewidth should depend on temperature since
electrons with energy within kT of the Fermi level can participate.
Morawitz [35] has predicted a T‘ dependence for this process. but Volokitin
disagrees [32]: he has argued that in. many cases the temperature
dependence is negligible. The processes of' phonon. broadening and
dephasing are expected to be strongly temperature-dependent in the 100-
500 K range. since the characteristic energies are adsorbate or substrate
vibrational energies. The temperature dependence for the phonon mechanism
has been theoretically predicted by several people [21.22.32.33]. It is
often assumed that inhomogeneous broadening is nearly temperature-

independent. since it does not involve dynamics. Its temperature

15
dependence could be uncertain. however. since the nature of the sources of
inhomogeneity can vary with temperature.
1.2.2 Summary of previous experiments

Vibrations at frequencies above 1000 cm'1 (high-frequency modes) have
been most widely studied. mainly because they are relatively easy to
measure by IRAS. They include most intramolecular vibrations.
particularly the C-0 stretch vibration which is at ~ 2000 cm'l. Even
though these modes reveal lots of dynamical and structural information.
they do not participate directly in surface processes such as sticking,
desorption, and diffusion. Indeed the adsorbate-substrate modes, which
are typically at low frequencies (300 - 1000 cm'1 ; called low-frequency
modes). are important for these processes. Primarily because these modes
are difficult to measure by infrared spectroscopy. there have been very
limited studies of linewidths and line shapes of these low frequency modes
[17-20,24,38.39].

The lifetime of high-frequency vibrations appears to be determined
primarily by the rate of generation of electron-hole pairs in the metal.
For CO on Cu(100). Ryberg [40] measured the C-0 stretch linewith as a
function of temperature and found that the linewidth and the line shape
are temperature-independent, indicating that pure dephasing is not an
important line broadening mechanism. He argued that the vibrational
lifetime determines the linewidth. and that the primary mechanism of
vibrational energy relaxation is the generation of electron-hole pairs in
the substrate. The lifetime of the mode is estimated to be T1 =3 1.2 ps.
using the measured linewidth of 4.5 cm'1 and the equation T1 - [21rc(Au)]‘1,

where c is the speed of light and AV is the full width at half maximum

16
(FWHM) of the line in cm“.

Recently. Morin et a1. [14] have directly measured a vibrational
lifetime of 2.0 ps for the C-0 stretch mode of CO on Cu(100) using SFG
spectroscopy. which would give a homogeneous linewidth of 2.7 cm”. They
interpret the energy relaxation process as caused by non-adiabatically
generating electron-hole pairs in the copper substrate. The non-adiabatic
energy transfer lifetime of this mode has been calculated by a density-
functional method [41] and by extrapolation of ab initio Hartree-Fock
electronic structure calculations for CO on Cu clusters [42]. Both
calculations give lifetimes in the l - 3 ps range, in good agreement with
the measured values.

Another candidate for lifetime broadening is the l-i-substrate wagging
mode of H on W(100) and H on Mo(100) [4] . The first. overtone of the mode
has the line shape predicted by Langreth [30] for the electron-hole pair
mechanism and its linewidth is temperature-independent.

For CO on Ni(lll). CO on Ru(100). and H on Si(100). the observed
linewidths of the C—0 and Si-H stretch modes have been attributed to
dephasing [4.5]. primarily on the basis of the strong temperature—
dependence of the linewidth. In the case of CO, the measured data have
been satisfactorily fit to a phenomenological model with several
adjustable parameters. For H on Si(100). a first-principles calculation
was used to obtain the linewidth within a factor of 2.

For low-frequency modes (lower than ~ 1000 cm'l). phonon-mediated
decay is more likely to be an important broadening mechanism as discussed
in section 2.1.1. However. there are no conclusive experimental

measurements so far to show the linewidth caused by the phonon mechanism,

l7
and recently Persson and Ryberg [25] have pointed out that all the
previous calculations have missed an important factor and the phonon
process is less likely to be an important line-broadening mechanism.

The first infrared measurement of a low-frequency adsorbate-substrate
mode was performed by infrared emission spectroscopy for CO on Ni(lOO) at
475 cm'1 [24] . The measured linewidth was 15 cm'1 at a temperature of 310
K. Unfortunately. the temperature-dependence could not be measured
because of the low oscillator strength of the mode and the very limited
temperature range accessible with emission spectroscopy, and the line
broadening mechanism was not clearly identified. Ariyasu et a1. [21] . and
Volokitin et a1. [22] . have calculated the linewidth contribution due to
phonon emission decay process for the C-Ni mode of CO on Ni(100) . Ariyasu
et al. have used only the two phonon process. while Volokitin et al. have
incorporated all phonon processes and found the three-phonon process
dominates. They have predicted the linewidth and slightly different but
strong temperature-dependences of the linewidth. But Persson [25] has
recently pointed out a serious error in all these previous calculations.
and has ruled out the possibility of the phonon emission decay for a line-
broadening mechanism.

The linewidth of the adsorbate-substrate mode of CO on Pt(111). at
467 cm'l, was measured at several temperatures by the same emission
spectroscopy [38]. This mode has an oscillator strength 5-10 times
greater than that of the corresponding mode of CO on Ni(100) . The results
were inconclusive. however. due to non-reproducibility of the measurements
and to the limited temperature range available.

Recently. the adsorbate-substrate mode of CO on Pt(111) has been

18

measured by Hoge et a1. [20] . Malik and Trenary [l9] . and Ryberg [18,25] .
Ryberg has measured the narrowest linewidth. 2.5 cm‘1 at 100 K, compared
to 3 - 5 cm'1 at 100 R in the other experiments. suggesting that the other
two measurements, as well as the earlier emission measurement, were
dominated by inhomogeneous broadening. They also show different
temperature dependences of the linewidth; Hoge et al. reported almost no
temperature dependence. Malik and Trenary found a strong temperature
dependence. from 3 cm‘1 at 80 K to 8 cm”1 at 300 K. and Ryberg found'a
linear temperature dependence from 2.5 cm‘1 at 100 K to 4.1 cm'1 at 250 K.
Persson and Ryberg [25] have argued that the line broadening is caused
mainly by dephasing via anharmonic coupling to the low- frequency parallel
frustrated translation and that the vibrational energy decays through
excitation of electron-hole pairs.

There have also been a few infrared measurements of the low-frequency
mode of adsorbed atomic oxygen: an emission measurement [38] of O on
Pt(111). at 480 cm'l. a diode laser reflection-absorption measurement [39]
.of O on Al(lll). at 580 cm“. and an FT-IRAS measurement [43] of O on
polycrystalline Ag, at 351 cm'l. In each case the line is very broad.
=38 cm'1 on Pt(111). £100 cm'1 on Al(lll). and =33 cm'1 on polycrystalline
Ag. and very weak. Little can be concluded regarding the line broadening

mechanism .

1.3 0-0 stretch mode of 02 on Pt(111)

In each of the experiments discussed above. there remain ambiguities
in determining the line broadening mechanism even for the high-frequency
mode. The difficulty arises mostly from the fact that inhomogeneous

broadening cannot be ruled out conclusively. Probably. a measurement of

19

the isotopic dependence of the linewidth is the best way to sort out the
inhomogeneous broadening, since the mass dependence of the inhomogeneously
broadened linewidth is predicted as a - - 0.5 [5] while other line-
broadening mechanisms are expected to exhibit a stronger mass dependence
[32.33.37]. A measurement of the isotope effect on the linewidth usually
requires very high signal-to-noise ratio of the instrument. since the
change of linewidth is often very small; the inhomogeneous linewidths for
12C160 and for 12C130 on Pt(111). for example. would differ only by 0.2 cm”.
In fact. there have been few successful measurements of the isotope effect
on the linewidth. Persson and Ryberg [37] once claimed that the isotope
effect of the C-H stretch modes of any) on Cu(100) was measured to be
a - - 1. which was taken as an evidence for electron-hole pair broadening.
Later these modes were remeasured by the same authors [44] and the
linewidths were found to be much narrower than in the original
measurement. with little difference in linewidth between the C-H and C-D
modes. The difference in linewidth would be 4.3 cm'1 if the line is
broadened by electron-hole pair damping. or 1.8 cm'1 if the inhomogeneous
broadening is dominant. ‘The only clear experiments have been for adsorbed
hydrogen. for which the mass and frequency change are so large that it
cannot be assumed that the chemical interactions and decay process are the
same for both isotopes.

We have chosen 0; on Pt(111) as a first system to study the line
broadening mechanism. Here. we have to admit that this is not the kind of
experiment the instrument was designed for --- it is not a very low
frequency mode and it is not an adsorbate-substrate mode either. But.

oxygen on Pt(111) is important. technologically for heterogeneous

20
catalysis. such as the oxidation of CO. NO and NH3. as well as
scientifically for the study of dissociative chemisorption. The 0-0
stretch band is also a little bit easier to measure; the oscillator
strength is larger than that of most adsorbate-substrate modes. but much
smaller than C-O stretch. and it is at moderate frequency, ~ 875 cmTK

Oxygen on Pt(111) has been widely studied by many surface
experimental techniques; temperature-programmed desorption (TPD) [45-49].
low energy electron diffraction (LEED) [47]. X-ray photoelectron
spectroscopy (XPS) [50], near-edge X-ray-absorption fine-structure
spectroscopy (NEXAFS') [51]. and molecular beam scattering [52]. The
system has been relatively less studied by surface vibrational
spectrosCOPY: only one reflectionrabsorption-infrared spectroscopy
measurement [53] and two electron energy loss spectroscopy measurements
[46,47] have been reported.

At low temperature (5 ~100 K). oxygen adsorbs molecularLy on the
Pt(lll) surface, with its bond axis parallel to the surface [46,47]. This
molecular oxygen is a metastable precursor state to dissociation into
atomic oxygen. which forms an ordered p(2x2) overlayer upon heating above
150 K. Despite the orientation of the 0-0 bond. the O-O stretch vibration
is observed at 875 cm'1 with a relatively large dynamic dipole, indicating
that large adsorbate-substrate charge transfers occur as the bond
stretches. The 0-0 stretch frequency of the chemisorbed Oz. 875 cm'l, is
surprisingly lower than the corresponding frequency of gas phase 02, 1556
cm’l. In fact. the double bond of free 02 is reduced to a single bond upon
chemisorption on Pt(111) surface. This reduction is explained by a large

static charge transfer from the metal substrate to the antibonding 21r'

21
orbital of the adsorbed 02 molecule. possibly changing the orbital from the
half filled state to the filled state [45.47].

Canning and Chesters [53] have observed the 0-0 stretch band of
molecularly adsorbed 02 on a recrystallized Pt foil at 875 cmfl, with a
relatively large linewidth 20 - 25 cm'1 at the saturation coverage 0 - 0.4.
They have measured the frequency shift due to an isotopic mixture and
confirmed that the band is really due to molecularly adsorbed oxygen on
the surface. They have also observed the lowest frequency 845 cm’1 at a
_coverage 0.2. and have explained that the overall frequency shift of 30
cm’1 is caused by dipole-dipole coupling.

Persson [54] has argued that the line broadening of the O-O stretch
vibration is caused by the generation of electron-hole pairs in the
substrate, based on an analysis of the line shape using Langreth's model.
The possibility of inhomogeneous broadening is excluded. by showing that
the same line width has been deduced from EELS and IRAS data. but the
experimental evidence is indirect and.inconclusive. Furthermore. the IRAS
spectra were measured on a recrystallized foil. without surface analytical
capabilities.

The dominant line-broadening mechanism can be determined
experimentally by measuring the temperature-dependence and/or the mass
dependence of the linewidth, and by examining the line shape. As
discussed in 1.2.1. the line will be either highly asymmetric or a
symmetric Gaussian.if it is inhomogeneously'broadened” while the line will
be asymmetric with a specific lineshape and an extended tail on either
side if electron-hole pair damping dominates the broadening [30].

Unfortunately. we cannot investigate the temperature-dependence of the

22

linewidth of the 0-0 stretch mode. at 875 cm”. since the desorption of 02
molecules already starts to occur at 100 K. The isotopic dependence of
the linewidth is, however. expected to be relatively large; the linewidths
for 1602 and for 1302 would differ by 1.2 cm'1 if inhomogeneous broadening
is dominant. or by 2.5 cm'1 if electron-hole pair damping dominates the
linewidth. The difference of 1.2 cm'1 or 2.5 cm'1 can be measured by IRAS.
if the system has high noise-to-signal ratio and excellent baseline
substraction. and the measurement will shed light on further understanding
of the line broadening mechanism. The measurement of the isotopic
dependence of the linewidth for 02 on Pt(111) has been thus chosen as the
first project of our new instrument.

The Pt-O vibration of the atomic oxygen on Pt(111) has been measured
at 480 cm'1 by EELS [46.47] and IR emission spectroscopy [38]. For
molecularly adsorbed 02 on Pt(111). the Pt-Oz vibration has been measured
at 380 cm'1 by EELS [46.47]. but it has never been measured by infrared
spectroscopy. Because of the low temperature required, emission
spectroscopy could not be used for the Pt-Oz band. These bands are
expected to be very weak and difficult to observe with conventional IRAS
instrument. With a new apparatus spanning the frequency range 350 to 3500
cm'l. these modes will be observed so that the entire process of adsorption

and desorption of such molecules as 02 can be traced with high resolution.

1.4 Summary

In this thesis work. a new far-infrared reflection-absorption
spectroscopy (IRAS) system has been built using a grating spectrometer and
a thermal source. The system has been specifically designed for maximum

sensitivity in the 330 - 1000 cm'1 spectral range (though it is usable up

23
to >3500 cm'l) with better than 2 cm'1 resolution. It incorporates a
liquid-nitrogen-cooled grating spectrometer and other cooled optics to
minimize ambient blackbody radiation. The spectroscopic system is coupled
to an ultrahigh vacuum surface analysis chamber. Most of the thesis is
about the design. construction and.evaluation of the IRAS instrument, with
the preliminary measurements using the system. The rest of the thesis is
organized as follows. In chapter 2. the feasibility and design criteria
of the IRAS system are presented. The typical IRAS signal on a platinum
surface and background photon noise are calculated and analyzed.
Characteristics of some optical elements and techniques are discussed with
a viewpoint of maximizing the signal-to-noise ratio. The design of our
IRAS system is discussed. specifically for operation at frequencies less
than 1000 cmfih In.chapter 3. I describe the details of the apparatus and
how’we have built the system. In chapter 4. the performance of the system
is evaluated and compared with the design goal and other systems. along
with the calibrations of the detector and spectrometer. and with the
'quantitative analysis and measurements of the system noise. The
preliminary measurements of CO on Pt(111) and.02 on Pt(111) are presented

in chapter 5 and conclusions follow in chapter 6.

24

References

10.

11.

12.

13.

14.

G.A. Somorjai. Chemistry in Two Dimension: Surfaces. (Cornell
University Press. Ithaca. New York. 1981).

A. Zangwill. Physic5' at Surfaces. (Cambridge 'University' Press,
Cambridge. 1988).

6.8. Higashi. Y.J. Chabal. G.W. Trucks and K. Raghavachari, Appl.
Phys. Lett. 56. 656 (1990).

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

R.G. Tobin, Surf. Sci. 183. 226 (1987).

A.M. Bradshaw and E. Schweizer. in .Advances in Spectroscopy:
Spectroscopy of Surfaces, edited by P.E. Hester (Wiley. New York.
1988) pp. 413 - 483.

F.M. Hoffmann. Surf. Sci. Reports 3, 107 (1983).

J.W. Gadzuk and A.C. Luntz. Surf. Sci. 144. 429 (1984).

I. Ueba. Progr. Surf. Sci. 22. 181 (1986).

E.J. Heilweil. M.P. Casassa. R.R. Cavanagh. and J.C. Stephenson, J.
Chem. Phys. 82. 5216 (1985).

J.D. Beckerle. R.R Cavanagh. M.P. Casassa. E.J. Heilweil, and J.C.
Stephenson. J. Chem. Phys. 95. 5403 (1992).

R.R. Cavanagh. J.D. Beckerle. M.P. Casassa. E.J. Heilweil. and J.C.
Stephenson. Surf. Sci. 269/270. 113 (1992).

A.L. Harris, L. Rothberg, L.H. Dubois. N.J. Levinos. and L. Dhar.
Phys. Rev. Lett. 64, 2086 (1990).

M. Morin. N.J. Ievinos. and A.L. Harris, J. Chem. Phys. 96. 3950

(1992).

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

25
P. Guyot-Sionnest. P. Dumas, Y.J. Chabal and G.S. Higashi, Phys. Rev.
Lett. 64. 2156 (1990).
P.L. Richards and.R.G. Tobin. in.Vibrationa1 Spectroscopy of Molecules
on Surfaces. edited. by .J.T. ‘Yates. Jr. and. T.E. Madey (Plenum
Publishing Co. New York, 1987).
C. Hirschmugl. G.P. Williams, F.M. Hoffmann and Y.J. Chabal, Phys.
Rev. Lett. 65. 480 (1990).
R. Ryberg, Phys. Rev. B40. 5849 (1989); B40. 8567 (1989).
I.J. Malik and M. Trenary. Surf. Sci. 214, L237 (1989).
D. Hoge. M. Tushaus, E. Schweizer and.A.M. Bradshaw. Chem. Phys. Lett.
151. 230 (1988)
J.C. Ariyasu. D.L. Mills. R.G. Lloyd. and J.C. Hemminger, Phys. Rev.

328. 6123 (1983); Phys. Rev. B30. 507 (1984).

A.I. Volokitin. O.M. Braun and V.M. Yakovlev, Surf. Sci. 172. 31
(1986).
V.P. Zhdannov. Surf. Sci. 197. 35 (1988).

S. Chiang. R.G. Tobin. P.L. Richards and P.A. Thiel. Phys. Rev. Lett.
52. 648 (1984).

B.N.J. Persson and R. Ryberg. Phys. Rev. B40, 10273 (1989).

P. Dumas. Y.J. Chabal, and G.S. Higashi, Phys. Rev. Lett. 65. 1124
(1991).

R. Disselkamp. H.-C. Chang and.G.E. Ewing. Surf. Sci. 240. 193 (1990);
C. Node. R.R. Richardson and G.E. Ewing. J. Chem. Phys. 92, 2099
(1990).

B.N.J. Persson. Solid State Commun. 27. 417 (1978).

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

26
B.N.J. Persson and M. Persson, Solid State Commun. 36, 175 (1980);
Surf. Sci. 97, 609 (1980).
D.C. Langreth, Phys. Rev. Lett. 54. 126 (1985).
Z. Crljen and D.C. Langreth. Phys. Rev. B35. 4224. (1987).
A.I. Volokitin. Surf. Sci. 224. 359 (1989).
B.N.J. Persson. J. Phys. 017. 4741 (1984).
B.N.J. Persson and R. Ryberg, Phys. Rev. B32. 3586 (1985).
H. Morawitz, Phys. Rev. Lett. 58. 2778 (1987).
P. Jacob. Y.J. Chabal and K. Raghavachari, Chem. Phys. Lett. 187. 325
(1991).
B.N.J. Persson and R. Ryberg. Phys. Rev. Lett. 48, 549 (1982).
R.G. Tobin and P.L. Richards. Surf. Sci. 179. 387 (1987).
V. M. Bermudez. R. L. Rubinovitz and J. E. Butler, J. Vac. Sci.
Technol. A6, 717 (1989).
R. Ryberg, Phys. Rev. B32, 2671 (1985).
T.T. Rantala and A. Rosen. Phys. Rev. BB4. 837 (1986).
M. Head-Gordon. J.C. Tully, J. Chem. Phys. 96. 3939 (1992).
X.-D. Wang. W.T. Tysoe, R.G. Greenler. and.K. Truszkowska. Surf. Sci.
257. 335 (1991).
R. Ryberg. J. Chem. Phys. 82. 567 (1985).
J.L. Gland. Surf. Sci. 93. 487 (1980).
J.L Gland. B.A. Sexton and G.B. Fisher. Surf. Sci. 95, 587 (1980).
H. Steininger. S. Lehwald and H. Ibach. Surf. Sci. 123. l (1982).
A. Winkler. X. Guo. R.R. Siddiqui. P.L Hagans and J.T. Yates. Jr..
Surf. Sci. 201. 419 (1988).

C.T. Rettner and C.B. Mullins. J. Chem. Phys. 94. 1626 (1991).

50.

51.

52.

53.

54.

27
A.C. Luntz, J. Grimblot and D.E. Fowler, Phys. Rev. BB9, 12903 (1989).
W. Wurth, J. Stdhr. P. Feulner. X. Pan. K.R. Bauchspiess. Y. Baba. E.
Hudel, G. Rocker. and D. Menzel, Phys. Rev. Lett. 65, 2426 (1990).
A.E. Wiskerke, F.H. Geuzebroek. A.W. Kleyn and B.E. Hayden. Surf. Sci.
272. 256 (1992).
N.D.S. Canning and M.A. Chesters. J. Elect. Spect. Relat. Phenom. 29,
69 (1983).

B.N.J. Persson. Chem. Phys. Lett. 139, 457 (1987).

2. Feasibility of the IRAS Experiment

In this chapter. I discuss the characteristics, feasibility and
designs of infrared spectroscopic systems that are required for the
studies of adsorbate vibrations at frequencies below 800 cmfih
Quantitative analyses are performed. to show that such a system is
feasible. using a conventional thermal infrared source. together with a
modern infrared detector. a cooled spectrometer and cooled optics. A.

specific design of such an infrared reflection-absorption spectroscopic

system is presented.

2.1 Comparison of IRAS with other vibrational spectroscopies

Surface vibrational spectroscopy has been one of the most important
techniques to study surface dynamics and adsorbates on solid surfaces.
Many different kinds of surface vibrational spectroscopy have been
developed for this purpose [1]. For the present discussion. I compare
three techniques below. that is. electron energy loss spectroscopy [2],
infrared emission spectroscopy [4.5]. and infrared reflection-absorption
spectroscopy [1.6.8.9].

Electron energy loss spectroscopy (EELS) uses the inelastic
scattering of an electron beam. A monochromatic low energy electron beam
(-5 eV) is incident on the surface and the scattered electrons are
detected with an energy-selective detector so that the energy loss due to
adsorbates is measured. EELS is one of the most widely used surface
vibrational spectroscopic tools because of high sensitivity as well as a
wide spectral range (200 - 4000 cm“). Also it can detect modes that are

not infrared-active. Virtually all adsorbate vibrations of interest

28

29

except the ones below 200 cm"1 can be and have been measured by this
technique. The resolution, however, is typically 40 - 80 cm'1 (resolution
of 8 - 16 cm“1 has been achieved recently [3]) which is not good enough to
measure the linewidths and frequency shifts, which are often <10 cm'l. and
of which accurate measurements are very important for the study of surface
dynamics. Another drawback of EELS is that it is very difficult to
evaluate intensities quantitatively, because the scattering processes are
not well understood.

Infrared emission spectroscopy [4,5] routinely achieves a resolution
of l - 5 cm'l. and covers much of the relevant frequency range. from 350
to >2500 cm'l. with good sensitivity. In the emission technique. the
thermal equilibrium radiation emitted by the sample surface itself is
collected. dispersed by a liquid-helium-cooled monochromator. and detected
with a sensitive photoconductive detector. Comparing a spectrum measured
with a clean surface with one measured with an adsorbed overlayer, the
small enhancement of emissivity due to the adsorbate vibration can be
detected. Since the sample itself is used as the radiation source. the
sample temperature needs to be high enough to maintain adequate
sensitivity. It has to be higher than 200 K even at 500 cm“. At higher
frequencies. above the peak of the blackbody curve. the emitted intensity
falls exponentially with decreasing temperature. and even higher
temperatures are required to get the necessary sensitivity. Since many
adsorbates begin to desorb at temperatures below 400 K, and often below
300 K. the temperature range available to experiment is very narrow; Even

for such a relatively stable system as CO on Pt(111) the usable

30
temperature range for the emission technique extends only from 200 to
275 K.

Infrared reflection-absorption spectroscopy (IRAS) [1.6.8.9] uses a
separate infrared radiation source and thus does not impose any
constraints on the sample temperature. Infrared radiation from the source
is reflected off the sample, sent to a dispersive monochromator or Fourier
transform infrared (FTIR) instrument and collected by a detector. The
spectrometer can come before the sample. The adsorbed layer on the
surface absorbs a small amount of radiation so that the reflectance of the
surface is changed. The small change in the reflectance is detected by
comparing a spectrum measured with a clean surface with one measured with
an adsorbed layer. IRAS routinely achieves resolution of 1 - 5 cm'l.
Conventional IRAS. which uses a room temperature spectrometer and optics.
however, is difficult to extend to frequencies less than 800 cm'l. We have
built a new system which maintains the sensitivity in the low frequency
region by using a sensitive detector and reducing the background photon
noise from the environment by cooling the spectrometer and the relevant

optics. This issue is discussed in detail in sections 2.5 - 2.7.

2.2 Surface selection rule on metal surfaces

In an IRAS experiment. the electromagnetic field of the infrared
radiation interacts with the dynamic dipole moment associated with a
particular normal vibrational mode of the molecules and a small proportion
of the radiation is absorbed by the vibrational excitation of the
molecules. Since the dielectric response of the substrate dominates the
electromagnetic field on a surface at infrared frequencies. it is

important to investigate the dielectric properties of substrates to

31
optimize the experimental conditions.

On metal surfaces. the s-polarized component of incident infrared
radiation is almost perfectly screened and only the p~polarized component
survives and is enhanced by a factor of nearly 2 for angles of near-
grazing incidence. as was first recognized by Francis and Ellison [7]. As
shown in Figure 2.2.1 the s- and p-polarized electric fields of the
radiation are perpendicular and parallel to the plane of incidence.
respectively. Figure 2.2.2 shows the dependence of the electric field
strength E/Eo on the angle of incidence 0 for both s- and p-polarized light
at a platinum surface for u - 500 and 2100 cm'1 [8]. Here. E0 is the
amplitude of the incident beam and E the resultant amplitude of the
electric field at the surface. The p-polarized component B, can be further
split into components EIn and EPI’ perpendicular and parallel to the
surface. respectively. The field strength of Em increases with increasing
angle of incidence 0, reaching a maximum between 80 and 90° but falling
rapidLy to zero at 90°. On the other hand s-polarized light and the
.parallel component of p-polarized light are screened.out to be negligible,
particularly at 500 cm“, as is also shown in Figure 2.2.2.

Since the parallel component of the field is screened at the metal
surface. any molecular dynamic dipole moment parallel to the metal surface
is screened out. This is visualized by an induced image dipole in Figure
2.2.3. For the dipole perpendicular to the surface the apparent dipole
moment is enhanced since the image dipole is in the same direction. while
the parallel dipole moment is cancelled out. It is. therefore. expected
that only vibrational modes with dynamic dipole perpendicular to the

surface should be observed in IRAS.

32
Recently. however, a frustrated rotation mode [10] has been observed
with IRAS. Persson and Volokitin [11] have theoretically shown that
parallel translation modes can be observed in IRAS, as anti-absorption
peaks or strong Fano lineshapes. due to indirect excitations of electrons

in the metal substrate.

2.3 Incidence angle dependence of the surface signal

Most IRAS analysis is based on the three-layer model introduced by
Greenler [12], who investigated the conditions necessary to give the
highest absorption in a multiple reflection technique. Later. however, it
was shown that a single reflection was sufficient to obtain good infrared
reflection spectra [13]. Since then the single reflection technique has
been extensively used and refined for measuring IRAS of adsorbates on
metal surfaces. The macroscopic theory of IRAS has been refined by
McIntyre and Aspnes [14], Dignam et a1 [15]. Bagchi et a1 [16]. Persson
[17] . and Langreth [18] . Recently Tobin [19] used a macroscopic theory of
reflectivity developed by Langreth [18] to incorporate the effect of the
dielectric response of the substrate into the IRAS analysis. For present
purposes it does not matter which analysis is used to calculate the
adsorbate signal. since we are not trying to do any detailed quantitative
work. We follow the derivation by Persson.

Classical electromagnetism is used to calculate the reflectance of
infrared radiation on a metal surface. The macroscopic response of a
solid surface is described by a dielectric function 600) which is a
prOportionality factor between the electric displacement D and an applied
electric field E. In a linear medium the relation between the two is:

D- 6E . (2.3.1)

33

First we consider a two phase system with a plane boundary of vacuum
and bare substrate. The infrared beam is described by a plane
electromagnetic wave through a medium. Suppose p-polarized light is
incident at an angle 0 on a surface from the vacuum side as shown in
Figure 2.3.1. The incident light is described by an electric field:

E. - Eo (xcoso + fisinO) ei‘k't’m’ . (2.3.2)
The reflected field is given by :

Er - Eo rp (-§icoso + fisinO) ei‘k'T'm" , (2.3.3)

where r is given by the Fresnel equation for p-polarized light [1]:

P

ecoso - (6 - sin20)1/2
r - , (2.3.4)

ecoso + (e - sin20)1/2

 

So the reflectance is given by:
11,- lrplz. (2.3.5)
The intensity of the reflected wave is given by:
c c

Er'E;I-
8x 8x

 

 

|::,,|21~:.,2 . (2.3.6)

The field in the z-direction, which is perpendicular to the surface, at
the surface (z-O) is:
E“ - E0 sin0 (1+rp) e1‘kx'mo’“) . (2.3.7)

Now suppose the surface is covered by a very thin (d<<A) layer
characterized by a surface polarizability a... A surface dipole moment
density P, is induced by an applied field E, such that:

P,-a, Ell . (2.3.8)

Note that the dimensions of P are (esu cm/cmz). of E are (esu/cmz) and of
a, are (cm). In general the surface polarizability is complex and contains

the effects of coupling within the film. In the case of an array of

34
uncoupled oscillators. each with a polarizability a (cm?) with a surface
density N, (oscillators/cmz). we can write 03 - Ngz.
We want to calculate the change in the reflected intensity AI due to
the presence of the film. If the substrate does not transmit the light,
we can divide this change into two parts:

AI - - (Power/area absorbed by film) - A(Power/area absorbed by
substrate).

(2.3.9)
The second term is due to the change in the substrate absorption caused by
the presence of a dielectric layer on the surface. and is ignored here
(this term is included in references 16. 18, 19).
The work done on the dipoles by the electric field is dW - E-dP. So
the power per unit area absorbed is:
Al - - (E-dP/dt)

Re(E..-dP:/dc)

 

2 c030
to sin”
- - —— £02 |1 + rl,|2 Im(a.) . (2.3.10)
2 c030 '
The 1/c050 factor enters because a, and P1 are defined as quantities per
‘unit area of the substrate. while the intensity is defined as a power per

unit area perpendicular to the k-vector of the light. The fractional

change of the reflectance is therefore expressed as:

 

 

 

AR AI 41w sin?0 ]l+rp|2
— - - - Im(a.)
R I c c080 Irpl2
sinzo |1+rp|2
- - 32n3v Im(a,) . (2.3.11)

 

 

coso |2rp|2

 

 

 

 

 

 

wh.
alt

COL]

35
where u - w/2ttc is measured in cm'l.
To calculate the integrated fractional signal. let's consider the

case of simple decoupled damped oscillators. The surface polarization is

then:

 

e’z/tt
a,(w) - N, a, + , (2.3.12)
((002 — (.02) + iwl‘

where a. is the electronic polarizability. e"r the effective dynamic charge,
wo the resonance frequency. 1" the linewidth. and p the reduced mass. Using

this equation the integration is evaluated:

AR sinzo Il-t-rp I 2 w e'Z/p
—— dw - 41rN, dw — Im
o

R c050 [rpl2 c (Loo-(0)2 + 1101‘

 

 

 

 

41rN,e"z s inzfi I l+rp| 2 1x2
-—-—— dx
0

pc c059 Irpl2 (1-xz)2 + 12x2

(2.3.13)
where x - w/wo and 1 - F/wo. For narrow lines, 1 << 1. the integrand is
1rd(x)/2 and the value of the integral is 1r/2. So for uncoupled

oscillators the fractional adsorbate signal is:

 

 

I AR 21r2N,e"2 sinzo |1+rp|z
— du - - . or
R uc c030 Irplz
1' AR 1rN,e"z sinzfi [Lt-rpl2
-— dU - - . (2.3.14)
R ucz c030 erI2

which is measured in cm'l. In fact. the equation (2.3.14) is quite general
although it is derived for uncoupled oscillators. It applies for dipole
coupled oscillators if a renormalized e' is used [20].

Figure 2.3.2 shows the reflectances of a Pt surface at 500 cm'1 (solid

36
line) and at 2000 cm'1 (dotted line). for p-polarized light. over a range
of angles of incidence. For angles less than 80 degrees. the reflectances
are very high (2 80 1), as expected for a metal in the far infrared. As
a result. the reflected background intensity is high. and the fractional
signal due to the adsorbate is correspondingly small. In Figure 2.3.3.
the integrated fractional signals. calculated using parameters appropriate
to the stretching vibrational mode of CO on Pt(111). are shown for 500 cm"1
(solid line) and for 2000 cm'1 (dotted line). with the same e' but
different 11's. Approaching grazing incidence the reflectance drops
sharply. and the fractional signals show strong peaks. Real low frequency
modes have lower e"s. Note that at 500 cm‘1 the peak appears closer to

grazing.

2.4 Requirements for an IRAS system

The adsorbate-substrate vibrations we are interested in generally lie
at low frequencies. and they are usually much weaker than the internal
vibrations of molecules. Table 2.4.1 shows peak IR intensities for a
-number of molecular and atomic adsorbates on metal surfaces. Clearly, the
instrument must have a sensitivity of better than 0.02 X to study
adsorbate-substrate modes of atomic and molecular adsorbates on metal
surfaces.

As we discussed in 2.3. the integrated fractional signal shows a
strong peak approaching grazing incidence. but the amount of radiation
reflected off the sample will drop sharply since the solid. angle of
reflection at the sample becomes smaller. We should maintain a certain
level of radiation to minimize its contribution to system noise. So.

there is a trade-off.

37

At least two spectral scans are required for an IRAS measurement.
The first scan is measured with the clean surface and the second scan with
adsorbates on the surface. The small change of the reflectance, which is
caused by a vibrational interaction of the adsorbate with infrared
radiation. is then obtained by comparing the two spectral scans.
Non-reproducibility of the scans. other than effects due to adsorbates.
can show up as noise or drift in the comparison spectrum. The
reproducibility of scans should be better than 1 x 10", the required
sensitivity.

As discussed in Chapter 1. a resolution of l - 5 cm'1 over frequency
region 350 - 3500 cm'1 is required to study lineshapes and line broadening
mechanisms. The system should have the capability of measuring the
temperature dependence of the linewidth and peak frequency precisely. in
order to compare to theoretical predictions of the characteristic
dependence of each process on temperature. The baseline of comparison
spectra is required to be very flat and smooth to measure the effect of
isotopic substitution on the linewidth, which is often _<_ 1 cm‘1 (see

section 1.3 and 5.3).

2.5 Calculation of photon noise

With a highly sensitive detector. the fundamental limit on
sensitivity comes from photon noise. provided that non-reproducibility of
the spectra can be made low enough. In this section. we thus calculate
photon noises from the source and ambient radiation.

The number of photons from a blackbody source at temperature T can be
calculated from statistical mechanics [21] . The Planck distribution gives

the average number n, of photons in each possible energy and polarization

38

 

state:
1
ns - . (2.5.1)
ehcy/kT _ 1
where u is measured in cm'l. There are 41w2du photon states per unit

volume per polarization and with frequency in the range between v and v +
du. So the mean number of photons of a given polarization per unit volume

in the frequency range between v and v + du is:

1

n..- 4trv2du . (2.5.2)
ehcu/kl' _ 1

 

Since photons travel isotropically. the number of photons per unit volume
with k-vectors in the direction of an infinitesimally small solid angle d0
is nudO/41r. Integrating over the effective solid angle 0.“ - f c030 d0.
the number of photons per unit time radiating from a blackbody with an
area A is nycAn.“/41r. Taking into account the reflectance of the sample
R.p and the transmittance of the optical system 1. the photon rate at the
detector is:
cvzdu

N -' Rme.£f h , (2.5.3)
e cu/kT _ 1

 

where we are assuming a single polarization since only p-polarized light
is involved in exciting stretching modes of IRAS as discussed in 2.2.

In general. for an optics system of an aperture area A and a
subtended solid angle 0. the quantity At), which is called the throughput,
étendue. or luminosity. is invariant throughout a lossless system [22].
For surface IRAS, the maximum available throughput at the sample surface

is given by:

39

An.“ - it h d c059 lsin(¢v/2))2

nhd c059

4(f/#)2

z a h d (cosfl)3. (2.5.4)
where 0 is the incidence angle. w the beam divergence. h the sample
height. d the sample width and f/# the focal number of the beam at the
sample [9].
Now if the detector is photon-noise-limited. the noise equivalent

photon rate [photons/sec thql can be calculated by [4]:
AN 1/2
NEN - [ ——-— (2.5.5)
n

Using the two equations. (2.5.3) and (2.5.5), we can calculate the
source photon-noise-limited sensitivity of the experiment. NEN/N , which
has dimension of Hz‘“q. This quantity is plotted in Figure 2.5.1 as the
line labeled source photon noise. It represents the smallest fractional
signal that can be detected in a photon-noise-limited experiment. with a
signal-to-noise ratio of unity. in an 1 Hz bandwidth. In addition to the
source noise. the photon noises of ambient thermal radiation from the
room-temperature and liquid-nitrogen-cooled environments are calculated
using same optical throughput and plotted in Figure 2.5.1. Here the
ambient photon rate is integrated over frequency from the detector cutoff
to infinity.

In the calculations for Figure 2.5.1. the Drude model for dielectric
function is used with following free electron parameters [23]:

N - 5.77 X 1021 cm‘3 : conduction electron density,

rc - 7.83 X 10*” sec : mean free time.

40
The Drude model is a good approximation for Pt at frequencies below about
1000 cm-1. The other parameters used are as follows:
Sample dimensions: 1 x 3 cm?
1 - 0.10 for source photon rate calculation
1 - 1.0 for ambient photon rate calculations
0 - 0.50
V/AV - 400 to get the required resolution
T - 1300 K for source temperature
T - 300 K for room temperature environment
T - 100 K for liquid-nitrogen cooled environment
An.“ - 3.2 x 10'3 cmzsterad for source photon rate calculation. from
the sample reflectionnwith sample size

A0.“ - 2.0 x 10'2 cmzsterad for ambient photon rate calculations,
from the detector aperture

detector cutoff - 350 cm'1
angle of incidence 0 - 86°.

From Figure 2.5.1 we can tell several things. First. the ambient
noise, whether it is cooled by liquid nitrogen or at room temperature. is
generally bigger. particularly in the low frequency region (< 1000 cm”) ,
than source photon noise. This is mainly due to the fact that the
background noise is integrated over frequency. which also makes the
difference between 300 K and 100 K independent of frequency. Second. the
photon noise from a liquid nitrogen cooled object is smaller by an order
of magnitude than that from a room temperature object. For high frequency
measurements. the cutoff frequency of a detector can be high enough to
reduce the ambient photon noise to a negligible level (see Figure 2.7.1).
Third, in the case of liquid-nitrogen-cooled environment we can extend the
frequency range to 300 cm"1 while it is extremely difficult to extend the
frequency below 800 cm'1 with required sensitivity, if we use room
temperature optics. Fourth, a thermal source is bright enough to get the

required sensitivity, if we cool the background to 100 K. There is no

41
fundamental need for brighter. non-thermal sources until we get below

300 cm"1 .

2.6 General considerations of the IRAS instrument

A generic IRAS system. as shown in Figure 2.6.1, consists of an
infrared source. sample, chopper. spectrometer. detector and relevant
optics. The infrared source could be a conventional thermal source.
tunable infrared laser. or synchrotron radiation. In order to do the
vibrational spectroscopic measurements of adsorbates on clean single
crystals. the sample should be in a UHV chamber with infrared windows.
The windows should preferably be transparent in the visible region to ease
optical alignments and have a wide infrared transmission range. The
chopper could be placed anywhere in the optical path as long as the beam
size is small enough to be chopped. For FTIR the modulation usually comes
from the moving mirror. The spectrometer could be a grating. FTIR or
circular variable filter (CVF). The spectrometer can. and usually does.
precede the sample. For a diode laser source, no spectrometer may be
-needed [24]. The detector could be a thermal detector. bolometer or
photoconductor. Some of the optical components crucial to a successful
surface infrared instrument are discussed below.

2.6.1 Radiation source

The conventional infrared sources are thermal blackbody sources. the
globar and the Nernst glower. These thermal sources have been widely used
in IRAS experiments. The spectral photon rate distribution from a
blackbody at a frequency v (cm’l) in an interval AV is given by the Planck

distribution law [21]:

42

e 2cv2Au

NM - . (2.6.1)
ehcv/k'l.‘ _ 1

 

where 6 is the emissivity and T the source temperature. Note that in the
far-infrared (hcu/kT<l), the photon rate distribution becomes:

N(u) - eZusz/h (Au/v) . (2.6.2)
and it is linear in T and proportional to uz for fixed resolving power
u/Au.

As long as the source intensity and optical system are stable. a
brighter source provides better sensitivity since the noise is independent
of intensity if ambient noise dominates and is proportional to the square
root of the intensity if source photon noise dominates. So it is
desirable to increase the operating temperature as much as possible. The
spectral distribution falls off sharply as the frequency gets lower. as is
seen in the equation (2.6.2).

The Nernst glower is made of a mixture of zirconium and.yttrium oxide
and the globar is a silicon carbide rod” The globar is often favored over
the Nernst glower since it is more stable in general. The globar is
usually operated at around 1200 ~ 1500 anith DC current in a water-cooled
or air-cooled housing. At far infrared frequencies the 1300 K blackbody
source is brighter than the room temperature environment merely by a
factor of 4.3.

A non-thermal source that is much brighter than the room temperature
environment could be better to improve the sensitivity if it is
sufficiently stablen Tunable infrared lasers and.synchrotron.sources have
been used in IRAS. Bermudez et al. [25] (see also section 4.8) have used

tunable diode lasers. and successfully measured vibrations of'O onHA1(111)

43
at 580 cm’1 and 890 cm'l. though at relatively low resolution (30 - 4O
cm'l). The Brookhaven National Laboratory (BNL) group [10] (see section
4.8) has used synchrotron radiation and measured low frequency vibrational
modes of CO on Cu(100). Ni(100). and Cu(lll).
2.6.2 Spectrometer

Although FTIR is currently more popular. grating spectrometers have
been widely used for IRAS [6.9] . They are usually of Czerny-Turner design
(see Figure 3.4.4). The resolution of the spectrometer is determined by
the dispersive power of the grating. the slit width and the focal length
of the focusing mirror. The intensity throughput is determined by the
slit size and the focal ratio of the collimator. which is called the focal
number (f/#) of the spectrometer. The throughput of the spectrometer,
which determines the maximum amount of radiation that can be collected
from a blackbody source, and the resolution are therefore inversely
related. A larger slit width gives higher throughput. but degrades the
resolution. It is necessary to maximize the throughput to increase the
sensitivity while maintaining adequate resolution for the desired
experiments. But there is no point in increasing the throughput beyond
the value set by the sample. Since the energy resolution is almost
linearly proportional to the frequency and each grating is made with a
blaze angle where the efficiency of the grating is greatest. each grating
can be used over a restricted spectral range (typically a factor of two)
with a reasonably high resolution and intensity throughput maintained. It
is also necessary to use a cutoff filter to eliminate higher order
diffraction. Three or four gratings and filters are required to cover the

spectral range of 300 - 3000 cm“.

44

Circular variable filters (CVF) can be used for monochromation of
transmitted radiation. A CVF is made by laying out optical coatings of
increasing thickness on a quartz or germanium disk in circular segments.
Monochromation of radiation is performed by rotating the filter while the
radiation is passing the filter through a slit. Golden et al. [26] have
used a CVF for IRAS with rather low resolution (40 - 80 cm'l) in the range
of 1000 - 3000 cm'l. The low resolution. which is determined by the
bandwidth of the filter and slit width. is a serious drawback in a CVF
systems The advantage of a CVF is short optical paths. ease of operation.
and relatively high transmittance (larger than 30 1).

Another candidate for the spectrometer is Fourier Transform Infrared
Spectroscopy (FTIR) using interferometry: 'Traditionally, FTIR.instruments
are known to have two advantages over dispersive. monochromators. the
throughput advantage and the multiplex advantage [22]. An interferometer
receives information from the entire range of a given spectrum during each
time element of a scan. whereas a grating spectrometer receives
information only in a narrow ‘band. If the noise is limited by
instrumental or ambient photon noise that is independent of the signal
level. the multiplex advantage gives an improvement of a factor Nb”. for
a spectrum containing N resolution elements. Most IRAS experiments.
however. are limited (or nearly limited) either by run-to~run
nonreproducibility or by signal photon noise. In such cases any multiplex
advantage is negligible [4].

The "throughput advantage” refers to the fact that for FTIR
instruments there is no tradeoff between throughput and resolution; very

high resolution can.be achieved with no sacrifice of intensity, simply by

45
increasing the optical path length. For IRAS. however. the sample
restricts the usable throughput, and resolution better than 1 cm"1 is
rarely necessary; there is little difficulty in achieving adequate
resolution while accepting the full usable throughput.
2.6.3 Detector

It is crucial to use a sensitive detector in order to measure
reflectance changes at the level of $10". The sensitivity of a detector
is measured by the noise-equivalent power NEP (or the noise-equivalent
photon rate NEN). which is defined as the amount of radiant power (photon
rate) that must be incident to produce a signal equal to the noise of the
detector. The NEP (NEN) is thus measured in WA? (photons/(8.515)).
Typical IRAS requires NEP better than 10‘10 W//H_z to achieve the required
sensitivity of the system [6].

Infrared detectors are usually classified into tw0 broad categories -
thermal detectors and photon detectors. Thermal detectors essentially
respond to the incident radiant power to produce a electrical signal
proportional to the power. Thermocouples. Golay cells and bolometers are
thermal detectors. Thermal detectors are generally less sensitive than
photon detectors, though certain bolometers are highly sensitive and
photon-noise limited. The NEP of a thermal detector is about 10" W/fi
when operating at room temperature environment. The sensitivity improves
a lot when operating at cold temperatures and low backgrounds. The most
sensitive thermal detectors available are semiconductor bolometers. Such
bolometers have NEPs of «10'15 W/fil-z when operated at liquid helium
temperatures and low backgrounds [4] . Thermal detectors are much slower

than photon detectors, since their response time is limited by the thermal

46
relaxation time of the macroscopic element.

Photon detectors (also called quantum detectors) respond to incident
photons that generate electrical carriers in the detector proportional to
the number of photons. Photoconductors and photovoltaic detectors belong
to the photon detector class. Photon detectors respond much faster than
thermal detectors. but only to radiation with frequencies higher than the
gap or impurity energy. due to their quantum nature. The liquid-nitrogen-
cooled HngTe (MOT) and InSb photoconductive detectors are most widely
used in the mid infrared frequency range (800 - 5000 cm'l) and have a
typical NEP of 10'12 W/flE. Liquid-helium-cooled germanium or silicon
photoconductors have been developed to be used at far infrared frequencies
with NEP of 10'16 1.1/AB or better when operated at low backgrounds.

2.6.4 Modulation

Signal modulation and phase sensitive detection are commonly used in
infrared spectroscopy to separate a signal from superimposed background
noise. A rotating chopper is most widely used in a dispersive instrument.
and the moving mirror in a FTIR interferometer. In addition, three more
modulation schemes, wavelength modulation. polarization modulation and
double-beam modulation, can be applied to surface infrared spectroscopy.
To choose between various modulation schemes. it is very important to
understand what limits the N/S ratio of the system for a specific
experiment.

Wavelength modulation can be achieved by incorporating a vibrating
mirror (with a frequency f) in the monochromator. For small modulation
amplitude. the signal after demodulation is proportional to the first

derivative of the intensity with wavelength. It can also be used as

47

second derivative technique at a modulation frequency 2f [27]. Because of
this derivative nature this modulation can. be useful only if the
spectrometer efficiency does not have sharp structure in the frequency
range of interest [9]. Wavelength modulation has been successfully used
in IRAS [6.27]. However. since it cannot be placed very close to the
source. it fails to discriminate effectively against stray ambient
radiation. so it should be used as a double modulation scheme with
intensity modulation.

Polarization modulation exploits the fact that only the p-polarized
component of the infrared.beam excites the adsorbates on the metal surface
[6.9]. If we switch the polarization of the infrared beam impinging on
the sample, we can get a surface signal after demodulation that is
proportional to the difference of intensities for two polarizations. The
polarization of the beam can be switched by a rotating polarizer or a
photoelastic modulator. But we do not adopt the polarization modulation
because there are some problems. First, little enhancement of the surface
signal relative to the substrate sinal is expected. particularly in the
far infrared. Since the reflectance of the sample for the p-polarization
is smaller at the grazing angle than that for the s-polarization. the
demodulated signal is not much smaller than the total intensity. Second.
the efficiency of a grating monochromator depends strongly on both
polarization and wavelength in a complicated way. Even though this problem
can be circumvented.in.principle by using a fixed polarizer set between 3.
and.p-polarization. this type of modulation remains restricted to a narrow
spectral range or a less polarized spectrometer [1.9]. Third. there are

many practical difficulties in operating,a rotating polarizer system [1].

48
such as mechanical noise in the rotation mechanism. a quadrature signal
and degradation of the polarizer with use. Many of these practical
problems can be overcome by using a photoelastic modulator. but
photoelastic modulators are not available for frequencies below 700
cm'1 [1].

Double-beam modulation can be very useful for a surface infrared
spectroscopy if proper matching of the sample and the reference beams is
achieved. It is extremely difficult to match the optical paths of the two
beams in the UHV system [6.9] . Consequently. the double-beam modulation

is not attempted in our system.

2.7 Design of the IRAS system

The main experimental issue in IRAS is to measure a small signal on
a large background, simply because there is a very small number of
molecules available on surfaces. Therefore. in order to measure molecular
vibrations at submonolayer coverage. very high sensitivity and stability
(N/S S l X 10" l/ffi) are required. In sections 2.2 and 2.3. we have
shown that the polarization of the incident light and the angle of

incidence are very important in determining the sensitivity of an IRAS

experiment. Second. the frequency resolution of the IRAS system is
required to be 1 - 5 cm"1 over the spectral range from 350 cm'1 to
3000 cm".

At lower frequencies, there are three major problems. First. thermal
radiation from the room-temperature environment becomes important. As
discussed in section 2.5. the ambient photon noise makes it extremely
difficult to extend the frequency below 800 cm'l, if we use room

temperature optics and a thermal source. Second, the reflection geometry

49
at the sample required for adequate surface sensitivity becomes more
restrictive at low frequencies. as shown in section 2.3. Third. the
vibrational modes found in the low frequency region are typically weaker
than the high frequency modes. Even higher sensitivity than 1 x 10" l/lI-E
is required to measure those low frequency modes. In this section. we
discuss the optimal design considerations of the instrument for IRAS
experiment. particularly at frequencies below 800 cm'l.
2.7.1 Angle of incidence and throughput

As we discussed in 2.5.. the maximum available throughput at the
sample surface is given by

An.“ - it h d c050 [Sln(¢/2))2

z n h d (c030)3. . (2.7.1)
where 0 is the incidence angle. 16 is the beam divergence. h is the sample
height. and d is the sample width [9]. As shown in Figures 2.3.2 and
2.3.3. the reflectance drops sharply and the fractional adsorbate signal
shows a strong peak close to grazing angle. The optical throughput,
however. decreases sharply as 0 approaches grazing incidence. as is
obvious from the equation (2.7.1). While a conventional IRAS system
designed for frequencies above 1000 cm‘1 might use an f/4 beam centered at
82°. we have chosen the incidence angle to be 86° with a focal number 7.2
for the far infrared [8.9].

The sample size is also limited by several factors -- availability of
single crystal, capability to clean the sample and uniformity of
adsorbates on the surface should be considered. Once the sample size. the
incidence angle and the focal number are determined. the effective f/# of

the sample reflection is related to the incidence angle as

50
l/2(f/#) — c030. and an equivalent slit width at of the spectrometer at the
sample is determined by m/d - coso [9] .

To accept all the light reflected off the sample. the following
optics should be matched to the f number of the sample reflection. If the
transfer optics has a magnification of 0.54. then a spectrometer of f/4
can be used. Now the sample width is 3 cm so that the minimum slit width
of the spectrometer is to be 1.2 mm to accept all the radiation reflected
from the sample.

Even though a relatively large sample (1 cm x 3 cm) is used. the
maximum throughput A0...“ of the sample reflection is 3.2 x 10"3 (cmzsterad).
Due to the limited throughput and therefore power arriving at the
detector. a highly sensitive detector should be used.

2.7.2 Spectrometer

We have decided to use a grating spectrometer of a conventional
Czerny-Turner design of f/4.2 (see Figure 3.1). A dispersive grating
spectrometer is chosen over a Fourier transform (FTIR) spectrometer for
several reasons. Traditionally. as mentioned earlier, FTIR instruments
are said to have two advantages over dispersive monochromators. the
throughput advantage and the multiplex advantage [22] . In the case of
reflection-absorption spectroscopy the throughput is limited by the
sample. not by the spectrometer. as discussed above. Since the required
resolution is typically 1 - 5 cm'l. a grating spectrometer which can accept
the full useful throughput can be designed with a reasonable f/#. As
discussed in 2.5, background noise dominates over signal photon noise. The
multiplex advantage of the FTIR can thus be significant to reduce the

background photon noise effect. In practice, however. sensitivity is

51

always determined by stability -- even at 2000 cm‘1 nobody gets the
background-noise-limited system realized. For broadband measurements,
FTIRs can be more useful since they can record a large spectral region
much faster than a dispersive instrument. But it is not so attractive in
IRAS experiments. since usually'a single. narrow frequency region contains
useful information. The grating spectrometer is. moreover, mechanically
simpler. and easier to evacuate and cool than an FTIR.

The spectral resolution of the instrument is determined by the
dispersive power of the grating (i.e. the number of grooves/mm). the
effective focal length f of the focusing mirror and slit width w:

w v

-Au _ , (2.7.2)
2ftan0

 

where 0 is the grating angle. The optical throughput of the spectrometer

is determined by the slit area A, and the focal number f/#:

A, it
An.“-————. (2.7.3)
AME/if)2
The smaller the slit width. the higher the resolution but the smaller the
throughput. The larger the focal length the higher resolution but the

smaller the throughput. There is thus a trade-off.

l 1

To get the resolution of 1 cm' at a frequency 400 cm’ . we need a
resolving power R - v/Au of 400. The beam should have a height h and
effective width dcoso. and beam divergence u. to illuminate the sample.
Here h and d are the height and width of the sample. and the angle of
incidence 0 and the beam divergence u are chosen above. Using a 1.5 mm

wide slit and a 60 grooves/m grating ( ~30° blaze angle ). the focal

length f need to be 65 cm from the equation (2.7.2) to get the required

52
resolving power. To match slit dimensions and the focal number of sample
reflection. the focal number of the spectrometer is chosen as 4.2 with the
magnification M of the transfer optics. 0.54. Since the focal length is
25.6”, the diameter of the collimator is determined to be 6.0".

The Czerny-Turner designmwith paraboloidal.mirrors was selected.after
extensive computer ray-tracing analyses of many other grating mountings.
including Ebert. Littrow, Littrow with Cassegrain optics. Czerny-Turner
with out-of—plane slits. and Czerny-Turner with spherical mirrors. All of
the other designs gave inferior aberration. performance and/or were
complicated to build.

2.7.3 Minimizing ambient radiation

As mentioned in section 2.5. the blackbody radiation from room
temperature objects is so strong that it is very difficult with thermal
sources to extend the spectral range down to the far infrared frequencies
below 800 cm'l. Particularly in the far infrared. the brightness of the
ambient thermal photons increases rapidly such that thermal sources are
only a factor of 4.3 brighter than room temperature, as shown in Figure
2.7.1. Fluctuations in this ambient photon flux very often dominate the
system noise. To reduce this noise. cooling the relevant optics is a
necessity [4].

There are three classes of radiation to be considered: modulated
narrow-band. , unmodulated narrow-band. and unmodulated broad-band
radiation. Details are discussed below. with a view point of noise
contribution.

l). Modulated narrow-band radiation

Any fluctuation of ambient radiation that is modulated will appear in

53

the signal with no attenuation. since it is indistinguishable from signal
radiation. Such ambient radiation could be from the chopper blade. and
from elements between the chopper and the source. so that the chopper
needs to be located close to the source. and the chopper and the source
housing must be cooled. The radiation that is reflected from the chopper
blade could be such a fluctuation noise source. Such reflected radiation
should be reduced to a negligible level.

2). Unmodulated narrow-band radiation

This class of radiation comes from elements between the chopper and
the spectrometer. such as transfer optics. windows and sample. The
radiation is not modulated. and does not contribute directly to the signal
radiation. In addition, it will not be strong enough to increase the
photon noise significantly since it must pass through the monochromator
before reaching the detector. Therefore. it is not necessary to cool
those elements. This is the reason why we put the monochromator after the
sample. But we should not neglect the possibility of indirect
contributions to the modulated signal by the reflection off the chopper
blade of thermal emission from the sample. or windows.

3). Unmodulated broad-band radiation

Radiation from the spectrometer wall. transfer optics after the
spectrometer polarizer and filter belongs here. The radiation is not
monochromatized and thus strong enough to be a major noise source.

The monochromator at room temperature can generate a large amount of
ambient radiation much greater than the signal radiation. At all
positions of the grating. therefore at all frequencies. the thermal

radiation from the walls of the monochromator would be able to reach the

54

detector by diffraction, most strongly in zero order. The photon flux of
the radiation is estimated using equation 2.5.3 for N. with T - 300 K. and
6 - l. and integrating over frequency from 350. the detector cut-off
frequency. to 800 cm". the cut-off frequency of a cooled filter. Using
1 - 1.0. since thermal radiation will reach the detector in all orders of
diffraction. the photon flux is calculated to be 9.8 x 101‘ photons/sec,
more than two orders of magnitude greater than the flux of 500 cm'1
radiation reaching the detector from the source. which is 2.4 x 1012
photons/sec. It is clear that the monochromator should be cooled.

There is a transfer optics between the spectrometer and the detector,
which is an ellipsoidal mirror on a mirror mount adjustable from outside
(see section 3.5). Any radiation from the transfer optics between the
spectrometer and the detector. if it reaches the detector. can be strong
enough to contribute to the photon noises since it is not attenuated. The
transfer optics needs to be cooled by liquid nitrogen and the optical path
may need to be shielded with a liquid-nitrogen cooled baffle.

2.7.4 Detector

We have assumed that a photon-noise-limited detector is used in the
experiment such that the system noise is not limited by the detector
noise. An extrinsic Si:B photoconductive detector from Infrared
Laboratories is photon-noise-limited in light levels as low as 100
photons/sec. so detector noise is certainly negligible in the much higher
background levels encountered in our experiment. If we were really
background photon noise limited. it would be better to use a detector with

a cutoff above 500 cm’1 for higher frequency measurements. since then the

100 K background would be negligible (see Figure 2.7.1). As I show in

55
Chapter 4. however. nonreproducibility currently limits the sensitivity,
so a different detector would offer little benefit. The Si:B detector is
sensitive to radiation at frequencies greater than 350 cmfl. so a single
detector is able to cover the frequency range of interest. The spectral
range of the instrument can extend to a lower frequency if a far infrared

detector is used. but background photon noise will increase a lot.

2. 8 Summary

In this chapter. we have shown that a low frequency IRAS experiment
is feasible, and that the required sensitivity ~l x.]£Y‘ and resolution
~ 1 - 5 cm'1 can be achieved with a conventional thermal source if we cool
relevant optics. A specific design of such a system has been presented.

The next chapter describes the details of our actual system.

56

References

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

2. H. Ibach and D.Ln Mills. Electron Energy Loss Spectroscopy and surface
Vibrations. (Academic Press, New York, 1982).

3. G. Kisters. J.C. Chen. S. Lehwald and H. Ibach. Surf. Sci. 245. 65
(1991), H. Ibach. M. Balden, D. Bruchmann and S. Lehwald. Surf. Sci.
269/270. 94 (1992).

4. P.l“ Richards and.R.G. Tobin, in.Vibrational Spectroscopy'of Molecules
on Surfaces. edited by J.T. Yates Jr. and T.E. Madey (Plenum. New
York, l987)pp. 417-463.

5. S. Chiang. R.G. Tobin and P.L. Richards. J. Vac. Sci. Technol. A2.
1069 (1984).

6. F.M. Hoffmann. Surf.Sci. Reports 3. 107 (1983).

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

8. A.M. Bradshaw and E. Schweizer. in Advances in Spectroscopy:
Spectroscopy of Surfaces, edited by P.E. Hester (Wiley. New York.1988)
pp. 413-483.

9. B.E. Hayden. in Vibrational Spectroscopy of Molecules on Surfaces,
edited by J.T. Yates, Jr. and T.E. Madey (Plenum, New York, 1987) pp.
267-344.

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

11. B.N.J. Persson and A.I. Volokitin. Chem. Phys. Lett. 185. 292 (1991).

12. R.G. Greenler. J. Chem. Phys. 44. 310 (1966). l

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

14. J.D.E. McIntyre and D.E. Aspnes. Surf. Sci. 24, 417 (1971).

15.

l6.

l7.

l8.

19.

20.

21.

22.

23.

24.

25.

26.

27.

57
M.J. Dignam. M. Moskovits. and R.W. Stobie, Trans. Faraday Soc. 67,
3306 (1971).
A. Bagchi. R.G. Barrera. and A.K. Rajagopal. Phys. Rev. 820, 4824
(1979)
B.N.J Persson. Solid State Commun. 30. 162 (1979).
D.C. Langreth, Phys. Rev. B39. 10 020 (1989).
R.G. Tobin. Phys. Rev. B45. 12 110 (1992).
B.N.J. Persson and R. Ryberg, Phys. Rev. 324. 6954 (1981).
F. Reif, Fundamentals of Statistical and’Thermal Physics (McGraw'Hill,
Inc.. New York. 1965) pp. 381-338.
R.J. Bell, Introductory Fourier Transform Spectroscopy (Academic
Press. New York. 1972).
A.P. Lenham and D.M. Treherne. in Optical Properties and Electronic
Structure of Metals and Alloys. edited by F. Abeles (North-Holland,
Amsterdam. 1966) pp. 196-201.
D. Lambert. Appl. Opt. 27. 3744 (1988).
V.Ms Bermudez. R.L. Rubinovitz. and.J.E. Butler. J. Vac. Sci. Technol.
A6. 717 (1988).
W.G. Golden. D.S. Dunn. and.J. Overend. J. Phys. Chem. 82, 843 (1978).

a. Ryberg. Phys. Rev. 340, 8567 (1989), Surf. Sci. 114, 627 (1982).

58

Table 2.4.1 Peak IR intensities and frequencies for the adsorbate-
substrate and internal vibrations for a number of molecular and atomic
adsorbates on metal surfaces. External modes have weaker signals and
lower frequencies, in general.

 

 

 

 

 

 

 

 

 

 

Adsorbate Substrate Peak Peak References
intensity frequency
CO Pt(1ll) 0.15 1 475 cm'1 27, 28, 29
CO Ni(100) 0.04 470 30
CO Cu(100) 0.07 345 10
0 Ag ribbon 0.16 351 31
0 Pt(1ll) 0.06 460 29
S Cu(lOO) <0.02 ~300 31
C-0 stretch Pt(lll) 6.0 2100 8. 23
C-0 Ag ribbon 1.8 2000 32
0-0 Pt(lll) 0.42 870 33

 

 

 

 

 

 

 

59

IR In
IR Out

 

 

 

 

 

 

 

Figure 2.2.1 Polarization of infrared light. The electric field components of the
polarizations parallel and perpendicular to the plane of incidence are denoted by Ep and

E3, respectively.

 

 

 

 

(a) (b)
-i . - . d
2'0 500 cm ' 2 o 2100 cm"
I. "‘ «III E
5 EDI. 1.5 pL
O
3 1.0- 1.0-Ly" ___€oj."°
kl ' . - “w
‘0
\e
0.5- 0.5— x.
5.- ‘.~.~ ‘ ‘ \.
5'30 1 "- l l
100-1
a, 80- 304
In
8
- 60-
N.
c 20-
2 1,0.
4
a.“ 10
T 20-
‘ r F I
0 30 60 90 0 3O 60

 

 

 

 

 

 

 

 

 

 

 

Angle of incidence. Bldeq Angle of incidence. Bldeq

Figure 2.2.2 The electric field E/Eo and (E/Eo)2/cos6 for platinum at (a) 500 cm-1
and (b) 2100 cm-1 as a function of the angle of incidence 9 [Ref.8, p.420].

61

 

((5)

b. . -‘

 

 

77/1/777/ /'/V.'/F ////f/V\f7/f///f fli/l7/l/l/l7
. .l l 1.. - ‘Vs‘ infill 9,9," ,A’
I I' '01 ‘~-‘ ‘ " "
: l‘.‘. I" 24’7”? ““~“~“'~-
. “‘ ’00 v I’ ‘ I \ ______ "J‘ I ‘\
| ‘\ T-fitt- ,' r ‘4 ‘ -------- ’ 1’ '
‘ ‘VI\<‘ ' \ ~~~~~ ’ ’
\\ ’4 g I’ \ T--- I
‘.’I \‘.'I \A‘---------”,1
(a) (b)

Figure 2.2.3 A schematic of dipoles on a metal surface. showing the screening of a
dipole oriented parallel to the surface and the enhancement of a perpendicular dipole
[Ref.9, p.318].

 

Figure 2.3.1 P-polarized light is incident at an angle 0 on a surface from
vacuum side. 6(0) is the dielectric flmction of the substrate.

 

 

Ill-[Ir

63

 

 

 

 

 

 

L—
r-
8 r-—
q -
e
é -
o _
2
H -
$30.4- __
0.2—
- “.1
0.0-LllllllllllllLJlllLlllLllllgld
60 65 70 75 so 85 90

Incident. Angle (deg)

Figure 2.3.2 Reflectance of a platinum surface for p-polarized light at 500 cm-1 (solid
line) and at 2000 cm'1 as a function of the angle of incidence, calculated using a Drude
model for the dielectric function with Drude parameters of conduction electron density
N = 5.77 x 1021 cm‘3 and mean free time tc = 7.83 x 10’15 sec [24].

 

0.4 III Tllllllrllllll IIIITIIIT

0.3 '—
O.2 "—

0.1 *—

I
n
o'-
a
.-
.-
0"
.u
l‘.
0".
0'.

rlL'lllllltlllllll

Integrated Fractional Signal (cm‘i)

.0
3
.-
.-
.0
av'
......
I
1"
nnI'
IO'

 

 

' l
oo‘ixnlnlllllLllllllllllllJlll

60 65 70 75 80 85 90
Incident. Angle (deg)

 

Figure 2.3.3 Integrated fractional IRAS signal from the adsorbate-substrate stretch
vibration (solid line) and the internal vibration (dotted) line of CO on Pt(l l 1), using a
non-interacting harmonic oscillator model with the same e' but different u's. The
difference of the two signals comes mostly from the difference in u' s. Note that the

signal has a peak at near grazing angle and at 500 cm‘1 the peak occurs closer to
grazing.

65

 

 

 

 

-1 ‘ I I I I I I I I I I I I I I I If I I I I I fr?
1° l l I l l
10'2 g
E I
go 10‘3 l
"g 300 K ambient. noise
g 10'4
c
I: l‘ 100 K ambient noise
:5 10'5
0
a.
10-6 source photon noise
10.? L L4 11441111111111141LLILIJ l 1 lg

 

500 1000 1500 2000 2500 3000
Frequency (cm")

Figure 2.5.1 Photon-noise-limited sensitivity of IRAS experiments on platinum surface
as a function of frequency for three cases; source photon-noise—limited case, 100 K and
300 K ambient photon-noise-limited cases.

 

 

 

 

- _.-..._.-.-.- _.-......... [._.-.- —B_.—v-. ...... .-.-.I—.—v ........ p... _l-.-.-.— _._.—.- --—.-._ -.- —.—.—»-.4 ...... I_.--._-.-.

source optics sample optics spectrometer optics detector

 

 

 

 

 

 

 

 

 

 

 

 

X chopper or other modulation

Figure 2.6.1 Schematic of a generic IRAS system. Source could be thermal or non-
thermal. Chopper can go anywhere. For FTIR, modulation usually comes from the
moving mirror. Spectrometer could be grating, FTIR or CVF, but it has to have
particular A!) and Av. Spectrometer can, and usually does, precede sample [1,6,8,9].
For diode laser source, no spectrometer may be needed [24].

67

 

 

   

1017 : I I I f I I I I I I I I I I I I I I I I I I I I I I I I I I
F I -
_ 1300 K ‘
A P- —
I P \.
i3
8 1016 e“ “‘5
.‘J I- -1
IE " _,
U ..
I
8 1 o 15 _:
m _
x I 300 K I
It. _ -1
G _1
0 : :
.8 I
.c: _ 100 K :
Du . _
l l n l l L l l l l l l l L J L l l l J l l l l

 

 

 

 

O 500 1000 1500 2000 2500 3000
Frequency (cm‘I) '

Figure 2.7.1 Photon flux per unit frequency for a single polarization from a
blackbody at 100 K, 300 K and 1300 K, as a function of frequency. The vertical
dotted line indicates the detector cut-off frequency.

3. Description of the Apparatus

The basic optical layout of the system is shown in figure 3.1.
Chopped radiation from a globar, which is surrounded by a liquid-nitrogen-
cooled housing, is focused onto a sample in the ultrahigh vacuum chamber
through a transfer optics and a differentially pumped CsI window. The
incidence angle is 86° from the normal to the sample surface. The
reflected radiation is collected, and refocused onto the entrance slit of
a liquid-nitrogen-cooled grating monochromator through another
differentially pumped CsI window and a transfer optics. The spectrally
dispersed output is then refocused onto a liquid-helium-cooled Si:B
photoconductive detector through a transfer optics which is also cooled
with liquid nitrogen. The output signal is demodulated with a PAR 5209
lock-in amplifier and is collected by an IBM compatible computer, which
also controls the grating position, and permits precise data manipulation.
In order to avoid atmospheric absorption, and to allow the cooling of the
components, the entire optical path is evacuated to a pressure of less

than 10'5 Torr .

3.1 Radiation source in a liquid-nitrogen cryostat

We use a conventional silicon carbide globar (Oriel 6361) as a
source. Several other commercial thermal IR sources were tested, but were
found to be inferior in brightness, stability or powerconsumption. The
globar is in‘a cylindrical copper housing, which is accommodated in a
home-built liquid-nitrogen cryostat. The globar is heated resistively to
approximately 1300 K by an HP model 62648 power. supply (typical power

consumption; 8 A x 12 V). The source temperature stabilizes in about two

68

69

hours after the power supply is turned on. The fractional intensity
fluctuation noise of the source is 3 X 10'5 l/a/l—lz, which is measured with
the source cryostat directly coupled to the detector with no filter.

3.1.1 Cryostat

The schematic diagram of the cryostat is shown in Figure 3.1.1. The
outer shell is made of stainless steel in a cylindrical shape of 7.0"
(17.8 cm) diameter and 18" (45.7 cm) height. The optical components and
source housing are bolted down on a 0.5” (1.3 cm) thick copper plate,
which forms the bottom of the liquid nitrogen can. The radiation shield,
l/32" (0.079 cm) aluminum sheet, surrounds the optical components and the
copper source housing. The entire inner assembly is supported by three
3/8" (0.95 cm) diameter stainless steel rods, which are kinematically
mounted on.the bottom flange using,0.25' (0.64 cm) diameter ball bearings.
The supporting rods are thermally insulated with teflon bases from the
room temperature bottom flange.

The temperatures of the optical components and the radiation shield
are monitored using silicon diodes. Silicon diodes are widely used as
inexpensive but reliable low temperature sources in the temperature range
77 - 300 K [1,2]. The effective resistance of a silicon diode increases
as the temperature decreases under a low forward current bias (10 -
100 uA). Silicon diodes, type 1N4l42, are epoxied with a low temperature
epoxy, Stycast, or clamped on the mirror, base plate, radiation shield and
source housing. The diodes are biased with a constant current of 20 “A.
The voltage across the diode is monitored by a lab computer, using an
analog/digital input/output board (A/D I/O board; Analog Devices, model

RTI-815) and a chart recorder program written by Professor R. G. Tobin.

70

It takes about 4 - 5 hours to cool the system to a stabilized temperature
of 100 K. Figure 3.1.2 shows the temperature as a function of time for
each monitoring point. Note that mirror cooling is the limitation. When
the globar is turned on at 1300 K, the temperature of the source housing
rises to approximately 120 K.

The source cryostat is pumped out by an Ultek ion pump to about
1 x 10'" Torr and is cooled by liquid nitrogen. An automatic liquid-
nitrogen feeding device is installed. Approximately 2.5 liter of liquid
Initrogen is transferred at a time and lasts about 40 minutes when the
globar source is operated at 1300 K.

3.1.2 Globar source in a copper housing

The schematic diagram of the source housing is shown in Figure 3.1.3.
The globar [1/4" (0.64 cm) diameter, 4" (10.2 cm) long] is connected to
electrical copper wires with copper couplings and set screws. The wires
are electrically insulated from the copper housing and the cold plate with
a flanged teflon.bearing on the top and.with a machined.Macor piece on the
bottom. The estimated power delivered from the globar to the housing is
about 80 W for a source temperature of 1300 K. The wall thickness of the
housing is 1/8' (0.32 cm), which is so determined for the heat to be
transmitted effectively to the bottom cold plate.

3.1.3 Ellipsoidal and plane mirrors

There are two mirrors in.the source cryostat; an.off-axis ellipsoidal
and a plane mirror. The ellipsoidal mirror is a gold coated front surface
mirror, custom made by Interoptics Co, 60' off-axis angle between rays
from center of mirror and foci, 1.75" (4.45 cm) diameter, 2.000”

(5.080 cm) and 4.000” (10.160 cm) effective focal lengths from the center

71

of mirror to the foci. Hence the magnification is 2.0. The optical
layout for the source is shown in Figure 3.1.4. Infrared light from the
source is reflected and focused with a magnification of two, by the
ellipsoidal mirror to an exit slit where the tuning fork chopper (see
below) is located. The Optic axis is 3.00" (7.62 cm) from the cold base
plate. The ellipsoidal mirror is mounted on.a modified.NeWport model MM-Z
mirror mount with a rising block. The mirror can be adjusted from the
outside of the cryostat; a square hole is made on each adjusting screw and
1/8" (0.32 cm) diameter rod with matching square tip is used to adjust.

The plane mirror, 1/2" (1.27 cm) diameter circular shape from Newport
00., is mounted on a turret mirror mount, Newport model TM-75. The mirror
is used for reflecting a laser beam coming through a window from the
outside of the cryostat into the infrared optic axis. The laser beam is
used for aligning the whole system and for calibrating the spectrometer.
We can move the mirror along a rail on the cold plate using a 1/8"
(0.32 cm) rod with a coupling from the outside of the cryostat. The
mirror is usually out of the infrared optical path and is moved in only
for alignment.

3.1.4 Tuning fork chopper

Since it is required that the chopper should be operated in vacuum
and at liquid nitrogen temperature, a motor driven mechanical chopper is
not acceptable in our system. Instead we use a custom made tuning fork
chopper, Multi-Scanning System Corp. Type RC-2, operating at a fixed
800 Hz chopping frequency. The aperture width is 1 mm (0.039") at rest
and 2 mm (0.078") at full opening. The tuning fork chopper requires very

low power and more importantly no lubrication, which is required for

72
operating at vacuum and low temperature.

The tuning fork assembly consists of the fork, a pick up coil, a
driving coil, a magnet and a mounting base. The coils and magnet are
located near the vanes. The motion of the vanes generates an AC voltage
in the pick up coil which is amplified by a solid state amplifier in a
controller box. The amplified voltage is applied to the drive coil in the
proper phase to cause the vanes to begin vibrating at the fork's own
frequency. The signal voltage available from the amplifier is synchronous
with the vane motion and is used for a reference signal in demodulation.
The amplitude stability of the chopper is crucial, since it relates
directly to signal amplitude. The fractional noise of the amplitude is
~5 x 10-5 l/JIE which is measured by a lock-in amplifier feeding the
chopper controller signal as an input and reference signal. We had to

modify the drive electronics to achieve this level of stability.

3.2 Transfer optics

There are two transfer optics, mirror images of each other; one in
'between the source cryostat and the UHV sample chamber, and the other in
between the UHV sample chamber and the spectrometer. The first transfer
optics literally transfer lights from source to sample and the second one
from sample to spectrometer.

Each transfer optics consists of two off-axis paraboloidal mirrors:
30° off-axis angle and 3.000" (7.620 cm) diameter for both mirrors, and
17.722” (45.014 cm) and 9.646” (24.500 cm) effective focal lengths
respectively. Both mirrors, custom made by Interoptics Inc., are gold
coated front surface mirrors. They are mounted on NeWport MM-2 mirror

mounts, microrails and translator carriers which ease the linear

73
alignment. One mirror can be adjusted from outside the vacuum can, the
other can be aligned only when it is open to atmosphere. The chambers are
pumped out to a pressure of 10'5 torr or lower to eliminate atmospheric
absorption.

The first transfer optics images the chopper aperture onto the sample
with a magnification of 1.8 and the second images the sample onto the
entrance slit of the spectrometer, demagnified by the same factor.
Together they enable the infrared radiation to illuminate the whole area
of the sample with an adequate size of the infrared windows and to be
collected by a spectrometer with a relatively large focal number f/4. It
is important to have the infrared radiation interact with as many
adsorbates as possible for better sensitivity. The sample size is,
however, limited by other factors, such as availability, cost and
cleanness control. The infrared radiation needs to illuminate the whole
area of a given sample. On the other hand, the effective solid angle of
the infrared radiation on the sample is limited by the geometry of the
sample chamber, that is, the diameter of the infrared window and UHV

chamber .

3.3 UHV system

The sample chamber is a custom-built stainless steel ultrahigh vacuum
(UHV) chamber, 12" (30.5 cm) diameter and 25" (98.4 cm) height. The
chamber is divided into three levels: infrared level in the lowest,
surface analysis level at 8" (20.3 cm) from IR level, and a manipulator on
the top. The manipulator can travel 2!: 1" (2.54 cm) in x- and
y-directions, and 10" (25.4 cm) in z-direction, and can be rotated 360°

with a differentially pumped rotary seal (both from Thermionics Lab.

74
Inc.). On the infrared level there are two differentially pumped CsI
windows and two leak valves connected to the gas doser. On the surface
analysis level we have an ion sputtering gun, an Auger electron
spectroscopy (AES) system, low energy electron diffraction (LEED) optics
and a quadrupole mass spectrometer.

The UHV chamber is pumped by a Perkin-Elmer TNBX-lOOO ion pump system
equipped with a molecular sieve sorption pump on the roughing manifold and
a titanium sublimation pump. A turbomolecular pump is also installed on
the roughing manifold, and is used for fast initial pumping after exposure
to atmosphere, and.during argon ion sputtering to clean.the sample because
argon gas is not easily pumped out by the ion pump.

The vacuum pumping process begins with rough pumping by the molecular
sieve sorption pump. Once the pressure in the chamber reaches 1 x 10‘1
Torr or less, we close the roughing valve to the sorption pump and open
the roughing valve to the turbo pump. When the pressure reads about
1 x 10'6'Dorr or lower, we close the roughing valve to the turbo pump and
the bakeable.manifold roughing valve which seals the UHV chamber from the
roughing manifold. The ion pump is turned on, or the poppet valve is
opened slowly if only the sample chamber is at atmosphere. Pumping
overnight brings the pressure down to 5 x 10'2m Torr or lower. The base
pressure goes down to 4 X 10'11 Torr after bake-out at 110%: for about
48 hours. The bake-out is performed by two Calrod heaters around the
lower chamber, enclosed by a bakeout blanket, and by an internal quartz
heat lamp and several heating tapes on the sample chamber, which is

covered with aluminum foil. Temperatures on the chamber are monitored by

75
using a switch panel with a Omega Engineering model 199 meter and K-type
thermocouples.

3.3.1 Surface analysis tools

Experimental studies of surfaces all require sample preparation. We
want to study single crystal surfaces whose surface composition.and atomic
structure are uniform» ordered, and free of' defects. Carbonaceous
compounds are easily deposited on the surface when the sample is exposed
to atmosphere, or impurities from the bulk of the sample may diffuse to
the surface upon heating and segregate at the surface [3]. In order to
prepare clean surfaces it is necessary to get rid of those impurities by
chemical treatment, by heating the sample in flowing gas, like oxygen or
hydrogen gas, or by ion sputtering of inert gas, like argon or neon gas.
A Varian Ion Bombardment Gun, Model 981-2043, is installed for ion
sputtering.

For examining surface cleanness Auger electron spectroscopy (AES) is
the most widely used. The sensitivity of AES is about 1 percent of a
monolayer [3]. A.Perkin-Elmer Model 10-155 Cylindrical-Mirror Analyzer is
installed for AES analysis. The analyzer detects Auger electrons ejected
from specimens excited with an electron beam which is generated by the
electron gun in the analyzer. In routine operation, the optimum sample
position is obtained by adjusting the sample position to the correct
elastic peak position, which is set at 2 KeV. The emission current of the
filament is typically set at 0.2 mA which gives a current of 5 pA.between
the sample and ground.

A reverse view Low Energy Electron Diffraction (LEED) optics

(Princeton. Research Instrument Co.) is used to investigate ordered

76

adsorbate structures and to examine the sample cleanness. It focused,
monoenergetic electron beam from the electron gun strikes the sample which
is atomically clean and ordered. Diffracted beams emerge in various
directions with various intensities from the sample and after passing
through several grids reach the collector screen, which is coated with a
phosphor. The visible LEED pattern on the screen is viewed from behind
the collector screen through a view port in the center of the mounting
flange.

A Dycor quadrupole mass spectrometer, model MAlOOM, is installed for
residual gas analysis in the UHV sample chamber and temperature programmed
desorption (TPD) measurements. The mass spectrometer is controlled by a
computer through R8232 interface to measure TPD and other residual gas
analyses.

3.3.2 Sample mount

The sample mount consists of a stainless tube, which serves as a
liquid nitrogen reservoir, a Conflat flange with three copper feedthroughs
for heat transmission, and a pair of copper bars where the sample is
mounted. The sample [1 cm width x 3 cm length x 0.12 cm thickness
(0.39” x 0.39" x 0.047")] has 0.025” (0.06 cm) slits cut into both ends
and is clamped.by thin tantalum clamps. Figure 3.3.1 shows the lower part
of the sample mount. Thin sapphire washers are placed between the sample
and the copper bars. The sapphire washers provide electrical isolation of
the sample from the copper bars, which carry the heater current. Further,
the sapphire washers actually help to cool and heat the sample
effectively, since they are good thermal conductors at low temperature,

but poor at high temperature.

77

By cooling with liquid nitrogen the sample temperature goes down to
90 K in about 15 minutes from room temperature. The sample can be heated
radiatively up to 1400 K by a tungsten filament 1 ~ 2 mm (0.04 ~ 0.08")
behind the sample. The heating can be done manually or controlled by a
computer. The sample mount can be translated in three directions and can
be rotated 360° by a manipulator with a differentially pumped rotary
platform as mentioned above.

3.3.3 Infrared window

There are two differentially pumped CsI windows on the UHV chamber;
one toward the source, the other toward the spectrometer. These are the
only windows in the optical path. Figure 3.3.2 shows a schematic diagram
of the zero-profile differentially pumped window assembly [4], which is
adapted from a design by Hollins and Pritchard [5]. Vacuum seals are
provided by three viton O-rings. The recess in the flange and the O-ring
holder are designed not to touch the window. The window is pressed on
only by the O-rings. This allows thermal expansion of the windows without
.cracking.

The interspace between the lower O-ring and the top inner O-ring is
differentially pumped through a 1/4" (0.64 cm) OD tube welded on the side
wall of the UHV flange. The interspace is pumped out with a mechanical
pump in the UHV foreline. ‘The windows have been baked at ~100°C repeatedly
and a base pressure of ~5 x 10‘“ Torr is achieved in the UHV chamber. The
differential pumping port has often been left valved off for more than a
few weeks with no observable increase in the base pressure. Since CsI is

easily damaged by moisture in atmosphere, the window is heated to about

78
40W: whenever it is exposed to the atmosphere, and is also kept under
vacuum as much as possible.
3.3.4 Gas doser

A. home-built cosine gas doser is installed approximately 0.4"
(1.0 cm) in front of the sample in IRAS position for gas dosing. The gas
doser consists of a 10” (25.4 cm) long 1/8” (0.32 cm) OD stainless tube
with a rectangular box [0.25” wide x 0.25" high x 1.2" long (0.64 cm x
0.64 cm x 3.05 cm), 0.010” (0.025 cm) wall thickness] which has 5 holes of
diameter 0.050” (0.127 cm), so that each hole acts as a cosine doser. The
gas molecules come out of the holes by effusion, since the diameter of the
holes is much smaller than the mean free path of the gas molecules. If
the diameter of the holes in the closet is sufficiently (2. 3-5 times)
larger than the wall thickness, the gas flux from each.hole is distributed
as a cosine of the angle measured from its center line [6], and the device
is called a cosine doser. The calculated uniformity of the gas flux is
shown in Figure 3.3.3. The gas flux at the edge of the sample is 71 X of
that at the center. The enhancement factor is approximately 20, which
means that the effective gas flux at the sample surface is larger by a
factor 20 directly in front of the doser than that at other position in
the sample chamber. The enhancement factor was measured.by comparing the
gas doses (measured at the ion gauge, far from the doser) needed to give
the same adsorbate coverage, as measured by temperature programmed
desorption (TPD), for the two methods. The response time of the doser is
2 ~ 3 seconds. Thanks to the high enhancement factor a large amount of
gas dosing can be performed with a much lower background pressure of the

UHV chamber. It is also indispensable for dosing gases of low sticking

79
coefficient, such as oxygen on a Pt surface. Oxygen molecules are easily
replaced by CO molecules in the background if the background pressure is
too high, since the sticking coefficient of CO is much higher than that of
oxygen.

Two leak valves, Vacuum Generators model MD6, are connected to the
gas doser to let gas in for dosing or cleaning the sample. The MD6 leak
rate is continuously controllable between 10'5 and 10'10 mbar l/sec. The
fully closed leak rate is below 1031 mbar l/sec. One of the two leak
valves is connected to a gas manifold, which makes it easy to change
gases. The gas manifold is pumped through the UHV turbomolecular pump.
Whenever it is necessary to change gases a few cycles of pumping and
filling with gas to be used are performed.before filling the manifold with
gas for dosing or cleaning. The other leak valve is used for dosing

isotope gas from a Matheson laboratory bottle.

3 .4 Grating Spectrometer

3.4.1 Cryostat and structure

The spectrometer and cryostat were home built in the MSU physics
department machine shop, except for the outer shell and spectrometer base
plate, which were machined at Raycon Machine Tool and welded at the NSCL.
A vertical cross section of the system is shown in Figure 3.4.1. The
system consists of a spectrometer shell, spectrometer top cover, liquid
nitrogen (1N5) can, spectrometer base plate and radiation shield. The
spectrometer shell is made of stainless steel; bottom 1/2' (1.27 cm)
thick, side wall 3/8" (0.95 cm) thick. The spectrometer shell sits on

four heavy duty casters with V-shaped steel wheels which roll over two

80
rails on the table and thus ease assembling and disassembling. For
mechanical rigidity there are three 1/2" (1.27 cm) square bars running on
the bottom and two on each side wall.

The spectrometer top cover has a top plate and a cylindrical shell
over the liquid nitrogen can. The top plate, which is made of 1/2"
(1.27 cm) thick stainless steel, is bolted down on the spectrometer shell
with a large O-ring seal. The cylindrical shell over the LN; can is
clamped down on the top plate with an.O-ring seal. A 1/2" (1.27 cm) thick
aluminum top plate seals the cylindrical shell with an O-ring. The top
plate above the LN: can has three holes through which three LNZ
feedthroughs run from the liquid nitrogen can.

The liquid nitrogen can is made of 12.0" (30.5 cm) OD x 10.3"
(26.2 cm) high x 1/16" (0.16 cm) wall thickness stainless steel tube. The
top plate, 1/8” (0.32 cm) thick stainless steel plate, has three 1N2
feedthroughs welded on. Each feedthrough consists of a 3/8" (0.95 cm) OD
stainless steel bellows (Cajon model 321-6-181) and a flange with an
O-ring seal and 8 blank tapped holes to be held on the top plate of the
outer shell. The feedthroughs are used for liquid nitrogen filling,
venting and level checking. The bottom plate of the LN; can is made of
3/4' (1.9 cm) thick copper plate cut down concentrically by 3/8” (0.95 cm)
to make four ring-shaped patterns in order to increase the surface area
inside the can. Under the ring-shaped patterns forty-one evenly spaced
1/4-20 blind tapped holes are made to bolt down the spectrometer base
plate.

The spectrometer base plate is made of 1/2" (1.27 cm) thick aluminum

plate and painted black. All the optical components are either screwed

81
down or heat sunk onto the base plate; two [paraboloidal mirrors, two
folding mirrors, the grating table, slit wheels and the polarizer mount.

Similarly to the source cryostat, the assembly of the spectrometer
base plate and the liquid nitrogen can is supported by three 3/4" (1.9 cm)
diameter invar rods, which are kinematically mounted on the bottom of the
spectrometer shell using 5/16" (0.79 cm) diameter ball bearings and
thermally insulated with teflon bases from the room temperature bottom of
the spectrometer shell. The invar rods are used to minimize thermal
contraction and thus to get better alignment of the cold grating drive
shaft with the warm step motor shaft.

The spectrometer base plate is clamped by three 1/4” (0.64 cm) steel
rods from the outside of the spectrometer shell for mechanical stability.
Each rod has 5/16-24 thread in the middle and one end tipped with a 1/4"
(0.64 cm) diameter ball bearing. ‘The detailed clamp mechanism.is shown in
Figure 3.4.2. A quick disconnect with a 1/4" (0.64 cm) hole is modified
to have a 5/16-24 tapped hole through which the rod is screwed, and is
soldered to a modified pipe threaded brass blank. Three modified quick
disconnects are screwed in the side walls of the spectrometer shell. Each
rod presses an angled stainless plate attached on the side wall of the
spectrometer base plate so that the direction of force is toward a point
approximately 1' (2.5 cm) below the center of the base plate. They also
exert a downward force to press the kinematic mounts onto their bearings.

Four electrical feedthroughs are installed on the top plate of the
spectrometer shell. Each electrical feedthrough is a hermetically sealed
Bendix 6-pin electrical connector soldered on a brass flange with an

O-ring seal. One of these is used for connections to silicon diodes to

82
monitor the temperature inside the cryostat. The others are used for
connections to the Inductosyn transducers (see 3.4.3), and to limit
swiches on the grating table.

The spectrometer is pumped out by a turbomolecular pump (200 l/s
pumping speed) backed with a mechanical pump. The effective pump speed
for a vacuum chamber to reach a high vacuum pressure is essentially
determined by the outgassing rate and area of the inner surfaces of the
chamber. The effective pumping speed S.“ to maintain a required pressure
P is given by [7]:

R8 A

S.“ - . (3.4.1.1)
P

 

where Rs is the outgassing rate of the material and A the surface area.
The inner surface area of the spectrometer is approximately 7.5 x 10" cm2,
including the radiation shield and internal components. Using Rtl- 2.5 x
10’8 mbarls'lcm'2 and P - l x 10-5 Torr, S.“ - 141 l/s is obtained. A pump
size of 200 l/s is chosen with a safety margin. The pressure of the
spectrometer is down to l x 10'5 Torr or below approximately 3 hours after
beginning to pump out.

The spectrometer is cooled with liquid nitrogen after it is pumped
down to high vacuum. It takes approximately 24 hours to cool the
spectrometer as shown in Figure 3.4.3. The total volume of the LNélcan is
approximately 20 liter and.the hold time is about 12 hours. Approximately
10 liter is transferred at a time with an automatic feeder and lasts about
5 hours. A pressure of 5 x 10'7 Torr is usually maintained when the

spectrometer is cold.

83

3.4.2 Optical components

As mentioned above, all the optical components of the spectrometer
are installed on the cold base plate. The optical layout of the
spectrometer is shown in Figure 3.4.4. Light coming through the entrance
slit is reflected off the folding,mirror and the collimating mirror to the
diffraction grating. The diffracted beam is reflected off the camera
mirror and another folding mirror to the exit slit through the polarizer.

3.4.2.1 Entrance slit

The entrance slit assembly of the spectrometer consists of a slit
wheel and a driving mechanism. The slit wheel is a 4" (1.0 cm) diameter
3/16" (0.48 cm) thick aluminum spur gear which has eight 3/4" (1.91 cm)
diameter 0.094” (0.239 cm) deep holes with 1/4" (0.64 cm) diameter through
holes at each center, where slit elements are installed. Six slit
elements of slit widths 0.25, 0.50, 1.0, 1.5, 2.0 and 2.5 mm are
installed” The slit elements, made of brass, are 0.090” (0.229 cm) thick,
3/4" (1.91 cm) diameter circle and have 1/8" (0.32 cm) wide groove along
which the actual slits are made. The slits are made in the Physics
department machine shop using an electric static discharge machining tool.

In order to change the slit, the slit wheel is rotated through a
pinion gear and a pair of bevel gears, the shaft of which is rotated by a
knob from outside the spectrometer. The coupling between the inner cold
shaft and the room temperature shaft is engaged only when it is used, to
keep heat transfer minimized.

3.4.2.2 Folding mirrors
Two folding mirrors are used in the spectrometer to change beam

direction by approximately 90°; located at 1.25" (3.18 cm) from the

84
entrance slit and the exit slit respectively. The folding mirrors are
elliptical plane ndrrors, bought from Edmund Scientific Company, with
dimensions 1.50" x 2.12” (3.81 cm x 5.38 cm) for a mirror near the
entrance slit and 1.88” x 2.65” (4.78 cm x 6.73 cm) for a mirror near the
exit slit. They are gold coated front surface mirrors on 3/8" (0.95 cm)
and 1/2" (1.27 cm) thick glass substrates, respectively. Each mirror is
clamped down on an aluminum block using two teflon brackets and four
screws for mechanical stability and good heat sinking. The aluminum block
is mounted on a brass mirror mount with three pairs of push-pull screws
[8], and is heat sunk through two 1/4” (0.64 cm) wide copper braids to the
cold base plate.
3.4.2.3 Paraboloidal mirrors

Two off—axis paraboloidal mirrors are used in the spectrometer as the
collimating mirror and the camera mirror. The mirrors are custom-made by
Interoptics Limited, cut from aluminum substrates with diamond machining
and coated with gold, with a figure accuracy of 1 wave at 6328 A. The
collimating mirror is 6” (15.2 cm) in diameter with an effective focal
length of 25.422” (64.572 cm) and an off-axis angle of l4.80°. ‘The camera
mirror is 8” (20.3 cm) in diameter with the same focal length and off-axis
angle. The diameter of the camera mirror is so determined that no light
from the grating is lost (see below).

Each mirror is bolted on a 3/8” (0.95 cm) thick brass plate with two
screws and two dowel pins. The plate is mounted onto an L-shaped aluminum
bracket with three pairs of push-pull screws such that the mirror can be
adjusted with those screws. The L-shaped bracket has three screw slots,

which allow us to move the mirror back and forth in the direction of

85
optical axis. The bracket is screwed onto the base plate. For better
heat-sinking of the mirrors, a 1/2” (1.27 cm) wide copper braid is used
for each mirror: one end is soldered on the side wall of the brass plate
on each mirror and the other end on a copper piece which is screwed down
to the base plate. Nevertheless, it requires 18 hours to cool the
mirrors.
3.4.2.4 Diffraction gratings

Three different gratings, which are purchased from Milton Roy
Company, are used to cover the full spectral range of interest, from
330 cm'1 to 3000 cm'l, with useful resolution and efficiency. The
specifications of the three gratings are shown in Table 3.4.1. The
gratings are replicas on aluminum substrates with dimensions 8.66" x 6.50"
x 1.38" (22.0 cm x 16.5 cm x 3.5 cm). The grating size is determined such
that no light reflected from the collimator is lost at the grating angles
to be used. The height and width should be equal to, or greater than
6"/cos40° z 8" (20.3 cm) and 6" (15.2 cm) respectively.

The gratings are installed one at a time and the spectrometer must be
opened to change the gratings. The grating is mounted on a grating table
which is made of two 1/4” (0.64 cm) thick aluminum plates to be adjustable
like a mirror mount, and is held down by a big aluminum clamp. The
grating table is rotated on a pair of precision angular contact ball
bearings with a 5/8" (1.59 cm) diameter shaft. Since the grating is
massive, a good heat-sink is required to cool the grating effectively.
Tens of 1/4" (0.64 cm) wide copper braids are soldered on a 1/8" (0.32 cm)
thick copper plate [2" (5.1 cm) wide x 4" (10.2 cm) high] which is screwed

onto back of the grating. The other side of copper braids are soldered on

86
a copper block, which is screwed onto the cold base plate. Even so, it
requires 24 hours to cool the grating.
3.4.2.5 Exit slit and filters

The exit slit wheel assembly is just the same as the entrance slit
assembly except that filters are mounted on the slits. The filters are GE
varnished on the slit elements. For the identification of the slit
positions, one of the silicon diodes used for temperature measurement is
mounted on a slit position with a switch, which is turned.on when the slit
position is selected. The filters are mounted at the slits rather than at
the detector, after we have found that the filters cause oscillation
features in the spectrum if they are cooled to liquid helium temperature.

3.4.2.6 Polarizer

An infrared polarizer is installed in front of the exit slit of the
spectrometer. The infrared polarizer is a Cambridge Physical Sciences
model IGP225 polarizer which consists of an array of parallel aluminum
wires (0.2 micron wide and grid spacing is approximately 0.4 micron) on a
2 mm (0.079") thick KRS-S substrate. The polarizer will pass radiation
whose E vector is perpendicular to the grid lines. The polarizer is
mounted on a 1.625" (4.128 cm) diameter and 0.25" (0.64 cm) thick ring,
which has 1.0" (2.54 cm) diameter clear aperture, and it is set for p-
polarization. It cannot currently be rotated when the spectrometer is
under vacuum.

3.4.3 Grating drive mechanism

Rigidity and reproducibility of the grating drive and mount are

absolutely critical, since errors appear as noise in infrared spectra” ‘We

have put lots of care and effort into the design and construction of the

87

grating drive and mount. In order to get the resolution of 1.0 cm'1 at 400
cm'l, the required angular step size is only O.l°. The effect of
non-reproducibility of the grating angles from one scan to the next
depends on the slope of the intensity versus angle I(0) curve as follows:
AI 1 I dI I
_ .. _ __ A9, (3.4.3.1)
I I d0
where A0 is an angular error. So in order to achieve the fractional noise
level of <1 x 10", the angular error A0 should satisfy the condition:

1 x 10"

as < , (3.4.3.2)
(dI/d0)/I

 

and it has to be determined empirically.

The original design of the grating drive mechanism was that the
grating is driven by a stepping motor (Superior Electric, Model M062) and
a stepping motor controller (Maxwell Electronics, Inc. Model SMC-202A)
with a 360:1 worm gear. The absolute grating position is not measured
except at 0°. The angular position of the grating is determined by the
number of steps of the stepping motor. The best angular accuracy achieved
by this way is approximately 1 x 10"3 degree, and the noise from the
angular error dominated the system noise. Some of the problems are due to
the fact that we can not use lubrication on the worm gears, since they are
operated in vacuum and at cold temperature. We use a low friction coating
of molybdenum-disulfide compound on the gears, but there is still some
wear. We made many efforts to improve the angular repoducibility without
major improvements. I

We have decided to use a device called an Inductosyn rotary position

transducer [9] to measure the absolute angular position of the grating,

88

and a feedback system to control the stepping motor. The device can
measure the grating position up to 3.0 x 10'5 degree, which is actually
limited by the 16-bit A/D converter. The transducer consists of a rotor
and stator; each has a hairpin pattern of metal with a period of 2° on the
face. The rotor and stator are mounted under the worm gear and on the
bearing holder of the grating table assembly, as shown in Figure 3.4.5.
The rotor is excited with a 10 kHz sine wave, and then a sine wave is
induced on the stator. The relative phase of the induced sine wave shifts
360° for every 2° rotation of the rotor. This phase shift is detected
electronically and converted to the grating angle. This angular
information is used as a feedback to control the stepping motor. The
inductosyn significantly improves the reproducibility of the scans. One
drawback of the feedback system is that it is much slower.

The stepping,motor can rotate the grating drive shaft either directly
with a geared pulley or through a 100:1 speed reducer (Allied Device model
DU312) using a clutch (Autotronics Inc. model MBC 8-3). Depending on
whether the clutch is turned on or not, the speed reducer is engaged or
disengaged to the driving shaft. The clutch is turned off when the
grating is rotated more than 1°.

Figure 3.4.6 shows a schematic diagram of the grating drive
mechanism. The shaft from the clutch is connected to a ferrofluid
feedthrough (Rigaku.Co. model RMS LS-6) with.a universal lateral coupling,
which is coupled to a worm driver shaft with a bellows coupling and a
detachable coupling mechanism. The detachable coupling is designed to
minimize the backlash in one direction, using the spring tension of the

bellows coupling. The assembly of the stepping motor, speed reducer,

89
clutch and ferrofluid feedthrough is mounted on the side wall of the

spectrometer, and can be detached when we open the spectrometer.

3.5 Cooled Transfer Optics

There is a transfer optics between the spectrometer and the detector,
which consists of an ellipsoidal mirror on a mirror mount, a vacuum can,
a copper radiation shield and an LN; dewar, as shown in Figure 3.5.1. The
ellipsoidal mirror is a gold coated front surface mirror (from Interoptics
Co.) , 90° off-axis angle between rays from center of mirror and foci, 2.25"
(5.72 cm) diameter, 5.800" (14.732 cm) and 7.250" (18.415 cm) effective
focal lengths from center of mirror. Hence the magnification is 1.25. A
3/4" (1.91 cm) ID edge-welded bellows is welded onto the bottom of the LN;
dewar and brazed to a copper block which is screwed down on the radiation
shield. The mirror mount is attached to the radiation shield and can be
adjusted from the outside of the vacuum can. The radiation shield is held
by screws tipped with spring loaded ball bearings against of the inner
wall of the vacuum can. The transfer optics can be cooled by liquid
nitrogen in order to reduce the room temperature background loading on the

detector.

3.6 Detector

The detector is an extrinsic Si:B ptotoconductor sensitive to
frequencies greater than 330 cm'l, and with a size of 1.0 mm x 1.0 mm x
0.4 mm (0.039" x 0.039" x 0.016”) and installed in a liquid-helium dewar
model HD-3, both from Infrared Laboratory. The detector is mounted on a
circular cavity of 4 mm (0.157”) diameter at the exit aperture of a

Winston cone. The Winston cone is a parabolic cone whose internal surface

90

is gold-coated to enhance reflectivity and thermal cooling. It is
designed such that any light entering one end of the tube within.the focal
ratio of the cone undergoes successive reflections on the wall, and
eventually emerges from the other end. It is very effective in collecting
far-infrared radiation.and delivering to a small area, such as a detector,
although it does not form an image. The Winston cone has an entrance
aperture of 10 mm (0.39”) diameter at a focal ratio of 4.4 and exit
aperture of 1.0 mm (0.039”).

A transimpedence amplifier (TIA) is used with two switchable feedback
resistors, 1 M0 and 60 M0, which are heat sunk to the cold plate of the
helium cryostat. A latching relay (Teledyne type 722-9) is used for
switching the cold feedback resistors. The relay is activated by a short
pulse from a spring-loaded center-off switch for low power dissipation.
The TIA consists of two stages, as shown in Figure 3.6.1. The first stage
of the amplifier is a dual JFET type at low temperature, and the second
stage is a conventional operational amplifier at room temperature. A
matched pair of J230 JFETs are used in the preamplifier along with a thin
film heater mounted on a thin fiberglass substrate to allow the JFETs to
operate at 77K.

Under a constant voltage bias on the detector, the detector current
is proportional to the number of photons per second N incident upon the
detector. In general the detector current can be written as;

I - eGnN , (3.6.1)
where I is the detector current, e the electronic charge, G the
photoconductive gain, 0 the quantum efficiency of the detector, and N the

incident photon rate. The JFETs are biased to maintain the node between

91

the detector and load resistor to be a virtual ground. Since the gate
impedance of the JFET is so high, the detector current is forced to flow
through the feedback resistor R1. The input test voltage is given by:

Vn - - 111,, (3.6.2)
and its DC level gives a measure of the total (background and signal)
photon flux. Since the second stage amplifier is AC coupled, the output
voltage gives only the chopped part of the signal and is given by:

Von - AslnR, , (3.6.3)
where As is the gain of the second stage amplifier, switch-selectable as

2 or 10.

3.7 Data Acquisition

An IBM.AT compatible computer is used to collect and analyze data for
the experiment. A block diagram of the data acquisition system is shown
in Figure 3.7.1. In order to measure an infrared spectrum of the sample,
we need to get the detector signal as a function of the grating angle,
which is converted to the spectral frequency. The infrared.beam, which is
modulated by the tuning fork chopper, arrives at the detector and is
converted to an electrical voltage signal by the transimpedence amplifier
as discussed in 3.6. The detector signal is demodulated by a lock-in
amplifier with a reference signal from the chopper controller. The lab
computer controls the grating drive mechanism with a grating drive
program, written by Prof. R. G. Tobin, and also permits precise data
manipulation. It communicates with the stepping motor controller and
Inductosyn to read and change the grating angle with feedback, and then
reads the data from the PAR. 5209 lock-in amplifier' over the GPIB

interface. In a typical data acquisition, we integrate for 1 second at a

92
data point with the lock-in amplifier time constant of 0.030 second, and
move to the next point (change the grating angle) and wait 0.2 second for
the grating mount to settle down before starting the next data-taking. It
takes approximately 20 - 40 minutes for a typical spectral scan of 150
points. Most of the time is spent setting the grating position, and we
are working to improve the speed and efficiency of the grating control

system.

3.8 Optical Alignment

3.8.1 Optical alignment of the spectrometer

Optical alignment of the spectrometer is performed on the table at
atmosphere with both a laboratory He-Ne laser (Uniphase model 1101) and a
white light source (Oriel model 6343). In the first place, the He-Ne
laser is placed on a optical rail to illuminate the entrance slit with
both folding mirrors taken out. The position of the optical rail is
adjusted such that the laser beam gets through both the entrance and exit
slits to make sure that the laser beam is in the optical path. Both
”folding mirrors are placed in position and the first one is adjusted so
that the laser beam strikes the center of the collimating mirror. Next,
the collimating mirror is adjusted so that the laser beam strikes the
center of the grating or a plane mirror in place of the grating, which is
at 0°, and subsequently hits the center of the camera mirror. If the laser
beam misses the center of the camera mirror we need to adjust the grating
table using six adjusting screws. Now adjusting the last folding mirror
should direct the laser beam into the exit slit.

After finishing the alignment with the laser, we place a white light

93

source on the optical rail such that the point image of the source, which
is usually 1.0 mm (0.039") diameter or so, strikes the center of the
entrance slit and the expanded.beam is centered.on the collimating mirror.
The image on the mirror can be seen by placing a sheet of ruled tracing
paper in the beam. The focal ratio of the white light source is so
adjusted that the beam diameter is big enough to cover most of the
collimating mirror. The beam reflected from the collimating mirror is
examined to be a parallel beam at a distance as long as 10 feet from the
collimator with the grating taken out. A plane mirror as big as the
grating installed in place of the grating often makes the alignment
procedure easier. Then the position of the camera mirror is so adjusted
that the reflected beam is focused at the exit slit. The focused image on
the exit slit is examined.by eye. A fine tuning of the camera mirror and
the folding mirror is usually good enough to obtain a sharp image of 1.0
mm (0.039”) diameter or smaller at the center of the exit slit. Fine
adjustment of the grating table is often necessary to make the vertical
position of the image stay at the center of the slit as the grating angle
is increased. The vertical position can be maintained at the center
within 1 0.1 mm as the grating angle is increased to as high as 60°.

The alignment is rechecked with the laser beam and the point source
after the spectrometer is assembled in the vacuum shell and pumped out.
The alignment is not usually changed under vacuum. For this reason it is
not necessary to be able to adjust any of the spectrometer mirrors under
vacummL However, the alignment was monitored and found to change slightly
during cooling of the spectrometer. When the spectrometer is stabilized

at liquid nitrogen temperature, the vertical alignment is shifted downward

94
by 0.5 mm. A deliberate upward misalignment by 0.5 mm is therefore made
before cooling the spectrometer to make up the shift.

3.8.2 Optical alignment of the whole system

After the spectrometer is aligned at the room temperature, the whole
system is assembled as shown in Figure 3.1, excluding the detector to
observe the image by eye. The He-Ne gas laser is placed in front of the
window of the source cryostat with the small plane mirror pushed in. The
laser beam should be directed perpendicularly to the window, which is
confirmed by looking at the reflected beam from the window back to the
laser head. The laser beam passes between the chopper blades in the
source cryostat and arrives at the sample through reflections from the
paraboloidal mirrors in the transfer optics. Now'the laser beam reflected
from the sample comes through the second transfer optics and into the
entrance slit of the spectrometer. Fine adjustments of the paraboloidal
mirrors in the both transfer optics are often necessary to direct the
laser beam to the center of the entrance slit. The image is then observed
at the point of detector through another transfer optics.

After the alignment is done with the laser beam, the detector is
installed. The spectrometer and the source cryostat are pumped out and
cooled with liquid nitrogen. With the infrared source turned on, we often
need fine adjustments of the mirrors in the source cryostat and three
transfer optics, and the angular position of the sample, to obtain a
maximum signal.

In this chapter the mechanical details and operations of our system
have been described. The next chapter will describe the characterization

and performance of the system.

95

References

l.

N. R. Waltham, G. F. Clark and B. K. Tanner, Phys. Educ., 16, 104
(1981).

M. Ganapati Rao, Temperature, 1205 (1982).

G.A. Somorjai, Chemistry in two dimensions: Surfaces, Cornell
University Press, Ithaca, New York, 1981, pp. 37 - 84.

R.G. Tobin, C. Chung, and J.S. Luo, J. Vac. Sci. Technol. to be
published.

P. Hollins and J. Pritchard, J. Vac. Sci. Technol. 17, 665 (1980).
C. T. Campbell and S. M. Valone, J. Vac. Sci. Tech. A3, 408 (1985).
Vacuum Technology: Its Foundation, Formula and.Tab1es, pp. 39 - 43, in
Product and 'Vacuum 'Technology Reference Book 'by Leybold. Vacuum
Products, Inc. (1991).

J. F. James and R. S. Sternberg, The Design of Optical Spectrometers,
Chapman and Hall Ltd., London (1969), pp.188 - 189.

B. M. Gentry, L. Larrabee and C. Laurence Korb, Appl. Opt. 23, 2401
(1984). Inductosyn is available from Farrand Controls, Inc.,

Valhalla, N.Y.

96

Table 3.4.1 Specifications of the diffraction gratings

 

 

 

 

 

Grating number Number of Blaze angle Blaze frequency
grooves per mm (cm'l)
l 60 28° 41' 625
2 100 26° 45' 1111
3 300 36° 52' 2500

 

 

 

 

 

97

 

.358 Emmi—.8 on 055% 085ch .5693 v5 .35 :5 15 85:5 Lemon—2.8%
GEES .2259“ c2528 .838 2:. .6393 E88588... Bags be 3m: 05 do 59?. _8_EO _.m 2:9...

$.58".

 

 

>qoomocmuuam :o_LQcomuo _osco;_ Luggage vac cm~mco_o¢

388:8QO E58; $9: .mEoEm 838-5 3: ;

co_Lucha_o cocmum_w >Ocmcm gem Loamo>co moczom m_

U885 82% case

 

\lllllumar

 

 

 

 

 

comummoa umc0c_:_
m”_m wm_ooulmzm

 

 

 

 

 

seem

 

 

 

538523 9:95
Eosasoszaa:

3339 :2. 838.23 S

u_qcem _oum>cu m_o:_m ;L_a cwpsczu
m_m>_o:< ouoaczm ssaoo> ;o_;oLL_:

 

 

 

x as - a as; Beam
:8 m; ”Ezeamm
TE 83 - 8m ”38.. 858%

555 EoumoEBam “Giza Swizz -mfi

 

 

 

 

 

98

Liquid nitrogen Electrical
/ feedthrough / feedthrough

.v :L, LT
“E III
Eli—:3 t. -L

if W ““1

 

 

 

_

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Liquid
/_ nitrogen can
/
F
\— Cold plate
. tical
/— Supporting rod .==;: /— a?“
Kinematic mount = E

 

 

 

 

 

 

 

 

 

117—— Teflon base

.......................
........ ...............
.......................
.......................
.......................
.......................
.......................
II‘IIIILIIAAJAJAIILILJL

to ion pump

 

 

 

 

 

 

Figure 3.1.1 Schematic of the source cryostat

mmbmoxuo oOLSOm 23 tom mot/.50 wczooo N. _ .m “:st

A2505 o5?

. v m m a o

 

 

)

in"?!

 

— q q — d — d — u

>
I n .v I l .. .e b s
:1 etc .‘I 0'. (9 file) i ’ titli“ -l . . '3

    

maize: moaDOm I

Eme H5339: I
.5925 I

3505 cameos? III

Baa omen. I

fifim fixo I

 

 

OOH

com

com

()1) arnieredmai

100

.\ Flansed teflon bearins

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

 

 

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

 

 

/— Setscrew

/—— Globar

 

 

 

 

 

 

3? /— Macor insulator
3?? m Cold plate

 

T
v

 

 

 

 

 

 

l lllllll
\ . ‘ .'.

 

 

 

 

 

Figure 3.1.3 Schematic of the source housing

101

83% gen 05 we «woke..— véd arm

 

 

 

 

 

 

 

someone
gen newsflash hone—u 48% ”355
32a 28 _
0
g
a a .
.wv -. ................................. m ..... I ...........

 

 

 

 

 

 

 

 

 

 

338 .558
B.E.—E 133350

deuce—«2 aces— new
mesh:— 3 03968

 

 

Goran—:5 .83: EH0 new flea II\.

102
cold copper legs -- connected to

\per feedthroughs

. . 5w“

 

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

42:25:52: thermocouple ;;;;;;;;;:;;; Mam” Pi°°°

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

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

............
.............. um amp
............
. ........ . . . . .

.,..... ..._.:.:.:.:. .'.'.'.'.'.'.,.,....

0.. ....... . . . . . . .....
.‘9.0.. . . . L; . £ . . . . . . . .Q.Q.g..
I": I: .. .. ..
' Z l-i‘ “2'! ’ '-
.. .
a: .'l' 3.1. .
... 15‘ z
A. - .

.
.
I. I I'I .CIDI‘

 
 

 

 

 

TOP VIEW

 

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

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

.......

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

 

 

sample

sapphire washer

I ?
Wm mm mm

 

 

tungsten filament heater

Figure 3.3.1 Schematic of the lower part of the sample mount.

103

threaded ring

f O ' ring holder atmosphere side

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

/ V J

g .-:?:1;3:1:3:3:3:i:4;i.1:3:3.?:i:1:3:3;1:3:3:':322323:?:313:3:3:3:?:3:3:3:3:?:13:1: //

¢ efsiz§é§2§3§5§3§3§2§2§3§5§2§2§;§if2§2§:§2§s§s§2§2§3§3§3§2§5§3§3§2§5§3§3§3§3§3§3§2§s§:3;/

/ ............................................

a + ”M //

UHV‘t side
retaining ring differential
pumpins port
4. 5" Conflat flange CsI window

Figure 3.3.2 Differentially pumped CsI window. Vacuum seals are provided
by three viton O-rings. The CsI window is pressed on only by the O-rings

Nomalized flux

104

 

0.5 -

0.3 -

0.2 r

 

 

 

 

0.0 . 1 1 1
—1.5 —0.5 0.5

Distance (cm)

Figure 3.3.3 The calculated uniformity of the gas doser.

1.5

105

non—caches? on..— .«e deacon each. in?!» aid 953m

 

\\

omen deuce
ages" emu—“Bonu—

=enn 8389333 \

 

\

 

 

henchmen \

 

 

 

303a defiance."
men 5233 .8an

ope—Q 93 393.98
3.3a omen seven—38% «aid

33.. 38 388

use couch—nu Boo:

 

 

......................................
...u........na.”"umunwmnmuummuw.mn.unmwuumumenswewumwww.n.UEwemomma?“”wan“.u “nun“.

 

 

i... M

33 5.0

 

 

Bebe «fleas—u

106

spectrometer base plate

  

cryostat top plate
angled bracket

1/4" pipe thread
quick disconnect

clampscrewtipped
withballbearing

brazed
radiation shield

spectrometershell

Figure3.4.2Diagramoftheclampmechsnism. Threescrewsareusedto
clamp the spectrometer base plate to enhance the mechanical stability.

107

cm

LeboEobocmm on: com mo>czo mczooo med madman

$.5on one;

ON

0“

 

 

. .

 

mocha mLmEmo
EoEm :oScLUeL
mama omen,
Eon? imm
mcficcm

cocci: wcwEoH

 

 

on:

com

()1) ainqeiedmei

108

coax—6:89 2: no 26?. .828 v.96 2:3..—

 

 

 

 

 

 

 

 

 

 

(3

«E —‘B.
Sui-he‘s. hurt; I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

109

 

spectrometer base plate .

Figure 3.4.5 Grating table assembly with Inductosyn rotary position
transducer. '

110

 

 

Figure 3.4.6 Schematic of the grating drive

111

LN2 dewar

 
 
 
 
 
 
 
 
  
     

edge-welded bellows
Quick clamp
copper block

ellipsoidal mirror
mirror mount

adjusting
knob

    

screw tipped with
spring loaded ball
bearing

vacuum can

radiation shield

Figure 3.5.1 Schematic of the liquid-nitrogen-cooled transfer optics.

112

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

 

 

 

 

RELAY afsmscr
# 100

I :~J\/W\/\/\——+15

E g s

i I? : I .

“WW—\i 3 1'1.—
_/._/WW_.-E e INPUTTEST
- -.-._R.f-__.-._. "

 

 

 

 

 

 

 

 

 

 

 

 

DETECTOR

 

 

 

 

 

 

-c-u--------------- ..........................................................................................

Figure 3.6.1 The circuit diagram of the Si:B detector amplifier.

 

 

89593 593

£3 3888 as

.838. genes: as. to sauce seem Se team

 

 

m3.

 

 

anon

3:888
.838 3.33-

 

 

 

 

 

113

 

 

 

 

 

.8938 8.3.83
moan an

 

 

8:882

 

 

 

gov

 

 

 

e3. .335

 

 

 

couscous

hoe—econ

 

 

 

 

 

 

 

.80st
£338
.888 mafia—3n

 

 

 

83089539

 

 

 

 

 

 

 

4. System Performance

We have built our instrument specifically for measuring low frequency
surface vibrational modes at submonolayer coverage, which generally have
much weaker signals, often by orders of magnitude, than high frequency
modes. The central issue is therefore the capability to detect a small
signal on a large background, which requires a sensitivity better than
~1 x 10" in the frequency region below 1000 cm“. In fact, better
sensitivity (~1 x 10-5) is achieved with FTIRs in the high frequency region
(~2000cmf1), which is possible because the detectors are not sensitive to
low frequency radiation, so they are not subject to background photon
noise. This chapter presents a quantitative analysis of the performance

and limitations of our system, along with comparisons with other systems.

4.1 Calibration of spectral frequency

Accurate frequency measurements are important in IRAS to locate the
absorption band accurately. The calibration of frequency can be performed
usually by two methods; well-known absorption spectra are measured and
compared with their standard spectra, or spectra of different orders for
a known monochromatic source are measured and compared with calculated
ones. We have used the latter method to calibrate the spectral frequency.

He-Ne laser spectra of several orders are measured and used for the
calibration of the spectral frequency. We place the laser in front of the
window in the source cryostat and adjust the mirrors and sample as we do
for the system alignment. An empty slit is selected at the filter wheel
to allow the laser beam to pass through the system. As the grating is

rotated from 0° to high angles, sharp laser spectra of many orders are

114

115

measured at the detector.

The diffracted angle of the He-Ne laser line of n-th order is
expressed as the grating equation:

nA - 2dsin0ncosd, (4.1.1)
where A is the wavelength of the laser beam; 6328 A, n the order of
diffraction, d the groove spacing of the grating, 0n the grating angle at
the n-th order peak, and ¢ the half angle between the incident and
diffracted beams of light at the grating, as shown in Figure 4.1.1. We
carefully measure laser spectra of several orders along with on, and plot
nA versus 2dsin0m to get cosd from the slope, using least squares fitting,
as shown in Figure 4.1.2. In this way, we have determined cos¢ both for
the cold and the room temperature spectrometer; cos¢ is 0.982 1 0.003 for
the cold spectrometer and 0.986 :t 0.003 for the room temperature
spectrometer respectively. The expected value of cos¢ is 0.9869 from the
design value of ¢ - 9.302° at room temperature. The uncertainties are
estimated from the root sum square of the products of the standard
deviation of each data point multiplied by the effect which the data point
{2dsin0n} has on the determination of the parameter cosd, assuming that the
uncertainties are instrumental and that the standard deviations for the
data points are equal [1] . The apparent change of cos¢ with temperature
is mainly due to the change of the groove spacing d of the grating. The
change of cos¢ due to the change in d is estimated to be 3.85 x 10'3 as the
temperature changes from room temperature to 80 K, using the thermal
contraction AL/L - 3.9 x 10’3 for the aluminum substrate. This agrees well

with our measured change of 4 x 10‘3.

116

4.2 Spectrometer resolution

As discussed in 2.4, the angular resolution d0 of the monochromator
is determined by the slit width w and the effective focal length f of the
camera mirror:

d0 - w/(2f), (4.2.1)
where a factor of 1/2 comes in because the angle between the incident and
diffracted beams changes by an amount 2d0 as the grating angle 0 is
changed by an amount d0, for a given wavelength. It is assumed in the
equation (4.2.1) that the number of grooves illuminated is much larger
than the resolving power, and that aberrations are small compared to the
slit width w. The first condition is easily satisfied in our system - the
number of grooves illuminated is always >9000, compared to a typical
resolving power ~400. The second condition was verified by a ray-tracing
analysis, by visual inspection of the image, and by direct resolution
measurements.

The angular resolution of the spectrometer was determined for each
'slit width by measuring the full width at half maximum for the zeroth
order peak, using a He-Ne laser as the light source. The laser beam was
expanded to fill the optics. The zeroth order peaks for different slits
are shown in Figure 4.2.1 and the angular resolution is plotted as a
function of slit width in Figure 4.2.2, along with the calculated values
using the equation (4.2.1). The measured resolution behaves as if the
effective focal length were 74 cm instead of 65 cm. The reasons for the
disagreement are not well understood, but possibly because the slit was
not filled completely with the laser beam, or the effective widths of the

slits are reduced by ~ 10 1 due to black paint coating.

117
The frequency resolution.af the monochromator du can be obtained from

the grating equation, the focal length, and the slit width. Let us start
with the grating equation, for a fixed angle 2d between the incident and
diffracted beams:

nA - 2dsin0cos¢ , (4.2.2)
where A is the wavelength of the light, n the order of diffraction, d the
groove spacing of the grating, 0 the angle of the grating, and d the half
angle between the incident and diffracted beams at the grating (see Figure
4.1.1). Differentiating with respect to 0, we get an equation:

n dA - 2dcos€cos¢ d0 . . (4.2.3)
Dividing equation (4.2.2) by equation (4.2.3) gives:

dA du d0

 

 

A v tanfi

 

(4.2.4)

Assuming that we use a grating of 300 grooves/mm (60 grooves/mm) the
grating angle 0 is 48.S9° (36.87°) for u - 2000 cm‘1 (500 cm“). With a
2 mm wide slit and thus the measured angular resolution 0.075° (see Figure
4.2.2), the frequency resolution.du and.the resolving power R (- u/du) are

2.3 cm'1 and 870 at 2000 cm'1 (du - 0.9 cm’1 and R - 570 at 500 cm’l).

4.3 Detector calibration

The detector receives a beam of infrared radiation and converts it
into an electrical signal. As mentioned above, we use an extrinsic Si:B
photoconductive detector. Photons incident on the detector excite
electrons from the valence band to an impurity band, which is formed by
the boron impurities, leaving holes in the valence band so that the
electrical resistance changes as the number of incident photons changes.

Due to this quantum nature, photoconductive detectors are sensitive and

118
respond faster than thermal detectors, but they have a well-defined
spectral limit beyond which they will not respond. The long-wavelength
cutoff is determined by the semiconductor band gap for the intrinsic case
or by the depth of the impurity band for the extrinsic case.

The responsivity of a detector is a measure of its sensitivity; the
responsivity of a photoconductive detector can be expressed as the ratio
of the detector photocurrent to the incident radiation power wattage or
the number of incident photons per second. Since the responsivity is
different from one detector to another even for the same type of detector,
it is important to calibrate the detector in order to understand the
sensitivity of a system and the nature of noises.

The calibration setup for the Si:B detector is shown in Figure 4.3.1.
The blackbody source is home-made of a copper cone (1.5" diameter, 3"
long) painted black with a high temperature black paint, and fitted into
a ceramic cone surrounded by a Nichrome heater coil. The heater is
plugged in a Variac and the temperature is measured with a K-type
thermocouple attached to the copper cone. The source aperture is an
aluminum sheet with a 0.25" diameter hole (or 0.125" diameter hole). A
rotating chopper is placed behind the source aperture to chop the
radiation. The output signal voltage is measured with a PAR S209 lock-in
amplifier using a reference signal from the chopper controller.

The voltage output (peak-to-peak) at the amplifier of the detector
can be expressed as:

vout ' Mgnfpchop . (4-3-1)
where 3 is the current responsivity of the detector (here the detector

includes the Winston cone and cavity), As the gain of the 2nd stage

119
amplifier (see section 3.6), Rt the feedback resistance and Pchop the
chopped power from the radiation source. Here the detector photocurrent
IPflmp due to the radiation power and bias voltage flows only through the
feedback load resistance and is measured as a voltage me at the output
of the amplifier circuit.

In extrinsic photoconductive detectors, a fraction of the incident
photons produces either conduction electrons or holes, which generate a
photocurrent under the influence of an external bias voltage. The
conduction electrons and holes are called charge carriers or carriers and
the number of carriers generated.per incident photon is called the quantum
efficiency 0. All the carriers generated by incident photons do not
necessarily reach the electrode. Some carriers may decay by recombination
before reaching the electrode. The ratio of the number of carriers
reaching the electrode to those generated by incident photons is called

the photoconductive gain G and can be expressed by [2]:
G - ___ , (4.3.2)

where 1 is the carrier lifetime and t1. the transit time of a carrier from
one electrode to the other one. The photocurrent is therefore given by
the equation:

I - Gnflchope , (4.3.3)
where G is the photoconductive gain, 0 the quantum efficiency (0 includes
the transmittance of the Winston cone and cavity in our analysis), Nam?
the chopped photon rate from the radiation source, and e is the electronic
charge. The photo current flows through the feedback load resistor. The

output voltage at the amplifier of the detector is therefore given by:

120

v0“, - AstG’nNchope . (4.3.4)

While As and R, are known from the amplifier circuit of the detector,
the chopped power and photon rate are calculated from the source
temperature and the geometry of the calibration setup. Since the
radiation source is modulated by a room temperature chopper and therefore
the detector sees alternately the radiation source and the room
temperature chopper blade, the chopped power (photon rate) is the
difference of the power (photon rate) from the blackbody radiation source
and that from the chopper blade. The power emitted by a blackbody into an
effective solid angle 0,“, in both polarizations, at frequencies between
ml and 4:2 is given by the equation (see section 2.5):

2“ (02 U3
P(T) - an,“ —— ___ dw
(2K)3c2 wl e “V“ - 1

- A0.“ dx , (4.3.5)

 

2h (kT)‘ x2 x3
(2103c2 ft‘ Lg e‘ - 1
where A and T are the area and temperature of the blackbody, and
x - fiw/kT. The corresponding photon rate emitted from a blackbody is
calculated by the equation:

2 (kT)3 [x2 x2

(2«)3cz #3

dx . (4.3.6)

 

N(T) "' An"! x 1
x1 3 ‘

The throughput is calculated from the geometry of the calibration setup
(see Figure 4.3.1):
A1A2
A0.“-————, (4.3.7)
d2
where A1 and A2 are the detector and source aperture areas respectively,

and d is the distance between the detector and source apertures. Once Vent

121
is measured at the detector output and Pchop is calculated, the current
responsivity of the detector is computed from the equation (4.3.1). The
quantity Gr; can be obtained from the equation (4.3.4) if Vout is measured
at the detector and Nchop is calculated using the equation (4.3.6).

The responsivity a of our detector has been measured for three
different source temperatures and the results are shown in Table 4.3.1,
along with the quantity Cry. The measured responsivity 3 is 0.14 - 0.15
(A/W) and the measured value of Ga is 0.92 x 10'2 - 1.01 x 10'2. The
transmittance of the filter and window is incorporated in calculating the
photon rate and radiation power, while the efficiency of the detector
optics, Winston cone and cavity, is included in G!) and a as mentioned
above. Some parameters of the calibration setup used for measurements are
as follows:

aperture of the detector (8 mm dia.) : A1 - 0.50 cm2

aperture of the source (1/8" dia.) : A2 - 7.9 x 10'2 cm2
distance between the two apertures : d - 10.9 cm

gain of the 2nd stage amplifier : A! - 10

filter frequency : u - 336 ~ 725 cm'1
filter transmittance : t: - 0.72

KRS-S window transmittance : tw - 0.70

chopping frequency : f - 92 Hz

detector bias voltage : Vb“, - 6.0 V

In the rest of this section, we discuss the noise-to-signal ratio of
the calibration setup for the detector, in order to separate the quantum
efficiency 1) and the photoconductive gain G from the measured quantity On.
The discussion about noise here is solely for the calibration setup of the
detector. Noise for the IRAS system will be discussed in sections
4.5 - 4.7.

In an arbitrarily given time interval t, the number "fit of carriers

are generated in the detector by incident photons. Since the photons

122

arrive at the detector in a random fashion, Poisson statistics can be used
to calculate the root-mean-square (RMS) fluctuation in this number of
charges, that is, A(nNt)m - (nNt)1/2. The time averaged RMS noise current
is e(nN/t)“q. Since the postdetection noise bandwidth is given by l/(2t)
for averaging over a time t, the corresponding noise current per unit post
detection bandwidth is:

AIm - Ge(4nN)1/2 (A/ffiz), (4.3.8)
for an extrinsic photoconductive detector, where another factor 2 has come
in owing to the generation-recombination statistics of the carriers in the
detector element (extrinsic semiconductor) [3]. Here, N should include
both the photon rate N, from the source and the photon rate N35 from room
temperature background in our calibration setup.

From the equations (4.3.3) and (4.3.8), the noise-to-signal ratio
(N/S) of the detector is calculated as:

N 2m, + aw)“:

___ - . (4.3.9)
S "IIZN'

 

The equation (4.3.9) can be rewritten for the quantum efficiency:

4m. + N»)
n - . (4.3.10)
Elias/8)2

 

If we measure N/S ratio of the detector, we can calculate the quantum
efficiency from the equation (4.3.10). Noise and signal are measureduwith
a PAR 5209 lock-in amplifier. The background photon rate N36 and the
source photon rate N;.are calculated with the equation (4.3.6), using the
following parameters:
AD.“ - 3.32 x 10" (cmzsterad) for N. calculation;
from the calibration setup

An.“ - 2.04 x 10-2 (cmzsterad) for “so calculation;
from the filter aperture of the detector with f/4.4

123
T3 - 572 K

room temperature - 300 K

filter frequency - 336 ~ 725 cm'1

filter transmittance - 0.72

window transmittance - 0.75 .

The calculated and measured results are as follows:

source photon rate ; N, - 3.13 x 101‘ (photons/sec)

background photon rate ; Nae - 7.78 x 1015 (photons/sec)

measured Gn ; Gn - 0.93 x 10'2

measured noise ; Nm - 3.9 x 10" (VA/il—z')

measured signal ; Sm“ - 2.57 (V)

The measured N/S ratio is 1.52 x 10*’(1//fi2) at the detector output, and
the quantum efficiency n is thus 0.14 from the equation (4.3.10). It is
obvious from the equation (4.3.9) that the N/S ratio will be much better
if we reduce the background photon rate. The quantum efficiency n is in
general a function of frequency and the value obtained above is therefore
a kind of average over the relevant frequencies in the calibration
experiment. The photoconductive gain G is not expected to depend on
frequency since it has to do with the generated carriers in the detector
element, but it does depend on bias voltage and its dependence can be
measured. It needs to be corrected for other values of bias voltage.

It is traditional to define the noise equivalent photon rate NEW as
the photon rate that must be incident on.a detector for the output current
to be equal to the RMS fluctuation per unit bandwidth in output
current [3]. For a photoconductive detector this condition is
Gne(NEN) - Ge(4nN)“Q. So the noise equivalent photon rate is:

NEW - (413/:7)”2 . (4.3.11)
Similarly, a noise equivalent power NEP can be defined as the incident

signal power required to obtain an output signal equal to RMS output

noise. The noise equivalent power is thus expressed as:

124
NEP - hcu(4N/n)l/2 . (4.3.12)
Using these equations the noise equivalent photon rate and power are

computed for our detector. The results are as follows:

NEN - 4.8 x 108 (photons/sec//§E), and
NEP - 5.1 x 10'12 (was), (4.3.13)

where a median frequency v - 530 cmfl'is used for calculating NEP. Note
that these numbers are measured in high background. They will improve in
actual experimental conditions. The estimated NEW and NEP under actual
experimental conditions are 3 x 107 photons/sec/AE and 2 x 10'13 W/fiE
respectively (see section 4.5.2).

The quantity Ga and the quantum efficiency 0 are needed to estimate
the optical efficiency of the system and noise (see sections 4.4 - 4.6).
As mentioned above the quantity Gn is in general a function of frequency
and bias voltage. From the relations Vb“, - - Rf/RD Vu- (see equation
4.5.1.10 and Figure 3.6.1) and Vu- - RINemn) (equation 3.6.2), it is
expected that the quantity 6» is linearly proportional to bias voltage
Vb,“ as; Gr) - - [RD/(thfieflvbiu, provided that RD and R, do not depend on
in“... Though the actual dependence is not linear (R0 is not constant) as
measured by VI, versus Vbu“ the deviation is not significant for Vb“, -
2 - 6 V, so we have estimated Gn with a linear approximation for biases
other than that measured. As for the frequency dependence, the measured
60 is a kind of averaged value over the spectral range in the calibration

measurement, and is used for other frequencies with no correction.

4.4 Optical efficiency of the system

The optical efficiency fun of the system is defined as the ratio

of the number of photons per second arriving at the detector to the photon

125
rate from the radiation source accepted by the throughput of the system.
The efficiency can be estimated from the expected losses of the optical
components in the system. It can be expressed as:
e... - R...” R. T. If T. e. . (4.4.1)
where Ru is the reflectivity of a gold coated mirror, R, - 0.98; R, the
reflectivity of the sample, R, - 0.60 (calculated value, see Figure 2.3.2);

T the transmission coefficient of the polarizer, T - 0.7; T,, the

P P

transmission coefficient of a CsI window, T; - 0.8; T, the transmission
coefficient of the filter, T, - 0.75; and ‘3 the grating efficiency,

5 - 0.8; parameters are obtained from the specification sheets Or

a
catalogs. Note that we consider only p-polarization. Similarly, we can
define the optical efficiency of the spectrometer and transfer optics as
the ratio of photon rate at the detector to that entering the
spectrometer:

e, - 12,5 1",, I, e, . (4.4.2)
Substituting known parameter values into the equation.(4.4.l) and.(4.4.2),
the estimated optical efficiency Gun of the system is 12.1 I, and the
spectrometer efficiency 6, is 38.0 I for one polarization.

0n the other hand, we can determine the measured efficiency of the
system from the output signal voltage and detector parameters. From the
equation (4.3.4), the measured signal photon rate Nu“ at the detector is
expressed as:

25%.“)...

Nu... - . (4. 4. 3)
GfleRIA‘

 

where (me)uu is root-mean-square voltage output, G the photoconductive

gain, 0 the quantum efficiency, R, the feedback resistance andAt the gain

126

of the second stage amplifier. At 1000 cm'l, the measured signal voltage
Wont)“, is 233 mV with 8.5 pm filter and the following parameters:

Rf - 6 x 107 0,

V1318! - 2.5 V,

G17 - 4.2 x 10-3,

As - 10,
where Cu is determined by a linear extrapolation from the measured value
at Vb“, - 6.0 V. The measured photon rate Nu,“ is therefore 1.6 x 1012
photons per second.

Now the photon rate accepted by the throughput of the system can be
calculated using the equation (4.3.6), with following parameters:

Tum,“ - 1200 K,

An.“ - 3.2 x 10'3 cmzsterad,

v - 1000 cm'l,

Au - 1.5 cm'l,

e - 0.8 ; emissivity of the globar.
The calculated photon rate No.1 is 4.5 x 1013 photons per second for
p-polarization. Taking the ratio of the measured photon rate to the
calculated one, we find the measured efficiency of the system:
(steam. - 3.7 1. Compared with the expected efficiency of 12.1 1,
' approximately 70 1 of the radiation is lost due to alignment error, lower
than expected component efficiencies, or other reasons we do not

understand. Possibly, the quantity 0» is a lot smaller at 1000 cm'1 than

at the lower frequencies (336 - 725 cm'l) where Gr) has been measured.

4.5 Calculated noise

The required fractional noise level is S l x 10" 1M1; at frequencies
less than 1000 cm'1 to study low frequency surface vibrational modes at
submonolayer coverage. Possible sources of noise are the detector and its

amplifier circuit, photon fluctuations from source and background,

127
instabilities of the source intensity and the chopper amplitude, sample
movements, nonreproducibility of grating angles, movement of inner optical
parts of the spectrometer relative to the shell, and noise from electrical
measuring instruments.

The expected noise of the system is estimated from two major sources;
the noise from the detector circuit and photon noise from the system. In
the following section, the calculated noise is compared.with the measured
noise of the system. We have found that the dominant source of noise is
scan-to-scan nonreproducibility.

4.5.1 Noise of the detector circuit

A noise equivalent circuit of the detector and preamplifier is shown
in Figure 4.5.1 [4]. The various noise sources represented in this model
are:

eG - voltage noise of JFET input stage,

is - current noise of JFET input stage,

if - Johnson current noise of the feedback load resistor,

in - Johnson current noise of the detector.

R: is the feedback load resistance with its capacitance C: and R0 the
detector resistance with capacitance CD. Note that the JFET, the feedback
load resistor and the detector are at liquid helium temperature, as shown
in Figure 3.6.1. Let the complex impedances of the feedback load resistor

and the detector be 2, and 20 respectively, then 2, and Zn are:

 

 

R:
z, - , (4.5.1.1)
1 + inoctllf
RD
zD - . (4.5.1.2)

1+imCDRD

128

CD consists of the dielectric capacitance Cd“ of the detector and JFET
gate capacitance (33' Cd“ is calculated with the formula: Cd... - eA/d,
where e is the dielectric constant of the detector, A the cross-sectional
area, and d the length of the detector. Using 6 - 11.7, A - 4.0 x 10'3 cm2
and d - 0.1 cm, Cm“ is 0.52 pF. The JFET gate capacitance C8 is 4 pF,
from the specification sheet. CD is therefore 4.52 pF. C: can be
estimated from measurements of the voltage gain of the circuit with an AC
signal injected at the bias point, see Figure 3.6.1, as a function of
frequency, using gain - - Zf/ZD. C, is approximately 0.5 pF. The roll-off
frequencies (21rRC)"1 for R, and the detector itself are 5.3 K32 and 2.5
KHz, respectively. At a chopping frequency of 800 Hz, they are basically
resistive, though the detector combined with the gate capacitance of JFET
will behave capacitively. Under operating conditions the detector is
under constant voltage bias, so its capacitance does not affect circuit
response.

We assume that the noise sources are independent, so that we can
analyze them separately and find out their contributions to noise at the
input test of the circuit. The amplifier gain a is assumed to be
sufficiently large throughout the noise analysis that the amplifier input
is a true virtual ground.

The noise at the input test point due to the voltage noise of the
JFET input stage is: I

20 + Z,
Wyn)“- - —— es . (4.5.1.3)
Zn
The noise current of the JFET input stage flows through the load impedance

to appear at the input test so that the noise is:

129

(vgcn)II ' 1922- (4.5.1.4)
Combining these two noise sources, the noise due to the JFET input stage
is expressed as:

(vfin)Ir - ((1 + z,._./z,,)2e,,2 + (15292)“? (4.5.1.5)

The Johnson current noises of the detector and the feedback load
resistance are expressed as:

1,, - (4kT/RD)1/2 , (4.5.1.6)
and

if - (4kT/R£)1/2, (4.5.1.7)
respectively. Similarly to the current noise of JFET input stage, these
current noises appear at the input test as 102, for the detector resistance
and 1,2, for the feedback load resistance, respectively. The noise at the
input test due to these two sources is expressed as:

(an)Ir - (4kT(1/RD + 1/R,))1/zz, . (4.5.1.8)
The total noise (Vnnhr due to the detector circuit is a quadratic addition
of these noises:

Wan)“ - ((V‘“)r,2 + (VJ“)I,2)1/z. (4.5.1.9)

Substituting some typical numbers, we can estimate the predicted
noises from the various noise sources. The ratio of the detector
resistance to the feedback load resistance under operating conditions is
measured from the dc input test and dc bias voltage of the detector
circuit:

Rt/Rn " ' Vines/VI! - (4-5-1-10)
Using the bias voltage and measured input test voltage when the
spectrometer, transfer optics and the detector are cold, Rt/RD - 0.30.

Since Rf - 6.0 x 107 0, RD is 2.0 x 108 0. The upper limit of the voltage

130

and current noises of the JFET input stage are 2.0 x 10'8‘V//fiz and 6.4 x
104‘ A/JEE respectively, from the specification sheet. From equations
(4.5.1.1) and (4.5.1.2), zq/ZD z 0.31 + 0.091 at 800 Hz. The estimated
noise from the JFET input stage, which is at 4.2 K, at the input test is
thus (using the equation 4.5.1.5):

(v83)IT - 2.6 x 10'8 V//H—z. (4.5.1.11)
From the equation (4.5.1.8), the Johnson noise from the detector and
feedback load resistance, which are both at 4.2 K, at the input test is:

(thr - 1.34 x 10'7 V//H_z', (4.5.1.12)
where the Johnson noise due to the feedback resistance is actually
dominant. Adding these two noises quadratically, the total noise due to
the detector circuit is:

(an)IT - 1.37 x 10'7 V/lfi. (4.5.1.13)

From this estimation, we expect the Johnson noise to dominate the
detector circuit noise even though the feedback load resistor is cooled to
liquid helium temperature. In fact, room temperature feedback resistors
were used in the original detector amplifier circuit, which has been
modified to include a cold preamplifier and cold feedback resistors to
reduce the Johnson noise. Detector circuit noise is now'much smaller than
the background photon noise (see below).

4.5.2 Photon noise

The expected photon noise can be calculated for background limited
operation of the detector. The noise current in the detector due to the
fluctuations of incident photons is given by the equation (4.3.8). The
noise current flows through the feedback resistor and appears as a noise

voltage at the input test. The noise voltage is therefore expressed as:

131

(Vp“)n- - 2Ge(nN)1/ZZ£ . (4.5.2.1)

Substituting some typical numbers, we estimate the noise voltage at
the input test to be:

(Vp“)n- - 1.4 x 10" V/JEE ; from 100 K background, and

(Vp“)n- - 2.7 x 10'7 V/fi ; from the source photon rate.
The background photon rate N36 and source photon rate NS used to estimate
the photon noise are:

N36 - 4.2 x 1013 photons/sec

NS - 1.6 x 1012 photons/sec , (4.5.2.2)
which are calculated using the typical parameters:

C - 3.0 x 10'2

n - 0.138

thRt-6x1070,

A0,“ - 8.5 x 10‘3 cmzsterad

r - 1.0 for the background calculation,
T36 - 100 K

V - 350 - an

AD.“ - 3.2 x 10'3 cmzsterad

r - 0.03 for the source photon rate
Tum,“ - 1300 K calculation.

Au - 2 cm'1
11 - 1000 cm'1

Background photon noise is much bigger than source photon noise, and
is still dominant, even though the spectrometer and relevant optics are
cooled to 100 K“ This is mainly due to the fact that the background noise
is integrated over frequency with efficiency 1 - 1.0, while the signal is
monochromatic, with efficiency'r - 0.03 that comes from the product of the
total efficiency of the system fun - 0.037 and the globar emissivity e -
0.8 (see section 4.4). Note that the blackbody radiation of 100 K has a
peak at about 250 cm and that the detector has a low frequency cut-off,

350 cm“, (see Figure 2.6.1). All the calculated noises are summarized in

132

Table 4.5.1, along with the measured noises.

4.6 Measured noise

4.6.1 Noise from the detector circuit

As discussed in section 4.5, the noise at input test consists of
noise from the detector circuit and that from photon fluctuations if other
noise sources are not significant. If we block the entrance aperture of
the detector by a metal plate that is cooled to liquid helium temperature
and make photon noise negligible, the noise due to the detector circuit
can be estimated from the measured noise at the input test. The noise is
measured either by an HP spectrum analyzer, or by a computer with a
spectrum analyzer program, a preamplifier and a band pass filter, or by
the lock-in amplifier. They give consistent results. The measured noise
at the input test is:

(Vr“)u - 2.8 x 10’7 WIFE, (4.6.1.1)
with Vb“. - 3.0 V and VI, - 44 mV. Since the photon noise of 4.2 K
background is negligible, the measured noise should mainly come from the
detector circuit. This noise is larger than the estimated one, 1.4 x 10'7
V/Jfiz, by a factor of two, possibly because the detector is under voltage
bias, 3.0 V, giving rise to excess noise.

4.6.2 Photon noise

It is not easy to isolate photon noise from noise of the detector
circuit in measurements, but we can measure the photon noise just by
measuring the total noise at input test voltage since background photon
noise is expected to be dominant. The photon noise can.be estimated also

from the measured input test voltage, since that voltage is proportional

133
to the total photon rate (see equations 3.6.1 and 3.6.2). Using equation
(4.5.2.1) and the relation Vn- - ZteGnNBG, the photon noise is expressed as:

(Vp“)n- - 2(ceztvu)1/2. (4.6.2.1)
When the spectrometer and the transfer optics are cooled by liquid
nitrogen, the measured input test voltage Vu- is -0.76 V at the bias
voltage Vb,“ - 2.5 V and frequency v - 1000 cm”. The estimated photon
noise from the equation 4.6.2.1 and measured Vn- is therefore 9.34 x 10'7
V//H_z. The calculated photon noise is 1.39 x 10" V/flE, which is slightly
higher than the estimated noise from the measured input test voltage,
indicating that the‘ actual background level is slightly lower than
expected, but lower by about 20 1 than the noise measured with a spectrum
analyzer, 1.8 x 10'6 V/flE.

We have measured the photon noise from the room temperature
background when the spectrometer and transfer optics are warm. The
measured photon noise is 1.08 x 10‘5 V/JIE. The estimated noise from the
input test voltage is 9.9 x 10'6 V/fil-Z and the calculated photon noise from
room temperature background is 1.0 x 10'5 V//ll_z. The agreement is very
good. Maybe, the agreement for the room temperature results is better,
because the G!) values are measured for the characteristic frequencies of
300 - 500 K blackbodies.

For the purpose of comparison, we have also measured the photon noise
from the gate valve between the transfer optics and detector. The input
test voltage is -9.75 V at a bias of 2.42 V. So the estimated photon
noise from the input test voltage is 3.4 x 10" V/JFIE. The measured noise
at the input test is 2.8 x 10'6 V/fi, which is in good agreement with the

estimate. But this is smaller than that of room temperature background by

134
a factor of three, suggesting that the 4.2 K ambient radiation of the
detector cryostat is reflected back to the detector from the gate valve.
With the cold spectrometer and the room-temperature transfer optics,
the measured noise is 2.1 x 104’V//fiz, which is larger by 50 1 than the
calculated 100 K background noise. The extra noise most likely comes from
stray radiation. This is the reason why we have modified the transfer
optics to be cooled by liquid nitrogen, and installed liquid-nitrogen
cooled baffles in the optical path both between the spectrometer and the
transfer optics, and between the transfer optics and the detector. With
this modification, however, the measured.noise decreases only by 10 2, not
so much as expected, by a factor 2, from the measured input test voltage.
This discrepancy may come again since we use 00 values measured for 300 -

500 K blackbodies, as mentioned above.

4.6.3 Behavior of system noise

As we have discussed in section 4.5, we expect the noise at the
detector output to be background photon noise limited. From the equations
(4.5.2.1) and (4.6.2.1), the noise intensity (Vp“)n- should have'a square-
root dependence on the input test voltage Vnror the photon rate N at the
detector. A log-log plot of (Vp°)n- versus V“- or N should show a straight
line with a slope of 1/2, if the detector is operated under the photon
noise limited condition. We have measured the noise at the input test as
a function of input test voltage, which is shown in Figure 4.6.1, with a
bias voltage of 10 V. The photon flux, and therefore the input test
voltage, are varied by rotating the liquid-helium-cooled filter wheel in
the detector. The noise shows a photon noise limited behavior, except

near V11- - 10 V.

135

1 we measure a signal level of ~ 270 mV

At a frequency of- 900 cm'
with a source temperature of 1300 K. Using the signal value and the noise '
numbers we have discussed so far, we can estimate the sensitivity for each
case. First, the noise-to-signal (N/S) ratio would be 1.0 x 10‘5 l//H_z if
we were limited by source photon noise. This is the best we could
possibly do with the present source, detector, and the optical throughput
and efficiency of the system. We have not achieved this level of
sensitivity. Second, the background-photon-noise limited N/S ratio is
6 x 10'5 l/flE. This is the best our current instrument could do with
perfect reproducibility, and is what we currently seek to achieve. This
level of N/S ratio is routinely achieved in time scans at a fixed grating
angle.

A noise spectrum measured at the input test is shown in Figure 4.6.2.
The noise spectrum has a l/f noise feature at low frequencies, and a long
tail extending to 800 Hz. We have chosen 800 Hz as the modulation
frequency to avoid this noise. Tuning fork choppers have a tradeoff
between high frequency and high amplitude. It is related to the maximum
stress that can be applied in the vibrating arms. If we tried to go to
higher frequency we would not have enough amplitude. There are no sharp
features due to vibrations in the noise spectrum, which means that the
system is very stable and insensitive to the mechanical vibrations.

So far we have limited our discussion to the noise with no grating
movements involved. In real infrared spectroscopic measurements, we need
to compare a spectrum taken with adsorbates on the surface with one of a

clean surface. In order to measure the noise that matters in the real

surface spectroscopic measurements, we need to take and compare two

136
sequential spectra with clean sample surface. In this way we have reached
a noise-to-signal (N/S) ratio of 2 x 10" 1/fil—z' near 850 cm'1 and the extra
noise comes mostly from non-reproducibility of one scan to the next. In
Figure 4.6.3, two noises are shown; one from a time scan and the other
from a real frequency scan. In the real scan noise, the spikes from the
non-reproducibility dominate the noise. This is discussed in the next

section.
4.7 Improving noise-to-signal ratio

In addition to the noise sources we discussed in section (4.5) and
(4.6), namely detector noise and background noise, there are other
possible noise sources in the system. In this section I describe those
noise sources and other efforts that we have made to improve the N/S ratio
of the system.

1) Instability of the radiation source can be caused by instability
of the power supply, or mechanical movements of internal optical parts
relative to the external optics.

2) Instability of the chopper or chopper controller can be a problem.
While phase jitter is the only problem of the conventional rotating
chopper, amplitude stability is crucial for a vibrating chopper. The
fractional noise of the chopper can be measured by a lock-in amplifier
feeding the controller signal as an input and reference signal. We have
modified the circuit of the chopper controller to improve the fractional
noise from 2.0 x 10" l/JHE to 2.0 x 10‘5 1//sz:. At this level it is well
below background photon noise.

3) Though mechanical vibration of the sample does not contribute

directly to the noise because of chopping, it can affect the

137

reproducibility. The sample is required to travel more than 8 inches in
the sample chamber between the IRAS level and the surface analysis level.
The total length of the sample mount is 30 inch long which is made of 20
mil thick stainless steel tube. The manipulator is a cantilever model
which has three lead screws for better stability.

We know that l) - 3) do not dominate because they would show up in
time scans. The dominant noise comes only when we scan the spectrometer.

4) Non-reproducibility of the optical efficiency of the spectrometer
can be a noise source. The optical efficiency of the spectrometer can
change from scan to scan due to mechanical motion of the internal
spectrometer optics relative to the external cryostat, which may be caused
by the grating drive force or by the vibration of vacuum pumps. As
discussed in section 3.4, three clamping screws have been introduced
through the walls of spectrometer Chamber'to keep the spectrometer optical
table from moving around. Also the grating can tilt. A tilt of a few x
10'5 radians can shift the image at the exit slit vertically by ~5 x.10“
of its height. We have modified the grating mount and drive to include a
pair of precision bearings at the worm shaft of the grating table and a
pair of angular contact bearings at the worm driving shaft, to achieve
more mechanical rigidity.

5) The dominant noise source turns out to be the non-reproducibility
of grating angle and/or efficiency. This noise source appears only when
two spectra are measured and compared each other. In each spectral scan
the signal intensity is measured as a function of grating angle over a
range of angles. If the angles in one scan are not exactly reproduced in

the next scan, the result appears as noise in the comparison of two

138
spectra.

Suppose that instead of returning to a particular angle 0 on the
second scan, the grating returns to a slightly different angle 0 + A0.
Since the light intensity I reaching the detector depends on the angle,
the intensity measured in the second scan is different from that in the
first by AI - (dI/d0)A0. Since A0 varies randomly from point to point in
the spectrum, these angular errors appear as signal fluctuations when the
two spectra are compared (see section 3.4.3). It is expected that the
noise is greater as the slope of the intensity is larger, which is exactly
observed. We can estimate the non-reproducibility of the grating angle
from the noise.

As discussed in 3.4.3, the original grating drive was a worm-to-worm
gear combination with a 360-to-l gear ratio and a bellows coupling to a
stepper motor. The angular non-reproducibility was approximately 2 arc
sec with the N/S ratio of 4 x 10" l//Hz around 850 cm“. This noise
cannot be eliminated by any modulation, and was not improved.by averaging
scans since it was not random. We made many efforts to enhance the
mechanical rigidity. These efforts did help, but were not enough. We
decided to install a high-resolution angular transducer and a feedback
control system (see section 3.4.3) and the N/S ratio has been improved to
2 x 10" l/flE around 850 cm'l. But the nonreproducibility of the grating
angle and/or efficiency from a scan to the next'is still the dominant
noise source. We have installed a half height slit in the entrance slit
wheel to investigate the problem.

Since the system noise is limited by the scan-to-scan

nonreproducibility, the system noise is linearly proportional to the

139
signal level. That implies the following:
a) increasing the signal level does not help - brightness of the source
is not a limitation,
b) further reducing background photon noise does not help - it is not
a limitation, either,
c) FTIR would not necessarily have the multiplex advantage [3].
We are exploring several possible avenues to improved sensitivity:
a) further improvements in the rigidity of the grating mount and drive,
b) double modulation, combining intensity modulation with wavelength
modulation. Higher signal intensity may be needed to take advantage
of this technique,
c) use of a detector array to obtain true multiplexing and avoid the

necessity of scanning the grating.

4.8 Comparison of our system with others

When we started our project in 1988, no IRAS systems were available
for the spectral range below 800 cmPK. Since then, however, several groups
have developed IRAS systems which extend the spectral range into the far-
infrared region (as low as a few hundred cm'l), maintaining a reasonable
sensitivity. Table 4.8.1 summarizes some features of those systems.
While two groups, the Brookhaven National Laboratory (BNL) group and the
Bermudez group, are using brighter sources to improve the N/S ratio,
others are using conventional blackbody sources, such as globar sources,
and room temperature spectrometers, mostly FTIR.

Using a vacuum FTIR and synchrotron radiation as a radiation source,
the BNL group has, on rare occasions, been able to achieve a N/S ratio of

6 x 10'5 in a frequency region around 100 - 800 cm'1 [5]. Synchrotron

140
radiation instabilities dominate the system noise and a more typical N/S
ratio attainable is ~2 x 10". They have measured the metal-carbon
stretching band for CO on Cu(111) and Cu(100).

Bermudez et a1. [6] have developed a diode laser IRAS system using a
ZnSe photoelastic modulator (PEM). Two diode lasers are used to cover the
spectral range from 600 cm'1 to 1020 cm'l. Spectra are measured without a
monochromator by varying the diode temperature. The resolution is limited
by the energy range over which multimode laser emission occurs at a given
temperature and current, and is ~10 cmfi'at best. The best N/S ratio is
~3 x 10“,‘which is seemingly limited by the lasing instability. Later,
they were able to obtain a N/S ratio of ~1 x 10" in a frequency range 600
- 1000 cm'l, using an FTIR spectrometer with a HngTe detector and likely
a conventional globar source [7].

Ryberg [8] has measured the metal-carbon stretching band for CO on
Pt(111), using a vacuum grating spectrometer with wavelength modulation
and a W - filament source. Apparently he has achieved a N/S ratio of l x
IIY‘ or better, and likely background photon fluctuation noise dominates
the system noise. According to our calculation, the source temperature
should be higher than 1500 K in order to achieve that N/S ratio.

The Trenary [9] and Bradshaw [10] groups have developed very similar
systems and measured the metal-carbon stretching band for CO on Pt(111).
They use vacuum FTIR's, globar sources and similar sample sizes. The N/S
ratios of these systems are likely limited by background photon
fluctuation noise.

The Greenler group [11] has developed an IRAS system using a grating

spectrometer, globar source and thermocouple detectorz They'have measured

141
the Ag-C stretching mode of C0, and the 0-0 and Ag-O stretching bands of
oxygen on a polycrystalline silver foil. Considering that they are using
a ribbon sample which allows them to use a large sample and therefore
large throughput, the fractional noise of the system is not so impressive,
~3 x 10" (estimated from the published data), and is probably limited by
detector noise.

To summarize, two groups, BNL and Bermudez, use brighter infrared
radiation sources to improve N/S ratio to overcome the background photon
fluctuation noise, while four other groups use conventional blackbody
radiation sources and must have developed very stable systems,
mechanically and. optically, so that the noise-to-signal ratios are
approaching the background-photon-fluctuation noise limit. However, the
former two groups have noise sources from the source itself, limiting the
N/S ratio, and the other groups are expected to have intrinsic
difficulties to further improve N/S ratio from the background-photon-
fluctuation noise.

0n the other hand, we have designed and built our system to minimize
the background photon fluctuation noise by cooling the spectrometer and
relevant optics. The present N/S ratio of our system is 2 x 10" l/JFE and
is limited by nonreproducibility of grating movement and/or spectrometer
efficiency. If we improve the reproducibility of the spectrometer, the
N/S ratio can be improved by a factor of 4.

4.9 Summary

We have measured and analyzed the performance of the system. Most of

the characteristics, such as the spectral accuracy, the frequency

resolution and optical efficiency, are satisfactory, but the sensitivity

142

does not completely meet the design target, 51 x 10" 1M? The
sensitivity of the system achieved to date is 2 x 10"ZL//HE at ~850 cm‘l,
which is good enough to measure the external and internal vibrations of
many adsorbates on metal surfaces, but needs to be improved for the study
of other weak external modes such as atomic adsorbates. The sensitivity
is limited by scan-to-scan nonreproducibility, rather than random noise.
If we measure noise at a fixed frequency, therefore at a fixed grating
angle, as a function of time, the fractional noise drops to 6 x. 10'5 l/flE.
This noise comes mostly from 100 K background radiation.

The measured detector noise at input test is 2.8 x 10'7 V/flE, which
is not significant compared with the 100 K background noise, 1.8 x 10"
V//HE, as expected, but is higher than the calculated noise by a factor of
2 probably due to the bias voltage. The source photon noise estimated
from the signal intensity is 2.7 x 10'7 V/flfi, which is not significant
either.

The sensitivity of our system can be improved to a level of 6 x 10'5
l/JHE if we improve the reproducibility of the spectrometer. There is no
far-IRAS system that routinely operates at this level of sensitivity. The
IRAS apparatus at Brookhaven National Laboratory has a sensitivity of
approximately 2 x 10" l/JITE, which is apparently limited by the stability
of the synchrotron radiation source, while others have the room-
temperature-background limitations. Other possibilities for overcoming
non-reproducibility include wavelength modulation or using a linear array

detector.

143

References

1.

10.

ll.

P.R Bevington, Data Reduction and Error Analysis for the Physical
Sciences, (McGraw-Hill Book Company, New York, 1969).

S. M. Sze, Physics of Semiconductor Device, 2nd edition (John Wiley &
Sons, Inc. New York, 1981) pp. 744 - 748.

P. L. Richards and R. G. Tobin, in Vibrational Spectroscopy of
Molecules on Surfaces, edited by J. T. Yates, Jr. and T. E. Madey
(Plenum Publishing Co. New York,l987) pp. 417 - 463.

E. L. Dereniak, R. R. Joyce and R. W. Capps, Rev. Sci. Instrum. 48,
392 (1977).

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

V. M. Bermudez, R. L. Rubinovitz and J. E. Butler, J. Vac. Sci.
Technol. A6, 717 (1989).

V. M. Bermudez and A. S. Class, Langmuir 5, 316 (1989).

R. Ryberg, Phys. Rev. 40, 5849 (1989), Phys. Rev. 40, 8576 (1989),
Phys. Rev. 40, 10273 (1989), Surf. Sci. 114, 627 (1982).

I. J. Malik and.K. Trenary, Surf. Sci. 214, L237 (1989); V. K. Agrawal
and M. Trenary, J. Chem. Phys. 95, 6962 (1991).

D. Hoge, M. Tflshaus, E. Schweizer and A. M. Bradshaw, Chem. Phys.
Lett. 151, 230 (1988)

X.-D. Wang, R. G. Greenler, Surf. Sci. 226, L51 (1990), Phys. Rev. 43,
6808 (1991), X.-D. Wang, W. T. Tysoe, R. G. Greenler and K.

Truszkowska, Surf. Sci. 257, 335 (1991), Surf. Sci. 258, 335 (1991).

144

Table 4.3.1 Detector Calibration
The responsivity of the detector has been measured for three different
temperatures, using the calibration setup shown in Figure 4.3.1.
efficiency of the detector optics is included in the responsivity and the

amplifier gain A, used is 10.

The

 

 

 

 

 

 

 

 

 

 

 

Ts (Avout)pp Rf Pchop N chop 3 3’7

(K) (V) (K0) (#W) X101‘(S'1) (A/V) X104
380 0.67 500 0.88 0.83 0.15 1.01
518 0.37 100 2.67 2.48 0.14 0.93
573 0.47 100 3.45 3.20 0.14 0.92

 

 

145

Table 4.5.1 Summary of calculated and measured noises.
See sections 4.5 and 4.6 for details.

 

 

 

 

 

 

 

Calculated Estimated Measured noise
noise (V//HE) noise from Vfr at input
(V/ffi) (VA/HE)
Detector circuit 1.4 x 10'7 N.A. 2.8 x 10"7
noise
Photon noise: 1.4 x 10" 9.3 x 10'7 1.8 x 10'6
100 K background
Photon noise: 300 1.0 x 10'5 9.9 x 10‘6 1.1 x 10'5
K background
Photon noise: 1300 2.7 x 10'7 ......................
K source
100 K plus 300 K .......... 2 3 x 10'6 2 l x 10"6
BC from transfer
optics

 

 

 

 

 

146

 

 

 

 

 

 

 

 

 

.003 huwmcouca
Hvac comm - omm uouospcooouona 00H000-~zq \.uooam
\7588 L: x N .5 a x 3 m .3 93 -- 55on 838 636385 pm:
7.8 83 - 8... 56:6 Ea: 5E 6.66% E
\7500me .79." x m 0.20 30 GA .0» Home." 00030 nouns—08:00:06 0: woman—yon
Tao ooom - cos .
\Taooome .Toa x H 330 uouodpaooouoza .8.»qu 9830.: .3
Taoooouw Pea x m .020 .8 m6 2% .0080 unnoaw \MHE 850a> 30£mvoum
7.5 83 - 2: 8.5:. m< .62.. .53 63:63
Hmummuohaom omuam was >80 _oH_
\9roH x ml Sow x a0~.H w oaasoooauocu Hugoam ..uoomm wcwueuw Hoacoouu
.780 aces - coo poaooo
xmwc .mwp H0003 Houuwa wca>0a _m.
\RioH x Hz 80 m~.H .ww Houoaoaon um -- Honoaw \MHBh assoe> huocouh
dTao comm - own Aevxmfip uouodpcooouosn. unoaoawm .006 SuwSoH0>m3 _n_
\Rioa x at .mwp 30 H .m.um sonou - 3 \wcHumum azaou> wuonxm
dv.8 ooos - oma nouuwe wcu>0a
\Smoacfivfg x N 363 8.08 \Eou 038:3 7: 9.8»
3.6.05 Tea x w .8 H x 30 N .mw uouoaoaon muwm 80.89.3083 MHE 8508' A.z.m
owcou hocozvoum oocopwocw oouaom soauaascoa
\0Humu m\z ouum mamamm m0 mamas 80000000 uswaa \uouolouuooAM

 

 

 

 

 

 

 

 

.mmwoc vasoumxomn weapon
00 mouuao 00m Houoaouuoomm 00H000-:ow0uuwc-0wscaa 0 magma mg sz .cowuauomne uwuonmnoaum onu weapon ou
assom> 0H 00008090 haumoa .uouoaouuoomm eunumuoaaou-aoou 0:0 moousom HoaHoSu Hoaoauce>800 mafia: one muonuo
magma .xuwmsoucw Hmawam 0:0 0>0HQEH 00 moouaom Houswwun mafia: one .9008& nopdsuon 0:0 naouw qzm .masouw 039

maoumam m5: m0 803.8980 H.m.¢ OHQMB

147

tic axis
/ °p

light in light out

\ grating normal
6 \

   

 

We grating angle
Z crating

Figure 4.1.1 Light at oblique incidence onto a grating, showing the .
definition of the angles ¢ and 0.

148

 

(Hm)

n3.

 

 

 

 

' 0.0 1.0 2.0 3.0 4.0 5.0
2dsinO (le)

Figure 4.1.2 Calibration of spectral frequency for cold spectrometer. The calibration
curve for room temperature spectrometer is not shown since it is indistinguishable. A
nominal value of d is used for 2dsin0.

149

 

 

 

 

 

 

 

 

1]? 0.006 m r l t l l I
'E 02mmfln
5 ......... 0.5mm
. 0.005 F / ....... - 1.0mm _
_o , ------- 1.5mm
S... / ——- 2.0mm
3 /
>. 0.004 — / i -
f. I i
{3 I :'
3 0.003 — I :' —
.5 I :
I :'
0.002 » / -
. / f
/ I
0.001 1 / f . f 3. ~
/ I” I. ‘ .
{'l’ ,I'A f ~\‘
0.000 “2,481.34 A~ '1 . JAM—
-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15
Angle (degrees) A

Figure 4.2.1 Measurement of the angular resolution of the spectrometer. Spectra of a
He-Ne laser are measured for different slits. Peaks are not centered at zero degrees
The

since the grating angle was not calibrated when the measurements were made.
intensities are arbitrarily scaled for convenience.

150

 

 

 

 

0.10 - —
Calculated -

7; 0.08 - —
Q)
Q) ,- e
80 -".. Measured
m 0.06 — -+
3
2 .
E 0.04 ~ ,s’ ' -
a.

0.02 — - -_

r . '
0.00 l l A l i l l l l
0.0 0.5 1.0 1.5 2.0 2.5

Slit Width (mm)

Figure 4.2.2 Measurement of angular resolution of the spectrometer. Both calculated
and measured values are shown. The measured resolution behaves as if the effective
focal length were 74 cm instead of 65 cm.

151

/— LHe
Source aperture A 2

 

]/

 

 

  

/’ Window

/

 

 

Blackbody source

 

 

 

 

 

i”
a

 

 

 

 

 

 

 

 

 

/ / L6 d
/ Spectral filter
/ with aperture A I
— Winston cone

 

Chopper wheel

/
L Detector

Figure 4.3.1 Calibration setup for the detector. The distance between the source
aperture and the detector is 10.9 cm. The temperature of the blackbody is measured

with a K-type thermocouple attached to the copper cone.

152

 

 

 

 

 

 

 

 

 

 

Input Test

 

,\ .
/

 

 

 

 

 

 

 

 

 

 

O Bias Test

Figure 4.5.1 Equivalent circuit of detector and preamplifier

 

153

 

(V6 ”)0 (V/VHZ)

l0‘6_

 

 

 

l.l.l llllllll.LA¥J 1 1111111

10“ 100 101
V|T(V)

 

Figure 4.6.1 Measured noise as a function of input test voltage. The line has a slope of
1/2 indicating the photon noise limited operation of the detector with 11 = 0.1.

154

.8085? =88:er 3 0.6 3.238»
82” 0: can 205 “m5 302 .30 $55 mm 80: e: 20:3 .Nm 8w 3 88:0 mm accuses—a. ”530:0 05.
.t0 880:0 05 53, :5 sweets: 2%: 858 55 “m3 “:9: 05 an 8388 858% came: 4.. ~04 8:me

Anzv accesvosm

 

 

000—. can.— 000. 000 3.0
_ _ _ _ a _ 4. 00.0
I ........... n ............ Mo ...................................... u ............ I ....................... l
j ........... . ............ . ........................ t ........................ l ON.O
I. ............................................................... W ...................... I1 0'0. 0
m /
m s
I ........... c ............ o ...................................... m ............ I oooooooooo l o
m u
m i
I ........... . ............ . ....................................... . ............ t ....................... l 0%.. )
m H
m z
.I ........... . ............ s ...................................... m ............ s ..................... nl (
I ............................................................... m ................................ n] 00..
r ........... . ............ . ..................................... ............ . ................... l
m l x
r _ _ _ r _ r n u

 

 

 

 

155

 

1.5E-3 T T fl f . T ' f . I . I T T
l
— time scan
0 1013-3 ‘ . --------- real scan ‘
<1 : r.
5.0E—4 ~ g , ‘
3: ~53 ; 555: , 5'; 3. .l ‘
:: : ' ‘ :0. :3 '; :31. :2 o ::.:'.: i 0... .0... q. ’0...
0.0EO - 3.1 .. 5;; . " .::..-°' : - _
—5.0E-4 r i i

 

 

 

_10E_3 . l 1 ~. 1 . l . l l l .
780 800 840 860 880 900 920 940
frequency (cm-1)

820

Figure 4.6.3 Noise from a ”time scan” and ”real scan". The frequency label of the x-
axis applies to the real scan. It takes approximately 3 minutes to measure the time
scan. Deviation from the mean value is shown for the time scan and a comparison of
two subsequent spectral scans is shown for the real scan.

5. Measurements on Oz/Pt( 111) and CO/Pt(111)

In this chapter, I present vibrational spectroscopic measurements of -
CO on Pt(111) and 02 on Pt(111). They have been widely studied, partly due
to the practical application to catalysts for oxidation of carbon
monoxide. In particular, carbon monoxide on Pt(111) has been extensively
studied by both infrared and electron energy loss spectroscopies. But
infrared measurements of low frequency modes, O-O, C-Pt, and O-Pt, are
very limited despite their scientific as well as technological importance.
We have chosen CO on Pt(lll) and 02 on Pt(111) as the first two systems to

measure by our new IRAS system.

5.1 Sample preparation and characterization

The single crystal platinum sample was purchased from Aremco, Inc.
The original sample was approximately rectangular, with dimensions 3 cm
wide x 1 cm high x 0.1 cm thick. We cut off the sample edge at one corner
by 0.3 cm x 1.5 cm in a triangular shape because of misorientation by
approximately 9° from the [111] direction while the other region is
misoriented by 2.7°. The sample was repolished mechanically with 0.5
micron diamond paste. The normal to the sample face happens to be
oriented at 2.7° from the [111] direction, as determined with the back
reflection Laue X-ray technique.

The crystal was examined with low energy electron diffraction (LEED) .
No crystallographic faults could be detected with LEED. Figure 5.1.1
shows a LEED pattern with a doublet that is from the ordered step
structure. Since the intra-doublet separation is inversely proportional

to the distance between steps, the terrace width can be obtained by

156

157
comparing the doublet separation with a major spot separation. The
terrace is 21 atoms wide as calculated from the doublet separation of the
LEED pattern. The intensity distribution within the doublets can be
obtained as a function of step height from kinematic analysis [2]. For
the {00} spot, the voltages Vm,(in.Volts) at which a singlet spot exists
with maximum intensity are given by the equation [2,3]:

150

 

V°o(Singlet Max.) - S2 , (5.1)

Adz

where d is the step height in A and S is an integer. For half-integer
values of S doublet spots of equal intensity exist. Table 5.1 shows our
measured values of V°o(Singlet Maximum) and V°°(Equal Doublet), and
estimated step heights in A. It is clear that the steps are monatomic,
since the expected step height is 2.12 A (the radius of a Pt atom is
1.221A). From the terrace width and step height, the normal to the sample
surface is at 2.73° from the [111] direction, which is in very good
agreement with that determined from Laue X-ray diffraction.

The primary contaminants in the crystal were carbon and calcium. The
calcium was removed by argon ion sputtering. Sputtering is performed at
room temperature for about 10 - 20 minutes with Ar pressure at 5 x 10'5
Torr. The sputtering energy is 400 eV and the ion current reaching the
sample is approximately 5 nA. The sample is then annealed at 1000 K for
30 minutes after sputtering. The carbon is removed by heating the sample
in oxygen (1100 K, S x 10’8 Torr with the closer) for several hours, and the
sample is then heated to 1375 K for 10 - 30 minutes to remove residual
oxygen on the surface. As the carbon level decreases to less than 102 of

a monolayer, lower oxygen pressure (1 x 10’8 Torr) is used for removing

158
carbon. The last few percent of a monolayer of carbon are removed by
dosing the sample with 0.5 - 1.0 L of oxygen through the closer [I Langmuir
(L) - 1 x 10'6 Torr for 1 sec. , the closer enhances the effective dosing by
a factor of ~20 ; see section 3.3.4], and then heating the sample to
1375 K for S - 10 minutes. During the oxygen treatments, calcium often
comes out of the bulk and is removed by argon ion sputtering. Sputtering
always results in new carbon contamination, and additional oxygen
treatments are necessary in turn. After numerous cycles of cleaning,
residual concentrations of both carbon and oxygen could be reduced to $3 1
of a monolayer (noise level). Figure 5.1.2 shows an Auger spectrum of
such a surface. About 2 - 3 percent of carbon usually appears in the

Auger spectrum and is thought to be picked up from the Auger filament.

5.2 Measurements of vibrational spectra for CO on Pt(11 1)

In this section, we present measurements of the C-0 and Pt-C
stretching modes for CO on Pt(111). Carbon monoxide on Pt(111) is one of
the most extensively studied chemisorption systems by a wide variety of
experimental surface techniques. There have been reported numerous
studies using infrared reflection absorption (IRAS) [3-8] and electron
energy loss spectroscopy [9-13] ,as well as low energy electron diffraction
(LEED) [4,12], temperature programmed desorption (TPD) [4,5,12],
ultraviolet photoelectron spectroscopy (UPS) [l3], and inelastic helium
scattering [I4] . With such a broad experimental background carbon
monoxide on Pt(111) could serve as a reference system. It might be
natural that CO on Pt(111) is chosen to see the performance of our system.

On the Pt(lll) surface CO is adsorbed non-dissociatively with the

carbon bonded to the surface in a linear (on-top) or a bridged

 

159

configuration depending on the coverage. At a fractional coverage
0 S 0.33 and temperature of 150 - 170 K, a diffuse (fix/3)R30° overlayer
structure is reported, which with increasing coverage is continuously
transformed into c(4x2) at 0 - 0.5, and then into (fIV—ZxfififlllT overlayer
structure at 0 - 0.58 [12]. In our experiment a sharp c(4x2) pattern is
observed at 90 K with 0.5 L of C0 exposure with doser. The sample is
dosed with C0 gas using the closer at 90 K and annealed at 290 K for 3
minutes and then cooled back to 90 K. This observation of a sharp c(4x2)
pattern means that the sample cleanness is well controlled.

In temperature programmed desorption (TPD), one main desorption
maximum appears between 400 and 600 K. Figure 5.2.1 shows a TPD curve for
a CO exposure of 0.5 L. Note that CO desorption around 650 K is
negligible compared to the main peak. It is generally believed that
desorption above 600 K is due to desorption from imperfections like kinks
and steps. Therefore the defect concentration on our (111) surface is
considered rather small, but we know it is at least 5 1 from the LEED
analysis.

IRAS spectra are shown in Figure 5.2.2 for the C-0 stretching mode of
on-top CO on Pt(111) with different exposures and in Figure 5.2.3 for the
Pt-C stretching mode. The spectra are measured using the room-temperature
spectrometer with cos¢ - 0.987 for the C-0 mode and using liquid-nitrogen-
cooled spectrometer with cos¢ - 0.982 for the CoPt mode, respectively (see
section 4.1).

The absorption intensity and linewidth of the on-top C-O band are
about 6.5 2 at 2109 cm'1 and 3.8 cm'1 at the saturation coverage, measured

with instrument resolution 2.9 cm'l. Compared with other works, these

 

160
values are in good agreement; Luo [15] has reported an intensity of 5.5 ~
7 2 at the peak frequency 2106 cm‘1 with a linewidth 3.5 cm'l, and Tushaus
et a1. [7] have reported an intensity z 5 I at the peak frequency 2104 cm'1
with a linewidth 4.1 cm'l. The small linewidth is also a good evidence of
a clean surface.
The measured peak intensity and linewidth of the C-Pt band are 0.13%

and 8.8 cm'1 at the peak frequency 466 cm'1

, at the saturation coverage.

Due to poor noise-to-signal ratio, the spectrum has been smoothed and a
third order polynomial baseline was subtracted. The resolution was then
5 cm'l. These spectra were measured before we installed the Inductosyn,
and the noise came mainly from the nonreproducibility of grating angles.
These values are quite similar to those of Malik and Trenary [16]; the
peak intensity was ~0.l I at the frequency 466 cm'1 and the linewith was
6 - 8 em”, but Ryberg [17,18] has reported a much narrower linewidth ~2.5
cm'1 at the frequency 463 cm'l, at the coverage 0.5, claiming that he has
minimized inhomogeneous broadening by careful sample preparation. These
measurements show that our system - though not yet fully optimized for

very low frequencies - is capable of measuring adsorbate-substrate modes

at submonolayer coverage.

5 .3 Measurements of 0-0 stretching band of 02 on Pt(l 1 1)

Oxygen on Pt(111) is an excellent model system for the study of
dissociative chemisorption. Oxygen is adsorbed molecularly below 100 K
and is adsorbed atomically above 150 K. The molecular state that we have
measured is a metastable precursor to dissociation upon heating above
150 K, leaving atomic oxygen on the surface [19,20,22] . The 0-0 bond of

the chemisorbed 0; molecule is oriented parallel to the surface and is

161

greatly 'weakened from. the gas ‘phase; the 0-0 stretch frequency is
decreased from 1556 cm'1 to 875 cm'1 [19] . Even though the 0-0 bond lies
parallel to the surface, this mode has a relatively large dynamic dipole
moment perpendicular to the surface, indicating that there is a large
charge transfer between the metal and the molecule as the bond is
stretched. This mode is, however, a lot weaker than the C-0 stretch mode,
especially because of its large linewidth.

Steininger et a1. [19] have reported a (3/2 x 3/2)R15° overlayer LEED
pattern for 02 molecules on the Pt(lll) surface at 100 K. We could barely
observe the LEED pattern, but it is too diffuse to tell the structure and
disappears in 10 - 20 seconds from the LEED screen, probably because the
structure is destroyed by the electron beam from the LEED filament. After
annealing to 300 K, a sharp (2 x 2) LEED pattern is observed, which is
formed by atomic oxygen adsorbates on the surface. A sharp (2 x 2) LEED
pattern is also observed after dosing the sample with 1.0 L of oxygen at
room temperature.

At 90 K, saturation coverage is achieved after an exposure to about
0.75 L of oxygen through the doser. In the TPD spectrum two desorption
peaks appear as shown in Figure 5.3.1; one sharp and strong peak at 150 K
and a rather broad peak with its maximum at ~800 K. The first desorption
peak is from molecularly adsorbed oxygen and the second desorption results
from atomic oxygen“ This has been verified by other groups [20,21] using
isotopic mixing experiments. After dosing the sample with a mixture of
1‘02 and 1802, complete isotopic mixing (160130) has been observed in TPD for
the high temperature state while no isotopic mixing has been found for the

low temperature state. We have observed a CO peak picking up and the 02

162
peak decreasing in the TPD spectrum if we keep the sample at 100 K even
for 5 - 10 minutes after dosing the sample with 0.75 L of oxygen. Oxygen
molecules stick to the surface so weakly that even at 100 K they are
replaced by carbon monoxide in the UHV background. This process, however,
does not occur at an observable rate if the sample is kept at 90 K for
hours.

Figure 5.3.2 shows IRAS spectra for several oxygen exposures. The
spectra have been measured at 90 K after dosing oxygen at 90 K. The 040
stretching band is observed at 858 cm'1 at lowest coverage, but shifts
upward to 875 cm‘1 with increasing coverage. The half-width of the bands
is approximately 20 cm'1 and shows little dependence on the coverage. The
sharp spikes are due to angular errors, and have now been eliminated by
use of the Inductosyn transducer.

Persson [23] has argued.that the 0-0 vibration is likely to be damped
via excitation of electron-hole pairs near the Fermi level in the metal,
showing that the asymmetric line shape predicted by Langreth [25] (see
below) for electron-hole damping is reproduced quite well by fitting to
the experimental data of IRAS by Canning and Chesters [24], and of EELS by
Steininger and coworkers [19]. The conclusion also seems quite plausible
considering that there is a strong charge coupling between the metal and
the 0; molecule. The IRAS measurements were, however, performed using a
polycrystalline ribbon sample which also lacked good surface
characterization, while the resolution of EELS (50 cm“) is too large to
examine the linewidth. In particular, Persson excludes the possibility of
inhomogeneous broadening based on indirect and inconclusive experimental

evidence.

 

163

Langreth [25] has derived the vibrational line shape of an adsorbate
on a metal surface due to electron-hole damping and has shown that it has
to be asymmetric with a tail on either side of the resonance frequency.
It is assumed that the adsorbate has an induced resonance level close to
the Fermi level so that electrons move in and out of the resonance level
as the adsorbate vibrates and the position of the resonance level moves up
and down accordingly. This leads to a breakdown of adiabaticity on the
order of wr, where w is the vibrational frequency and r is the tunneling
time of an electron in the resonance level. Tunneling of adsorbate
electrons and holes into the substrate, thus creating electron-hole pairs
in the substrate, causes the damping of the adsorbate vibration. The
dynamic dipole of the adsorbate has two components; p - p*'+ u', where p1
is the contribution of ion core and.uP the valence electron contribution,
and u' should have an imaginary part so that p - pl + p2 - p1(l + iwr).
The line shape is expressed as:

Zulzwrmu-wzrz) - 2wr(w2-w,2)

IM(U) "" a (5'3'1)
(wz - 0,2)2 + (0212

 

where w, is the resonance frequency, 1 the vibrational linewidth. The
equation 5.3.1 is a generalization of the well-known Fano line shape. The
asymmetry in the line shape is determined by the parameter wr, and the
tail extends to either side of the resonance frequency w, depending on the
sign of wr. The asymmetry parameter wr can be expressed as [25-27]:

«or - (tp.u)(p‘/p1' + l)’1 , (5.3.2)
where p. is the number of electrons that move in and out of the resonance
level per cycle and 111' is the real part of u‘. The tail therefore extends

toward the high frequency side if p‘ and pf have the same sign or they

164
have different sign with [pi] < [”1“] , otherwise it extends toward the low
frequency side.

Although the equation 5.3.1 has been derived for vibrational modes of
a single adsorbate perpendicular to the surface, the same equation applies
to a layer of dipole-coupled adsorbates with renormalized parameters [26] ,
and to vibrational modes parallel to the surface [27]. Persson [23] has
derived the same Fano line shape for the electron-hole damping of
adsorbate vibrations on metal surfaces using a somewhat different
approach.

The isotope dependence of the linewidth 1 dominated by electron-hole
pair damping has been predicted to be characterized by a - -1. Here a is
defined as a - (aln 1/aln m,), where m, is the reduced mass of adsorbate
(see section 1.2). The linewidth 1 for electron-hole pair damping is
expressed as [25-27]:

1 - 21rm,.(6¢p.)z , (5.3.3)
where 66 is change in the energy of the induced resonance level of the
adsorbate caused by stretching of the bond, and p. is the number of
electrons that move in and out of the resonance level per cycle. Since 56
~ Q0 ~ (m,w,)1/2, the linewidth is inversely proportional to the reduced
mass: 1 ~ mt'l.

As discussed in section 1.3, the mass dependence is probably the
best way to sort out the inhomogeneous broadening. We are therefore
trying to measure the isotopic effect on the linewidth, in order to
further understand the line broadening mechanism. We expect to see that
the linewidth for “02 is smaller by 1.2 cm'1 than that for 1‘02 if it is

dominated by inhomogeneous broadening, or by 2.5 cm'1 if it is dominated

165

by electron-hole pair damping. In Figure 5.3.3, the solid curve shows the
best fit, using the equation 5.3.1, to the measured IRAS spectrum for M0;
on Pt(111) at the saturation coverage N'- 6 x 101‘<nf1. From the fitting,
the extracted parameters are; the linewidth 1 - 21.6 cmfl, the asymmetry
parameter wr - 0.17, the effective charge e'l- 0.27e, and the vibrational
polarizability av - 0.047 A3. Compared with those from the measurement of
Canning and Chesters [23,24] for 1302 on a platinum foil; 1 - l9 cm'l,
wr - 0.12, e' - 0.39e, and.a~ - 0.096 A3 [The equation (1) of Ref. 24 for
the integrated absorption is off by a factor 4, and thus the numerical
values of e' and av in Ref. 24 and 25 are larger by the factors 2 and 4
respectively than they should be.], our IRAS spectrum shows a slightly
more asymmetric line profile and a smaller vibrational polarizability (IV
by a factor 2. There is a difference of 2.6 cm‘1 in between the linewidth
for 1302 from Canning and Chesters' measurement and that for 1‘02 from our
measurement, suggesting that electron-hole pair damping dominates the
linewidth of the 0-0 stretch mode.

At this point, however, we can not draw a conclusion about the line
broadening, since the linewidth for 1802 was measured on a different sample
and further its reproducibility is not known. The spectra of our
measurements also show a fluctuation of linewidth 1 ~ 2 cm'l, which is
probably caused by remaining inadequacies in baseline substraction, sample
cleanness and noise-to-signal ratio. These need to improve for better
reproducibility. Next, the linewidths for 1802 will be measured and
compared with those for 1‘02 to sort out the line broadening mechanism.

In addition to this first project, the investigation of the line-

broadening mechanism for the 0-0 stretch mode of 02, we are planning to

166
study the adsorbate-substrate modes of molecular and atomic adsorbates,
eg. O-Pt, Oz-Pt, C-Pt, and S-Pt, alkali metal adsorption on metals,
adsorption on semiconductors and broadband IR absorption measurements of
metal surfaces, which require high resolution and high signal-to-noise

ratio.

 

167

References

10.

ll.

12.

'13.

14.

15.

l6.

17.

18.

19.

20.

M. Henzler, Surf. Sci. 19, 159 (1970).

B. Lang, R.W. Joyner and G.A. Sommerjai, Surf. Sci. 30, 440 (1972).

A. Crossley and D.A King, J. Chem. Phys. 68, 1344 (1978).

G. Ertl, M. Neumann and K.M. Streit, Surf. Sci. 64, 393 (9177).

B.E. Hayden and A.M. Bradshaw, Surf. Sci. 125, 787 (1983).

R.G. Tobin, R.B. Phelps and P.L. Richards, Surf. Sci. 183, 427 (1987).
M. Tushaus, E. Schweizer, P. Hollins and A.M. Bradshaw, J. Electron.
Spectrosc. Relat. Phenom. 44, 305 (1987).

E. Schweizer, B.N.J. Persson, M. Tushaus, D. Hoge and A.M. Bradshaw,
Surf. Sci. 213, 49 (1989).

H. Froitzheim, H. Hopster, H. Ibach and S. Lehwald, Appl. Phys. 13,

147 (1977).

H. Hopster and H. Ibach, Surf. Sci. 77, 85 (1978).

N.R. Avery, J. Chem. Phys. 74, 4202 (1981).

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

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

A.M. Lahee, J.P. Toennies and Ch. Woll, Surf. Sci. 177, 371 (1986).

J.S. Luo, PhD Thesis, Michigan State University (1992). P. 144.

I.J. Malik and M. Trenary, Surf. Sci. 214, L237 (1989).

R. Ryberg, Phys. Rev. B40, 8567, (1989); B40, 5849 (1989).

B. Persson and R. Ryberg, Phys. Rev. 340, 10273 (1989).

H. Steininger, S. Lehwald and H. Ibach, Surf. Sci. 123, 1 (1982).

J.L. Gland, Surf. Sci. 93, 487 (1980).

21.

22.

23.

24.

25.

26.

27.

168
A. Winkler, X. Guo, H.R. Siddiqui, P.L Hagans and J.T. Yates, Jr..
Surf. Sci. 201, 419 (1988).
J.L Gland, B.A. Sexton and G.B. Fisher, Surf. Sci. 95, 587 (1980).
B.N.J. Persson, Chem. Phys. Lett. 139, 457 (1987).
N.D.S. Canning and M.A. Chesters, J. Elect. Spect. Relat. Phenom. 29,
69 (1983).
D.C. Langreth, Phys. Rev. Lett. 54, 126 (1985).
Z. Crljen and D.C. Langreth, Phys. Rev. B35, 4224 (1987).

A.I. Volokitin, Surf. Sci. 224, 359 (1989).

169

Table 5.1 Voltages for singlets of maximum intensity and doublets of
equal intensity and step height in A from equation 5.1.

 

 

 

S V°°(Singlet Max.) V°°(Equal d (A)
Doublet)
If

2 34 2.10
2.5 59 1.99
3 78 2.08
3.5 ----

4 112 2.31
4.5 158 2.19
5 p 193 2.20
5.5 234 2.20
6 275 2.22
6.5 311 2.26
7 343 2.31

 

 

 

 

 

 

 

170

 

Figure 5.1.1 LEED pattern of the crystal with a doublet that is from the ordered step
structure. The step is monatomic and the terrace is 21 atoms wide. The normal to the surface
is at 2.73° from the [111] direction.

171

.89. has: a 9» c5
combs. Ba 2893 90 may :38 «o .528 Bow a “a5 Bez .8atam 2:3.— 520 8.2 5.58% Bwa< .2 «46 Semi

Se. >035 coboefi

 

 

 

 

 

 

 

 

 

 

0mm cor cm? 00.. can OOH omw com om; on: a:
m A q _ q a a an. _ d
E e
a fi fi
.. a
a E r.
e
a .. .
t t a t
e a r. e
z a
p D.
3 N
\ a m
K’s/>712 N
s s ..
a. 00 U nu...
0 fix
0 one $0.21??—
L L. _ fr _ _ _ L _

 

 

172

 

 

 

 

 

 

r- a
A r ‘
E
3. ,- i
E , .
3
i
s t i
U)
8 r .
r 1
' 1
WW!
I a m J; J A _L a a L
0.1 0.3 0.5 0.7 0.9 1.1

Temperature (K)

Figure 5.2.1 A TPD spectrum measured after dosing the Pt(lll) sample with 0.5 L of
CO gas using the doser.

 

173

 

0.005 L

0.02 L

0.1L

0.5 L

2

D
wlw
' I ' I v v .
%%fi 4
i L A 1 r 1

 

 

 

2050 2080 2110 2140 2170 2200
Frequency (cm-1)

Figure 5.2.2 Spectra for GO stretch mode of on-top CO on Pt(111). CO gas
exposures are in Langmuir and are referred to ion gauge readings. The effective
exposures at the surface were larger by a factor 20 because of the doser. A linear base
line is subtracted for each spectrum .

690833 mm on: 83 a can 858.5. mm 8.58% 25. .286 9.282% 560 05 .«o 8.58% mg m.~.m Semi

A 2-83 honoeuoum

 

174

 

m _m E? mt. «we _mv c _v

_ 4 _ _ . _ . J L WflGOAUI
1 :86-
1 Seed. W
- mocod-

.. \2/

/ \itt \/\./

1 886

 

L » F F _ . _ . p . mocc.c

 

175

 

 

 

 

 

 

 

 

r- -<

r- a

.g t- q

'3 r- 'i
3

‘2 T '“
.9
m

g i‘ X10 1

r- 4

[- ‘1

2 L l

0.2 0.4 0.6 0.6 513.0

Temperature (K) x

Figure 5.3.1 A TPD spectrum measured after dosing the sample with 0.8 L of 02 gas.

The sharp strong peak at 160 K is from molecularly adsorbed oxygen and the rather
broad peak at about 800 K is from atomic oxygen on the surface.

176

 

 

 

 

 

0 l
8 030L -
a
‘9. r
o
.3 osor ,
<
0.701.
025%
810 840 870 900 930 960
Frequency (cm-1)

Figure 5.3.2 IRAS spectra of the O-O stretchung band for different 02 exposures.
Spikes are caused mainly by the nonreproducibility of the grating angles (see also
section 4.7)

177

.8588— ocam 05 o. E v.8 .8838
85889. 85 8 8888 22?. :0 89.88 88888 NOE 8.5 .8589? 5.2% 0.0 05 no 858% mg 5. m.m.m 2am."—

A.-.Eov 88:70.."—

 

 

 

 

 

omc _ com sea 3% cf. can .

i . . _ 1 _ . _ . $25 a.

\ r nnccd.

\ f A r c256.

a. . 5.: _ .1386. W
.J .45. "A‘ .a :4 fit...
1 c—ocd.
L _ _ _ e _ r L » “cocoa

6. Conclusion

Many interesting surface vibrational experiments require high
resolution, high signal-to-noise performance and excellent baseline
subtraction, especially in the low frequency region. Instruments with
those capabilities are not readily available. We have thus built a new
far- infrared reflection- absorption spectroscopy system designed
specifically for the frequency region 350 - 1000 cm”, though it can be
used up to 3500 cm'l. It incorporates a liquid-nitrogen-cooled grating
monochromator and other cooled optics, a Si:B photoconductive detector and
a UHV chamber equipped with LEED, Auger and TPD.

We have measured the 0-0 stretch mode of molecularly adsorbed 02 on
Pt(111) as a function of coverage and the molecule-substrate mode of CO on
Pt(111), showing that the system is working well. 3 The noise-to-signal
(N/S) ratio achieved to date is 2 x 10" l/fl-E and the dominant noise
source is the nonreproducibility from one scan to the next. An
improvement of the N/S ratio by a factor of 4 is expected by improving the
scan-to-scan reproducibility.

As a first experiment, the isotope effect on the linewidth of the O-O
stretch mode will be measured. The shifts in the linewidth are expected
to be 1.2 cm'1 if inhomogeneous broadening is dominant, or 2.5 cm'1 if
electron-hole pair broadening is dominant. With the resolution, N/S ratio
and baseline subtraction performance of our new system, those shifts can
be measured.

A broadband infrared absorption measurement is planned as the next

experiment. The broadband infrared absorption is known to have a

178

179
characteristic dependence on. the frequency [for' metals, and. to have
correlations with surface resistivity, depending on whether there are
adsorbates or not [1]. In addition, we will perform measurements of
adsorbate-substrate vibrations of atomic and.molecular adsorbates, alkali
metal adsorption on metal surfaces, and adsorption on semiconductors,
which require high resolution, very high signal-to-noise ratio and

excellent baseline subtraction.

180

Reference

1. K.C. Lin, R.G. Tobin, P. Dumas, C.J. Hirschmugl and G.P. Williams,

Phys. Rev. B, to be published, and references therein.

 

"00000000000