|
SURFACE CHARACTERIZATION AND CHEMISORPTION PROPERTIES
OF POLYCRYSTALLINE SYSTEMS:
Sn02, Pt/SnO2 and Zr
By
DAVID FULLEN COX
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1984
Copyright 1984
by
David Fullen Cox
ACKNOWLEDGMENTS
The author would like to thank Gar Hoflund for the
guidance and encouragement he furnished in his role as
research advisor. Special thanks go to Gar also for
many hours spent in the discussion of academic and purely
nonacademic matters and for invaluable assistance rendered
in the author's quest to master the French (not to mention
the English) language, pate de foie gras! Thanks go
also to Herb Laitinen for the benefit of his expertise
on tin oxide in all its forms and for providing the
laboratory facilities used for sample preparation. Thanks
go to Paul Holloway for his helpfulness in allowing the
use of his SIMS apparatus and for his patience in enduring
the associated visits. A special thank you also goes
to Dick Gilbert at the University of Nebraska for a
multitude of software and hardware contributions which
played such a major role in all the results presented
here.
The author thanks G.B Hoflund and H.A. Laitinen
for financial support supplied through sponsored research
grants. Thanks go also to the Department of Chemical
Engineering and the State of Florida for financial support.
Lastly, the author acknowledges the Florida Guaranteed
Student Loan Program for unilateral support during the
final six months of his degree program.
iii
TABLE OF CONTENTS
PAGE
ACKNOWLEDGMENTS iii
LIST OF TABLES vi
LIST OF FIGURES vii
ABSTRACT xi
SECTION
ONE GENERAL INTRODUCTION 1
Motivation 1
Format 4
TWO AN XPS INVESTIGATION OF TIN
OXIDE SUPPORTED PLATINUM 8
Introduction 8
Experimental 10
Results and Discussion 13
Conclusions 21
THREE AN ELECTRONIC AND STRUCTURAL
INTERPRETATION OF TIN OXIDE
ELS SPECTRA 23
Introduction 23
Experimental 25
Background 26
Results and Discussion 29
Conclusions 56
FOUR A STUDY OF THE DEHYDRATION OF
TIN OXIDE SURFACE LAYERS 58
Introduction 58
Experimental 61
Results and Discussion 62
Conclusions 71
PAGE
FIVE AN OBSERVATION OF WATER ADSORPTION
ON TIN OXIDE USING ESD AND
GRAZING-EXIT-ANGLE XPS AND AES 73
Introduction 73
Experimental 74
Results and Discussion 81
Conclusions 90
SIX THE INTERACTION OF POLYCRYSTALLINE
ZIRCONIUM WITH 02, N2, CO AND N20 91
Introduction 91
Experimental 92
Results and Discussion 93
Conclusions 115
SEVEN GENERAL CONCLUSIONS AND
RECOMMENDATIONS FOR FUTURE
RESEARCH 117
Pt Sn Oxide 117
Zirconium 119
APPENDICES
A A BRIEF DESCRIPTION OF THE
EXPERIMENTAL TECHNIQUES 121
X-Ray Photoelectron Spectroscopy (XPS) 121
Auger Electron Spectroscopy (AES) 124
Electron Energy-Loss Spectroscopy (ELS) 126
Electron-Stimulated Desorption (ESD) 128
B COMPUTER-INTERFACED DIGITAL PULSE
COUNTING CIRCUIT 131
Introduction 131
Circuit Description 132
Time-of-Flight Modification 139
Acknowledgments 140
REFERENCES 142
BIOGRAPHICAL SKETCH 147
LIST OF TABLES
PAGE
TABLE
3-1. Electronic Configurations of Atomic
Tin and Oxygen and the Stoichiometric
Oxides. 40
5-1. Variation in O/Sn Ratio With Emission
Angle. 83
LIST OF FIGURES
FIGURE PAGE
2-1. Pt 4f XPS spectrum of an electro-
chemically platinized substrate.
The lower binding energy doublet
is characteristic of Pt metal,
and the higher binding energy
doublet is characteristic of
a Pt species chemically bonded
to be the tin oxide substrate. 14
2-2. Pt 4f XPS spectra of a sample
prepared by platinum chemisorp-
tion. Spectrum (A) is obtained
immediately after pumpdown,
(B) after high temperature oxida-
tion, (C) after high temperature
reduction and (D) after a high
temperature anneal in vacuo. 17
3-1. Energy level diagram representing
the SnO2 band structure. The
locations of the major occupied
(unoccupied) valence and core
(conduction) states involved
in the energy loss spectrum
are shown. The approximate
locations of the SnO VBM and
CBM are indicated by dashed
lines. 30
3-2. N(E) ELS spectrum of tin oxide
after a high temperature vacuum
anneal. The high primary beam
energy (Ep = 1500 eV) at normal
incidence results in primarily
a bulk sensitivity. The indicated
features are characteristic
of well-annealed SnO2. The
low energy features are due
to VB ----> CB transitions,
and the high energy loss features
are core ----> CB transitions. 33
vii
PAGE
3-3. Variation in the N(E) ELS spectrum
with primary beam energy, Ep.
The set of spectra represent
a depth profile of the annealed
material. The growth of the
27 eV feature is due to an
increasing oxygen deficiency
as the spectra become more surface
sensitive. 36
3-4. EN(E) ELS spectrum for Ep = 50
eV. The main loss feature at
13 eV shows that the annealed
material is essentially SnO
at the surface. 37
3-5. ELS spectra for a sample annealed
at 6000C. The angle of incidence
for the primary beam is 45.
With Ep = 1500 eV it is seen
that the bulk is primarily SnO2.
The Ep = 300 eV spectrum shows
SnO at the surface and evidence
of a defect structure between
bulk SnO2 and the surface. 44
3-6. Valence band XPS spectra after
a (A) 6000C anneal and (B) 2
KeV argon-ion bombardment. The
spectra are more bulk than surface
sensitive. 46
3-7. ELS spectra following 2 KeV argon-
ion bombardment. The valence
band features show the bulk
to be SnO21ike while the core
features reveal a significant
concentration of defects. The
VB features in the more surface
sensitive spectrum illustrate
an amorphous structure at the
surface due to sputtering. 49
3-8. Valence band XPS spectra following
(A) a short 5000C anneal after
sputtering and (B) subsequent
oxygen exposure. The presence
of a mixture of SnO and SnO2
is indicated in (A). 52
viii
PAGE
3-9. ELS spectra corresponding to Figures
3-8(A) and (B) respectively. 54
4-1. Valence band XPS spectra after
(A) hydration by exposure to
atmospheric humidity, (B) a
5000C vacuum anneal for 45 minutes
and (C) a 6000C vacuum anneal
for 30 minutes. 64
4-2. ELS spectra of (A) subsurface
and (B) surface regions after
hydration due to atmospheric
humidity. 65
4-3. ELS spectra of (A) subsurface
and (B) surface regions after
a 5000C vacuum anneal. 68
4-4. ELS spectra of (A) subsurface
and (B) surface regions after
a 6000C vacuum anneal. 70
5-1. Variation in path length with
emission angle. 76
5-2. Deflection circuit for desorption
event initiation. 79
5-3. Time-of-flight spectrum for mass
analysis. 86
5-4. Ion kinetic energy distribution
after sputtering. 88
5-5. Ion kinetic energy distribution
following water exposure. 89
6-1. AES spectra taken after (A) 2
hours of heating and (B) 14
hours of heating below the HCP-to-
BCC transition temperature. 94
6-2. AES spectra of state 1 zirconium
after room temperature exposure
to (A) nitrogen and (B) nitrous
oxide. 98
PAGE
6-3. XPS spectra showing the zirconium
3d peaks for (A) clean zirconium,
(B) N2 exposure, (C) N20 exposure
and (D) oxygen exposure 99
6-4. XPS spectra showing the zirconium
3s peak for (A) clean zirconium,
(B) nitrogen exposure and (C)
N20 exposure. The nitrogen
Is peak appears at 396 eV in
(B) and (C) 100
6-5. XPS spectra showing the oxygen
Is peak after (A) 02 adsorption
on zirconium and (B) N20 adsorp-
tion on zirconium 102
6-6. AES spectrum for clean zirconium
after heating near the melting
temperature for 3 hours. The
AES 175 eV peak is greatly dimi-
nished which is characteristic
of state 2 zirconium 104
6-7. AES spectrum taken after clean
state 2 zirconium is exposed
to CO contamination from the
electron beam for 8 hours. The
carbon peak shows characteris-
tics of both graphite and carbidic
carbon 106
6-8. XPS spectrum of the carbon Is
peak corresponding to the AES
spectrum shown in Figure 6-7.
Both graphitic and carbidic
carbon are present 107
6-9. AES spectra after exposing state
2 zirconium to CO at (A) room
temperature and (B) high tempera-
ture but allowing the sample
to cool during the exposure 108
6-10. XPS spectra of the carbon Is peak
corresponding to the AES spectra
shown in Figure 6-9. The room
temperature adsorption produces
approximately equal amounts
of graphitic and carbidic carbon
as shown in spectrum (A) while
the high temperature adsorption
results in predominantly carbidic
carbon as shown in spectrum
(B) 109
PAGE
6-11. (A) AES spectrum taken after expos-
ing state 2 zirconium to nitrogen
at 5x10-6 Torr for 15 minutes
at room temperature. A small
amount of carbon and oxygen
contamination accumulated during
the long exposure and subsequent
AES run. (B) AES spectrum taken
after exposing state 2 zirconium
to nitrogen initially at high
temperature and then allowing
the sample to cool during the
exposure. (C) AES spectrum
taken after allowing the sample
to remain in vacuum for 3 days
at room temperature. State
2 zirconium has transformed
into state 1 zirconium. (D)
AES spectrum taken after exposing
the state 1 zirconium of spectrum
(c) to nitgrogen at 5x10-6 Torr
for 5 minutes at room temperature. 111
A-1. Photoemission process. 122
A-2. KL1L2 Auger decay process. 125
A-3. Electron energy-loss process. 127
B-l. On-board timer schematic showing
jumper selectable system clock
rates. 133
B-2. Schematic of the control logic
section. 135
B-3. (A) Timing and (B) event counter
schematics. 137
B-4. Layout and wiring diagram. All
unmakred (pull up) resistors
are 10K ohms. 138
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
SURFACE CHARACTERIZATION AND CHEMISORPTION PROPERTIES
OF POLYCRYSTALLINE SYSTEMS:
Sn02, Pt/SnO2 and Zr
By
DAVID FULLEN COX
December, 1984
Chairman: Gar B. Hoflund
Major Department: Chemical Engineering
X-ray photoelectron spectroscopy (XPS) is used to
characterize platinum supported on tin oxide. A feature
in the platinum 4f XPS spectrum associated with the bond
formed between supported platinum and the tin oxide
substrate is identified. The bond is believed to form
with surface lattice oxygen resulting in a Pt-O-Sn surface
species. This substrate-bonded species appears to act
as a nucleation site for cyrstallite growth in both the
electrochemical desposition of platinum and in the
sintering of supported platinum.
It is demonstrated that electron energy-loss
spectroscopy (ELS) is an acceptable technique for
xii
distinguishing between the different oxides of tin. The
major features in the N(E) loss spectrum are interpreted
as due to collections of optically allowed interband
transitions. It is shown that depth profile information
about tin oxide may be obtained by varying the primary
electron beam energy. Combined ELS and valence band
XPS results indicate that a significant amount of
structural information may be inferred from -the size,
shape and/or position of the N(E) ELS features. Core
level features are found to be quite sensitive to the
presence of defects in an SnO2 lattice with some
specificity as to the type of defect.
The chemisorption properties of polycrystalline
zirconium have been found to vary dramatically depending
on the thermal history of the sample. Chemisorption
on this surface is found to be suppressed by heating
for prolonged periods of time above the HCP-to-BCC phase
transition temperature at 11350K. The chemisorption
behavior can be correlated roughly with the appearance
or disappearance of a zirconium MVV Auger peak. A slow
phase transition at the surface is postulated as the
cause of the variation in chemisorption properties.
xiii
SECTION I
GENERAL INTRODUCTION
Motivation
The primary motivation for the work presented here
is an interest in the catalytic properties exhibited
by tin oxide supported platinum. It has been demonstrated
that platinized tin oxide surfaces display higher catalytic
activities than platinum electrodes in the electrochemical
oxidation of methanol [1-5] and for the reduction of
oxygen in alkaline [6] and 85% phosphoric acid solutions
[7]. A similarity between the electronic properties
of platinum in these supported systems and in the
industrial platinum-tin bimetallic reforming catalyst
supported on alumina has been demonstrated recently [8].
The bimetallic catalyst is known to exhibit improved
stability and higher average catalytic activity than
platinum supported on alumina [9,10].
Tin oxide and its modified forms also exhibit a
significant catalytic activity. Tin oxide has been shown
to be active for the catalytic oxidation of CO [11,12]
and the reduction of NO [12-14]. Chromia-doped tin oxide
is very active for the reduction of NO with CO, H2 and
C2H4 [13]. Antimony-doped tin oxide is known to be active
toward the selective (partial) oxidation of propylene
to acrolein [15] and the oxidative dehydration of butene
to butadiene [16]. Much of this behavior is believed
to be linked to the oxidation and reduction of active
sites on the catalyst surface.
The catalytic properties of the tin oxide support
are believed to play an important role in platinized
tin oxide catalysts. For example, Tseung and Dhara [7]
have postulated that a spillover mechanism may be involved
in the electrochemical reduction of oxygen. Their results
suggest that adsorbed oxygen migrates from the supported
platinum to the tin oxide surfaces before undergoing
reduction. The possible importance of spillover mechanisms
in platinum/tin oxide catalyzed reactions and the apparent
redox behavior of tin oxide surfaces both indicate that
a successful surface characterization of the supported
system must include a determination of the tin oxidation
state.
The main focus of the work presented here is the
application of ultrahigh vacuum (UHV) surface probes
to a fundamental characterization of tin oxide and
platinized tin oxide catalysts. It is hoped that this
characterization will aid in understanding the
physiochemical properties affecting the preparation and
the catalytic behavior of these systems. To this end,
a realistic system is studied which utilizes
polycrystalline tin oxide as the support material. The
polycrystalline nature of the support presents immediate
problems in terms of structural characterization because
the usual surface-structure-sensitive technique, low-energy
electron diffraction (LEED), is not applicable. For
this reason the approach primarily has been to use electron
spectroscopies as probes of the electronic properties
of the materials. The direction taken in the work
presented here has been influenced by some earlier studies
which have already been reported in the literature [17-19].
The earlier results will not be repeated here but will
be referenced when appropriate to the discussion of present
results.
One unforeseen result of the study of tin oxide
has been a broadening of interests to include structural
effects in electronic spectra (see Sections III and IV).
Out of this interest has grown a study of the chemisorption
properties of clean polycrystalline zirconium.
Observations in the literature of anomalous effects on
the Auger electron spectrum of zirconium due to gas
adsorption combined with conflicting results on the uptake
rate of adsorbates are responsible for the selection
of zirconium for study. Though not directly related
to the characterization of platinized tin oxide, these
zirconium results are also presented here.
Format
The results presented here are divided into
independent sections, each of which is complete within
itself. Each section deals with one aspect of the surface
characterization of a platinum-tin oxide catalyst with
the exception of Section VI which deals with the apparent
structural dependence of the chemisorption properties
of polycrystalline zirconium.
Section II presents results on the characterization
of supported platinum on tin oxide. X-ray photoelectron
spectroscopy (XPS) has been used in this characterization
to investigate the valence (oxidation) states of the
supported platinum. In situ thermal and chemical
treatments are used to help identify the various platinum
oxidation states and to investigate the nature of the
chemisorption bond formed between platinum and the tin
oxide substrate.
While XPS has proven useful in determining the valence
states of supported platinum, it has been of limited
use in characterizing the oxidation state of tin in the
support. Though core level XPS can distinguish between
metallic tin and tin oxide, it can not distinguish between
the different oxides of .tin, SnO and SnO2. In view of
the apparent importance of the determination of the tin
valence state, the majority of effort has been devoted
to finding a suitable characterization technique for
the tin oxide support. The chosen technique is electron
energy-loss spectroscopy (ELS).
Section III is a discussion of the use of ELS in
the characterization of tin oxide. An interpretation
of the spectra is given in terms of the excitation of
interband transitions. Using ELS in conjunction with
valence band XPS, it is demonstrated that a significant
amount of structural information about polycrystalline
tin oxide may be inferred from changes in the electronic
structure as probed by ELS.
An understanding of the interaction of water with
tin oxide surfaces is believed to be of primary importance
in elucidating the chemisorption properties of platinum
on tin oxide. This belief is supported by considerable
evidence that the active chemisorption sites are surface
hydroxyl groups [6,19]. In Section IV ELS is applied
to the study of the dehydration of a tin oxide sample
after exposure to atmospheric humidity. The flexibility
of analysis depth provided by this technique affords
a particularly worthwhile characterization of the
subsurface region.
Because of a phenomenon known as electron-stimulated
desorption (ESD), ELS is not useful in studying the
interaction (i.e. adsorption) of water with the surface
of tin oxide. The incident electron beam used for ELS
actually removes the adsorbed species of interest from
the surface. This phenomenon has been observed in the
earlier studies [17-19] which show that significant surface
modification can result from an impinging electron beam.
However, the ESD phenomenon can itself be used to probe
the interaction between surface and adsorbate.
Section V presents some preliminary observations
of the adsorption of water on tin oxide.
Grazing-exit-angle XPS provides a measure of the water
adsorbed from the background vacuum in the UHV system.
Preliminary results are also presented which demonstrate
the potential of ESD in characterizing the interaction
of water with tin oxide surfaces.
Section VI presents a study of the chemisorption
properties of polycrystalline zirconium. The study
focuses on the chemisorption behavior as a function
of the thermal history of the sample. A slow phase
transition in the surface region is postulated as the
cause of a suppression in chemisorption after high
temperature annealing. A zirconium feature in the Auger
electron spectrum is shown to be an indicator of the
chemisorption properties of the surface.
Section VII presents a summary of conclusions for
the entire study along with recommendations for future
research. Two appendices are included after Section
VII. Appendix A contains a brief description of the
physical processes involved in the experimental techniques
7
used in this work. A brief introduction to XPS, AES,
ELS and ESD is given. Appendix B describes the
computer-interfaced digital pulse counting circuit used
for data collection in XPS, ELS and time-of-flight ESD
measurements.
SECTION II
AN XPS INVESTIGATION OF TIN OXIDE
SUPPORTED PLATINUM
Introduction
The study of tin oxide supported platinum is motivated
by the interesting catalytic properties displayed by
mixed Pt-Sn systems. It has been shown that platinized
tin oxide exhibits a catalytic activity 50 to 100 times
greater than that of platinum electrodes for the
electrochemical oxidation of methanol [1-5]. Watanabe
et al. [6]. have studied the effects of platinum loading
on the catalytic activity for oxygen reduction in alkaline
solution. Their results show that the catalytic activity
of highly dispersed platinum on tin oxide may exceed
that of platinum electrodes by a factor of four or more
[6]. An interest in impure H2 fuel cells using 85%
phosphoric acid at 1500C has prompted Tseung and Dhara
[7] to study supported Pt on antimony-doped tin oxide
because of the corrosion resistance and electrical
conductivity exhibited by this system. Their results
show a significant increase in the catalytic activity
for oxygen reduction over that of platinum black. In
addition, the similarity between platinized tin oxide
9
and the Pt-Sn bimetallic reforming catalyst has been
demonstrated recently [8].
The catalytic effects of low coverages of strongly
bound oxygen on Pt single-crystal surfaces has been
demonstrated by Smith, Biberian and Somorjai [20]. They
interpret dramatic, oxygen-coverage-dependent changes
in activity and selectivity for hydrogenation and
dehydrogenation reforming reactions as being due to the
formation of surface Pt oxides. Interestingly, it has
been shown that the formation of Pt-Sn alloys generally
results in lower catalytic activities [21-24]. These
observations suggest that oxygen may be partially
responsible for the catalytic properties exhibited by
Pt/tin oxide systems.
Several X-ray photoelectron spectroscopy (XPS) studies
have been performed on Pt-Sn systems [4,5,8,25]. These
studies all agree that tin is largely present as tin
oxide while platinum is present in the metallic form
and as Pt2+ and Pt4+ in the form of oxides and hydroxides.
These previous XPS studies have examined intimately mixed
systems of tin oxide and platinum oxides and metal. In
the present work a supported platinum system is studied.
The use of this system in conjunction with in situ chemical
and thermal treatments has allowed the assignment of
a Pt "oxidation state" characteristic of the chemical
bond formed between the metal and the tin oxide substrate.
Experimental
The tin oxide substrates are prepared by the thermal
hydrolysis of Sn (IV) from a solution containing 3 M
SnC14 5H20, 1.5 M HC1 and 0.03 M SbC13. The solution
is sprayed onto the hot surface of a titanium foil held
at 5000C in air. The formation of tin oxide occurs
according to:
SnC14 + 2H20 -----> Sn02 + 4HC1
The resulting planar film is a polycrystalline, n-type,
Sn02 semiconductor with the rutile structure. The antimony
is incorporated into the film at a concentration
approximately twice that of the spray, i.e. 2% [26].
This dopant is known to be in solid solution with the
tin oxide [27], and it acts as a donor which raised the
conductivity of the film to a level adequate for
electrochemical studies [26]. Spraying is continued
0
until a tin oxide layer approximately 6000 A to 7000
0
A thick is obtained as determined by the colors of the
interference fringes of the layer. After cooling in
air, the samples are polished with 0.25 pm diamond paste.
The use of alumina as a polishing compound is avoided
because of the overlap of platinum 4f and aluminum 2p
peaks in XPS [8].
A previous Auger electron spectroscopy (AES) and
XPS investigation has shown that the surface of an
antimony-doped tin oxide film prepared in the described
manner may contain a number of surface contaminants [17].
Among these surface contaminants are carbon, chlorine,
potassium, sodium, calcium and sulfur in varying amounts.
Argon ion bombardment and high temperature oxygen
treatments have proven to be effective in removing this
surface contamination, but the effects on the tin oxide
surface (see Section III) and platinum oxidation state
(as shown below) are substantial. Hence, an understanding
of surfaces such as those being used in electrochemical
studies [6,28] may require analysis in the presence of
several types of surface contamination.
Two different techniques are used in the present
study to prepare the supported Pt. The first of these
techniques is electrochemical in nature. In essence,
the platinum is plated from a 5x10-5 M solution of H2PtCl6
buffered at a pH of 6.8. The process is carried out
for varying amounts of time at -0.5 V versus SCE. The
second technique utilizes a chemisorption mechanism.
The substrate is pretreated by exposure to a 10 M NaOH
solution at 900C for 30 minutes. The pretreated substrate
is washed in distilled water and then exposed to an 800C
solution of 0.01 M KOH and 500 ppm Pt (IV) from Na2Pt(OH)6.
The Pt loading is dependent on the exposure time for
both preparation techniques. Regardless of the procedure
used, the samples are washed in distilled water after
platinization and solvent cleaned before mounting in
the vacuum system.
An important consideration in the preparation of
these supported platinum catalysts is the relative rate
of crystallite nucleation to growth. The indications
are that the growth of crystallites is favored over
nucleation in the electrochemical preparation [29]. The
chemisorption technique, however, has been shown to be
capable of producing highly dispersed (>90%) platinum
at low loadings [6]. The alkaline pretreatment is believed
to hydroxylate the surface, thereby increasing the number
of active sites available for Pt chemisorption [6,19].
Electron beam effects on these surfaces can be
dramatic. The removal of carbon, chlorine, oxygen,
hydrogen and sodium by electon-stimulated desorption
(ESD) has been observed previously [17-19]. These beam
effects somewhat limit the usefulness of AES as an
analytical tool on these surfaces making XPS the preferred
technique because of its less destructive nature. However,
an understanding of the ESD phenomenon in terms of an
interatomic Auger decay model [30] shows promise in helping
to unravel the chemistry of these surfaces (see Section
V and ref. 19).
All XPS spectra were collected with a Physical
Electronics double-pass CMA using Mg Ka X-rays as an
excitation source. A pass energy of 50 eV (AE/E = 0.014)
13
was used throughout. All binding energies are referencecd
to the tin 3d 5/2 peak at an assumed energy of 486.4
eV [31]. It has been shown that there is no change in
this core level binding energy for the different oxides
of tin [32-34] making this peak an excellent reference.
The base pressure in the vacuum system for this study
was lx10-9 Torr. Details of the vacuum system have been
given previously [17].
Results and Discussion
Figure 2-1 shows the Pt 4f XPS spectrum of an
electrochemically platinized sample. The plating process
was carried out for 40 minutes at -0.5 V versus SCE.
A fairly high platinum loading of about 40 pg/cm2 is
obtained by this process as estimated from a previous
Rutherford backscattering (RBS) study [35]. Deconvolution
of the spectrum reveals the presence of two platinum
species. In the assignment of these features platinum
chloride species are neglected. A check for surface
chlorine contamination using standard sensitivity factors
[31] showed the concentration to be low (Cl/Pt < 0.1).
The lowest binding energy doublet in Figure 2-1
has the 4f 7/2 peak at 71.2 eV and the 4f 5/2 peak at
74.5 eV. These features are assigned to Pt metal in
agreement with the work of Katayama [5,25]. The binding
energy reported here is about 0.4 eV higher than that
77 76 75 74 73 72 71 78
BINDING ENERGY CEV)
Figure 2-1. Pt XPS spectrum of an electrochemically plati-
nized substrate. The lower binding energy doublet is
characteristic of Pt metal, and the higher binding energy
doublet is characteristic of a Pt species chemically bonded
to the tin oxide substrate.
generally reported for the bulk Pt metal [31]. This
observation of a higher binding energy for supported
clusters crystallitess) over that of bulk metal is in
agreement with general expectations [36]. This shift
is most likely the result of differences in the reference
levels (work functions) of the bulk metal and the tin
oxide support and/or a decrease in the final-state
extra-atomic relaxtion energy as a result of the change
from bulk metal to small cluster [37].
The doublet at higher binding energies in Figure
2-1 has a 4f 7/2 peak at 72.3 and a 4f 5/2 peak at 75.5
eV. This doublet is shifted about 1.1 eV above Pt metal
and about 0.5 eV below the position expected for a Pt(OH)2
species [5,25,31]. The higher oxides of platinum all
fall to significantly larger binding energies which removes
them from consideration (see below). Examination of
a second substrate electrochemically plated for only
1/4 the time (i.e. 10 minutes) gives a Pt 4f spectrum
(not shown) characterized primarily by this high binding
energy species observed in Figure 2-1. These results
suggest that the higher binding energy doublet in Figure
2-1 may be associated with a platinum species directly
bonded to the tin oxide substrate. Further, the appearance
of Pt metal at longer plating times demonstrates that
this substrate-bonded Pt species acts as a nucleation
site for the growth of metallic crystallites by
electrochemical deposition. These results are consistent
with a model of nucleation and crystallite growth suggested
by earlier work on the electrodeposition of platinum
on tin oxide [29].
Figure 2-2 shows the Pt 4f peaks for a sample prepared
by the chemisorption technique. The pretreated tin oxide
substrate was exposed to the 800C Na2Pt(OH)6 solution
for one hour. The platinum loading is approximately
2 .g/cm2. Because the Pt loading is small, the
signal-to-noise ratio does not justify a spectrum
deconvolution. However, the use of in situ chemical
and thermal treatments allows a manipulation of the Pt
valence state for a more complete determination of the
supported species. Because only a trace of chlorine
was detected, the possibility of platinum chloride species
was again discounted.
Figure 2-2a shows the spectrum obtained immediately
after pumpdown. The position of the doublet indicates
that the substrate-bonded species is the predominant
form of platinum obtained by the chemisorption procedure.
However, the peak widths also suggest the presence of
small amounts of Pt metal at lower binding energies and
Pt(OH)2 at slightly higher binding energies. The
observation of the substrate-bonded species as the primary
form of platinum is consistent with earlier work showing
nucleation is preferred over crystallite growth during
chemisorptive platinization [6].
' I LU I I I 1 1 1f 1 111 I i
PT 4F
PT
PT(OH)m J
PT-O-SN
'I 'l l i I I r i I I I, I I i,
BINDING ENERGY (EV)'
Figure 2-2. Pt XPS spectra of a sample prepared by plati-
num chemisorption. Spectrum (A) is obtained immediately
after pumpdown, (B) after high temperature oxidation,
(C) after high temperature reduction and (D) after a
high temperature anneal in vacuo.
Figure 2-2b shows the effect of a high temperature
(6000C) in situ oxidation in 11 Torr of 02 for 30 minutes.
The oxygen treatment shifts the Pt 4f XPS peaks to higher
binding energies. The presence of the higher oxides,
PtO and Pt02, is clearly indicated by structure on the
high binding energy side of the specturm. The PtO features
are shifted approximately 2.9 eV higher with respect
to Pt metal while the PtO2 features are shifted about
3.9 eV [31]. Though there is no evidence of Pt metal
in Figure 2-2b, the shoulder at 72.3 eV is clear evidence
of the persistence of the substrate-bonded Pt species.
The observation of this species in conjunction with PtO
and Pt02 confirms that the substrate-bonded species is
not simply a stoichiometric platinum oxide.
A 5000C in situ reduction in lxl0-3 Torr of H2 for
30 minutes results in the spectrum shown in Figure 2-2c.
An XPS inspection of the Sn 3d core levels shows no sign
of a reduction of the substrate to bulk Sn metal. However,
the PtO and PtO2 species observed in Figure 2-2b have
undergone a complete reduction leaving primarily Pt metal.
The loss of the higher oxides coupled with the appearance
of a Pt 4f 7/2 peak at 71.2 eV confirms the earlier
assignment of this binding energy to Pt metal. Evidence
of the substrate-bonded species is also found in Figure
2-2c in the form of a shoulder on the high binding energy
side of the 4f 5/2 peak.
19
A subsequent 8000 anneal in vacuo has little effect
on the XPS peak positions as shown by Figure 2-2d. The
platinum remains primarily in the metallic state with
a small contribution due to the substrate-bonded species.
The presence of this substrate-bonded species after high
temperature annealing confirms that these features are
not due simply to a platinum hydroxide or hydrate species.
Decomposition or dehydration of such species would be
expected at significantly lower temperatures.
Before the oxidation-reduction cycle the platinum
from the chemisorption preparation is largely present
in the substrate-bonded form. This observation suggests
a high dispersion as found by Watanabe et al. for samples
prepared in a similar fashion [6]. The high temperature
oxidation-reduction cycle results in a sintering of the
supported species as shown by the large fraction of Pt
metal in Figure 2-2c. The remaining presence of the
substrate-bonded species suggests that a fraction of
these species acts as nucleation sites for the crystallite
growth as was observed in the electrochemical platinization
process.
The constancy of the XPS peak positions for the
substrate-bonded species obtained by either the
electrochemical or chemisorption process indicates a
similarity in the species formed regardless of the
20
procedure used. It has been shown above that this species
may not be identified as simply a PtOx or Pt(OH)y species.
Likewise, the binding energy shift of this species with
respect to Pt metal is not consistent with that observed
in the formation of Pt-Sn alloys [38]. These observations
suggest that the bond formed with the surface occurs
through surface lattice oxygen. The formation of a Pt-O-Sn
substrate-bonded Pt species is postulated. The tenacity
displayed by this species in resisting complete reduction
by chemical and thermal treatments is characteristic
of a species exhibiting such a strong interaction with
the substrate. Komiyama et al. have observed a similar
resistance to reduction by ion bombardment of strongly
interacting rhenium species on iron oxide [39].
Previous work on samples prepared by the chemisorption
technique offers insight into the mechanism of formation
of the substrate-bonded species. Watanabe et al. Have
shown that an alkaline pretreatment of the substrate
prior to platinization results in an increased Pt uptake
[6]. It is believed that the pretreatment hydroxylates
the surface and provides an increased number of active
chemisorption sites for the platinum species in solution.
Earlier studies using secondary-ion mass spectrometry
(SIMS) [40] and ESD [19] lend support to the surface
hydroxylation model by showing significant increases
in surface hydrogen and oxygen after the alkaline
pretreatment. Platinum chemisorption is believed to
occur by replacement of the proton on the surface hydroxyl
group with the loss of a coordinated ligand from. the
platinum solution species. Under the pH conditions used
for chemisorption from a H2PtCl6 solution, the
chloroplatinate undergoes hydrolysis resulting in the
replacement of two chlorines by hydroxyl groups.
Chemisorption should occur via
Sn-OH + Pt(OH)2C142 ----> Sn-O-Pt(OH)C142- + H20
with a subsequent dehydroxylation and loss of chlorine
from the surface complex. For chemisorption from an
alkaline solution of Na2Pt(OH)6 the surface hydroxyl
group is ionized through the loss of the acidic proton.
Chemisorption is expected to occur via
Sn-O- + Pt(OH)62- ----> Sn-O-Pt(OH)52- + OH-
leaving the substrate-bonded species after dehydration
of the surface complex.
Conclusions
XPS has been used to study tin oxide supported
platinum prepared by electrochemical and chemisorption
techniques. Features in the Pt 4f spectrum have been
assigned to a species chemically bonded to the substrate.
The position of these features is independent of the
preparation technique used. In situ chemical and thermal
treatments confirm that this substrate-bonded platinum
is not simply a PtOx or Pt(OH)y species. The platinum
is believed to bond through surface lattice oxygen giving
a Sn-O-Pt surface species. High temperature reduction
results in a sintering of these species, but the inability
to completely reduce the platinum is indicative of the
strong chemical interaction between the platinum and
tin oxide.
A model for the chemisorption of platinum on tin
oxide is proposed. Surface hydroxyl groups are believed
to be the active chemisorption sites for platinum species
in solution. The chemisorption process is believed to
occur through the replacement of the hydroxyl group proton
with the loss of a coordinated ligand from the platinum
species.
SECTION III
AN ELECTRONIC AND STRUCTURAL INTERPRETATION
OF TIN OXIDE ELS SPECTRA
Introduction
The spectroscopic study of tin oxide surfaces is
complicated by the difficulty in distinguishing between
the two oxides of tin, SnO and SnO2. Several x-ray
photoelectron spectroscopy (XPS) studies have failed
to detect any changes in core level binding energies
between SnO and SnO2 [32-34]. Similar problems are
encountered using Auger electron spectroscopy (AES) where
no significant differences in kinetic energies or line
shapes are found [41]. As expected, however, the valence
band spectra of the two oxides do differ. An ultraviolet
photoelectron spectroscopy (UPS) study of tin oxidation
by Powell and Spicer [42] and a valence-band XPS study
by Lau and Wertheim [32] have shown these differences,
but interpretation difficulties associated with analysis
depth have proven to be substantial.
Electron energy-loss spectroscopy (ELS) is a technique
which offers flexibility of analysis depth and is sensitive
to changes in the valence band density of states. Powell
[41] has shown that ELS may be used to distinguish between
24
the two oxides of tin and has given a preliminary
interpretation of the spectra in terms of differences
in plasmon frequencies. For SnO2 a main loss feature
at 19.5 eV was identified while for SnO a main loss feature
was found at approximately 13.5 eV. A combined UPS
and high-resolution electron energy-loss spectroscopy
(HREELS) study of 3% Sb doped and undoped SnO2 has shown
the room temperature occupied conduction bands to be
very free-electron like [43] in agreement with a bulk
tight-binding band structure calculation [44]. For the
heavily doped sample an HREELS loss feature at 0.55 eV
was found. Based on the experimentally determined carrier
concentration and effective mass ratio, the 0.55 eV loss
feature was identified as a surface plasmon loss associated
with conduction band electrons from Sb donors. Since
valence band and core level electrons in Sn02 are not
free-electron like, the higher energy ELS losses in the
present study are not assigned to plasmon losses.
While an interpretation of the ELS spectrum would
be useful for distinguishing between the two oxides of
tin, an additional benefit may be derived due to the
usefulness of ELS measurements in the interpretation
of electron-stimulated desorption (ESD) threshold studies.
It has been shown that core level transitions can be
correlated with desorption thresholds and may specify
adsorbate binding sites [30,45,46]. In particular, the
ability to distinguish between transitions from Sn 4d
and 0 2s core levels which cannot be resolved using XPS
could be most useful in understanding the chemistry of
tin oxide surfaces.
Experimental
The polycrystalline tin oxide films used in this
study were prepared by spraying a solution of 3 M SnCl4
and 1.5 M HC1 onto a titanium foil maintained at 5000C
in air. Unlike the samples used in Section II and in
previous studies [17-19], a high purity (99.998%) anhydrous
SnCl4 reagent was used. The resulting samples were found
to have significantly less surface contamination. Trace
chlorine and carbon contamination was found to be removed
quickly in situ by heating at 5000C in 10 Torr of oxygen
for about 5 minutes. This procedure gave a clean oxide
surface as determined by AES.
The samples were annealed in vacuo initially and
were heated briefly and allowed to cool before each
individual measurement. Using angle-resolved ultraviolet
photoelectron spectroscopy (ARUPS) on an ion-sputtered
SnO2 (001) single crystal surface, Gobby [47] has shown
that the annealing process (5500C to 8350C) strengthens
the primary emission from the valence bands and increases
the sharpness and magnitude of the anisotropic emission
indicating a well ordered crystal. Similar annealing
effects in the sharpness and magnitude of ELS spectra
and on the magnitude of core level emission in XPS have
been observed for polycrystalline tin oxide samples in
this study.
All spectra were collected with a double-pass CMA.
Details of the vacuum system have been published previously
[17]. The ELS data were taken in the retarding (N(E))
mode to allow a comparison with the data of Powell [41].
All ELS spectra were collected with a pass energy of
25 eV ( A E/E = 0.014) with the exception of the 50 eV
primary beam measurement. This spectrum was recorded
in a nonretarding (EN(E)) mode to suppress the large
signal from secondary electrons at near zero kinetic
energies. All ELS spectra were collected using 100 nA
beam currents and pulse counting detection. The XPS
spectra were taken using a Mg K a source and a 50 eV
analyzer pass energy. The base pressure in the vacuum
system for this study was 1 x 10-10 Torr.
Background
For energy losses of the magnitude of electronic
excitations, the inelastic scattering event may be
described in terms of optical (dipole) selection rules
in cases where the primary electron energy is high enough
to justify the Born approximation. It is generally thought
that primary electron beam energies above 100 eV to 200
eV satisfy this criterion [48-51].
27
Consider a primary electron of momentum h K scattered
inelastically into a state h K' resulting in an interband
4.
transition between one-electron states, Ik,l> ---->
k',l'>. Momentum conservation requires AK = k' k
+ -> ->
+ G where AK E K K' and G is a reciprocal lattice
vector. Energy conservation requires h2( K 12- 1K' 12)
= 2m( Ek',l'-k,l) where Eq,l is the eigenenergy of the
4.
one-electron state q,l>. Not only must energy and
momentum be conserved, but the matrix element
+ -+ +
T =
must be nonzero [48,51,52]. Expansion in powers of
.* 4.
(AK*r) yields the selection rules. It has been shown [48,51]
that the monopole term vanishes due to orthogonality
and that retaining only the dipole (linear) term gives
T = i AK < k + AK, 1' r k, 1>
For small AKI Rudberg and Slater [48] have shown a fair
approximation at small energy losses or large IKI may
be obtained by considering only direct transitions, k
= k'. Hence, in the regime where the Born approximation
applies the selection rules are essentially optical in
nature. To a first approximation, the energy dependence
of the loss spectrum should be similar to that measured
in optical absorption [49]. Since the momentum transfer
-4
in the ELS transition, h A K, may be different than in
the optical process, it is expected that a broadening
in the energy dependence of the ELS features will occur
with respect to the optical features [48].
Because the results from this study are for
polycrystalline samples with a random grain size which
is small compared to the excitation volume, the orientation
of K with respect to the crystal axes may be assumed
to be random. Therefore, the ELS spectra presented here
represent an average over the entire Brillouin zone.
The present results should be most comparable to optical
absorption studies of polycrystalline samples.
It should be mentioned that a breakdown in dipole
selection rules is possible for low beam energies and
large energy losses. In this case the expansion of the
phase factor, exp(i A K *r), must be carried to the
quadratic (quadrupole) term to obtain an accurate
description. Ludeke and Koma [50] and Colavita et al.
[51] have taken advantage of this effect to identify
loss features due to quadrupole-allowed transitions between
dipole-unallowed states. No such identifications have
been made in the present work.
Using a generalization of the joint density-of-states
function for optical interband transitions which includes
finite momentum changes, Ludeke and Esaki [53] have shown
that the energy-loss distribution due to transitions
from narrow, filled initial states to empty conduction-band
final states may be proportional to the conduction-band
(CB) density of states. This density-of-states
interpretation requires the initial state to be isolated
with no additional scattering channel existing near the
same energy loss. An additional complication may arise
if there is a significant modulation of the scattering
cross section due to a partial filling of the conduction
bands from a competing scattering channel originating
from a different initial state. In spite of these problems
it should be possible to obtain some picture of the CB
density-of-states in tin oxide if the Sn 4d and O 2s
core levels couple to final states of significantly
different energy.
Results and Discussion
Figure 3-1 is an energy level diagram depicting
the band structure of SnO2. The character of the
electronic states in the valence and lower conduction
bands is due to Robertson [44]. The assignment of Sn
4f character to high lying conduction band states is
due to Gobby [47]. The width of the valence bands and
the location of the three major features therein are
from the available photoemission data [32,47]. The
position of the O 2s and, Sn 4d core levels are from XPS
measurements made in this laboratory with no attempt
SN 4F
SN5P 02P
SNSS
02P LONE PAIR
MIN. BONDING 02P
02P SNSS BONDING
02S SN4D
Figure 3-1. Energy level diagram representing the SnO2
band structure. The locations of the major occupied
(unoccupied) valence and core (conduction) states involved
in the energy loss spectrum are shown. The approximate
locations of the SnO VBM and CBM are indicated by dashed
lines.
25
-22
18
-12
9.5
3.6
-2
-4
--7.5
--9.5
(EV)
"-23
31
at deconvolution. Photoemission results [47] were used
to locate the states in the conduction bands which couple
strongly to various valence and core states as discussed
below. The cut-off position at the top of the conduction
bands was determined form the ELS spectrum in Figure
3-2 based on the interpretation of the high-energy loss
features given below. While all the states are represent
by single horizontal lines, some are quite broad and
may extend over 5 eV or more. The dashed lines in the
band gap and lower conduction bands represent the
approximate location of the SnO valence-band maximum
(VBM) and conduction-band minimum (CBM) respectively.
These assignments are due to photoemission results for
SnO [32] and optical absorption on highly defect laden
tin oxide films [54].
Annealing Effects
Figures 3-2 to 3-4 show the ELS data for a sample
annealed at 7500C in vacuo. Each of these spectra were
recorded for a normal incidence primary beam of specified
energy, Ep. The annealing process was carried out until
the background chamber pressure went through a clear
maximum (about 45 minutes). Giesekke et al. [55] have
shown using thermogravimetric analysis and electron
diffraction that the decomposition of tin (IV) hydroxide
proceeds through four distinct crystalline hydrogen
containing compounds before yielding SnO2 above 6000C.
The observed pressure maximum during the annealing process
is indicative, in part, of this dehydration. The
hygroscopic nature of tin oxide and the study of hydrated
surfaces is discussed in Sections IV and V.
Figure 3-2 is the loss spectrum for a 1500 eV primary
beam. This spectrum may be divided into two parts; the
higher energy loss features above about 28 eV and the
features at lower energy losses. The lower half of the
spectrum consists of two major features at 19.5 eV and
13 eV in agreement with the SnO2 spectrum reported by
Powell [41]. Additionally, extrapolation of the linear
portion of the leading edge of the loss spectrum to the
baseline gives a minimum energy loss of 3.6 eV. This
value is equal to the best available optically determined
band-gap energy for SnO2 single crystals [56,57] and
the calculated lowest energy direct-allowed one-electron
transition ( Fr ----> F ) found by Robertson [44]. Using
constant-intial-states (CIS) ARUPS measurements and
angle-integrated UPS for SnO2 (001), Gobby [47] has shown
that VB-to-CB transitions are dominated by excitations
form an initial state about 1.5 eV below the VBM to final
states near 10 eV, 13 eV and 19 eV to 22 eV higher in
energy as shown in Figure 3-1. Inspection of Figure
3-2 reveals a shoulder in the loss spectrum near 10 eV
as well as the two higher energy features. This 10 eV
loss feature also corresponds to a collection of VB-to-CB
VACUUM
ANNEAL
7SC
t rT t
46.1 36.2 19.5 3.6
v 12.9
z
ENERGY LOSS
SPECTRUM
Ep= 158 EV
NORMAL INCIDENCE
68 58 48 38 28 18
ENERGY LOSS (EV)
Figure 3-2. N(E) ELS spectrum of tin oxide after a high
temperature vacuum anneal. The high primary beam energy
(Ep = 1500 eV) at normal incidence results in primarily
a bulk sensitivity. The indicated features are
characteristic of well-annealed SnO2. The low energy
features are due to VB ---> CB transitions, and the high
energy loss features are core ---> CB transitions.
dipole-allowed transtions at the r point in the Brillouin
zone for bulk SnO2 as found by Robertson. It is concluded
that the lower energy loss features in Figure 3-2 are
due to collections of optically (dipole) allowed interband
(VB -> CB) transitions.
The loss features above 30 eV are strongly dependent
on the thermal history of the sample and are dominated
by core-to-conduction-band transitions from tin 4d and
oxygen 2s levels. Gobby [47] has shown that these core
levels couple to final states in two energy regimes.
Coupling to CBs which are 32 eV to 36 eV above the core
level is observed easily in UPS while coupling to the
lower CBs (the CB minimum lies approximately 26.6 eV
above the core levels) is not observable due to the
photoemission threshold and large background of secondary
electrons. At higher photon energies coupling to CBs
37 eV and higher relative to the core levels is observed.
This coupling begins to strengthen at 40 eV above the
core level, but higher energies were not used because
of a lack of photon intensity. However, a higher energy
CB final state was identified for an initial state feature
in the lower VBs. This final state falls about 45 eV
above the core level, and Gobby suggests that it is a
Sn 4f derived state (see Figure 3-1). In Figure 3-2
a range of energy-loss features from about 29 eV to 48
eV are visible. The strongest features fall near 36
35
eV and 46 eV in excellent agreement with the photoemission
results of Gobby.
On the basis of the similarities between the
photoemission results for single crystal SnO2 and the
energy-loss spectrum, Figure 3-2 is interpreted as being
characteristic of a well-annealed (though polcrystalline)
SnO2 material. Additionally, these similarities support
the conclusion that the main features observed in the
ELS spectrum are due to single inelastic events possibly
in conjunction with elastic scattering events. Because
of the long mean free path of electrons near 1500 eV,
the spectrum in Figure 3-2 (Ep = 1500 eV) is primarily
due to contributions from the bulk of the material.
Figure 3-3 shows the effect on the loss spectrum
of varying the primary beam energy from 1500 eV to 200
eV. Figure 3-4 shows the EN(E) loss spectrum for a 50
eV primary beam. Decreasing the beam energy decreases
the analyses depth due to a reduction in the electron
mean free path with kinetic energy. The set of spectra
in Figures 3-3 and 3-4, therefore, represent a depth
profile of the vacuum-annealed tin oxide material. In
Figure 3-3 the main change in the valence band region
is seen to be a growth of the 12 eV to 13 eV feature
relative to the 19 eV feature with decreasing beam energy.
This change is most apparent in Figure 3-4 where a feature
near 13 eV dominates the spectrum. Changes in the core
Ep = 1588 EV
Ep 188 EV
Ep = 688 EV
Ep = 488 EV
Ep = 200 EV
NORMAL INCIDENCE
i l 1 111111 11 l lllll l lilllil l ,,,ijji ij jjjjjjjjj jjjj
68 58 48 38 20 10
ENERGY LOSS (EV)
Figure 3-3. Variation in the N(E) ELS spectrum with
primary beam energy, Ep. The set of spectra represent
a depth profile of the annealed material. The growth
of the 27 eV feature is due to an increasing oxygen
deficiency as the spectra become more surface sensitive.
VACUUM
ANNUAL
758
111111111 111-[H 11 111iil 11 1111" 1 1ll ,,llllTl' l jTll If
48 38 28 18
ENERGY LOSS (EV)
Figure 3-4. EN(E) ELS spectrum for Ep = 50 eV. The
main loss feature at 13 eV shows that the annealed material
is essentially SnO at the surface.
z\
z
W
1 111111 1 I ll ff ll rrr i 1 1 1 1 11 11 1 1 1,1 l I I fT
VACUUM
ANNEAL
758C
Ep = 58 EV
13
NORMAL INCIDENCE
27
l l ll,, i lli, ,i l , ,i ,l ,i ,l l
------------
"
38
level region are more dramatic. The core level losses
may be resolved into two features. The large loss feature
at 46 eV is seen to decrease rapidly with beam energy
leaving a separate feature near 36 eV. Concurrent with
the loss of the 46 eV feature, the growth of a feature
at 27 eV is observed.
By comparison to the work of Powell [41], the changing
valence-band derived features in Figures 3-3 and 3-4
may be loosely interpreted as a change in the tin oxide
from a SnO2 compound in the bulk to a more SnO-like
material at the surface. Because the SnO-like feature
near 13 eV dominates the spectrum only for a 50 eV primary
beam, it appears that such a material exists in the near
surface region, possibly in the top few atomic layers.
This interpretation is reasonable in view of the well
documented oxygen loss from tin oxide surfaces during
high temperature annealing [47,58,59]. Such oxygen losses
have been observed frequently in this laboratory.
Decreases in surface O/Sn ratios from near 2 down to
1 on annealing have been monitored with AES and XPS.
The interpretation of changes in the core-to-CB
region of the spectrum leads to the same conclusion as
derived from the VB-to-CB features, but some discussion
of the symmetry of the initial and final states involved
is required. The band structure calculations of Robertson
[44] and Munnix and Schmeits [60] as well as the ARUPS
measurements of Gobby [47] show that the Sn02 valence
bands are mostly 0 2p like with only a small admixture
of Sn derived states. The lower conduction bands are
primarily Sn 5s and 5p like, and within 3 eV to 4 eV
of the CBM these states are 90% Sn 5s like [441. To
a first approximation the atomic character of these states
suggests that the electronic structure of SnO2 may be
considered to be ionic. Within this ionic approximation
the electronic configurations of the atomic and
stoichiometric oxide systems are those given Table 3-1.
For SnO2 the highest occupied states are oxygen
2p like, and the lowest unoccupied states are tin 5s
like in basic agreement with the band structure
calculations. Reduction of SnO2 to SnO populates the
Sn 5s states leaving the lowest unoccupied states more
Sn 5p like. Likewise, the removal of oxygen from SnO2
to form a nonstoichiometric oxide should result in a
mixing of Sn 5s states into the valence bands (possibly
as defect states) leaving a more Sn 5p like CBM. Such
a variation in symmetry near the conduction band minimum
should be apparent in the energy-loss spectrum. In
particular, a Sn 4d core-to-CBM transition will be dipole
unallowed for a Sn 5s dominated CBM, but dipole allowed
(Al=l) for a Sn 5p like CBM. Hence, the loss of oxygen
from Sn02 should result in a change in the Sn 4d
core-to-CBM transition from unallowed to allowed. Notice
Table 3-1. Electronic Configurations of Atomic
Tin and Oxygen and the Stoichiometric Oxides
Atomic Tin
Atomic Oxygen
Stannous Oxide, SnO
Sn : [Kr] 4d10 5s2 5p2
00 : is2 2s2 2p4
Sn2+ : [Kr] 4d10 5s2 5p
02- : 1s2 2s2 2p6
Sn4+ : [Kr] 4d10 5s 5p
02- : Is2 2s2 2p6
Stannic Oxide, SnO9
in Figure 3-1 that little change is expected in the energy
of the CBM between SnO2 and SnO.
The changing nature of states near the tin oxide
CBM may be seen clearly in Figures 3-3 and 3-4. The
46 eV loss feature may be interpreted as a transition
from the Sn 4d core to a high lying Sn 4f-like CB state
[47]. The 27 eV feature may be interpreted as a Sn 4d
core-to-CBM transition [61]. A feature near 27 eV has
been observed in the N(E) loss spectrum for both SnO
and Sn metal [41,62] but not for SnO2. For the case
of metallic tin, this feature may be viewed as a transition
from the Sn 4d core to empty states above the Fermi level.
The growth of the 27 eV loss feature in conjunction with
the decrease in the 46 eV feature may be interpreted
as a change in the CBs. The growth of the 27 eV loss
feature with decreasing beam energy is characteristic
of the changing nature of the CBM due to a deficiency
of oxygen in the surface region. This interpretation
is supported by the relative strengths of the two
transitions. The d ----> f transition is expected to
be stronger than the d ----> p transition [63].
The insensitivity of the 36 eV loss feature to
incident beam energy relative to the Sn 4d core features
discussed above suggests that the atomic origin of this
core derived feature is, significantly different. From
Figure 3-3 the main change in this feature is a gradual
decrease in intensity with decreasing beam energy. The
assignment of this 36 eV loss feature to an 0 2s core-to-CB
transition can explain this trend for a material exhibiting
a decreasing oxygen concentration on moving from the
bulk to the surface. This is precisely the situation
encountered in the present case.
Because the 0 2s and Sn 4d core levels couple to
CB final states of significantly different energy, the
energy-loss distribution due to these transitions may
be viewed as approximately proportional to the CB density
of states over a narrow range. There is certainly some
overlap between the 0 2s and Sn 4d transitions in the
neighborhood of the 0 2s feature. At the extremes,
however, near the 46 eV or 27 eV feature the
density-of-states interpretation should be valid although
substantial matrix element differences are likely between
these two regions. The observation from Figure 3-3
that changes in the core level features at high beam
energies are more dramatic than in the VB loss features
suggests the Sn 4d core features are more sensitive to
low concentrations of crystal structure defects than
the VB features. The Sn 4d core level losses may be
viewed as an strong indicator of the structural order
of the tin oxide material. Supporting evidence is found
from results on ion-sputtered samples.
Sputtering Effects
In order to increase the surface sensitivity of
the ELS measurement, the sample orientation was changed
to give the coaxial electron beam from the CMA an incident
angle of 450 with respect to the sample normal. This
change allowed reasonably surface sensitive measurements
with higher primary beam energies, and it eliminated
the problem of low-energy secondary electrons inherent
in the use of low electron beam energies (50 eV) for
N(E) measurements. Also, the probability of encountering
additional quadrupole-allowed features was minimized.
Assuming a straight line incident and exit path for an
electron scattered through a nominal angle of 137.70
(fixed by the CMA [64]), a very crude estimate of the
ELS analysis depth based on sample orientation and electron
mean free path can be made. Since the main features
observed in the ELS spectrum are due to single inelastic
events possibly in conjuction with elastic events, a
total path length of twice the mean free path of an
electron at the primary beam energy seems appropriate.
0
These assumptions lead to an estimate of 5 to 10 A (2
0
to 4 atomic layers) at Ep = 200 eV and 15 to 20 A (5
to 7 atomic layers) at Ep = 1500 eV. These estimates
should be viewed as qualitative at best.
Figure 3-5 shows the ELS data for a sample annealed
at 600C. The spectrum for Ep = 1500 eV shows the
structure characteristic of a well annealed Sn02 bulk
Ep = s158 EV
V Ep = 280 EV\
18 13
688C ANNEAL
45 DEGREE INCIDENCE
\lil I\ill\ \i il\ l\lil lil\ il l iil i li il iilll i n1 1 l ifillim
68 58 48 38 28 18
ENERGY LOSS CEV)
Figure 3-5. ELS spectra for a sample annealed at 6000C.
The angle of incidence for the primary beam is 450. With
Ep = 1500 eV it is seen that the bulk is primarily SnO2.
The Ep = 200 eV spectrum shows SnO at the surface and
evidence of a defect structure between bulk SnO2 and
the surface.
1111 11111IHI 11 1111m111111111 1 Im IIiiii II11 1im
material. The more surface sensitive spectrum for Ep
= 200 eV has a sharp structure near 13 eV which is
characteristic of SnO [41]. The broad feature centered
at 18 eV is not characteristic of either SnO or SnO2,
and it most likely comes from a subsurface
nonstoichiometric defect structure accompanying the change
in structure from SnO2 in the bulk to SnO at the surface.
The 27 eV feature is also present indicating a structure
which is oxygen deficient in comparison to SnO2.
Figure 3-6a is the valence band XPS spectrum for
the 6000C annealed sample. The resolution of the VB
XPS data is seen to be poor. This poor resolution is
due to a combination of very low signal intensity, the
x-ray line. width, x-ray satellite emission from Sn 4d
and O 2s core levels and data smoothing. In spite of
these difficulties, the general shape of the VB emission
is similar to that for SnO2 as found by Lau and Wertheim
[32]. The obvious lack of surface sensitivity in this
measurement is not unexpected. Because the kinetic energy
of the valence band photoelectrons is large ( > 1200
eV), the mean free path is correspondingly large.
Additionally, the sample orientation is such that the
angle between the surface normal and the cylinder axis
is very nearly equal to the nominal 42.30 acceptance
angle of the CMA [64]. Since photoemission from
polycrystalline materials is expected to peak at the
(B)
O/SN a 8.96
0/SN U 1.38
VALENCE BAND XPS
i l l I I L I I I 1 1 I I f I |
iS 18 5 0
BINDISG EERGY (EV)
Figure 3-6. Valence band XPS spectra after a (A) 6000C
anneal and (B) 2 KeV argon-ion bombardment. The spectra
are more bulk than surface sensitive.
I I I I I I I I I I I P1 I I II I I I 1 1I
surface normal, the VB XPS results shown here have their
largest contribution from high energy electrons at near
normal emission. Hence, Figure 3-6a is primarily due
to the bulk SnO2 material.
The annealed sample characterized by Figures 3-5
and 3-6a was ion sputtered with 2 KeV argon ions. Figure
3-6b illustrates the change in the VB XPS spectum. AES
and core level XPS show no evidence of a reduction to
metallic tin in this particular case, but the preferential
sputtering of oxygen is demonstrated by a drop in the
O/Sn ratio. Ion sputtering introduces a shoulder on
the VB emission near a binding energy of 2 eV to 3 eV.
A similar feature has been observed in ARUPS and
interpreted as emission from defect states associated
with a deficiency of oxygen [47]. Interestingly, this
sputter-induced feature lies near the same binding energy
as the highest lying SnO VB feature [32]. There is
even a fair correspondence between the SnO VBM as found
by Lau and Wertheim and the low binding energy edge of
the defect emission.
The assignment of the shoulder in Figure 3-6b to
defect states rather than SnO is justified by the ELS
spectra for the sputtered sample in Figure 3-7. The
more surface sensitive spectrum, Ep = 200 eV, shows a
broadening of the characteristic SnO feature at 13 eV.
The entire valence band portion of the spectrum becomes
48
broad and relatively featureless as a result of sputtering.
This broadening may be interpreted as a change from the
SnO structure at the surface to a more amorphous structure
caused by sputtering. For Ep = 1500 eV the loss spectrum
is sensitive to the bulk within the region probed by
the VB XPS measurements. While there is some broadening
and a small shift toward lower energy losses, the VB
features are still very much SnO2 like in approximate
agreement with the VB XPS spectrum shown in Figure 3-6b.
The core level loss features reflect the defect presence
much more strongly than the VB loss features. The absence
of the 46 eV feature and prominence of the 27 eV feature
confirm the change from a well-annealed SnO2 structure
to a more oxygen-deficient defect structure after
sputtering.
The damage induced by ion sputtering is heaviest
in the top few layer's of the solid as illustrated by
Figure 3-7. Thus, the amorphous structure at the surface
implied by the valence band features for Ep = 200 eV
is not unexpected. Sputtering damage in layers deeper
in from the surface may result from ion implantation,
knock-in and other ion-matrix phenomena, but the damage
in these deeper layers should be significantly less than
near the surface. Bearing in mind that ion bombardment
effects become less apparent as the experiment becomes
more bulk sensitive, a comparison of the results between
Ep = 1588 EV
18
45 27 13
I =
9Ep = 288 EV
z
2 KEV ARGON ION BOMBARDMENT
45 DEGREE INCIDENCE
ll Ii I I1 I iI l i i il i [ 1 11l1 1 1i 1il lii ll i il1 i 11 1 ili ii
68 58 48 38 28 18
ENERGY LOSS (EV)
Figure 3-7. ELS spectra following a 2 KeV argon-ion
bombardment. The valence band features show the bulk
to be SnO2 like while the core features reveal a signifi-
cant concentration of defects. The VB features in the
more surface sensitive spectrum illustrate an amorphous
structure at the surface due to a sputtering.
annealed and sputtered samples suggests that a significant
amount of qualitative structural information may be gained
from the N(E) energy-loss spectrum. The width and
center-of-gravity position of the valence band features
can be used as a gross indicator of the tin oxide
structure. A matrix characteristic of a stoichiometric
form of tin oxide is suggested by sharper, more well
defined valence band loss features near 19.5 eV for SnO2
and near 13 eV for SnO as was found by Powell [41]. A
broadening and shift in energy between these two
characteristic features suggests an increasing structural
disorder. The radical change in core level features
in comparison to VB features suggest a higher sensitivity
to lattice defects. In particular, the 46 eV feature
appears to be an excellent indicator of the SnO2 structure.
Even when the valence band features appear to be very
SnO2-like, the presence of defects is indicated by the
loss or decrease of the 46 eV feature relative to the
VB features. This interpretation of structurally related
changes in the ELS spectrum is strongly supported by
the combined LEED and ELS study of de Fresart et al. [58]
on SnO2 (110).
Oxygen Effects
It is shown above that a growth of the 27 eV feature
and loss of the 46 eV feature reflects a change in tin
oxide away from a well-annealed SnO2 material. Some
distinction between the origins of the changes in these
two core level features can be made. This distinction
requires a measure of the oxygen concentration which
is provided by core level XPS using standard sensitivity
factors [31]. To make comparisons with VB XPS useful,
attention is limited to the more bulk sensitive energy-loss
measurements for 1500 eV primary beam energies. The
quantitation of oxygen levels within the matrix by core
level XPS presents a problem due to a difference in
analysis depth with respect to VB XPS and ELS. This
problem is minimized by using the O/Sn ratios determined
in this manner as only a rough measure of the oxygen
concentration further into the bulk. The O/Sn ratios
are reported within an uncertainty of 0.03 which describes
the reproducibility of the measurements. No uncertainty
in the sensitivity factors is reported. In this regard
trends in the O/Sn ratios are more important than the
absolute values.
For the sputtered sample described by Figures 3-6b
and 3-7, the O/Sn ratio is 0.96. Annealing the sputtered
sample in vacuo at 5000C for 20 minutes repairs some
of the sputter-induced damage. Figure 3-8a shows the
effect on the VB XPS spectrum. A decrease in the defect
feature at low binding energies is observed, and a
splitting in the VB features at 4.5 eV binding energy
appears. This splitting is characteristic of a mixture
15 10 5 8
BINDING EMERSY (EV)
Figure 3-8. Valence band XPS spectra following (A) a
short 5000C anneal after sputtering and (B) subsequent
oxygen exposure. The presence of a mixture of SnO and
SnO2 is indicated in (A).
V
z
2
I1 1 1 1 I 1 1 1 I I I 1 1 I i 1 1 11 I 1 1 11
(A)
0/SN 1 1.34
O/N- 1.14
VALENCE BAND XPS
\1 1 \ \1 1 \ 11 1 1 1 1 1 1 1
of SnO and SnO2 [32]. The short anneal also increases
the O/Sn ratio to 1.14 presumably due to some oxygen
diffusion into the surface region from the bulk. The
energy-loss spectrum in Figure 3-9a also shows more
evidence of structural repair caused by annealing. The
largest valence band feature is sharper and centered
at 19.5 eV, and the presence of the high energy loss
feature near 45 eV is again slightly visible. Both of
these features indicate the presence of an Sn02 structure.
The presence of the 27 eV feature reveals an oxygen
deficiency relative to Sn02, and the size and shape of
the feature near 13 eV suggests the possibility that
SnO is present. However, the 13 eV feature is a
convolution of SnO and Sn02 features which yields little
information by casual inspection.
Subsequent treatment in situ with 11 Torr of 02
at 5000C for 15 minutes results in the addition of a
significant amount of oxygen to the matrix, O/Sn = 1.34.
Figure 3-8b shows the effect on the VB XPS spectrum.
The splitting which is apparent in Figure 3-8a is removed,
and the shape of the VB emission is predominantly that
of Sn02. The addition of oxygen also affects the ELS
spectrum as seen in Figure 3-9b. The change in VB features
is minimal. The main loss feature falls at 19.5 eV as
expected for an Sn02 material, and there is a decrease
in the feature near 13 eV relative to the 19.5 eV feature
(5)
t
18.5
t 13
27 4
4/>
Ep 15 EV
1 1 1 1 1 1 1 1.1 1 1 111 | | 1 1 111 11 1 1111 11 11 Il l
ifIfuI;;III r u r u I IIrrruI I
60 58 40 38 28 18
EMR6Y LOSS (EV)
Figure 3-9. ELS spectra corresponding to Figures 3-8(A)
and (B) respectively.
n[lir[i I[lllljnlilnrr[rrl 11I1I1I 11111 I1III1IIIII Ilrlnrllm
I
suggesting a loss of the SnO contribution to the spectrum.
The most apparent changes occur in the core level features.
A small increase in intensity of the feature near 36
eV is observed. This increase is consistent with the
assignment of this feature to 0 2s-to-CB transitions.
It can be seen that the 27 eV feature in Figure
3-9b is greatly diminished. The loss of this feature
by annealing in oxygen substantiates the earlier
interpretation that it is associated with a loss of oxygen
from the SnO2 structure and be may interpreted as due
to a change in symmetry of the states near the CBM. It
seems that the growth of the 27 eV feature reflects a
loss of coordinating oxygen or a lowering of the valency
of the tin. This loss may occur through the formation
of defects such as oxygen vacancies in a nonstoichiometric
or amorphous oxide, through the formation of stoichiometric
SnO or through the formation of metallic Sn.
Figure 3-9b demonstrates that the 46 eV and 27 eV
loss features are not strictly interdependent. The weak
intensity of the high-energy loss feature suggests a
sensitivity to defects other than those associated only
with a deficiency of oxygen in a SnO2 lattice. It is
postulated that the high-lying conduction-band final
states associated with this transition are strongly
dependent on the periodic potential of the SnO2 lattice
and easily perturbed by the presence of defects. This
56
strong dependence may occur if the states are less atomic
in nature than the valence and lower conduction bands
while still containing a fraction of tin 4f character
as suggested by Gobby [47].
Conclusions
The use of ELS combined with valence band
photoemission and results of band structure calculations
provides a powerful means for studying tin oxide surfaces.
In this study an assignment of the major features in
the tin oxide N(E) energy-loss spectrum is made. The
loss features are assigned to collections of optically
(dipole) allowed interband transitions based on a previous
photoemission study by Gobby [47]. It is found that
the low-energy portion of the spectrum may be associated
with valence-to-conduction-band transitions, and the
higher energy-loss features are due to
core-to-conduction-band transitions. Of these core level
features, it is possible to distinguish between transitions
from Sn 4d and 0 2s levels even though these features
cannot be resolved in XPS.
It is demonstrated for tin oxide surfaces that depth
profile information may be obtained using ELS. By varying
the primary electron beam energy and hence the analysis
depth, it is shown that a high temperature anneal results
in bulk SnO2 under an oxygen deficient structure which
is essentially SnO at the surface.
By using ELS in conjunction with valence band XPS,
it is found that a significant amount of structural
information may be inferred from the size, shape and
position of the N(E) ELS features. In particular,
distinctions can be made between SnO2, SnO and defect
or amorphous structures. The Sn 4d core level features
are found to be much more sensitive to defects in an
SnO2-like lattice than are the VB features. A loss feature
at 27 eV assigned to transitions from Sn 4d levels to
states near the CBM is associated with atoms in a lower
oxidation state or in a lattice deficient in oxygen
relative to SnO2. An SnO2 loss feature near 45 eV is
shown to be very sensitive to defects not necessarily
associated with oxygen vacancies or deficiencies, but
the specific type(s) of structural defects) associated
with the behavior of this high-energy-loss (45 eV) feature
has not yet been determined.
SECTION IV
A STUDY OF THE DEHYDRATION OF TIN OXIDE
SURFACE LAYERS
Introduction
The chemisorption properties of tin oxide surfaces
can be significantly influenced by the interaction with
water. Kaji et al. have demonstrated that it is possible
to fixate Cu (II) and Pd (II) complex ions on hydrated
tin oxide surfaces in the preparation of propylene
oxidation catalysts [65]. The modification of tin oxide
electrode surfaces by an alkaline pretreatment has been
shown to give enhanced cell emf responses to changes
in pH [28]. This enhancement is thought to occur through
the hydrolysis of surface Sn=O bonds to give Sn-OH surface
species. The specific adsorption of Fe (III) and Pb
(II) cations has been shown to occur on these hydrated
surfaces apparently by replacement of the proton on the
surface hydroxyl groups [66,67]. Similarly, the specific
adsorption of bromine and iodine anions on tin oxide
occurs only on hydrated surfaces [68,69]. Most recently
it has been shown that an increase in Pt uptake rates
during chemisorption from solution occurs on hydrated
tin oxide surfaces [6]. Pt dispersions can exceed 90%
on these surfaces, and the resulting catalytic activity
per surface Pt atom exceeds that of metallic Pt electrodes
for the electrochemical reduction of 02.
Secondary-ion mass spectrometry (SIMS) and
electron-stimulated desorption (ESD) were used in a
previous study to examine alkaline-pretreated tin oxide
for evidence of surface hydroxylation [19]. ESD
demonstrated higher yields of both H+ and 0+ after the
alkaline pretreatment suggestive of significant surface
hydroxylation. A small signal due to OH+ desorption
was also observed. Results using dynamic SIMS showed
no apparent differences in the bulk regardless of
pretreatment. It has become apparent, however, that
hydrogen is a major constituent in most tin oxide films,
and it appears that the actual film composition may be
best described as SnxOyHz. SIMS depth profiles of H+,
O+, OH+ and SnH+ species indicate an excess of hydrogen
and/or hydroxide or hydrated species at the surface of
tin oxide films [40]. A steep concentration gradient
within approximately the outer 30 A of the material
indicates that hydration is not limited strictly to the
outer atomic layer. While. the degree of hydration is
greatest for the alkaline-treated samples, significant
hydration occurs over the same depth for samples exposed
to atmospheric humidity only. This observation is
indicative of the hygroscopic nature of tin oxide surfaces.
The complexity of the interaction of water with
tin oxide is demonstrated by the work Giesekke et al.
on the decomposition of bulk tin (IV) hydroxide [55].
Using thermogravimetric analysis, it was determined that
the decomposition of SnO3H2 leads to the formation of
Sn205H2 above 2500C, Sn409H3 between 3250C and 3600C,
Sn8O06H2 at 5000C and SnO2 above 6000C. Electron
diffraction clearly shows that each dehydration product
is a different crystalline substance. Though an accurate
determination of the structures was not possible, a study
of proton magnetic resonance line shapes shows the
structures to be complex. None of the substances can
be described as simple hydrates or hydroxides.
In Section III it was shown that electron energy-loss
spectroscopy (ELS) is sensitive to electronic changes
in tin oxide and is useful as either a surface or
subsurface probe. With the aid of valence band x-ray
photoelectron spectroscopy (XPS), it was also shown that
certain changes in the ELS spectrum may be related to
structural changes in the material. ELS and valence
band XPS are used in the present study of the hydrated
layer formed on tin oxide by exposure to atmospheric
humidity. A preliminary observation of water adsorption
using grazing-exit-angle 'XPS and ESD is given in Section
V.
Experimental
The preparation of the polycrystalline tin oxide
film used in this study has been described in Section
III. Once prepared the sample was exposed to atmospheric
humidity for several months to allow hydration of the
near surface region. All spectra were collected with
a double-pass CMA. The ELS data were taken in the N(E)
mode to allow for a direct comparison with the data of
Powell [41]. A 100 nA, 0.1 mm diameter primary electron
beam was used. All ELS spectra were recorded with a
25 eV pass energy ( AE/E = 0.014) using pulse counting
detection. The XPS spectra were taken using a Mg Ka x-ray
source and a 50 eV analyzer pass energy. The base pressure
in the vacuum system for this study was lx10-10 Torr.
Details of the vacuum system have been given previously
[17].
The ELS spectra were taken using the coaxial electron
gun in the CMA at an incident angle of 450 with respect
to the sample normal. As shown in Section III, ELS
measurements sensitive to the top few atomic layers can
be obtained in this configuration using a 200 eV primary
beam energy (Ep). Using a 1500 eV primary beam energy
significantly decreases the surface sensitivity of the
ELS measurement. This higher beam energy makes ELS more
sensitive to the subsurface region with an estimated
analysis depth of approximately 20 A. The valence band
(VB) XPS spectra obtained with this sample orientation
are also more sensitive to the subsurface region, and
the VB XPS analysis depth is expected to be similar to
the high energy (Ep = 1500 eV) ELS measurements. Both
the subsurface VB XPS and ELS measurements are sensitive
to the same region in which previous SIMS results [40]
suggest that hydration occurs.
Results and Discussion
Prior to analysis, the sample was cleaned in situ
by heating to 5000C in 10 Torr of 02 for 5 minutes. This
procedure removed carbon and chlorine contamination while
leaving a trace amount of K on an otherwise clean oxide
surface as determined by Auger electron spectroscopy
(AES). This contamination is known to be segregated
at the surface [17], and it may be removed easily by
Ar+ bombardment. However, in order to preserve the
hydrated layer of the sample, no ion bombardment was
used. While the 02 treatment may have affected the very
near-surface region of the sample, the following data
reveal that the subsurface layers were not dehydrated.
Figure 4-1 shows the valence band XPS data for the
sample after various treatments. Figure 4-la is the
spectrum recorded after the in situ cleaning. Figure
4-2a illustrates the effect of a 5000C vacuum anneal
on the spectrum. The annealing process was continued
until the background chamber pressure went through a
distinct maximum (about 45 minutes). Figure 4-ic shows
the result of a similar (30 minute) 6000C anneal. Figures
4-2, 4-3 and 4-4 show the ELS spectra corresponding to
Figures 4-la, 4-1b and 4-lc respectively.
The valence band spectrum in Figure 4-la is
characteristic of a sample hydrated due to exposure to
atmospheric humidity. The large feature near 10 eV binding
energy (7 eV below the valence band maximum (VBM)) is
largely due to this hydration though the form of the
incorporated water is unknown. The 10 eV feature is
similar to that found for water adsorption on other oxides
[70].
Figure 4-2a shows the ELS spectrum (Ep = 1500 eV)
corresponding to Figure 4-la. The features at energy
losses less than about 30 eV are due to
valence-to-conduction-band transitions as discussed in
Section III. The main VB loss feature in Figure 4-2a
falls at 20 eV which is characteristic of a SnO2-like
material (see Section III and ref. 41). Additionally,
a large shoulder associated with the VB loss features
is observed. This feature has not been previously
reported, and it appears to be composed of two loss
features near 24.5 eV and 27 eV. This shoulder is not
characteristic of SnO2. These additional features may
be interpreted as transitions from the hydrate-induced
ANNEAL
(C)
t2I ANEAL
(A)
HYDRATED
VALENCE BAND XPS
I I I I I i I I I I l l i [ i l 1 I
15 t1 5 8
BINDING EDERY CEV)
Figure 4-1. Valence band XPS spectra after (A) hydration
by exposure to atmospheric humidity, (B) a 5000C vacuum
anneal for 45 minutes and (C) a 6000C vacuum anneal for
30 minutes.
2I7, (B) Epa zee88 EV
Figure 4-2. ELS spectra of (A) subsurface and (B) surface
regions after hydration due to atmospheric humidity.
lower valence band feature observed in Figure 4-la to
conduction band. states. This interpretation is in
agreement with the ultraviolet photoelectron spectroscopy
(UPS) measurements of Gobby [47]. The UPS results show
that the lower valence band feature couples strongly
to conduction band states in a range from 25 eV to 30
eV higher in energy. Features at energy losses greater
than about 30 eV are due to core-to-conduction-band
transitions. In particular, the broad feature centered
near 36 eV is due to a set of O 2s-to-CB transitions
(see Section III).
Figure 4-2b (Ep = 200 eV) shows the more surface
sensitive ELS spectrum of the hydrated sample. The main
VB loss feature falls near 19 eV suggestive of an Sn02-like
material. The lack of higher energy-loss VB features
in this spectrum indicates a surface which is dehydrated
relative to the subsurface layers. This dehydration
is most likely the result of ESD from the surface under
the influence of the primary electron beam and/or some
superficial dehydration due to the elevated temperature
used during the oxygen cleaning procedure.
The main VB loss features in Figures 4-2a and 4-2b
suggest that the sample is fully oxidized in both the
surface and subsurface regions. This is confirmed for
the surface region in Figure 4-2b by the lack of a 27
eV Sn 4d core level loss feature which would be
characteristic of a deficiency of oxygen relative to
SnO2 (see Section III). Although the presence of such
a 27 eV feature would be obscured in Figure 2a by the
hydrate-induced VB loss feature, the absence of any low
binding energy structure (2 eV to 3 eV) in Figure 4-la
reveals that no oxygen deficiency exists (see Section
III). It is apparent, however, that the structure of
the material is perturbed relative to a well-annealed
SnO2 rutile structure. This perturbation is evidenced
by the lack of a core level loss feature near 45 eV in
Figure 4-2 (see Section III). For the hydrated subsurface
layers, the perturbation is easily understood as due
to the addition of excess oxygen and hydrogen from the
water of hydration while at the surface beam damage is
the most likely cause.
Figure 4-lb shows the effect of a 5000C vacuum anneal
on the valence bands. The large feature near 10 eV in
Figure 4-la has been greatly reduced suggestive of a
dehydration of the subsurface region, and a spectrum
very similar to that characteristic of SnO2 remains (see
Section III and ref. 32). This change is also reflected
in the ELS spectrum in Figure 4-3a. While the main VB
loss feature remains near 20 eV, the features in the
25 eV to 27 eV region associated with the hydrated oxide
are substantially decreased. The valence band ELS features
in Figure 4-3a are quite SnO2 like in agreement with
(A) Ep s88 I VE
46 28
(B) Ep 2M88 EV
27 T
13
SM CANEAL
IIIJj 111 w n wmmliltllinfil iim ii inn j jnn i liffit
s o 40 38 28 18is
ENERGY LOSS CE)
Figure 4-3. ELS spectra of (A) subsurface and (B) surface
regions after a 5000C vacuum anneal.
the VB XPS spectrum of Figure 4-1b. Concurrent with
the change in VB features, the appearance of a small
core level loss feature near 45 eV is observed in Figure
4-3a. The weak presence of the 45 eV loss feature
represents the beginning of a change in the subsurface
oxide to a true SnO2 structure (see Section III). However,
a significant perturbation of this structure is still
apparent. At the surface the annealing has resulted
in an oxygen deficiency as illustrated by Figure 4-3b
(Ep = 200 eV). The broad valence band features with
increased intensity at 13 eV show that the surface has
changed from SnO2-like to a more SnO-like material. The
oxygen deficiency in the surface region is confirmed
by the appearance of the small feature near 27 eV (see
Section III).
Further annealing at 6000C has only a small effect
on the VB XPS spectrum shown in Figure 4-lc. The feature
near 10 eV binding energy is completely removed leaving
a spectrum characteristic of SnO2 (see Section III and
ref. 32). Similarly, the ELS spectrum in Figure 4-4a
shows the core level loss features characteristic of
a well-annealed Sn02 material indicating the nearly
complete dehydration of the subsurface region. The
annealing process has, however, effected a reduction
at the surface. The sharp sturcture near 13 eV in Figure
4-4b is 'characteristic of SnO as found by Powell [41].
Ep 1588 EV t
28
45
V Ep = 208 EV
z
18 13
688C ANNEAL
68 58 48 38 28 18
ENERGY LOSS CEV)
Figure 4-4. ELS spectra of (A) subsurface and (B) surface
regions after a 6000C vacuum anneal.
71
The broad feature centered near 18 eV is not characteristic
of SnO or Sn02, and it may be interpreted as due to a
nonstoichiometric defect structure accompanying the change
from subsurface (bulk) SnO2 to surface SnO (see Section
III).
The changes observed in the subsurface layers using
ELS clearly indicate a temperature dependence in the
decomposition of the hydrated near-surface region. The
dehydration product observed at 500C is a forerunner
to the formation of a true SnO2 compound near 6000C in
the subsurface region. These observations are in agreement
with the work of Giesekke et al. [55] on the thermal
decomposition of bulk tin (IV) hydroxide. The formation
of Sn8016H2 could account for the apparent perturbation
of the subsurface crystal structure evidenced by Figure
4-3a while causing only a small variation in the VB density
of states from that expected for SnO2.
Conclusions
The near-surface region of a hydrated polycrystalline
tin oxide film has been studied. A large increase in
the lower VB density of states has been observed for
hydrated subsurface layers using VB XPS and ELS. These
observations are in agreement with SIMS data [40] which
suggests that hydration 'due to exposure to atmospheric
humidity occurs to depths of at least 30 A.
humidity occurs to depths of at least 30 A.
72
The thermal decomposition appears to proceed in
a stepwise fashion. The subsurface hydrated layers yield
SnO2 near 6000C, but the surface undergoes a reduction
to SnO. A comparison with existing data on bulk tin
(IV) hydroxide decomposition leads to an interpretation
consistent with the formation of an intermediate
hydrogen-containing compound in the subsurface region
near 5000C.
SECTION V
AN OBSERVATION OF WATER ADSORPTION ON TIN OXIDE
USING ESD AND GRAZING-EXIT-ANGLE
XPS AND AES
Introduction
As discussed in Sections I and IV, the chemisorption
properties of tin oxide surfaces can be significantly
affected through the interaction with water. In
particular, the chemisorption of platinum on tin oxide
surfaces is believed to occur at hydroxylated surface
sites (see Section II and ref. 6). In Section IV, the
dehydration of tin oxide surfaces has been studied using
valence band x-ray photoelectron spectroscopy (XPS) and
electron energy-loss spectroscopy (ELS). These techniques
have proven particularly useful in studying the subsurface
layers of the material. Though ELS may be made quite
surface sensitive by lowering the primary beam energy,
the study of adsorbates on tin oxide with this technique
is made difficult by the phenomenon of electron-stimulated
desorption (ESD).
Some preliminary observations of water adsorption
on tin oxide are given in this section. ESD experiments
and grazing-exit-angle XPS and Auger electron spectroscopy
(AES) measurements provide this observation. Though
the data presented here is incomplete, it provides an
interesting comparison with the dehydration study in
Section IV, with previous ESD data on alkaline and
non-alkaline treated tin oxide surfaces [19] and with
the work of Giesekke et al. [55] on the dehydration of
bulk Sn (IV) hydroxide. The incomplete nature of the
ESD experiments is due to a prolonged (16 months and
counting) failure of the Physical Electronics double-pass
CMA. Grazing-exit-angle XPS and AES are used because
of the unavailability of a preferred technique, ultraviolet
photoelectron spectroscopy (UPS).
Experimental
The preparation of the polycrystalline tin oxide
film used in this study has been described in Section
III. After preparation, the sample is rinsed in distilled
water and solvent cleaned prior to insertion into the
vacuum system. Before analysis the sample is cleaned
in situ by heating to 5000C in 10 Torr of 02 for 5 minutes.
This procedure removes trace chlorine and carbon surface
contamination leaving a clean oxide surface as determined
by AES. The AES, XPS and ESD data were taken with a
PHI double-pass. CMA. The AES spectra are collected in
the nonretarding (EN(E)) mode using a 3 KeV, 200 mA/cm2
electron beam. For XPS the analyzer was run in the
retarding (N(E)) mode with a pass energy of 50 eV (AE/E
= 0.014). Details of the vacuum system have been published
previously [17]. The base pressure for this study was
5x10-10 Torr.
The surface sensitivity of the electron spectroscopies
(AES and XPS) can be improved by collecting the emission
at angles away from the sample normal, i.e. at a more
grazing exit angle. The path length, L, that an escaping
electron (photoelectron or Auger electron) must travel
through a solid is related to the signal attenuation
due to inelastic collisions. Signal attenuation as a
function of path length can be described by an exponential
decay law with a uniform attenuation length. The
attenuation length, X, is known as the mean free path.
Hence,
I o( exp(-L/X )
where I is the signal intensity. Figure 5-1 illustrates
the increased surface sensitivity obtained at grazing
exit angles for a perfectly flat surface. If an emitting
source (atom) is a fixed distance, D, below the surface,
the path length, L, traversed within the solid increases
with increasing exit angle, 0, as
L'= D/cos 0
L= D
L = D/COSe
Figure 5-1. Variation in path length with emission angle.
Therefore, at fixed D, the signal intensity decreases
with increasing 9 making the measurement more surface
sensitive. In general, experimentally observed increases
in surface sensitivity are not as large as expected from
the above analysis. Two possible reasons for the deviation
are surface roughness and a decay of total signal with
increasing 0 causing a reduction in the signal-to-noise
ratio [71].
For all measurements the sample was mounted with
approximately a 450 angle between the CMA axis and the
surface normal. This orientation directs the sample
normal into the 42.30 60 acceptance cone of the CMA
[64]. Grazing exit angles are chosen with the 120 angular
acceptance aperture on the angle resolving drum mounted
coaxially within the inner cylinder of the second stage
of the CMA [72]. Using the relationship derived by Gobby
[47], the exit angle for a given drum setting may be
found.
ESD experiments are performed by using the CMA in
a time-of-flight (TOF) mode which allows for a simultaneous
determination of the mass and energy of desorbing ions.
For these measurements the analyzer is operated at a
constant pass energy of about 80 eV. This pass energy
(kinetic energy of the analyzed ions) sets the flight
time of the ions through the analyzer (about 4 psec for
H+). Because the CMA only passes charged particles of
the proper kinetic energy, species of different masses
(but same charge) have different axial velocities through
the CMA. Hence, the flight time of an ion through the
analyzer is directly proportional to the square root
of the mass-to-charge ratio. Traum and Woodruff [73]
have discussed in-depth the analyzer characteristics
which effect the flight time and mass resolution. Unity
mass resolution is possible for mass-to-charge ratios
(m/e) of at least 20.
To operate the CMA in a TOF mode for ESD experiments,
a computer-interfaced digital pulse counting circuit
is used (see Appendix B). The TOF modification to the
pulse counter allows it to perform three functions:
(1) it initiates the desorption event,
(2) it delays for a programmed flight time
(3) and it measures (counts) the signal pulses.
Before beginning the TOF analysis, the coaxial
electron gun in the CMA is configured with a +50 V charge
on the lower deflection plate. This voltage deflects
the electron beam (typically below 200 eV) downward out
of the analysis area (focal region) of the analyzer.
The pulse counter circuit initiates the desorption event
by supplying a 300 nsec TTL pulse to the base of a n-p-n
power transistor in series with the deflection plate
and ground (see Figuae '5-2). This pulse "grounds" the
deflection plate and swings the electron beam into the
TTL PULSE INPUT
- -+5 V
5K OlHMS
DEFLECTION
PLATE
.HRF427A
Figure 5-2.
initiation.
Deflection circuit for
desorption event
T
analysis region for 300 nsec. The circuit delays for
a programmed flight time before enabling an event counter
which records signal pulses for a similar 300 nsec period.
The count is subsequently read into the computer where
it is stored, and the process is repeated. By scanning
the programmed delay time a TOF (i.e. m/e) spectrum is
obtained. The only real-time constraint is that a total
time span be observed between desorption events at least
equal to the flight time of the most massive species
in the spectrum. This delay clears the analyzer of ions
before the initiation of a new desorption event.
To analyze low energy positive ions like those
obtained in the ESD experiment, the CMA is operated in
an accelerating mode. The inner cylinder and accelerating
grid which are connected internally are set initially
at -70 V, and the sample is biased at +10 V. This
potential difference between the sample and accelerating
grid raises ions initially at zero kinetic energy up
to the 80 eV analyzer pass energy thereby allowing their
detection. By ramping the accelerating grid to more
positive potentials, ions of higher initial energy
(typically 10 to 20 eV) can be measured. If the TOF
analyzer is operated at a fixed flight time, an energy
distribution spectrum of a single desorbing species may
be obtained. In this experiment, part of the accelerating
potential is imposed on the sample to provide a voltage
difference with an outer grid which is grounded to the
magnetic shield of the CMA. In this way. any spurious
signal due to ESD from this grid is shifted to apparent
negative kinetic energies and is easily recognizable
[73].
Results and Discussion
XPS and AES
Measurements have been made following three different
in situ treatments. These treatments include a 5000C
and a 6000C annealing step as was studied in Section
IV. Measurements have also been made following a 2 KeV
argon ion bombardment. Three types of AES and XPS
measurements are reported. Angle integrated results
are obtained with the angle resolving aperture retracted.
Angle dependent results have been obtained at normal
emission (00 40) and at a 700 2.50 grazing exit angle.
The results are presented in terms of O/Sn ratios. For
XPS this determination is made using the area under the
0 Is and Sn 3d 5/2 peaks corrected with standard
sensitivity factors [31]. The AES measurements are made
in a similar fashion using the peak-to-peak heights of
the 0 KLL (512 eV) and Sn MNN (437 eV) transitions [74].
A problem encountered when using the angle resolving
aperture for these measurements is a drop in total signal
and in signal-to-noise ratio. In AES this drop is not
a significant problem because of the magnitude of the
signal, but in XPS the drop is so large that a minimum
seven hour period is required to accumulate enough signal
for a single O/Sn ratio determination. The measurements
were undertaken originally to demonstrate the depletion
of oxygen near the surface due to annealing (as observed
in Section IV), but water adsorption from the background
Vacuum is observed instead because of the extended period
of time required to collect the data. Assuming a
background of water at the base pressure for this study
(5x10-5 Torr), the surface receives a 12.5 Langmuir (lL
= lxl0-6 Torr-sec) dose over a seven hour period. For
a unity sticking coefficient this dose represents about
12 monolayers of water. The water adsorption from the
background vacuum observed here is manifested by an
increased O/Sn ratio in the most surface sensitive (700)
XPS measurement. The adsorption of CO is believed not
to be a contributing factor because no carbon or CO
desorption signal is observed in the subsequent ESD
experiments.
The XPS and AES results are given in Table 5-1.
Angle integrated measurements (with the aperture retracted)
are made quickly after a given treatment before any
significant H20 adsorption occurs. These results indicate
a drop in the O/Sn ratio with increased annealing
temperatures and argon ion bombardment. The angle
Variation in O/Sn Ratio With Emission Angle.
XPS
*1
AES
ANGLE INTEGRATED 1.4
588C NORMAL EMISSION 1.5 1.3
78 DEGREE 2.1
6C ANGLE INTEGRATED 1.3
ANNUAL NORMAL EMISSION 1.4 1.5 1.1 1.2
ANAL 78 DEGREE 1.8 2.3
2KE ANGLE INTEGRATED 1.8
TV NORMAL EMISSION 1.1 1.8
SU R 78 DEGREE 1.5
Table 5-1.
integrated results are consistently similar to, but
slightly lower than, those obtained at normal emission.
This observation illustrates that the signal intensity
is highest at the sample normal as expected for a
polycrystalline material. A similar observation has
been made in Section III regarding the lack of surface
sensitivity in valence band XPS spectra. It is worth
noting that angle integrated measurements taken several
hours after a given treatment show a small increase in
O/Sn ratio like that observed with the seven hour normal
emission measurements.
The increase in O/Sn ratio observed for 70 emission
illustrates a significant uptake of H20 at the surface
from the background vacuum. Indeed, the large O/Sn ratio
of 2.3 observed in one case, suggests the formation of
a hydrated surface. Regardless of the order in which
the data is taken (i.e. the total exposure), the 70
emission always shows a substantially higher O/Sn ratio
indicating H20 adsorption at the surface. Water adsorption
during the normal emission measurements also explains
the small increase observed relative to the angle
integrated measurements.
The AES results given in Table 5-1 show no variation
with exit angle, and the O/Sn ratio is generally lower
than that obtained by angle integrated XPS. The lower
O/Sn value relative to XPS is probably due, in part,
to the increased surface sensitivity of AES. The kinetic
energies of the AES peaks are more than 200 eV less than
are observed in AES even after several hours of exposure
to the background vacuum is due to the ESD phenomenon.
As observed in a previous study [19], the surface
concentration of desorbing species can be rapidly depleted
under an electron beam of high current density. It is
believed that water adsorbed on the surface is quickly
removed by the incident beam used for the AES analysis
and is therefore undetected.
ESD
The first TOF ESD measurements made in this laboratory
are reported here. The data was acquired during the
process of tuning the instrument for the first time.
Unfortunately, an electron gun failure ended this
familiarization procedure before a good rapport could
be developed with the experimental set-up. Therefore,
the results shown here do not represent the full
capabilities of the equipment.
Figure 5-3 is the TOF spectrum of a tin oxide surface
sputtered with 2 KeV argon ions. The desorption of H+
and 0+ is clearly visible at 4.4 usec and 17.6 usec,
respectively. The feature at 18.6 4sec is possibly due
to H20+ desorption, but it is most likely due to O+
desorption with different initial conditions than in
the 17.6 usec peak. These possibilities can be checked
by increasing the accelerating potential and compressing
the flight times in the entrance region of the analyzer.
0
-I
H
>-
5 1t '15 28 25
ION FLIGHT TIME CMICROSEC)
Figure 5-3. Time-of-flight spectrum for mass analysis.
The 120 angular aperture has been used to select species
desorbing at near normal angles. Traum and Woodruff
[73] have shown that a significant increase in mass
resolution is possible by using the 40 angular acceptance
aperture. With the smaller aperture it should be possible
to resolve 0+ and OH+ species.
Figure 5-4 shows the ion kinetic energy distribution
for the sputtered sample after exposure to a high current
density electron beam for 30 minutes. Figure 5-4a shows
the total ion yield, 5-4b the time-gated (mass resolved)
H+ ion yield and 5-4c the time-gated 0+ ion yield. It
is seen that the CMA may be used for a simultaneous mass
and energy determination.
Figure 5-5 shows the total ion kinetic energy
distribution for the same sample after exposure to the
background vacuum for two hours. No time-gated
distributions were obtained in this case. Though a power
supply problem encountered during the analysis prevents
an accurate determination of the true kinetic energy
scale, the total ion energy distribution is seen to be
very different after H20 adsorption. This variation
suggest that ESD will prove useful in distinguishing
between different forms of hydrogen and oxygen on the
tin oxide surface.
|
|
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EDCS6T0IQ_5E1EKK INGEST_TIME 2017-07-14T23:32:37Z PACKAGE UF00085803_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
Copyright 1984
by
David Fullen Cox
SURFACE CHARACTERIZATION AND CHEMISORPTION PROPERTIES
OF POLYCRYSTALLINE SYSTEMS:
Sn02, Pt/Sn02 and Zr
By
DAVID FULLEN COX
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1984
ACKNOWLEDGMENTS
The author would like to thank Gar Hoflund for the
guidance and encouragement he furnished in his role as
research advisor. Special thanks go to Gar also for
many hours spent in the discussion of academic and purely
nonacademic matters and for invaluable assistance rendered
in the author's quest to master the French (not to mention
the English) language, pate7 de foie gras! Thanks go
also to Herb Laitinen for the benefit of his expertise
on tin oxide in all its forms and for providing the
laboratory facilities used for sample preparation. Thanks
go to Paul Holloway for his helpfulness in allowing the
use of his SIMS apparatus and for his patience in enduring
the associated visits. A special thank you also goes
to Dick Gilbert at the University of Nebraska for a
multitude of software and hardware contributions which
played such a major role in all the results presented
here.
The author thanks G.B Hoflund and H.A. Laitinen
for financial support supplied through sponsored research
grants. Thanks go also to the Department of Chemical
Engineering and the State of Florida for financial support.
Lastly, the author acknowledges the Florida Guaranteed
Student Loan Program for unilateral support during the
final six months of his degree program.
iii
TABLE OF CONTENTS
PAGE
ACKNOWLEDGMENTS iii
LIST OF TABLES vi
LIST OF FIGURES vii
ABSTRACT xi
SECTION
ONE GENERAL INTRODUCTION 1
Motivation 1
Format 4
TWO AN XPS INVESTIGATION OF TIN
OXIDE SUPPORTED PLATINUM 8
Introduction 8
Experimental 10
Results and Discussion 13
Conclusions 21
THREE AN ELECTRONIC AND STRUCTURAL
INTERPRETATION OF TIN OXIDE
ELS SPECTRA 23
Introduction 23
Experimental 25
Background 26
Results and Discussion 29
Conclusions 56
FOUR A STUDY OF THE DEHYDRATION OF
TIN OXIDE SURFACE LAYERS 58
Introduction 58
Experimental 61
Results and Discussion 62
Conclusions 71
IV
PAGE
FIVE AN OBSERRVATION OF WATER ADSORPTION
ON TIN OXIDE USING ESD AND
GRAZING-EXIT-ANGLE XPS AND AES 73
Introduction 73
Experimental 74
Results and Discussion 81
Conclusions 90
SIX THE INTERACTION OF POLYCRYSTALLINE
ZIRCONIUM WITH 02, N2, CO AND N20 91
Introduction 91
Experimental 92
Results and Discussion 93
Conclusions 115
SEVEN GENERAL CONCLUSIONS AND
RECOMMENDATIONS FOR FUTURE
RESEARCH 117
Pt Sn Oxide 117
Zirconium 119
APPENDICES
A A BRIEF DESCRIPTION OF THE
EXPERIMENTAL TECHNIQUES 121
X-Ray Photoelectron Spectroscopy (XPS) 121
Auger Electron Spectroscopy (AES) 124
Electron Energy-Loss Spectroscopy (ELS) 126
Electron-Stimulated Desorption (ESD) 128
B COMPUTER-INTERFACED DIGITAL PULSE
COUNTING CIRCUIT 131
Introduction 131
Circuit Description 132
Time-of-Flight Modification 139
Acknowledgments 140
REFERENCES 142
BIOGRAPHICAL SKETCH 147
v
LIST OF TABLES
TABLE
3-1. Electronic Configurations of Atomic
Tin and Oxygen and the Stoichiometric
Oxides.
5-1. Variation in O/Sn Ratio With Emission
Angle.
PAGE
40
83
vi
LIST OF FIGURES
FIGURE
2-1.
2-2.
3-1.
PAGE
Pt 4f XPS spectrum of an electro-
chemically platinized substrate.
The lower binding energy doublet
is characteristic of Pt metal,
and the higher binding energy
doublet is characteristic of
a Pt species chemically bonded
to be the tin oxide substrate. 14
Pt 4f XPS spectra of a sample
prepared by platinum chemisorp
tion. Spectrum (A) is obtained
immediately after pumpdown,
(B) after high temperature oxida
tion, (C) after high temperature
reduction and (D) after a high
temperature anneal _in vacuo. 17
Energy level diagram representing
the Sn02 band structure. The
locations of the major occupied
(unoccupied) valence and core
(conduction) states involved
in the energy loss spectrum
are shown. The approximate
locations of the SnO VBM and
CBM are indicated by dashed
lines. 30
N(E) ELS spectrum of tin oxide
after a high temperature vacuum
anneal. The high primary beam
energy (Ep = 1500 eV) at normal
incidence results in primarily
a bulk sensitivity. The indicated
features are characteristic
of well-annealed SnC>2 The
low energy features are due
to VB > CB transitions,
and the high energy loss features
are core > CB transitions. 33
vii
PAGE
3-3. Variation in the N(E) ELS spectrum
with primary beam energy, Ep.
The set of spectra represent
a depth profile of the annealed
material. The growth of the
27 eV feature is due to an
increasing oxygen deficiency
as the spectra become more surface
sensitive.
3-4. EN(E) ELS spectrum for Ep = 50
eV. The main loss feature at
13 eV shows that the annealed
material is essentially SnO
at the surface.
3-5. ELS spectra for a sample annealed
at 600C. The angle of incidence
for the primary beam is 45.
With Ep = 1500 eV it is seen
that the bulk is primarily SnC>2.
The Ep = 300 eV spectrum shows
SnO at the surface and evidence
of a defect structure between
bulk Sn02 and the surface.
3-6. Valence band XPS spectra after
a (A) 600C anneal and (B) 2
KeV argon-ion bombardment. The
spectra are more bulk than surface
sensitive.
3-7. ELS spectra following 2 KeV argon-
ion bombardment. The valence
band features show the bulk
to be Sn02like while the core
features reveal a significant
concentration of defects. The
VB features in the more surface
sensitive spectrum illustrate
an amorphous structure at the
surface due to sputtering.
3-8. Valence band XPS spectra following
(A) a short 500C anneal after
sputtering and (B) subsequent
oxygen exposure. The presence
of a mixture of SnO and Sn02
is indicated in (A).
36
37
44
46
49
52
vm
PAGE
3-9.
ELS spectra corresponding to Figures
3-8(A) and (B) respectively.
54
4-1.
Valence band XPS spectra after
(A) hydration by exposure to
atmospheric humidity, (B) a
500C vacuum anneal for 45 minutes
and (C) a 600C vacuum anneal
for 30 minutes.
64
4-2 .
ELS spectra of (A) subsurface
and (B) surface regions after
hydration due to atmospheric
humidity.
65
4-3 .
ELS spectra of (A) subsurface
and (B) surface regions after
a 500C vacuum anneal.
68
4-4.
ELS spectra of (A) subsurface
and (B) surface regions after
a 600C vacuum anneal.
70
5-1.
Variation in path length with
emission angle.
76
5-2.
Deflection circuit for desorption
event initiation.
79
5-3.
Time-of-flight spectrum for mass
analysis.
86
5-4.
Ion kinetic energy distribution
after sputtering.
88
LO
1
in
Ion kinetic energy distribution
following water exposure.
89
6-1.
AES spectra taken after (A)
hours of heating and (B) 1'
hours of heating below the HCP-to-
BCC transition temperature.
2
4
94
6-2.
AES spectra of state 1 zirconium
after room temperature exposure
to (A) nitrogen and (B) nitrous
oxide.
98
IX
PAGE
6-3 .
6-4 .
6-5.
6-6.
6-7 .
6-8.
6-9.
6-10.
XPS spectra showing the zirconium
3d peaks for (A) clean zirconium,
(B) N2 exposure, (C) N2O exposure
and (D) oxygen exposure
XPS spectra showing the zirconium
3s peak for (A) clean zirconium,
(B) nitrogen exposure and (C)
N2O exposure. The nitrogen
Is peak appears at 396 eV in
(B) and (C)
XPS spectra showing the oxygen
Is peak after (A) O2 adsorption
on zirconium and (B) N2O adsorp
tion on zirconium
AES spectrum for clean zirconium
after heating near the melting
temperature for 3 hours. The
AES 17 5 eV peak is greatly dimi
nished which is characterisitic
of state 2 zirconium
AES spectrum taken after clean
state 2 zirconium is exposed
to CO contamination from the
electron beam for 8 hours. The
carbon peak shows characteris
tics of both graphite and carbidic
carbon
XPS spectrum of the carbon Is
peak corresponding to the AES
spectrum shown in Figure 6-7.
Both graphitic and carbidic
carbon are present
AES spectra after exposing state
2 zirconium to CO at (A) room
temperature and (B) high tempera
ture but allowing the sample
to cool during the exposure
XPS spectra of the carbon Is peak
corresponding to the AES spectra
shown in Figure 6-9. The room
temperature adsorption produces
approximately equal amounts
of graphitic and carbidic carbon
as shown in spectrum (A) while
the high temperature adsorption
results in predominantly carbidic
carbon as shown in spectrum
(B)
99
100
102
104
106
107
108
109
PAGE
6-11.
A-l.
A-2 .
A-3 .
B-l.
B-2 .
B-3 .
B-4 .
(A) AES spectrum taken after expos
ing state 2 zirconium to nitrogen
at 5x10 Torr for 15 minutes
at room temperature. A small
amount of carbon and oxygen
contamination accumulated during
the long exposure and subsequent
AES run. (B) AES spectrum taken
after exposing state 2 zirconium
to nitrogen initially at high
temperature and then allowing
the sample to cool during the
exposure. (C) AES spectrum
taken after allowing the sample
to remain in vacuum for 3 days
at room temperature. State
2 zirconium has transformed
into state 1 zirconium. (D)
AES spectrum taken after exposing
the state 1 zirconium of spectrum
(c) to nitgrogen at 5xl0-^ Torr
for 5 minutes at room temperature. Ill
Photoemission process.
122
KLqL2 Auger decay process.
125
Electron energy-loss process.
127
On-board timer schematic showing
jumper selectable
rates .
system
clock
133
Schematic of
section.
the
control
logic
135
(A) Timing and
schematics.
(B)
event
counter
137
Layout and wiring
unmakred (pull
diagram. All
up) resistors
are 10K ohms. 138
xi
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
SURFACE CHARACTERIZATION AND CHEMISORPTION PROPERTIES
OF POLYCRYSTALLINE SYSTEMS:
Sn02f Pt/Sn02 and Zr
By
DAVID FULLEN COX
December, 1984
Chairman: Gar B. Hoflund
Major Department: Chemical Engineering
X-ray photoelectron spectroscopy (XPS) is used to
characterize platinum supported on tin oxide. A feature
in the platinum 4f XPS spectrum associated with the bond
formed between supported platinum and the tin oxide
substrate is identified. The bond is believed to form
with surface lattice oxygen resulting in a Pt-O-Sn surface
species. This substrate-bonded species appears to act
as a nucleation site for cyrstallite growth in both the
electrochemical desposition of platinum and in the
sintering of supported platinum.
It is demonstrated that electron energy-loss
spectroscopy (ELS) is an acceptable technique for
xix
distinguishing between the different oxides of tin. The
major features in the N(E) loss spectrum are interpreted
as due to collections of optically allowed interband
transitions. It is shown that depth profile information
about tin oxide may be obtained by varying the primary
electron beam energy. Combined ELS and valence band
XPS results indicate that a significant amount of
structural information may be inferred from the size,
shape
and/or
position
of
the
N(E) ELS features.
Core
level
features are found
to
be quite sensitive to
the
presence of
defects
in
an
SnC>2 lattice with
some
specificity as to the type of defect.
The chemisorption properties of polycrystalline
zirconium have been found to vary dramatically depending
on the thermal history of the sample. Chemisorption
on this surface is found to be suppressed by heating
for prolonged periods of time above the HCP-to-BCC phase
transition temperature at 1135K. The chemisorption
behavior can be correlated roughly with the appearance
or disapperance of a zirconium MW Auger peak. A slow
phase transition at the surface is postulated as the
cause of the variation in chemisorption properties.
xm
SECTION I
GENERAL INTRODUCTION
Motivation
The primary motivation for the work presented here
is an interest in the catalytic properties exhibited
by tin oxide supported platinum. It has been demonstrated
that platinized tin oxide surfaces display higher catalytic
activities than platinum electrodes in the electrochemical
oxidation of methanol [1-5] and for the reduction of
oxygen in alkaline [6] and 85% phosphoric acid solutions
[7]. A similarity between the electronic properties
of platinum in these supported systems and in the
industrial platinum-tin bimetallic reforming catalyst
supported on alumina has been demonstrated recently [8].
The bimetallic catalyst is known to exhibit improved
stability and higher average catalytic activity than
platinum supported on alumina [9,10].
Tin oxide and its modified forms also exhibit a
significant catalytic activity. Tin oxide has been shown
to be active for the catalytic oxidation of CO [11,12]
and the reduction of NO [12-14]. Chromia-doped tin oxide
is very active for the reduction of NO with CO, H2 and
C2H4 [13]. Antimony-doped tin oxide is known to be active
1
2
toward the selective (partial) oxidation of propylene
to acrolein [15] and the oxidative dehydration of butene
to butadiene [16]. Much of this behavior is believed
to be linked to the oxidation and reduction of active
sites on the catalyst surface.
The catalytic properties of the tin oxide support
are believed to play an important role in platinized
tin oxide catalysts. For example, Tseung and Dhara [7]
have postulated that a spillover mechanism may be involved
in the electrochemical reduction of oxygen. Their results
suggest that adsorbed oxygen migrates from the supported
platinum to the tin oxide surfaces before undergoing
reduction. The possible importance of spillover mechanisms
in platinum/tin oxide catalyzed reactions and the apparent
redox behavior of tin oxide surfaces both indicate that
a successful surface characterization of the supported
system must include a determination of the tin oxidation
state.
The main focus of the work presented here is the
application of ultrahigh vacuum (UHV) surface probes
to a fundamental characterization of tin oxide and
platinized tin oxide catalysts. It is hoped that this
characterization will aid in understanding the
physiochemical properties affecting the preparation and
the catalytic behavior of these systems. To this end,
a realistic system is studied which utilizes
3
polycrystalline tin oxide as the support material. The
polycrystalline nature of the support presents immediate
problems in terms of structural characterization because
the usual surface-structure-sensitive technique, low-energy
electron diffraction (LEED), is not applicable. For
this reason the approach primarily has been to use electron
spectroscopies as probes of the electronic properties
of the materials. The direction taken in the work
presented here has been influenced by some earlier studies
which have already been reported in the literature [17-19].
The earlier results will not be repeated here but will
be referenced when appropriate to the discussion of present
results.
One unforeseen result of the study of tin oxide
has been a broadening of interests to include structural
effects in electronic spectra (see Sections III and IV).
Out of this interest has grown a study of the chemisorption
properties of clean polycrystalline zirconium.
Observations in the literature of anomalous effects on
the Auger electron spectrum of zirconium due to gas
adsorption combined with conflicting results on the uptake
rate of adsorbates are responsible for the selection
of zirconium for study. Though not directly related
to the characterization of platinized tin oxide, these
zirconium results are also presented here.
4
Format
The results presented here are divided into
independent sections, each of which is complete within
itself. Each section deals with one aspect of the surface
characterization of a platinum-tin oxide catalyst with
the exception of Section VI which deals with the apparent
structural dependence of the chemisorption properties
of polycrystalline zirconium.
Section II presents results on the characterization
of supported platinum on tin oxide. X-ray photoelectron
spectroscopy (XPS) has been used in this characterization
to investigate the valence (oxidation) states of the
supported platinum. In situ thermal and chemical
treatments are used to help identify the various platinum
oxidation states and to investigate the nature of the
chemisorption bond formed between platinum and the tin
oxide substrate.
While XPS has proven useful in determining the valence
states of supported platinum, it has been of limited
use in characterizing the oxidation state of tin in the
support. Though core level XPS can distinguish between
metallic tin and tin oxide, it can not distinguish between
the different oxides of .tin, SnO and SnC>2. In view of
the apparent importance of the determination of the tin
valence state, the majority of effort has been devoted
to finding a suitable characterization technique for
5
the tin oxide support. The chosen technique is electron
energy-loss spectroscopy (ELS).
Section III is a discussion of the use of ELS in
the characterization of tin oxide. An interpretation
of the spectra is given in terms of the excitation of
interband transitions. Using ELS in conjunction with
valence band XPS, it is demonstrated that a significant
amount of structural information about polycrystalline
tin oxide may be inferred from changes in the electronic
structure as probed by ELS.
An understanding of the interaction of water with
tin oxide surfaces is believed to be of primary importance
in elucidating the chemisorption properties of platinum
on tin oxide. This belief is supported by considerable
evidence that the active chemisorption sites are surface
hydroxyl groups [6,19]. In Section IV ELS is applied
to the study of the dehydration of a tin oxide sample
after exposure to atmospheric humidity. The flexibility
of analysis depth provided by this technique affords
a particularly worthwhile characterization of the
subsurface region.
Because of a phenomenon known as electron-stimulated
desorption (ESD), ELS is not useful in studying the
interaction (i.e. adsorption) of water with the surface
of tin oxide. The incident electron beam used for ELS
actually removes the adsorbed species of interest from
6
the surface. This phenomenon has been observed in the
earlier studies [17-19] which show that significant surface
modification can result from an impinging electron beam.
However, the ESD phenomenon can itself be used to probe
the interaction between surface and adsorbate.
Section V presents some preliminary observations
of the adsorption of water on tin oxide.
Grazing-exit-angle XPS provides a measure of the water
adsorbed from the background vacuum in the UHV system.
Preliminary results are also presented which demonstrate
the potential of ESD in characterizing the interaction
of water with tin oxide surfaces.
Section VI presents a study of the chemisorption
properties of polycrystalline zirconium. The study
focusses on the chemisorption behavior as a function
of the thermal history of the sample. A slow phase
transition in the surface region is postulated as the
cause of a suppression in chemisorption after high
temperature annealing. A zirconium feature in the Auger
electron spectrum is shown to be an indicator of the
chemisorption properties of the surface.
Section VII presents a summary of conclusions for
the entire study along with recommendations for future
research. Two appendices are included after Section
VII. Appendix A contains a brief description of the
physical processes involved in the experimental techniques
7
used in this work. A brief introduction to XPS, AES,
ELS
and ESD is
given.
Appendix
B describes
the
computer-interfaced
digital
pulse
counting circuit
used
for
data collection
in XPS
, ELS
and
time-of-flight
ESD
measurements.
SECTION II
AN XPS INVESTIGATION OF TIN OXIDE
SUPPORTED PLATINUM
Introduction
The study of tin oxide supported platinum is motivated
by the interesting catalytic properties displayed by
mixed Pt-Sn systems. It has been shown that platinized
tin oxide exhibits a catalytic activity 50 to 100 times
greater than that of platinum electrodes for the
electrochemical oxidation of methanol [1-5]. Watanabe
et al. [6]. have studied the effects of platinum loading
on the catalytic activity for oxygen reduction in alkaline
solution. Their results show that the catalytic activity
of highly dispersed platinum on tin oxide may exceed
that of platinum electrodes by a factor of four or more
[6]. An interest in impure H2 fuel cells using 85%
phosphoric acid at 150C has prompted Tseung and Dhara
[7] to study supported Pt on antimony-doped tin oxide
because of the corrosion resistance and electrical
conductivity exhibited by this system. Their results
show a significant increase in the catalytic activity
for oxygen reduction over that of platinum black. In
addition, the similarity between platinized tin oxide
8
9
and the Pt-Sn bimetallic reforming catalyst has been
demonstrated recently [8].
The catalytic effects of low coverages of strongly
bound oxygen on Pt single-crystal surfaces has been
demonstrated by Smith, Biberian and Somorjai [20]. They
interpret dramatic, oxygen-coverage-dependent changes
in activity and selectivity for hydrogenation and
dehydrogenation reforming reactions as being due to the
formation of surface Pt oxides. Interestingly, it has
been shown that the formation of Pt-Sn alloys generally
results in lower catalytic activities [21-24]. These
observations suggest that oxygen may be partially
responsible for the catalytic properties exhibited by
Pt/tin oxide systems.
Several X-ray photoelectron spectroscopy (XPS) studies
have been performed on Pt-Sn systems [4,5,8,25]. These
studies all agree that tin is largely present as tin
oxide while platinum is present in the metallic form
and as Pt^+ and Pt4+ in the form of oxides and hydroxides.
These previous XPS studies have examined intimately mixed
systems of tin oxide and platinum oxides and metal. In
the present work a supported platinum system is studied.
The use of this system in conjunction with in situ chemical
and thermal treatments has allowed the assignment of
a Pt "oxidation state" characteristic of the chemical
bond formed between the metal and the tin oxide substrate.
10
Experimental
The tin oxide substrates are prepared by the thermal
hydrolysis of Sn (IV) from a solution containing 3 M
SnCl4 5H20, 1.5 M HC1 and 0.03 M SbCl3. The solution
is sprayed onto the hot surface of a titanium foil held
at 500C in air. The formation of tin oxide occurs
according to:
SnCl4 + 2H20 > Sn02 + 4HC1
The resulting planar film is a polycrystalline, n-type,
Sn02 semiconductor with the rutile structure. The antimony
is incorporated into the film at a concentration
approximately twice that of the spray, i.e. 2% [26].
This dopant is known to be in solid solution with the
tin oxide [27], and it acts as a donor which raised the
conductivity of the film to a level adequate for
electrochemical studies [26]. Spraying is continued
O
until a tin oxide layer approximately 6000 A to 7000
O
A thick is obtained as determined by the colors of the
interference fringes of the layer. After cooling in
air, the samples are polished with 0.25 pm diamond paste.
The use of alumina as a polishing compound is avoided
because of the overlap of platinum 4f and aluminum 2p
peaks in XPS [8].
A previous Auger electron spectroscopy (AES) and
XPS investigation has shown that the surface of an
11
antimony-doped tin oxide film prepared in the described
manner may contain a number of surface contaminants [17].
Among these surface contaminants are carbon, chlorine,
potassium, sodium, calcium and sulfur in varying amounts.
Argon ion bombardment and high temperature oxygen
treatments have proven to be effective in removing this
surface contamination, but the effects on the tin oxide
surface (see Section III) and platinum oxidation state
(as shown below) are substantial. Hence, an understanding
of surfaces such as those being used in electrochemical
studies [6,28] may require analysis in the presence of
several types of surface contamination.
Two different techniques are used in the present
study to prepare the supported Pt. The first of these
techniques is electrochemical in nature. In essence,
the platinum is plated from a 5x10^ m solution of H2PtClg
buffered at a pH of 6.8. The process is carried out
for varying amounts of time at -0.5 V versus SCE. The
second technique utilizes a chemisorption mechanism.
The substrate is pretreated by exposure to a 10 M NaOH
solution at 90C for 30 minutes. The pretreated substrate
is washed in distilled water and then exposed to an 80C
solution of 0.01 M KOH and 500 ppm Pt (IV) from Na2Pt(OH)g.
The Pt loading is dependent on the exposure time for
both preparation techniques. Regardless of the procedure
used, the samples are washed in distilled water after
12
platinization and solvent cleaned before mounting in
the vacuum system.
An important consideration in the preparation of
these supported platinum catalysts is the relative rate
of crystallite nucleation to growth. The indications
are that the growth of crystallites is favored over
nucleation in the electrochemical preparation [29]. The
chemisorption technique, however, has been shown to be
capable of producing highly dispersed (>90%) platinum
at low loadings [6]. The alkaline pretreatment is believed
to hydroxylate the surface, thereby increasing the number
of active sites available for Pt chemisorption [6,19].
Electron beam effects on these surfaces can be
dramatic. The removal of carbon, chlorine, oxygen,
hydrogen and sodium by electon-stimulated desorption
(ESD) has been observed previously [17-19], These beam
effects somewhat limit the usefulness of AES as an
analytical tool on these surfaces making XPS the preferred
technique because of its less destructive nature. However,
an understanding of the ESD phenomenon in terms of an
interatomic Auger decay model [30] shows promise in helping
to unravel the chemistry of these surfaces (see Section
V and ref. 19).
All XPS spectra were collected with a Physical
Electronics double-pass CMA using Mg K a X-rays as an
excitation source. A pass energy of 50 eV (AE/E = 0.014)
13
was used throughout. All binding energies are referencecd
to the tin 3d 5/2 peak at an assumed energy of 486.4
eV [31]. It has been shown that there is no change in
this core level binding energy for the different oxides
of tin [32-34] making this peak an excellent reference.
The base pressure in the vacuum system for this study
was lxl0-9 Torr. Details of the vacuum system have been
given previously [17].
Results and Discussion
Figure 2-1 shows the Pt 4f XPS spectrum of an
electrochemically platinized sample. The plating process
was carried out for 40 minutes at -0.5 V versus SCE.
A fairly high platinum loading of about 40 \ig/cm2 is
obtained by this process as estimated from a previous
Rutherford backscattering (RBS) study [35]. Deconvolution
of the spectrum reveals the presence of two platinum
species. In the assignment of these features platinum
chloride species are neglected. A check for surface
chlorine contamination using standard sensitivity factors
[31] showed the concentration to be low (Cl/Pt < 0.1).
The lowest binding energy doublet in Figure 2-1
has the 4f 7/2 peak at 71.2 eV and the 4f 5/2 peak at
74.5 eV. These features are assigned to Pt metal in
agreement with the work of Katayama [5,25]. The binding
energy reported here is about 0.4 eV higher than that
14
Figure 2-1. Pt XPS spectrum of an electrochemically plati
nized substrate. The lower binding energy doublet is
characteristic of Pt metal, and the higher binding energy
doublet is characteristic of a Pt species chemically bonded
to the tin oxide substrate.
15
generally reported for the bulk Pt metal [31]. This
observation of a higher binding energy for supported
clusters (crystallites) over that of bulk metal is in
agreement with general expectations [36]. This shift
is most likely the result of differences in the reference
levels (work functions) of the bulk metal and the tin
oxide support and/or a decrease in the final-state
extra-atomic relaxtion energy as a result of the change
from bulk metal to small cluster [37].
The doublet at higher binding energies in Figure
2-1 has a 4f 7/2 peak at 72.3 and a 4f 5/2 peak at 75.5
eV. This doublet is shifted about 1.1 eV above Pt metal
and about 0.5 eV below the position expected for a Pt(OH)2
species [5,25,31]. The higher oxides of platinum all
fall to significantly larger binding energies which removes
them from consideration (see below). Examination of
a second substrate electrochemically plated for only
1/4 the time (i.e. 10 minutes) gives a Pt 4f spectrum
(not shown) characterized primarily by this high binding
energy species observed in Figure 2-1. These results
suggest that the higher binding energy doublet in Figure
2-1 may be associated with a platinum species directly
bonded to the tin oxide substrate. Further, the appearance
of Pt metal at longer plating times demonstrates that
this substrate-bonded Pt species acts as a nucleation
site for the growth of metallic crystallites by
16
electrochemical deposition. These results are consistent
with a model of nucleation and crystallite growth suggested
by earlier work on the electrodeposition of platinum
on tin oxide [29].
Figure 2-2 shows the Pt 4f peaks for a sample prepared
by the chemisorption technique. The pretreated tin oxide
substrate was exposed to the 80C Na2Pt(OH)5 solution
for one hour. The platinum loading is approximately
2 pg/cm^. Because the Pt loading is small, the
signal-to-noise ratio does not justify a spectrum
deconvolution. However, the use of in_ situ chemical
and thermal treatments allows a manipulation of the Pt
valence state for a more complete determination of the
supported species. Because only a trace of chlorine
was detected, the possibility of platinum chloride species
was again discounted.
Figure 2-2a shows the spectrum obtained immediately
after pumpdown. The position of the doublet indicates
that the substrate-bonded species is the predominant
form of platinum obtained by the chemisorption procedure.
However, the peak widths also suggest the presence of
small amounts of Pt metal at lower binding energies and
Pt(0H)2 at slightly higher binding energies. The
observation of the substrate-bonded species as the primary
form of platinum is consistent with earlier work showing
nucleation is preferred over crystallite growth during
chemisorptive platinization [6].
NCE>
17
Figure 2-2. Pt XPS spectra of a sample prepared by plati
num chemisorption. Spectrum (A) is obtained immediately
after pumpdown, (B) after high temperature oxidation,
(C) after high temperature reduction and (D) after a
high temperature anneal in_ vacuo.
18
Figure 2-2b shows the effect of a high temperature
(600C) in. situ oxidation in 11 Torr of O2 for 30 minutes.
The oxygen treatment shifts the Pt 4f XPS peaks to higher
binding energies. The presence of the higher oxides,
PtO and PtC>2f is clearly indicated by structure on the
high binding energy side of the specturm. The PtO features
are shifted approximately 2.9 eV higher with respect
to Pt metal while the Pt02 features are shifted about
3.9 eV [31]. Though there is no evidence of Pt metal
in Figure 2-2b, the shoulder at 72.3 eV is clear evidence
of the persistence of the substrate-bonded Pt species.
The observation of this species in conjunction with PtO
and Pt02 confirms that the substrate-bonded species is
not simply a stoichiometric platinum oxide.
A 500C in_ situ reduction in lxlO-^ Torr of H2 for
30 minutes results in the spectrum shown in Figure 2-2c.
An XPS inspection of the Sn 3d core levels shows no sign
of a reduction of the substrate to bulk Sn metal. However,
the PtO and Pt02 species observed in Figure 2-2b have
undergone a complete reduction leaving primarily Pt metal.
The loss of the higher oxides coupled with the appearance
of a Pt 4f 7/2 peak at 71.2 eV confirms the earlier
assignment of this binding energy to Pt metal. Evidence
of the substrate-bonded species is also found in Figure
2-2c in the form of a shoulder on the high binding energy
side of the 4f 5/2 peak.
19
A subsequent 800 anneal in. vacuo has little effect
on the XPS peak positions as shown by Figure 2-2d. The
platinum remains primarily in the metallic state with
a small contribution due to the substrate-bonded species.
The presence of this substrate-bonded species after high
temperature annealing confirms that these features are
not due simply to a platinum hydroxide or hydrate species.
Decomposition or dehydration of such species would be
expected at significantly lower temperatures.
Before the oxidation-reduction cycle the platinum
from the chemisorption preparation is largely present
in the substrate-bonded form. This observation suggests
a high dispersion as found by Watanabe et al. for samples
prepared in a similar fashion [6]. The high temperature
oxidation-reduction cycle results in a sintering of the
supported species as shown by the large fraction of Pt
metal in Figure 2-2c. The remaining presence of the
substrate-bonded species suggests that a fraction of
these species acts as nucleation sites for the crystallite
growth as was observed in the electrochemical platinization
process.
The constancy of the XPS peak positions for the
substrate-bonded species obtained by either the
electrochemical or chemisorption process indicates a
similarity in the species formed regardless of the
20
procedure used. It has been shown above that this species
may not be identified as simply a PtOx or Pt(OH)y species.
Likewise, the binding energy shift of this species with
respect to Pt metal is not consistent with that observed
in the formation of Pt-Sn alloys [38]. These observations
suggest that the bond formed with the surface occurs
through surface lattice oxygen. The formation of a Pt-O-Sn
substrate-bonded Pt species is postulated. The tenacity
displayed by this species in resisting complete reduction
by chemical and thermal treatments is characteristic
of a species exhibiting such a strong interaction with
the substrate. Komiyama et al. have observed a similar
resistance to reduction by ion bombardment of strongly
interacting rhenium species on iron oxide [39].
Previous work on samples prepared by the chemisorption
technique offers insight into the mechanism of formation
of the substrate-bonded species. Watanabe et al. Have
shown that an alkaline pretreatment of the substrate
prior to platinization results in an increased Pt uptake
[6]. It is believed that the pretreatment hydroxylates
the surface and provides an increased number of active
chemisorption sites for the platinum species in solution.
Earlier studies using secondary-ion mass spectrometry
(SIMS) [40] and ESD [19] lend support to the surface
hydroxylation model by showing significant increases
in surface hydrogen and oxygen after the alkaline
21
pretreatment. Platinum chemisorption is believed to
occur by replacement of the proton on the surface hydroxyl
group with the loss of a coordinated ligand from the
platinum solution species. Under the pH conditions used
for chemisorption from a H2PtCl6 solution, the
chloroplatinate undergoes hydrolysis resulting in the
replacement of two chlorines by hydroxyl groups.
Chemisorption should occur via
Sn-OH + Pt(OH)2Cl42- > Sn-O-Pt(OH)CI42- + H20
with a subsequent dehydroxylation and loss of chlorine
from the
surface
complex.
For
chemisorption
from an
alkaline
solution
of Na2Pt(OH)5
the surface
hydroxyl
group is
ionized
through the
loss
of the acidic
proton.
Chemisorption is expected to occur via
Sn-O" + Pt(OH) g2- > Sn-O-Pt(OH)52- + OH-
leaving the substrate-bonded species after dehydration
of the surface complex.
Conclusions
XPS has been used to study tin oxide supported
platinum prepared by electrochemical and chemisorption
techniques. Features in the Pt 4f spectrum have been
22
assigned to a species chemically bonded to the substrate.
The position of these features is independent of the
preparation technique used. Tn situ chemical and thermal
treatments confirm that this substrate-bonded platinum
is not simply a PtOx or Pt(OH)y species. The platinum
is believed to bond through surface lattice oxygen giving
a Sn-O-Pt surface species. High temperature reduction
results in a sintering of these species, but the inability
to completely reduce the platinum is indicative of the
strong chemical interaction between the platinum and
tin oxide.
A model for the chemisorption of platinum on tin
oxide is proposed. Surface hydroxyl groups are believed
to be the active chemisorption sites for platinum species
in solution. The chemisorption process is believed to
occur through the replacement of the hydroxyl group proton
with the loss of a coordinated ligand from the platinum
species.
SECTION III
AN ELECTRONIC AND STRUCTURAL INTERPRETATION
OF TIN OXIDE ELS SPECTRA
Introduction
The spectroscopic study of tin oxide surfaces is
complicated by the difficulty in distinguishing between
the two oxides of tin, SnO and Sn02. Several x-ray
photoelectron spectroscopy (XPS) studies have failed
to detect any changes in core level binding energies
between SnO and Sn02 [32-34]. Similar problems are
encountered using Auger electron spectroscopy (AES) where
no significant differences in kinetic energies or line
shapes are found [41]. As expected, however, the valence
band spectra of the two oxides do differ. An ultraviolet
photoelectron spectroscopy (UPS) study of tin oxidation
by Powell and Spicer [42] and a valence-band XPS study
by Lau and Wertheim [32] have shown these differences,
but interpretation difficulties associated with analysis
depth have proven to be substantial.
Electron energy-loss spectroscopy (ELS) is a technique
which offers flexibility of analysis depth and is sensitive
to changes in the valence band density of states. Powell
[41] has shown that ELS may be used to distinguish between
23
24
the two oxides of tin and has given a preliminary
interpretation of the spectra in terms of differences
in plasmon frequencies. For SnC>2 a main loss feature
at 19.5 eV was identified while for SnO a main loss feature
was found at approximately 13.5 eV. A combined UPS
and high-resolution electron energy-loss spectroscopy
(HREELS) study of 3% Sb doped and undoped SnC>2 has shown
the room temperature occupied conduction bands to be
very f ree-electron like [43] in agreement with a bulk
tight-binding band structure calculation [44]. For the
heavily doped sample an HREELS loss feature at 0.55 eV
was found. Based on the experimentally determined carrier
concentration and effective mass ratio, the 0.55 eV loss
feature was identified as a surface plasmon loss associated
with conduction band electrons from Sb donors. Since
valence band and core level electrons in SnC>2 are not
free-electron like, the higher energy ELS losses in the
present study are not assigned to plasmon losses.
While an interpretation of the ELS spectrum would
be useful for distinguishing between the two oxides of
tin, an additional benefit may be derived due to the
usefulness of ELS measurements in the interpretation
of electron-stimulated desorption (ESD) threshold studies.
It has been shown that core level transitions can be
correlated with desorption thresholds and may specify
adsorbate binding sites [30,45,46]. In particular, the
25
ability to distinguish between transitions from Sn 4d
and 0 2s core levels which cannot be resolved using XPS
could be most useful in understanding the chemistry of
tin oxide surfaces.
Experimental
The polycrystalline tin oxide films used in this
study were prepared by spraying a solution of 3 M SnCl4
and 1.5 M HC1 onto a titanium foil maintained at 500C
in air. Unlike the samples used in Section II and in
previous studies [17-19], a high purity (99.998%) anhydrous
SnCl4 reagent was used. The resulting samples were found
to have significantly less surface contamination. Trace
chlorine and carbon contamination was found to be removed
quickly in. situ by heating at 500C in 10 Torr of oxygen
for about 5 minutes. This procedure gave a clean oxide
surface as determined by AES.
The samples were annealed ini vacuo initially and
were heated briefly and allowed to cool before each
individual measurement. Using angle-resolved ultraviolet
photoelectron spectroscopy (ARUPS) on an ion-sputtered
SnC>2 (001) single crystal surface, Gobby [47] has shown
that the annealing process (550C to 835C) strengthens
the primary emission from the valence bands and increases
the sharpness and magnitude of the anisotropic emission
indicating a well ordered crystal. Similar annealing
26
effects in the sharpness and magnitude of ELS spectra
and on the magnitude of core level emission in XPS have
been observed for polycrystalline tin oxide samples in
this study.
All spectra were collected with a double-pass CMA.
Details of the vacuum system have been published previously
[17]. The ELS data were taken in the retarding (N(E))
mode to allow a comparison with the data of Powell [41].
All ELS spectra were collected with a pass energy of
25 eV ( A E/E = 0.014) with the exception of the 50 eV
primary beam measurement. This spectrum was recorded
in a nonretarding (EN(E)) mode to suppress the large
signal from secondary electrons at near zero kinetic
energies. All
ELS
spectra
were collected using
100
nA
beam currents
and
pulse
counting
detection.
The
XPS
spectra were
taken
using
a Mg K a
source and
a 50
eV
analyzer pass energy. The base pressure in the vacuum
system for this study was 1 x 10^- Torr.
Background
For energy losses of the magnitude of electronic
excitations, the inelastic scattering event may be
described in terms of optical (dipole) selection rules
in cases where the primary electron energy is high enough
to justify the Born approximation. It is generally thought
that primary electron beam energies above 100 eV to 200
eV satisfy this criterion [48-51].
27
Consider a primary electron of momentum h K scattered
inelastically into a state h K' resulting in an interband
transition between one-electron states, |k,l> >
V 1
V
Momentum
>
conservation requires AK = k'
->
- k
->
-> ->
+ G where AK = K
- K' and
G is a reciprocal lattice
vector.
Energy conservation
> ~y
requires h^( | K | 2- j k'
I2)
= 2m( £]< i
->*
,l'e k,l>
where ^ q,l
V-.
is the eigenenergy of
the
one-electron state
-7"
1q,i>.
Not only must energy
and
momentum be conserved, but the matrix element
- -v
T =
must be nonzero [48,51,52]. Expansion in powers of
(AKr) yields the selection rules. It has been shown [48,51]
that the monopole term vanishes due to orthogonality
and that retaining only the dipole (linear) term gives
T = i AK < k + AK, 1' | r | k, 1>
For small | AK | Rudberg and Slater [48] have shown a fair
approximation at small energy losses or large | K | may
be obtained by considering only direct transitions, k
= k'. Hence, in the regime where the Born approximation
applies the selection rules are essentially optical in
nature. To a first approximation, the energy dependence
of the loss spectrum should be similar to that measured
28
in optical absorption [49]. Since the momentum transfer
>
in the ELS transition, h A k, may be different than in
the optical process, it is expected that a broadening
in the energy dependence of the ELS features will occur
with respect to the optical features [48].
Because the results from this study are for
polycrystalline samples with a random grain size which
is small compared to the excitation volume, the orientation
of K with respect to the crystal axes may be assumed
to be random. Therefore, the ELS spectra presented here
represent an average over the entire Brillouin zone.
The present results should be most comparable to optical
absorption studies of polycrystalline samples.
It should be mentioned that a breakdown in dipole
selection rules is possible for low beam energies and
large energy losses. In this case the expansion of the
41
phase factor, exp(i A k *r), must be carried to the
quadratic (quadrupole) term to obtain an accurate
description. Ludeke and Koma [50] and Colavita et al.
[51] have taken advantage of this effect to identify
loss features due to quadrupole-allowed transitions between
dipole-unallowed states. No such identifications have
been made in the present work.
Using a generalization of the joint density-of-states
function for optical interband transitions which includes
finite momentum changes, Ludeke and Esaki [53] have shown
that the energy-loss distribution due to transitions
29
from narrow, filled initial states to empty conduction-band
final states may be proportional to the conduction-band
(CB) density of states. This density-of-states
interpretation requires the initial state to be isolated
with no additional scattering channel existing near the
same energy loss. An additional complication may arise
if there is a significant modulation of the scattering
cross section due to a partial filling of the conduction
bands from a competing scattering channel originating
from a different initial state. In spite of these problems
it should be possible to obtain some picture of the CB
density-of-states in
tin
oxide
if the
Sn
4d
and 0 2s
core levels couple
to
final
states
of
significantly
different energy.
Results
and Discussion
Figure 3-1 is
an
energy
level
diagram
depicting
the band structure of SnC>2. The character of the
electronic states in the valence and lower conduction
bands is due to Robertson [44]. The assignment of Sn
4f character to high lying conduction band states is
due to Gobby [47]. The width of the valence bands and
the location of the three major features therein are
from the available photoemission data [32,47]. The
position of the 0 2s and, Sn 4d core levels are from XPS
measurements made in this laboratory with no attempt
25
SN 4F
SN5P 02P
SN 5S
02P LONE PAIR
MIN. BONDING 02P
02P SN5S BONDING
-22
- 18
12
9.5
3.6
a
-2
-4
-7.5
-9.5
Figure 3-1. Energy level diagram representing the SnC>2
band structure. The locations of the major occupied
(unoccupied) valence and core (conduction) states involved
in the energy loss spectrum are shown. The approximate
locations of the SnO VBM and CBM are indicated by dashed
lines.
31
at deconvolution. Photoemission results [47] were used
to locate the states in the conduction bands which couple
strongly to various valence and core states as discussed
below. The cut-off position at the top of the conduction
bands was determined form the ELS spectrum in Figure
3-2 based on the interpretation of the high-energy loss
features given below. While all the states are represent
by single horizontal lines, some are quite broad and
may extend over 5 eV or more. The dashed lines in the
band gap and lower conduction bands represent the
approximate location of the SnO valence-band maximum
(VBM) and conduction-band minimum (CBM) respectively.
These assignments are due to photoemission results for
SnO [32] and optical absorption on highly defect laden
tin oxide films [54].
Annealing Effects
Figures 3-2 to 3-4 show the ELS data for a sample
annealed at 7 50C in_ vacuo. Each of these spectra were
recorded for a normal incidence primary beam of specified
energy, Ep. The annealing process was carried out until
the background chamber pressure went through a clear
maximum (about 45 minutes). Giesekke et al. [55] have
shown using thermogravimetric analysis and electron
diffraction that the decomposition of tin (IV) hydroxide
proceeds through four distinct crystalline hydrogen
containing compounds before yielding SnC>2 above 600C.
32
The observed pressure maximum during the annealing process
is indicative, in part, of this dehydration. The
hygroscopic nature of tin oxide and the study of hydrated
surfaces is discussed in Sections IV and V.
Figure 3-2 is the loss spectrum for a 1500 eV primary
beam. This spectrum may be divided into two parts; the
higher energy loss features above
about
28
eV
and
the
features
at lower energy losses.
The lower
half
of
the
spectrum
consists of two major features
at
19.5
eV
and
13 eV in agreement with the SnC>2 spectrum reported by
Powell [41]. Additionally, extrapolation of the linear
portion of the leading edge of the loss spectrum to the
baseline gives a minimum energy loss of 3.6 eV. This
value is equal to the best available optically determined
band-gap energy for SnC>2 single crystals [56,57 ] and
the calculated lowest energy direct-allowed one-electron
transition ( Ft > ft ) found by Robertson [44]. Using
constant-intial-states (CIS) ARUPS measurements and
angle-integrated UPS for SnC>2 (001), Gobby [47] has shown
that VB-to-CB transitions are dominated by excitations
form an initial state about 1.5 eV below the VBM to final
states near 10 eV, 13 eV and 19 eV to 22 eV higher in
energy as shown in Figure 3-1. Inspection of Figure
3-2 reveals a shoulder in the loss spectrum near 10 eV
as well as the two higher energy features. This 10 eV
loss feature also corresponds to a collection of VB-to-CB
34
dipole-allowed transtions at the r point in the Brillouin
zone for bulk SnC>2 as found by Robertson. It is concluded
that the lower energy loss features in Figure 3-2 are
due to collections of optically (dipole) allowed interband
(VB -> CB) transitions.
The loss features above 30 eV are strongly dependent
on the thermal history of the sample and are dominated
by core-to-conduction-band transitions from tin 4d and
oxygen 2s levels. Gobby [47] has shown that these core
levels couple to final states in two energy regimes.
Coupling to CBs which are 3 2 eV to 36 eV above the core
level
is
observed easily in
UPS
while coupling
to
the
lower
CBs
(the
CB minimum
lies
approximately
26.6
eV
above
the
core
levels) is
not
observable due
to
the
photoemission threshold and large background of secondary
electrons. At higher photon energies coupling to CBs
37 eV and higher relative to the core levels is observed.
This coupling begins to strengthen at 40 eV above the
core level, but higher energies were not used because
of a lack of photon intensity. However, a higher energy
CB final state was identified for an initial state feature
in the lower VBs. This final state falls about 45 eV
above the core level, and Gobby suggests that it is a
Sn 4f derived state (see Figure 3-1). In Figure 3-2
a range of energy-loss features from about 29 eV to 48
eV are visible. The strongest features fall near 36
NCE>
33
Figure 3-2. N(E) ELS spectrum of tin oxide after a high
temperature vacuum anneal. The high primary beam energy
(Ep = 1500 eV) at normal incidence results in primarily
a bulk sensitivity. The indicated features are
characteristic of well-annealed SnC>2. The low energy
features are due to VB > CB transitions, and the high
energy loss features are core > CB transitions.
35
eV and 46 eV in excellent agreement with the photoemission
results of Gobby.
On the basis of the similarities between the
photoemission results for single crystal Sn02 and the
energy-loss spectrum, Figure 3-2 is interpreted as being
characteristic of a well-annealed (though polcrystalline)
Sn02 material. Additionally, these similarities support
the conclusion that the main features observed in the
ELS spectrum are due to single inelastic events possibly
in conjunction with elastic scattering events. Because
of the long mean free path of electrons near 1500 eV,
the spectrum in Figure 3-2 (Ep = 1500 eV) is primarily
due to contributions from the bulk of the material.
Figure 3-3 shows the effect on the loss spectrum
of varying the primary beam energy from 1500 eV to 200
eV. Figure 3-4 shows the EN(E) loss spectrum for a 50
eV primary beam. Decreasing the beam energy decreases
the analyses depth due to a reduction in the electron
mean free path with kinetic energy. The set of spectra
in Figures
3-3
and 3-4, therefore
, represent a depth
profile
of
the
vacuum-annealed
tin
oxide material. In
Figure
3-3
the
main change in
the
valence band region
is seen
to
be
a growth of the
12
eV to 13 eV feature
relative to the 19 eV feature with decreasing beam energy.
This change is most apparent in Figure 3-4 where a feature
near 13 eV dominates the spectrum. Changes in the core
NCE>
36
Figure 3-3. Variation in the N(E) ELS spectrum with
primary beam energy, Ep. The set of spectra represent
a depth profile of the annealed material. The growth
of the 27 eV feature is due to an increasing oxygen
deficiency as the spectra become more surface sensitive.
ENCE)
37
Figure 3-4. EN(E) ELS spectrum for Ep = 50 eV. The
main loss feature at 13 eV shows that the annealed material
is essentially SnO at the surface.
38
level region are more dramatic. The core level losses
may be resolved into two features. The large loss feature
at 46 eV is seen to decrease rapidly with beam energy
leaving a separate feature near 36 eV. Concurrent with
the loss of the 46 eV feature, the growth of a feature
at 27 eV is observed.
By comparison to the work of Powell [41], the changing
valence-band derived features in Figures 3-3 and 3-4
may be loosely interpreted as a change in the tin oxide
from a SnC>2 compound in the bulk to a more SnO-like
material at the surface. Because the SnO-like feature
near 13 eV dominates the spectrum only for a 50 eV primary
beam, it appears that such a material exists in the near
surface region, possibly in the top few atomic layers.
This interpretation is reasonable in view of the well
documented oxygen loss from tin oxide surfaces during
high temperature annealing [47,58,59]. Such oxygen losses
have been observed frequently in this laboratory.
Decreases in surface O/Sn ratios from near 2 down to
1 on annealing have been monitored with AES and XPS.
The interpretation of changes in the core-to-CB
region of the spectrum leads to the same conclusion as
derived from the VB-to-CB features, but some discussion
of the symmetry of the initial and final states involved
is required. The band structure calculations of Robertson
[44] and Munnix and Schmeits [60] as well as the ARUPS
39
measurements of Gobby [47] show that the SnC>2 valence
bands are mostly 0 2p like with only a small admixture
of Sn derived states. The lower conduction bands are
primarily Sn 5s and 5p like, and within 3 eV to 4 eV
of the CBM these states are 90% Sn 5s like [44], To
a first approximation the atomic character of these states
suggests that the electronic structure of SnC>2 may be
considered to be ionic. Within this ionic approximation
the electronic configurations of the atomic and
stoichiometric oxide systems are those given Table 3-1.
For SnC>2 the highest occupied states are oxygen
2p like, and the lowest unoccupied states are tin 5s
like in basic agreement with the band structure
calculations. Reduction of Sn02 to SnO populates the
Sn 5s states leaving the lowest unoccupied states more
Sn 5p like. Likewise, the removal of oxygen from SnC>2
to form a nonstoichiometric oxide should result in a
mixing of Sn 5s states into the valence bands (possibly
as defect states) leaving a more Sn 5p like CBM. Such
a variation in symmetry near the conduction band minimum
should be apparent in the energy-loss spectrum. In
particular, a Sn 4d core-to-CBM transition will be dipole
unallowed for a Sn 5s dominated CBM, but dipole allowed
(Al=l) for a Sn 5p like CBM. Hence, the loss of oxygen
from SnC>2 should result in a change in the Sn 4d
core-to-CBM transition from unallowed to allowed. Notice
40
Table 3-1. Electronic Configurations of Atomic
Tin and Oxygen and the Stoichiometric Oxides
Atomic Tin
Atomic Oxygen
Stannous Oxide, SnO
Stannic Oxide, SnO?
gn : [Kr] 4d10 5s2 5p2
0 : Is2 2s2 2p4
Sn2+ : [Kr] 4d-*-0 5s2 5p
02- : Is2 2s2 2p6
Sn4+ : [Kr] 4d10 5s 5p
O2- : Is2 2s2 2p6
41
in Figure 3-1 that little change is expected in the energy
of the CBM between SnC>2 and SnO.
The
changing
nature
of
states near
the
tin oxide
CBM
may
be seen
clearly
in
Figures 3-3
and
3-4. The
46
eV loss feature may
be
interpreted as
i a
transition
from the Sn 4d core to a high lying Sn 4f-like CB state
[47]. The 27 eV feature may be interpreted as a Sn 4d
core-
-to-CBM transition
[61]. A feature near
2 7 eV
has
been
observed
in the
N(E) loss spectrum
for
both
SnO
and
Sn metal
[41,62]
but not for SnC>2.
For
the
case
of metallic tin, this feature may be viewed as a transition
from the Sn 4d core to empty states above the Fermi level.
The growth of the 27 eV loss feature in conjunction with
the decrease in the 46 eV feature may be interpreted
as a change in the CBs. The growth of the 27 eV loss
feature with decreasing beam energy is characteristic
of the changing nature of the CBM due to a deficiency
of oxygen in the surface region. This interpretation
is supported by the relative strengths of the two
transitions. The d > f transition is expected to
be stronger than the d > p transition [63].
The insensitivity of the 36 eV loss feature to
incident beam energy relative to the Sn 4d core features
discussed above suggests that the atomic origin of this
core derived feature is- significantly different. From
Figure 3-3 the main change in this feature is a gradual
42
decrease in intensity with decreasing beam energy. The
assignment of this 36 eV loss feature to an 0 2s core-to-CB
transition can explain this trend for a material exhibiting
a decreasing oxygen concentration on moving from the
bulk to the surface. This is precisely the situation
encountered in the present case.
Because the 0 2s and Sn 4d core levels couple to
CB final states of significantly different energy, the
energy-loss distribution due to these transitions may
be viewed as approximately proportional to the CB density
of states over a narrow range. There is certainly some
overlap between the 0 2s and Sn 4d transitions in the
neighborhood of the 0 2s feature. At the extremes,
however, near the 4 6 eV or 27 eV feature the
density-of-states interpretation should be valid although
substantial matrix element differences are likely between
these two regions. The observation from Figure 3-3
that changes in the core level features at high beam
energies are more dramatic than in the VB loss features
suggests the Sn 4d core features are more sensitive to
low concentrations of crystal structure defects than
the VB features. The Sn 4d core level losses may be
viewed as an strong indicator of the structural order
of the tin oxide material. Supporting evidence is found
from results on ion-sputtered samples.
43
Sputtering Effects
In order to increase the surface sensitivity of
the ELS measurement, the sample orientation was changed
to give the coaxial electron beam from the CMA an incident
angle of 45 with respect to the sample normal. This
change allowed reasonably surface sensitive measurements
with higher primary beam energies, and it eliminated
the problem of low-energy secondary electrons inherent
in the use of low electron beam energies (50 eV) for
N(E) measurements. Also, the probability of encountering
additional quadrupole-allowed features was minimized.
Assuming a straight line incident and exit path for an
electron scattered through a nominal angle of 137.7
(fixed by the CMA [64]), a very crude estimate of the
ELS analysis depth based on sample orientation and electron
mean free path can be made. Since the main features
observed in the ELS spectrum are due to single inelastic
events possibly in conjuction with elastic events, a
total path length of twice the mean free path of an
electron at the primary beam energy seems appropriate.
O
These
asssumptions lead
to an estimate of 5
to
10 A
(2
to
4
atomic
layers)
at
Ep = 2 00 eV and 15
to
20 A
(5
to
7
atomic
layers)
at
Ep = 1500 eV. These
estimates
should be viewed as qualitative at best.
Figure 3-5 shows the ELS data for a sample annealed
at 600C. The spectrum for Ep = 1500 eV shows the
structure characteristic of a well annealed SnC>2 bulk
NCE>
44
ENERGY LOSS CEV)
Figure 3-5. ELS spectra for a sample annealed at 600C.
The angle of incidence for the primary beam is 45. With
Ep = 1500 eV it is seen that the bulk is primarily SnC>2.
The Ep = 200 eV spectrum shows SnO at the surface and
evidence of a defect structure between bulk Sn02 and
the surface.
45
material. The more surface sensitive spectrum for Ep
= 200 eV has a sharp structure near 13 eV which is
characteristic of SnO [41]. The broad feature centered
at 18 eV is not characteristic of either SnO or Sn02,
and it most likely comes from a subsurface
nonstoichiometric defect structure accompanying the change
in structure from Sn02 in the bulk to SnO at the surface.
The 27 eV feature is also present indicating a structure
which is oxygen deficient in comparison to Sn02.
Figure 3-6a is the valence band XPS spectrum for
the 600C annealed sample. The resolution of the VB
XPS data is seen to be poor. This poor resolution is
due to a combination of very low signal intensity, the
x-ray line width, x-ray satellite emission from Sn 4d
and 0 2s core levels and data smoothing. In spite of
these difficulties, the general shape of the VB emission
is similar to that for SnC>2 as found by Lau and Wertheim
[32]. The obvious lack of surface sensitivity in this
measurement is not unexpected. Because the kinetic energy
of the valence band photoelectrons is large ( > 1200
eV), the mean free path is correspondingly large.
Additionally, the sample orientation is such that the
angle between the surface normal and the cylinder axis
is very nearly equal to the nominal 42.3 acceptance
angle of the CMA [64]. Since photoemission from
polycrystalline materials is expected to peak at the
n
46
BINDING ENERGY CEV)
Fiqure 3-6. Valence band XPS spectra after a (A) 600C
anneal and (B) 2 KeV argon-ion bombardment. The spectra
are more bulk than surface sensitive.
47
surface normal, the VB XPS results shown here have their
largest contribution from high energy electrons at near
normal emission. Hence, Figure 3-6a is primarily due
to the bulk SnC>2 material.
The annealed sample characterized by Figures 3-5
and 3-6a was ion sputtered with 2 KeV argon ions. Figure
3-6b illustrates the change in the VB XPS spectum. AES
and core level XPS show no evidence of a reduction to
metallic tin in this particular case, but the preferential
sputtering of oxygen is demonstrated by a drop in the
O/Sn ratio. Ion sputtering introduces a shoulder on
the VB emission near a binding energy of 2 eV to 3 eV.
A similar feature has been observed in ARUPS and
interpreted as emission from defect states associated
with a deficiency of oxygen [47]. Interestingly, this
sputter-induced feature lies near the same binding energy
as the highest lying SnO VB feature [32]. There is
even a fair correspondence between the SnO VBM as found
by Lau and Wertheim and the low binding energy edge of
the defect emission.
The assignment of the shoulder in Figure 3-6b to
defect states rather than SnO is justified by the ELS
spectra for the sputtered sample in Figure 3-7. The
more surface sensitive spectrum, Ep = 200 eV, shows a
broadening of the characteristic SnO feature at 13 eV.
The entire valence band portion of the spectrum becomes
48
broad and relatively featureless as a result of sputtering.
This broadening may be interpreted as a change from the
SnO structure at the surface to a more amorphous structure
caused by sputtering. For Ep = 1500 eV the loss spectrum
is sensitive to the bulk within the region probed by
the VB XPS measurements. While there is some broadening
and a small shift toward lower energy losses, the VB
features are still very much Sn02 like in approximate
agreement with the VB XPS spectrum shown in Figure 3-6b.
The core level loss features reflect the defect presence
much more strongly than the VB loss features. The absence
of the 46 eV feature and prominence of the 27 eV feature
confirm the change from a well-annealed Sn02 structure
to a more oxygen-deficient defect structure after
sputtering.
The damage induced by ion sputtering is heaviest
in the top few layer's of the solid as illustrated by
Figure 3-7. Thus, the amorphous structure at the surface
implied by the valence band features for Ep = 200 eV
is not unexpected. Sputtering damage in layers deeper
in from the surface may result from ion implantation,
knock-in and other ion-matrix phenomena, but the damage
in these deeper layers should be significantly less than
near the surface. Bearing in mind that ion bombardment
effects become less apparent as the experiment becomes
more bulk sensitive, a comparison of the results between
N
49
ENERGY LOSS CE^)
Figure 3-7. ELS spectra following a 2 KeV argon-ion
bombardment. The valencp band features show the bulk
to be SnC>2 like while the core features reveal a signifi
cant concentration of defects. The VB features in the
more surface sensitive spectrum illustrate an amorphous
structure at the surface due to a sputtering.
50
annealed and sputtered samples suggests that a significant
amount of qualitative structural information may be gained
from the N(E) energy-loss spectrum. The width and
center-of-gravity position of the valence band features
can be used as a gross indicator of the tin oxide
structure. A matrix characteristic of a stoichiometric
form of tin oxide is suggested by sharper, more well
defined valence band loss features near 19.5 eV for SnC>2
and near 13 eV for SnO as was found by Powell [41]. A
broadening and shift in energy between these two
characteristic features suggests an increasing structural
disorder. The radical change in core level features
in comparison to VB features suggest a higher sensitivity
to lattice defects. In particular, the 46 eV feature
appears to be an excellent indicator of the SnC>2 structure.
Even when the valence band features appear to be very
Sn02~like, the presence of defects is indicated by the
loss or decrease of the 46 eV feature relative to the
VB features. This interpretation of structurally related
changes in the ELS spectrum is strongly supported by
the combined LEED and ELS study of de Fresart et al. [58]
on SnC>2 (110 ) .
Oxygen Effects
It is shown above that a growth of the 27 eV feature
and loss of the 46 eV feature reflects a change in tin
oxide away from a well-annealed Sn02 material. Some
51
distinction between the origins of the changes in these
two core level features can be made. This distinction
requires a measure of the oxygen concentration which
is provided by core level XPS using standard sensitivity
factors [31]. To make comparisons with VB XPS useful,
attention is limited to the more bulk sensitive energy-loss
measurements for 1500
eV primary
beam
energies.
The
quantitation of oxygen
levels within
the
matrix by
core
level XPS presents a
problem due
to a
difference in
analysis depth with respect to VB
XPS
and ELS.
This
problem is minimized by using the O/Sn ratios determined
in this manner as only a rough measure of the oxygen
concentration further into the bulk. The O/Sn ratios
are reported within an uncertainty of 0.03 which describes
the reproducibility of the measurements. No uncertainty
in the sensitivity factors is reported. In this regard
trends in the O/Sn ratios are more important than the
absolute values.
For the sputtered sample described by Figures 3-6b
and 3-7, the O/Sn ratio is 0.96. Annealing the sputtered
sample in_ vacuo at 500C for 20 minutes repairs some
of the sputter-induced damage. Figure 3-8a shows the
effect on the VB XPS spectrum. A decrease in the defect
feature at low binding energies is observed, and a
splitting in the VB features at 4.5 eV binding energy
appears. This splitting is characteristic of a mixture
NCE>
52
SINDIH6 EhCRSY
Figure 3-8. Valence band XPS spectra following (A) a
short 500 C anneal after sputtering and (B) subsequent
oxygen exposure. The presence of a mixture of SnO and
Sn02 is indicated in (A). '
53
of SnO and SnC>2 [32]. The short anneal also increases
the O/Sn
ratio
to
1.14 presumably
due
to some oxygen
diffusion
into
the
surface region
from
the bulk. The
energy-loss spectrum in Figure 3-9a also shows more
evidence of structural repair caused by annealing. The
largest valence band feature is sharper and centered
at 19.5 eV, and the presence of the high energy loss
feature near 45 eV is again slightly visible. Both of
these features indicate the presence of an SnC>2 structure.
The presence of the 27 eV feature reveals an oxygen
deficiency relative to SnC>2r and the size and shape of
the feature near 13 eV suggests the possibility that
SnO is present. However, the 13 eV feature is a
convolution of SnO and Sn02 features which yields little
information by casual inspection.
Subsequent treatment in. situ with 11 Torr of O2
at 500C for 15 minutes results in the addition of a
significant amount of oxygen to the matrix, O/Sn = 1.34.
Figure 3-8b shows the effect on the VB XPS spectrum.
The splitting which is apparent in Figure 3-8a is removed,
and the shape of the VB emission is predominantly that
of Sn02* The addition of oxygen also affects the ELS
spectrum as seen in Figure 3-9b. The change in VB features
is minimal. The main loss feature falls at 19.5 eV as
expected for an Sn02 material, and there is a decrease
in the feature near 13 eV relative to the 19.5 eV feature
NCE>
54
EM-R6Y LOSS (EV)
Figure 3-9. ELS spectra corresponding to Figures 3-8(A)
and (B) respectively.
55
suggesting a loss of the SnO contribution to the spectrum.
The most apparent changes occur in the core level features.
A small increase in intensity of the feature near 36
eV is observed. This increase is consistent with the
assignment of this feature to 0 2s-to-CB transitions.
It can be seen that the 27 eV feature in Figure
3-9b is greatly diminished. The loss of this feature
by annealing in oxygen substantiates the earlier
interpretation that it is associated with a loss of oxygen
from the Sn02 structure and be may interpreted as due
to a change in symmetry of the states near the CBM. It
seems that the growth of the 2 7 eV feature reflects a
loss of coordinating oxygen or a lowering of the valency
of the tin. This loss may occur through the formation
of defects such as oxygen vacancies in a nonstoichiometric
or amorphous oxide, through the formation of stoichiometric
SnO or through the formation of metallic Sn.
Figure 3-9b demonstrates that the 46 eV and 27 eV
loss features are not strictly interdependent. The weak
intensity of the high-energy loss feature suggests a
sensitivity to defects other than those associated only
with a deficiency of oxygen in a Sn02 lattice. It is
postulated that the high-lying conduction-band final
states associated with this transition are strongly
dependent on the periodic potential of the SnC>2 lattice
and easily perturbed by the presence of defects. This
56
strong dependence may occur if the states are less atomic
in nature than the valence and lower conduction bands
while still containing a fraction of tin 4f character
as suggested by Gobby [47].
Conclusions
The use of ELS combined with valence band
photoemission and results of band structure calculations
provides a powerful means for studying tin oxide surfaces.
In this study an assignment of the major features in
the tin oxide N(E) energy-loss spectrum is made. The
loss features are assigned to collections of optically
(dipole) allowed interband transitions based on a previous
photoemission study by Gobby [47]. It is found that
the low-energy portion of the spectrum may be associated
with valence-to-conduction-band transitions, and the
higher energy-loss features are due to
core-to-conduction-band transitions. Of these core level
features, it is possible to distinguish between transitions
from Sn 4d and 0 2s levels even though these features
cannot be resolved in XPS.
It is demonstrated for tin oxide surfaces that depth
profile information may be obtained using ELS. By varying
the primary electron beam energy and hence the analysis
depth, it is shown that a high temperature anneal results
in bulk Sn02 under an oxygen deficient structure which
is essentially SnO at the surface.
57
By using ELS in conjunction with valence band XPS,
it is found that a significant amount of structural
information may be inferred from the size, shape and
position of the N(E) ELS features. In particular,
distinctions can be made between SnC>2, SnO and defect
or amorphous structures. The Sn 4d core level features
are found to be much more sensitive to defects in an
SnC>2-like lattice than are the VB features. A loss feature
at 27 eV assigned to transitions from Sn 4d levels to
states near the CBM is associated with atoms in a lower
oxidation state or in a lattice deficient in oxygen
relative to SnC>2. An SnC>2 loss feature near 45 eV is
shown to be very sensitive to defects not necessarily
associated with oxygen vacancies or deficiencies, but
the specific type(s) of structural defect(s) associated
with the behavior of this high-energy-loss (45 eV) feature
has not yet been determined.
SECTION IV
A STUDY OF THE DEHYDRATION OF TIN OXIDE
SURFACE LAYERS
Introduction
The chemisorption properties of tin oxide surfaces
can be significantly influenced by the interaction with
water. Kaji et al. have demonstrated that it is possible
to fixate Cu (II) and Pd (II) complex ions on hydrated
tin oxide surfaces in the preparation of propylene
oxidation catalysts [65]. The modification of tin oxide
electrode surfaces by an alkaline pretreatment has been
shown to give enhanced cell emf responses to changes
in pH [28]. This enhancement is thought to occur through
the hydrolysis of surface Sn=0 bonds to give Sn-OH surface
species. The specific adsorption of Fe (III) and Pb
(II) cations has been shown to occur on these hydrated
surfaces apparently by replacement of the proton on the
surface hydroxyl groups [66,67]. Similarly, the specific
adsorption of bromine and iodine anions on tin oxide
occurs only on hydrated surfaces [68,69]. Most recently
it has been shown that an increase in Pt uptake rates
during chemisorption frbm solution occurs on hydrated
tin oxide surfaces [6]. Pt dispersions can exceed 90%
58
59
on these surfaces, and the resulting catalytic activity
per surface Pt atom exceeds that of metallic Pt electrodes
for the electrochemical reduction of C>2.
Secondary-ion mass spectrometry (SIMS) and
electron-stimulated desorption (ESD) were used in a
previous study to examine alkaline-pretreated tin oxide
for evidence of surface hydroxylation [19], ESD
demonstrated higher yields of both H+ and 0+ after the
alkaline pretreatment suggestive of significant surface
hydroxylation. A small signal due to 0H+ desorption
was also observed. Results using dynamic SIMS showed
no apparent differences in the bulk regardless of
pretreatment. It has become apparent, however, that
hydrogen is a major constituent in most tin oxide films,
and it appears that the actual film composition may be
best described as Snx0yH2. SIMS depth profiles of H+,
0+, 0H+ and SnH+ species indicate an excess of hydrogen
and/or hydroxide or hydrated species at the surface of
tin oxide films [40]. A steep concentration gradient
O
within approximately the outer 30 A of the material
indicates that hydration is not limited strictly to the
outer atomic layer. While the degree of hydration is
greatest for the alkaline-treated samples, significant
hydration occurs over the same depth for samples exposed
to atmospheric humidity only. This observation is
indicative of the hygroscopic nature of tin oxide surfaces.
61
Experimental
The preparation of the polycrystalline tin oxide
film used in this study has been described in Section
III. Once prepared the sample was exposed to atmospheric
humidity for several months to allow hydration of the
near surface region. All spectra were collected with
a double-pass CMA. The ELS data were taken in the N(E)
mode to allow for a direct comparison with the data of
Powell [41]. A 100 nA, 0.1 mm diameter primary electron
beam was used. All ELS spectra were recorded with a
25 eV pass energy ( AE/E = 0.014) using pulse counting
detection. The XPS spectra were taken using a Mg Ka x-ray
source and a 50 eV analyzer pass energy. The base pressure
in the vacuum system for this study was lxl0~l Torr.
Details of the vacuum system have been given previously
[17] .
The ELS spectra were taken using the coaxial electron
gun in the CMA at an incident angle of 45 with respect
to
the sample
normal. As shown
in
Section
III,
ELS
measurements sensitive to the top
few
atomic
layers
can
be
obtained in
this configuration
using
a 200
eV primary
beam energy (Ep). Using a 1500 eV primary beam energy
significantly decreases the surface sensitivity of the
ELS measurement. This higher beam energy makes ELS more
sensitive to the subsurface region with an estimated
O
analysis depth of approximately 20 A.
The valence band
60
The complexity of the interaction of water with
tin oxide is demonstrated by the work Giesekke et al.
on the decomposition of bulk tin (IV) hydroxide [55].
Using thermogravimetric analysis, it was determined that
the decomposition of SnC>3H2 leads to the formation of
Sn2C>5H2 above 250C, Sn^gf^ between 325C and 360C,
Sn816H2 at 500C and SnC>2 above 600C. Electron
diffraction clearly shows that each dehydration product
is a different crystalline substance. Though an accurate
determination of the structures was not possible, a study
of proton magnetic resonance line shapes shows the
structures to be complex. None of the substances can
be described as simple hydrates or hydroxides.
In Section III it was shown that electron energy-loss
spectroscopy (ELS)
is
sensitive
to electronic changes
in tin oxide and
is
useful as
either a
surface or
subsurface probe.
With
the aid
of valence
band x-ray
photoelectron spectroscopy (XPS), it was also shown that
certain changes in the ELS spectrum may be related to
structural changes in the material. ELS and valence
band XPS are used in the present study of the hydrated
layer formed on tin oxide by exposure to atmospheric
humidity. A preliminary observation of water adsorption
using grazing-exit-angle XPS and ESD is given in Section
V.
62
(VB) XPS spectra obtained with this sample orientation
are also more sensitive to the subsurface region, and
the VB XPS analysis depth is expected to be similar to
the high energy (Ep = 1500 eV) ELS measurements. Both
the subsurface VB XPS and ELS measurements are sensitive
to the same region in which previous SIMS results [40]
suggest that hydration occurs.
Results and Discussion
Prior to analysis, the sample was cleaned in_ situ
by heating to 500C in 10 Torr of O2 for 5 minutes. This
procedure removed carbon and chlorine contamination while
leaving a trace amount of K on an otherwise clean oxide
surface as determined by Auger electron spectroscopy
(AES). This contamination is known to be segregated
at the surface [17], and it may be removed easily by
Ar+ bombardment. However, in order to preserve the
hydrated layer of the sample, no ion bombardment was
used. While the O2 treatment may have affected the very
near-surface region of the sample, the following data
reveal that the subsurface layers were not dehydrated.
Figure 4-1 shows the valence band XPS data for the
sample after various treatments. Figure 4-la is the
spectrum recorded after the in. situ cleaning. Figure
4-2a illustrates the effect of a 500C vacuum anneal
on the spectrum. The annealing process was continued
63
until the background chamber pressure went through a
distinct maximum (about 45 minutes). Figure 4-lc shows
the result of a similar (30 minute) 600C anneal. Figures
4-2, 4-3 and 4-4 show the ELS spectra corresponding to
Figures 4-la, 4-lb and 4-lc respectively.
The valence band spectrum in Figure 4-la is
characteristic of a sample hydrated due to exposure to
atmospheric humidity. The large feature near 10 eV binding
energy (7 eV below the valence band maximum (VBM)) is
largely due to this hydration though the form of the
incorporated water is unknown. The 10 eV feature is
similar to that found for water adsorption on other oxides
[70].
Figure 4-2a shows the ELS spectrum (Ep = 1500 eV)
corresponding to Figure 4-la. The features at energy
losses less than about 30 eV are due to
valence-to-conduction-band transitions as discussed in
Section III. The main VB loss feature in Figure 4-2a
falls at 20 eV which is characteristic of a Sn02~like
material (see Section III and ref. 41). Additionally,
a large shoulder associated with the VB loss features
is observed. This feature has not been previously
reportad, and
it
appears
to
be composed
of two
loss
features near
24.5
eV and
27
eV. This shoulder is
not
characteristic
of
SnC>2. '
These additional
features
may
be interpreted as transitions from the hydrate-induced
nce:>
64
Figure 4-1. Valence band XPS spectra after (A) hydration
by exposure to atmospheric humidity, (B) a 500C vacuum
anneal for 45 minutes and (C) a 600C vacuum anneal for
30 minutes.
ncet:>
65
i n 111 i 11111111 ji'i h m n | u i i f 1111 j 1111 i 11 j 11111 i u i n 111 i i n 11
un 11111111111111111111111 in 11111111111111111111111111 m 11111
68 50 46 30 20 10
ENERGY LOSS CEV)
Figure 4-2. ELS spectra of (A) subsurface and (B) surface
regions after hydration due to atmospheric humidity.
66
lower valence band feature observed in Figure 4-la to
conduction band, states. This interpretation is in
agreement with the ultraviolet photoelectron spectroscopy
(UPS) measurements of Gobby [47]. The UPS results show
that the lower valence band feature couples strongly
to conduction band states in a range from 2 5 eV to 3 0
eV higher in energy. Features at energy losses greater
than about 3 0 eV are due to core-to-conduction-band
transitions. In particular, the broad feature centered
near 36 eV is due to a set of 0 2s-to-CB transitions
(see Section III).
Figure 4-2b (Ep = 200 eV) shows the more surface
sensitive ELS spectrum of the hydrated sample. The main
VB loss feature falls near 19 eV suggestive of an Sn02-like
material. The lack of higher energy-loss VB features
in this spectrum indicates a surface which is dehydrated
relative to the subsurface layers. This dehydration
is most likely the result of ESD from the surface under
the influence of the primary electron beam and/or some
superficial dehydration due to the elevated temperature
used during the oxygen cleaning procedure.
The main VB loss features in Figures 4-2a and 4-2b
suggest that the sample is fully oxidized in both the
surface and subsurface regions. This is confirmed for
the surface region in Figure 4-2b by the lack of a 27
eV Sn 4d core level loss feature which would be
67
characteristic of a deficiency of oxygen relative to
SnC>2 (see Section III). Although the presence of such
a 27 eV feature would be obscured in Figure 2a by the
hydrate-induced VB loss feature, the absence of any low
binding energy structure (2 eV to 3 eV) in Figure 4-la
reveals that no oxygen deficiency exists (see Section
III). It is apparent, however, that the structure of
the material is perturbed relative to a well-annealed
SnC>2 rutile structure. This perturbation is evidenced
by the lack of a core level loss feature near 4 5 eV in
Figure 4-2 (see Section III). For the hydrated subsurface
layers, the perturbation is easily understood as due
to the addition of excess oxygen and hydrogen from the
water of hydration while at the surface beam damage is
the most likely cause.
Figure 4-lb shows the effect of a 500C vacuum anneal
on the valence bands. The large feature near 10 eV in
Figure 4-la has been greatly reduced suggestive of a
dehydration of the subsurface region, and a spectrum
very similar to that characteristic of SnC>2 remains (see
Section III and ref. 32). This change is also reflected
in the ELS spectrum in Figure 4-3a. While the main VB
loss feature remains near 20 eV, the features in the
25 eV to 27 eV region associated with the hydrated oxide
are substantially decreased. The valence band ELS features
in Figure 4-3a are quite Sn02 like in agreement with
NCE>
68
t i i j 111111 i i q i n i n i Fi ] 1111111111111111 j i j j 11111 j 111111111111
iiiiilmim nil limn ill miimlm n i mlm mm In in m
ee
S8
46 38 20
QCR6Y LOSS CEV:
16
Figure 4-3. ELS spectra of (A) subsurface and (B) surface
regions after a 500C vacuum anneal.
69
the VB XPS spectrum of Figure 4-lb. Concurrent with
the change in VB features, the appearance of a small
core level loss feature near 45 eV is observed in Figure
4-3a. The weak presence of the 45 eV loss feature
represents the beginning of a change in the subsurface
oxide to a true SnC>2 structure (see Section III). However,
a significant perturbation of this structure is still
apparent. At the surface the annealing has resulted
in an oxygen deficiency as illustrated by Figure 4-3b
(Ep = 200 eV). The broad valence band features with
increased intensity at 13 eV show that the surface has
changed from Sn02~like to a more SnO-like material. The
oxygen deficiency in the surface region is confirmed
by the appearance of the small feature near 27 eV (see
Section III).
Further annealing at 600C has only a small effect
on the VB XPS spectrum shown in Figure 4-lc. The feature
near 10 eV binding energy is completely removed leaving
a spectrum characteristic of Sn02 (see Section III and
ref. 32). Similarly, the ELS spectrum in Figure 4-4a
shows the core level loss features characteristic of
a well-annealed Sn02 material indicating the nearly
complete dehydration of the subsurface region. The
annealing process has, however, effected a reduction
at the surface. The sharp sturcture near 13 eV in Figure
4-4b is characteristic of SnO as found by Powell [41].
71
The broad feature centered near 18 eV is not characteristic
of SnO or SnC>21 and it may be interpreted as due to a
nonstoichiometric defect structure accompanying the change
from subsurface (bulk) SnC>2 to surface SnO (see Section
III) .
The changes observed in the subsurface layers using
ELS clearly indicate a temperature dependence in the
decomposition of the hydrated near-surface region. The
dehydration product observed at 500C is a forerunner
to the formation of a true Sn02 compound near 600C in
the subsurface region. These observations are in agreement
with the work of Giesekke et al. [55] on the thermal
decomposition of bulk tin (IV) hydroxide. The formation
of SngOggl^ could account for the apparent perturbation
of the subsurface crystal structure evidenced by Figure
4-3a while causing only a small variation in the VB density
of states from that expected for SnC>2
Conclusions
The near-surface region of a hydrated polycrystalline
tin oxide film has been studied. A large increase in
the lower VB density of states has been observed for
hydrated subsurface layers using VB XPS and ELS. These
observations are in agreement with SIMS data [40] which
suggests that hydration -due to exposure to atmospheric
O
humidity occurs to depths of at least 30 A.
n < e;>
70
ENERGY LOSS CEV)
Figure 4-4. ELS spectra of (A) subsurface
regions after a 600C vacuum anneal.
and (B) surface
72
The thermal decomposition appears to proceed in
a stepwise fashion. The subsurface hydrated layers yield
SnC>2 near 600C, but the surface undergoes a reduction
to SnO. A comparison with existing data on bulk tin
(IV) hydroxide decomposition leads to an interpretation
consistent with the formation of an intermediate
hydrogen-containing compound in the subsurface region
near 500C.
SECTION V
AN OBSERVATION OF WATER ADSORPTION ON TIN OXIDE
USING ESD AND GRAZING-EXIT-ANGLE
XPS AND AES
Introduction
As discussed in Sections I and IV, the chemisorption
properties of tin oxide surfaces can be significantly-
affected through the interaction with water. In
particular, the chemisorption of platinum on tin oxide
surfaces is believed to occur at hydroxylated surface
sites (see Section II and ref. 6). In Section IV, the
dehydration of tin oxide surfaces has been studied using
valence band x-ray photoelectron spectroscopy (XPS) and
electron energy-loss spectroscopy (ELS). These techniques
have proven particularly useful in studying the subsurface
layers of the material. Though ELS may be made quite
surface sensitive by lowering the primary beam energy,
the study of adsorbates on tin oxide with this technique
is made difficult by the phenomenon of electron-stimulated
desorption (ESD).
Some preliminary observations of water adsorption
on tin oxide are given in this section. ESD experiments
and grazing-exit-angle XPS and Auger electron spectroscopy
(AES) measurements provide this observation. Though
73
74
the data presented here is incomplete, it provides an
interesting comparison with the dehydration study in
Section IV, with previous ESD data on alkaline and
non-alkaline treated tin oxide surfaces [19] and with
the work of Giesekke et al. [55] on the dehydration of
bulk Sn (IV) hydroxide. The incomplete nature of the
ESD experiments is due to a prolonged (16 months and
counting) failure of the Physical Electronics double-pass
CMA. Grazing-exit-angle XPS and AES are used because
of the unavailability of a preferred technique, ultraviolet
photoelectron spectroscopy (UPS).
Experimental
The preparation of the polycrystalline tin oxide
film used in this study has been described in Section
III. After preparation, the sample is rinsed in distilled
water and solvent cleaned prior to insertion into the
vacuum system. Before analysis the sample is cleaned
in situ by heating to 500C in 10 Torr of O2 for 5 minutes.
This procedure removes trace chlorine and carbon surface
contamination leaving a clean oxide surface as determined
by AES. The AES, XPS and ESD data were taken with a
PHI double-pass CMA. The AES spectra are collected in
the nonretarding (EN(E) ) mode using a 3 KeV, 200 mA/cm2
electron beam. For XPS the analyzer was run in the
retarding (N(E)) mode with a pass energy of 5 0 eV ( AE/E
75
= 0.014). Details of the vacuum system have been published
previously [17]. The base pressure for this study was
5xl0-10 Torr.
The surface sensitivity of the electron spectroscopies
(AES and XPS) can be improved by collecting the emission
at angles away from the sample normal, i.e. at a more
grazing exit angle. The path length, L, that an escaping
electron (photoelectron or Auger electron) must travel
through a solid is related to the signal attenuation
due to inelastic collisions. Signal attenuation as a
function of path length can be described by an exponential
decay law with a uniform attenuation length. The
attenuation length, X, is known as the mean free path.
Hence,
I exp(-L/X )
where I is the signal intensity. Figure 5-1 illustrates
the increased surface sensitivity obtained at grazing
exit angles for a perfectly flat surface. If an emitting
source (atom) is a fixed distance, D, below the surface,
the path length, L, traversed within the solid increases
with increasing exit angle, 0, as
L' = D/cos 0
76
Figure 5-1. Variation in path length with emission angle.
77
Therefore, at fixed D, the signal intensity decreases
with increasing 0 making the measurement more surface
sensitive. In general, experimentally observed increases
in surface sensitivity are not as large as expected from
the above analysis. Two possible reasons for the deviation
are surface roughness and a decay of total signal with
increasing 0 causing a reduction in the signal-to-noise
ratio [71].
For all measurements the sample was mounted with
approximately a 4 5 angle between the CMA axis and the
surface normal. This orientation directs the sample
normal into the 42.3 6 acceptance cone of the CMA
[64]. Grazing exit angles are chosen with the 12 angular
acceptance aperture on the angle resolving drum mounted
coaxially within the inner cylinder of the second stage
of the CMA [72]. Using the relationship derived by Gobby
[47], the exit angle for a given drum setting may be
found.
ESD experiments are performed by using the CMA in
a time-of-flight (TOF) mode which allows for a simultaneous
determination of the mass and energy of desorbing ions.
For these measurements the analyzer is operated at a
constant pass energy of about 80 eV. This pass energy
(kinetic energy of the analyzed ions) sets the flight
time of the ions through the analyzer (about 4 nsec for
H+). Because the CMA only passes charged particles of
78
the proper kinetic energy, species of different masses
(but same charge) have different axial velocities through
the CMA. Hence, the flight time of an ion through the
analyzer is directly proportional to the square root
of the mass-to-charge ratio. Traum and Woodruff [73]
have discussed in-depth the analyzer characteristics
which effect the flight time and mass resolution. Unity
mass resolution is possible for mass-to-charge ratios
(m/e) of at least 20.
To operate the CMA in a TOF mode for ESD experiments,
a computer-interfaced digital pulse counting circuit
is used (see Appendix B). The TOF modification to the
pulse counter allows it to perform three functions:
(1) it initiates the desorption event,
(2) it delays for a programmed flight time
(3) and it measures (counts) the signal pulses.
Before beginning the TOF analysis, the coaxial
electron gun in the CMA is configured with a +50 V charge
on the lower deflection plate. This voltage deflects
the electron beam (typically below 200 eV) downward out
of the analysis area (focal region) of the analyzer.
The pulse counter circuit initiates the desorption event
by supplying a 300 nsec TTL pulge to the base of a n-p-n
power transistor in series with the deflection plate
and ground (see Figure 6-2). This pulse "grounds" the
deflection plate and swings the electron beam into the
79
TTL PULSE INPUT
+50 V
5K OHMS
1/
MRF427A
DEFLECTION
PLATE
Figure 5-2. Deflection circuit for desorption
initiation.
event
80
analysis region for 300 nsec. The circuit delays for
a programmed flight time before enabling an event counter
which records signal pulses for a similar 300 nsec period.
The count is subsequently read into the computer where
it is stored, and the process is repeated. By scanning
the programmed delay time a TOF (i.e. m/e) spectrum is
obtained. The only real-time constraint is that a total
time span be observed between desorption events at least
equal to the flight time of the most massive species
in the spectrum. This delay clears the analyzer of ions
before the initiation of a new desorption event.
To analyze low energy positive ions like those
obtained in the ESD experiment, the CMA is operated in
an accelerating mode. The inner cylinder and accelerating
grid which are connected internally are set initially
at -70 V, and the sample is biased at +10 V. This
potential difference between the sample and accelerating
grid raises ions initially at zero kinetic energy up
to the 80 eV analyzer pass energy thereby allowing their
detection. By ramping the accelerating grid to more
positive potentials, ions of higher initial energy
(typically 10 to 20 eV) can be measured. If the TOF
analyzer is operated at a fixed flight time, an energy
distribution spectrum of a single desorbing species may
be obtained. In this experiment, part of the accelerating
potential is imposed on the sample to provide a voltage
81
difference with an outer grid which is grounded to the
magnetic shield of the CMA. In this way any spurious
signal due to ESD from this grid is shifted to apparent
negative kinetic energies and is easily recognizable
[73] .
Results and Discussion
XPS and AES
Measurements have been made following three different
in situ treatments. These treatments include a 500C
and a 600C annealing step as was studied in Section
IV. Measurements have also been made following a 2 KeV
argon ion bombardment. Three types of AES and XPS
measurements are reported. Angle integrated results
are obtained with the angle resolving aperture retracted.
Angle dependent results have been obtained at normal
emission (0 4) and at a 70 2.5 grazing exit angle.
The results are presented in terms of O/Sn ratios. For
XPS this determination is made using the area under the
0 Is and Sn 3d 5/2 peaks corrected with standard
sensitivity factors [31], The AES measurements are made
in a similar fashion using the peak-to-peak heights of
the O KLL (512 eV) and Sn MNN (437 eV) transitions [74],
A problem encountered when using the angle resolving
aperture for these measurements is a drop in total signal
and in signal-to-noise ratio. In AES this drop is not
82
a significant problem because of the magnitude of the
signal, but in XPS the drop is so large that a minimum
seven hour period is required to accumulate enough signal
for a single O/Sn ratio determination. The measurements
were undertaken originally to demonstrate the depletion
of oxygen near the surface due to annealing (as observed
in Section IV) but water adsorption from the background
vacuum is observed instead because of the extended period
of time required to collect the data. Assuming a
background of water at the base pressure for this study
(5x10~5 Torr), the surface receives a 12.5 Langmuir (1L
= lxl06 Torr-sec) dose over a seven hour period. For
a unity sticking coefficient this dose represents about
12 monolayers of water. The water adsorption from the
background vacuum observed here is manifested by an
increased O/Sn ratio in the most surface sensitive (70)
XPS measurement. The adsorption of CO is believed not
to be a contributing factor because no carbon or CO
desorption signal is observed in the subsequent ESD
experiments.
The XPS and AES results are given in Table 5-1.
Angle integrated measurements (with the aperture retracted)
are made quickly after a given treatment before any
significant H2O adsorption occurs. These results indicate
a drop in the O/Sn ratio with increased annealing
temperatures and argon ion bombardment. The angle
83
Table 5-1. Variation in O/Sn Ratio With Emission Angle.
XPS
AES
508C
ANNEAL
ANGLE INTEGRATED 1.4
NORMAL EMISSION 1.5
78 DEGREE 2.1
1.3
600C
ANNEAL
ANGLE INTEGRATED 1.3
NORMAL EMISSION 1.4 -1.5
70 DEGREE 1.8-2.3
l.l 1.2
2K EV
SPUTTER
ANGLE INTEGRATED 1.0
NORMAL EMISSION 1.1
70 DEGREE 1.5
1.0
84
integrated results are consistently similar to, but
slightly lower than, those obtained at normal emission.
This observation illustrates that the signal intensity
is highest at the sample normal as expected for a
polycrystalline material. A similar observation has
been made in Section III regarding the lack of surface
sensitivity in valence band XPS spectra. It is worth
noting that angle integrated measurements taken several
hours after a given treatment show a small increase in
O/Sn ratio like that observed with the seven hour normal
emission measurements.
The increase in O/Sn ratio observed for 70 emission
illustrates a significant uptake of H2O at the surface
from the background vacuum. Indeed, the large O/Sn ratio
of 2.3 observed in one case, suggests the formation of
a hydrated surface. Regardless of the order in which
the data is taken (i.e. the total exposure), the 70
emission always shows a substantially higher O/Sn ratio
indicating H2O adsorption at the surface. Water adsorption
during the normal emission measurements also explains
the small increase observed relative to the angle
integrated measurements.
The AES results given in Table 5-1 show no variation
with exit angle, and the O/Sn ratio is generally lower
than that obtained by angle integrated XPS. The lower
O/Sn value relative to XPS is probably due, in part,
to the increased surface sensitivity of AES. The kinetic
energies of the AES peaks are more than 200 eV less than
85
are observed in AES even after several hours of exposure
to the background vacuum is due to the ESD phenomenon.
As observed in a previous study [19], the surface
concentration of desorbing species can be rapidly depleted
under an electron beam of high current density. It is
believed that water adsorbed on the surface is quickly
removed by the incident beam used for the AES analysis
and is therefore undetected.
ESD
The first TOF ESD measurements made in this laboratory
are reported here. The data was acquired during the
process of tuning the instrument for the first time.
Unfortunately, an electron gun failure ended this
familiarization procedure before a good rapport could
be developed with the experimental set-up. Therefore,
the results shown here do not represent the full
capabilities of the equipment.
Figure 5-3 is the TOF spectrum of a tin oxide surface
sputtered with 2 KeV argon ions. The desorption of H+
and 0+ is clearly visible at 4.4 usee and 17.6 psec,
respectively. The feature at 18.6 nsec is possibly due
to H20+ desorption, but it is most likely due to 0+
desorption with different initial conditions than in
the 17.6 usee peak. These possibilities can be checked
by increasing the accelerating potential and compressing
the flight times in the entrance region of the analyzer.
ION YIELD
86
Figure 5-3. Time-of-flight spectrum for mass analysis.
87
The 12 angular aperture has been used to select species
desorbing at near normal angles. Traum and Woodruff
[73] have shown that a significant increase in mass
resolution is possible by using the 4 angular acceptance
aperture. With the smaller aperture it should be possible
to resolve 0+ and 0H+ species.
Figure 5-4 shows the ion kinetic energy distribution
for the sputtered sample after exposure to a high current
density electron beam for 30 minutes. Figure 5-4a shows
the total ion yield, 5-4b the time-gated (mass resolved)
H+ ion yield and 5-4c the time-gated 0+ ion yield. It
is seen that the CMA may be used for a simultaneous mass
and energy determination.
Figure 5-5 shows the total ion kinetic energy
distribution for the same sample after exposure to the
background vacuum for two hours. No time-gated
distributions were obtained in this case. Though a power
supply problem encountered during the analysis prevents
an accurate determination of the true kinetic energy
scale, the total ion energy distribution is seen to be
very different after H2O adsorption. This variation
suggest that ESD will prove useful in distinguishing
between different forms of hydrogen and oxygen on the
tin oxide surface.
NCE}
88
0 5 10
ION KINETIC ENERGY CEV)
Figure 5-4. Ion kinetic energy distribution after sputter-
ing.
NCE>
89
Figure 5-5. Ion kinetic energy distribution following
water adsorption.
90
Conclusions
Grazing-exit-angle XPS demonstrates that a significant
uptake of H2O occurs on tin oxide surfaces. This uptake
appears to proceed to the point of surface hydration
as expected from similar observations in Section IV.
Grazing-exit-angle AES demonstrates that this hydration
formed under UHV conditions is limited to the outer layers
of the material because it is easily removed by electron
bombardment. The ESD results demonstrate that the
experimental set-up used here is viable. Observed
differences in ion kinetic energy distributions for
sputtered surfaces before and after H2O adsorption
illustrates the potential of the ESD technique as a tool
for studying chemisorption on tin oxide.
SECTION VI
THE INTERACTION OF POLYCRYSTALLINE ZIRCONIUM
WITH 02, N2, CO AND N20
Introduction
Very few surface studies have dealt with zirconium.
Foord, Goddard, and Lambert [75] have studied the
interaction of zirconium with CO, NO, N2, 02 and D2 using
Auger electron spectroscopy (AES), work function
measurements and thermal programmed desorption (TPD).
In a later study Danielson [76] attempted to quantify
the zirconium AES peak heights as a function of amount
of carbon and oxygen contamination. These studies both
demonstrate that, quite unlike most metals, heating causes
adsorbates to migrate from the surface into the bulk
rather than to desorb. This is very convenient with
regard to producing a clean surface, but much information
is lost about the gas-solid interaction because TPD cannot
be used. However, Foord et al. were able to determine
the diffusion coefficients for surface to bulk transport
of C, N and 0 over a range of temperature by monitoring
the surface composition as a function of time with AES.
Only deuterium shows desorption behavior as the sample
is heated.
91
92
The two previous studies disagree on an important
point. Foord et al. claim that the zirconium 175 eV
AES peak height is the most sensitive to chemisorbed
species, but Danielson's study shows the 175 eV peak
height to be almost completely insensitive to surface
carbon and oxygen. This present study relates the
previously observed phenomena to the past history of
the sample, particularly with respect to heating the
sample above the HCP-to-BCC phase transition at 1135K.
The results suggest that the phase transition in the
surface region occurs slowly at room temperature over
a time period of several days and that the chemisorption
properties of the zirconium toward nitrogen change
dramatically over the same period. This appears to be
time-dependent chemisorption on a clean metal surface
due to electronic changes in the valence band caused
by alterations in geometric structure. Poppa and Soria
[77] have recently reported similar reductions in the
amount of CO and H2 adsorbed on (111)-type palladium
layers after annealing at 600K. Low dose, argon ion
bombardment restored the high adsorption probability.
Experimental
The experiments were carried out in an ultrahigh
vacuum chamber which has a base pressure of 3xlO""H Torr.
A PHI double-pass cylindrical mirror analyzer was used
to perform AES and XPS.
93
The zirconium
sample
was a foil
of
approximate
dimensions 12x3x0.5
mm and
99.9% pure.
The
sample was
cleaned in a hydrofluoric acid solution in order to remove
most of the accumulated oxide layer. Next it was solvent
cleaned in ethanol and then mechanically supported between
stainless steel rods. The sample was heated resistively.
Results and Discussion
An AES spectrum showed the sample to be contaminated
initially with carbon, oxygen and nitrogen. These
contaminates were removed easily by heating below the
HCP-BCC phase transition to drive them into the bulk
as discussed in the earlier papers [75,76]. Figure 6-la
shows an AES spectrum taken after 2 hours of heating.
The small amount of nitrogen and oxygen which are on
the surface initially have disappeared, and the carbon
peak shape has changed from graphitic to carbidic after
heating [74]. The oxygen diffuses into the bulk faster
than the carbon which is consistent with the conclusion
of Foord et al. The carbon peak decreases so slowly
that it requires heating overnight to be reduced to the
height shown in Figure 6-lb.. The sample could be cleaned
in a few minutes by heating above the transition
temperature. Cleaning in this manner does not
significantly alter the 'sample because even as much as
20 layers of contamination adds only 1 ppm of bulk
PNCE^/DE ARBITRARY UNITS
94
100 200 300 400 500 600
KINETIC ENERGY CEV)
Figure 6-1. AES spectra taken after (A) 2 hours of heating
and (B) 14 hours of heating below the HCP-to-BCC transition
temperature.
95
contamination after migration from the surface into the
bulk.
The zirconium AES transitions lie below 200 eV.
Foord et al. state that the 92 eV peak is of the type
(MMN + MNN) thus involving only core level electrons,
and the 120 eV signal has three contributions (one from
an MNN transition, one from an MMN transition and one
from an MNV transition). The 150 eV signal arises from
an MNV transition exclusively, and the 175 eV transition
arises exclusively from an MW transition [75]. Thus,
it is reasonable to base relative peak heights on the
92 eV peak which involves only core level electrons and
is relatively insensitive to changes in the valence band.
It is important to notice that the heights of the 150
and 175 eV peaks relative to the 92 eV peak change after
the cleaning process. Foord et al. attribute much of
this variation to contamination by sulfur and chlorine
which would produce AES peaks at 150 and 181 eV,
respectively. Although a similar variation in the AES
peaks is found in this study, XPS demonstrates that both
sulfur and chlorine are absent. Therefore, it is likely
that sulfur or chlorine were not present in the earlier
study. Foord et al. present another explanation later
in their paper in which they claim that the peak heights
are sensitive to changes in the valence band caused by
chemisorption. This explanation is consistent with the
96
present study and with the work of Danielson and will
be discussed in more detail below.
Adsorption of N2, N2O, O2 and CO were performed
by cleaning the sample and bringing the system pressure
up to 5xl0-6 Torr for some period of time (typically
1 to 10 minutes). The sample either was allowed to cool
before gas exposure, allowed to cool during exposure
to the gas, or held at an elevated temperature during
the adsorption process. In general the higher sample
temperatures result in increased amounts of adsorbate
at the surface unless the temperature is so high that
the adsorbed gas diffuses into the bulk more rapidly
resulting in a low surface concentration of adsorbate.
It is observed qualitatively that the room temperature
sticking coefficients are very low. It is estimated
that about 0.1 monolayers of contaminant accumulate on
the surface over a 24 h period at a pressure of 1.0x10^
Torr. This crude observation suggests a sticking
coefficient which is less than 0.01 for 02/ CO, N2 and
N2O. A more rigorous determination of the sticking
coefficients of these gases would be difficult. Since
the gases do not desorb thermally, TPD cannot be used
to determine the amount of gas adsorbed. Similarly,
calibrated uptake measurements would be difficult to
perform because the sticking coefficients are so small.
Pumping and/or outgassing due to the chamber walls, ion
97
gauge, etc., would interfere with the measurements. One
known exception is hydrogen which does thermally desorb.
Lin and Gilbert [78] have measured its sticking coefficient
on zirconium finding that it exhibits a maximum of 6.5xl0-^
near 700K and falls to about 4x10^ below 350K and
above 1000K. These results are similar to the observation
presented here for CO, N2, N2O and O2.
Figure 6-2 shows AES spectra after room temperature
exposure to (a) nitrogen at 5xl0-^ Torr for 5 minutes
and (b) nitrous oxide at 1x10"^ Torr for 2 minutes.
Spectra are not shown for O2 and CO exposures because
CO gives a spectrum with a carbon peak identical to Figure
6-la and a typical oxygen peak as does O2. The influences
of the adsorbates on the zirconium XPS 3d peaks are shown
in Figure 6-3. The clean spectrum is shown in Figure
6-3a for comparison. The spectrum due to oxygen exposure
is shown in Figure 6-3d. It can be deconvoluted into
two spectra due to zirconium oxide and zirconium metal.
The spectrum in Figure 6-4b due to N2 exposure shows
a very small shift of about 0.2 eV with a slight broadening
of the peaks while Figure 6-4c due to N2O exposure shows
the combined effects of both nitrogen and oxygen exposure.
The changes caused by adsorption are even more apparent
in the zirconium 3s region shown in Figure 6-4. The
clean zirconium spectrum is shown in Figure 6-4a for
comparison purposes. Figure 6-4b is due to N2 exposure
DNCE2/DE ARBITRARY UNITS
98
Figure 6-2. AES spectra of state 1 zirconium after room
temperature exposure to (A) nitrogen and (B) nitrous
oxide.
NCET} ARBITRARY UNITS
100
Figure 6-4. XPS spectra showing the zirconium 3s peak
for (A) clean zirconium, (B) nitrogen exposure and (C)
N2O exposure. The' nitrogen Is peak appears at 396 eV
in (B) and (C).
NCE} ARBITRARY UNITS
99
BINDING ENERGY CEV)
Figure 6-3. XPS spectra showing the zirconium 3d peaks
for (A) clean zirconium, (B) N2 exposure, (C) N2O exposure
and (D) oxygen exposure.
101
and Figure 6-4c is due to exposure of clean zirconium
to nitrous oxide. Figure 6-4b shows a splitting of the
zirconium 3s peak into two peaks; one due to metallic
zirconium and a second due to nitride formation which
is shifted by 1.5 eV. Also, a nitrogen Is peak appears
at 396 eV which is characteristic of a nitride [33].
Attempts to produce a nitrogen peak characteristic of
molecular N2 or N2O were unsuccessful at room temperature
or
above.
Adsorption of N2O
(Figure
6-4c) shows a
similar
splitting
in the zirconium
3s peak
as Figure 6
-4b
due
to
nitride
formation and a
similar
nitrogen Is
peak
at
396
eV.
In addition, a
shoulder
has formed
on
the
zirconium 3s peak about 4.5 eV greater in binding energy.
This shoulder is attributed to oxide formation. The
shift is comparable to the shift in the zirconium 3d
peaks due to oxide formation [31]. Figure 6-5 compares
the XPS oxygen Is peak at 530.6 eV due to oxygen adsorption
(Figure 6-5a) with the oxygen Is peak due to N2O adsorption
shown in Figure 6-5b. They are essentially identical
indicating dissociative adsorption of N2O.
The sample temperature only affects the amount of
adsorbate at the surface which depends on both the
adsorption process and the surface-to-bulk diffusion
process. These results indicate that CO, N2, N2O and
O2 adsorb dissociatively on polycrystalline zirconium
at room temperature and above. This observation is in
NCEJ) ARBITRARY UNITS
102
11111111111111111 m jTm 1111111111111111
i 11 1 I 1 1 1 I 1 I 1 I I [ 1 I 1 I I II I 1 I 1 1 1 1 I 1 1 1 I I f I
550 540 530 520 510
BINDING ENERGY CEVO
Figure 6-5. XPS spectra showing the oxygen Is peak after
(A) O2 adsorption on zirconium and (B) N2O adsorption
on zirconium.
DNCE)/DE ARBITRARY UNITS
104
KINETIC ENERGY CV)
Figure 6-6. AES spectrum for clean zirconium after heating
near the melting temperature for 3 hours. The AES 175
eV peak is greatly diminished which is characteristic
of state 2 zirconium.
103
agreement with the conclusion of Foord et al. but provides
more direct evidence of dissociative adsorption.
Unusual adsorption phenomena is observed after heating
the sample at high temperature for prolonged periods.
An AES spectrum taken after cooling the clean sample
to room temperature is shown in Figure 6-6. This spectrum
results from heating near the melting point for about
3 hours. The largest change occurs in the 17 5 eV peak.
It now has become a very small feature whereas in the
AES spectra presented earlier the 175 eV peak is one
of the more prominent features. When the 17 5 eV peak
appears as in Figure 6-6 (i.e., the two peaks in the
170-185 eV region are of comparable size), the zirconium
will be said to be in "state 2," otherwise zirconium
will be referred to as "state 1."
Both CO and N2 adsorption experiments have been
performed after the high temperature heating period (i.e
on state 2 zirconium). The CO adsorption experiments
are performed in three different ways: (1) exposing the
sample to CO contamination from the electron beam for
long periods, (2) dosing the sample at room temperature
and (3) dosing the sample while it is initially hot and
allowing it to cool. It is observed that exposing a
clean sample to the electron beam causes a very slow
growth of carbon and oxygen peaks probably due to CO
cracking by the hot filament. Figure 6-7 shows an AES
105
spectrum of a clean sample after about 8 hours of beam
exposure. The carbon peak appears to contain both carbidic
and graphitic features. The corresponding XPS carbon
Is
peak
is
shown in Figure 6-8.
It
shows
two peaks;
one
at
284
eV
due to graphitic
carbon and
another at
281
eV
due
to
carbidic carbon.
Most
of the
carbon is
in the graphitic form.
CO exposures to clean state 2 zirconium at room
temperature and at high temperature are shown in Figures
6-9a and 6-9b, respectively. The exposure is for 8 minutes
at a pressure of 5x10^ Torr. In the high temperature
exposure, the sample was allowed to cool from above the
transition temperature during the exposure. The two
spectra are very different. At room temperature only
a small amount of CO adsorbs compared to the same CO
exposure for state 1 zirconium. However, much more CO
adsorbs during the high temperature exposure as evidenced
by the size of the carbon peak. The oxygen peak is very
small in Figure 6-9b because oxygen diffuses rapidly
into the bulk at elevated temperature. The corresponding
XPS carbon Is spectra are shown for room temperature
adsorption in Figure 6-10a and for high temperature
adsorption in Figure 6-10b. Both spectra contain peaks
due to graphitic and carbidic carbon. However, the
relative amounts are different. The room temperature
adsorption results in about equal amounts of the two
DNCO/DE ARBITRARY UNITS
106
Figure 6-7. AES spectrum taken after clean state 2 zircon
ium is exposed to CO contamination from the electron
beam for 8 hours. The carbon peak shows characteristics
of both graphitic and carbidic carbon.
NCE> ARBITRARY UNITS
107
Figure 6-8. XPS spectrum of the carbon Is pe§k correspond
ing to the AES spectrum shown in Figure 6-7. Both graphi
tic and carbidic carbon are present.
DNCE^/DE ARBITRARY UNITS
108
KINETIC ENERGY CEV)
Figure 6-9. AES spectra after exposing state 2 zirconiiim
to CO at (A) room temperature and (B) high temperature
but allowing the sample to cool during the exposure.
NCE) ARBITRARY UNITS
109
29S 290 28S 280 27S 270
BINDING ENERGY CEV¡>
Figure 6-10. XPS spectra of the carbon Is peak correspond
ing to the AES spectra shown in Figure 6-9. The room
temperature adsorption produces approximately equal amounts
of graphitic and carbidic carbon as shown in spectrum
(A) "while the high temperature adsorption results in
predominantly carbidic carbon as shown in spectrum (B).
110
types of carbon while the elevated temperature adsorption
results in mostly the carbidic form. These data suggest
that CO adsorbs dissociatively forming an oxide and a
graphitic layer of carbon followed by transformation
of the graphitic carbon into carbidic carbon. The
transformation occurs more rapidly at elevated temperatures
as does diffusion of the oxide and carbidic form of the
carbon into the bulk.
Nitrogen adsorption shows very unusual behavior
for state 2 zirconium compared to state 1 zirconium.
With state 1 zirconium it is easy to adsorb large amounts
of nitrogen at room temperature as evidenced by the large
AES nitrogen peak shown in Figure 6-2a. Figure 6-2a
results from an exposure of 5x10" Torr of nitrogen for
5 minutes. With state 2 zirconium, nitrogen does not
adsorb appreciably at room temperature. This can be
seen in Figure 6-lla which results from a room temperature
exposure to nitrogen at a pressure of 5x10^ Torr for
15 minutes. A small amount of carbon and oxygen
contamination has accumulated during this long exposure
and the subsequent AES run. Figure 6-llb shows a state
2 surface in which the nitrogen adsorption was carried
out on a hot surface which cooled during the exposure.
The exposure was 5xl0-6 Torr of nitrogen for 5 minutes.
More nitrogen adsorbs than during the state 2 room
temperature adsorption, but much more adsorbs during
a state 1 room temperature adsorption.
Ill
IS)
I-
H
z
3
>-
oc
<
O'
t-
H
CD
O'
<
UJ
Q
\
A
LlJ
v
Z
a
Figure 6-11. ('A) AES spectrum taken after exposing state
2 zirconium to nitrogen at 5xl0-^ Torr for 15 minutes
at room temperature. A small amount of carbon and oxygen
contamination accumulated during the long exposure and
subsequent AES run. (B) AES spectrum taken after exposing
state 2 zirconium to nitrogen initially at high temperature
and then allowing the sample to cool during the exposure.
(C) AES spectrum taken after allowing the sample to remain
in vacuum for 3 days at room temperature. State 2 zircon
ium has transformed into state 1 zirconium. (D) AES
spectrum taken after exposing the state 1 zirconium of
spectrum (C) to nitrogen at 5xl0-^ Torr for 5 minutes
at room temperature.
iee 200 300 m 500 600
KINETIC ENERGY CEV)
112
After allowing the state 2 zirconium sample to remain
in vacuum at room temperature for 3 days an unexpected
phenomenon occurs. The AES zirconium 175 eV peak increases
in size as shown in Figure 6-llc. Not only does it
resemble the state 1 17 5 eV peak of earlier scans, but
the sample again adsorbs nitrogen at room temperature
as shown in Figure 6-lld for a 5 minute exposure at 5x10
Torr of N2
Two possible explanations of this time-dependent
chemisorption phenomena are presented here. The first
is that the HCP-BCC transition occurs slowly at least
in the surface region. The change in the atomic positions
appears to cause major changes in the valence electronic
struture as reflected in the AES levels which involve
valence electrons. The fact that nitrogen adsorption
essentially does not occur when the AES zirconium 175
eV peak is absent indicates that the valence electrons
which participate in this AES transition are the valence
electrons which are primarily responsible for bonding
zirconium with nitrogen. An attempt was made to observe
the valence band using XPS, but the cross sections of
the valence electrons for photoemission using x-rays
were to small to yield meaningful information. A similar
statement applies to CO adsorption on zirconium but to
a lesser extent because some CO adsorbs at room temperature
when the AES 175 eV peak is diminished.
113
A second possible explanation is that adsorption
of hydrogen (i.e., the presence of zirconium hydride)
is responsible for the presence of the valence structure
which participates in the AES 175 eV peak. Hydrogen
desorbs from zirconium giving TPD peaks at about 700,
1000 and 1135K [78]. The third peak occurs at the phase
transition temperature and is the largest. This behavior
would explain why the AES 17 5 eV peak diminishes after
heating above the transition temperature and why the
175 eV peak grew back after 3 days of exposure to
background hydrogen in the system. Although the presence
of hydrogen generally cannot be detected with XPS or
AES, it should have been possible to detect the desorbing
hydrogen. However, none was observed. This explanation
also suggests that CO
and N2 adsorb
more
rapidly
on a
hydrided surface
than
on clean zirconium metal
which
seems unlikely.
Also,
heating state
1
zirconium
near
the melting point
for
short periods
of
time (about 1
minute) does not convert it to state 2 even though most
hydrogen initially present would be desorbed. Therefore,
the second possibility does not seem very probable but
will not be ruled out completely without further tests.
As stated previously, Foord et al. report that the
17 5 eV peak height is the most sensitive AES peak to
adsorption while Danielsn states that the 150 eV peak
height is the most sensitive and that the 17 5 eV peak
114
height is relatively insensitive to adsorption. This
present study actually agrees with both Foord et al.
and Danielson by showing the 175 eV peak to be the most
sensitive peak if the sample has not been heated above
the HCP-BCC transition temperature for prolonged periods
while the same 175 eV peak is relatively insensitive
to adsorption after extensive heating above the transition
temperature. This is consistent with the earlier papers
in that Foord et al. argon ion sputtered and typically
heated below the transition temperature while Danielson
routinely heated above the transition temperature at
1200 to 1800K during the adsorption experiments. In
essence Foord et al. studied state 1 zirconium while
Danielson studied state 2 zirconium. Danielson also
used very large carbon monoxide exposures of 1x10^ Torr
for 400 minutes at 1200K for adsorption which overcame
the low sticking coefficient problem for state 2
adsorption. The results of this present study indicate
that the adsorption properties of zirconium and the AES
relative peak heights depend upon the previous history
of the sample. These obervations imply that it is not
particularly useful to plot AES relative peak heights
of. zirconium as a function of adsorbate concentration
unless the zirconium sample is in a well-characterized
state.
115
Conclusions
Zirconium is reactive toward CO, O2, N2 and N2O,
but the room temperature sticking coefficient of these
gases on zirconium is low (C0.01). This low value is
probably due to the fact that the adsorption is
dissociative and requires activation energy to proceed.
No evidence has been found which indicates that these
molecules bond molecularly to zirconium at room temperature
or above although they may bond molecularly below room
temperature.
A clean zirconium surface can be in two different
"states" as reflected by the chemisorption properties
and the electronic structure given by AES. State 1 occurs
when the sample has not been heated for prolonged periods
(many hours depending upon the temperature) above the
transition temperature, and state 2 occurs after prolonged
heating above the transition temperature. State 1 is
characterized by a large AES zirconium 17 5 eV peak while
this peak is very small for state 2. This peak is due
to an MW transition thus indicating that the valence
electronic strutures of the two states are different.
State 1 adsorbs CO and N2 much more readily than state
2. It can be concluded that the valence electrons
responsible for the AES 175 eV peak also participate
in the adsorption process. The transformation between
state 1 and state 2 is reversible. The rate of
116
transformation of state 1 to state 2 is much faster than
for the state 2-to-state 1 transformation because it
occurs at high temperature. This behavior is also
characteristic of a phase change. The bulk phase
transformation from HCP to BCC occurs rapidly [78], but
apparently the transformation in the surface region occurs
slowly.
SECTION VII
GENERAL CONCLUSIONS AND
RECOMMENDATIONS FOR FUTURE RESEARCH
Pt Tin Oxide
A feature has been identified in the Pt 4f XPS
spectrum associated with the bond formed between supported
platinum and the tin oxide substrate. The bond is believed
to form with surface lattice oxygen resulting in a Pt-O-Sn
surface species. This substrate-bonded species appears
to act as a nucleation site for crystallite growth in
both the electrochemical deposition of platinum and in
the sintering of supported platinum.
Future attempts to characterize the metal-substrate
interaction may be aided by the study of a more
geometrically ideal system. Low-energy electron
diffraction (LEED) studies of the SnC>2 (110) face have
shown that several specific types of oxygen-deficient
defect structures may be obtained by annealing at different
temperatures [58]. By vapor depositing Pt in. situ on
a tin oxide single crystal the interaction with specific
defect surfaces and tin atoms of varying coordination
could be studied.
It has been demonstrated that the oxidation state
of supported platinum may be manipulated in. situ and
118
characterized by XPS. This observation should prove
useful in future molecular beam studies relating catalytic
activity to surface characterization. It is recommended
that the first such attempts should be directed at tin
oxide surfaces with high platinum loadings. High loadings
will benefit the XPS analysis by increasing the
signal-to-noise ratio and should benefit the molecular
beam studies by providing more reaction products.
An acceptable technique for characterizing tin oxide
has been developed. It has been demonstrated that ELS
is useful in characterizing both the surface and subsurface
regions of the tin oxide support. The interpretation
of the ELS data presented here also provides a basis
for the interpretation of ESD threshold measurements
in the study of chemisorption on tin oxide surfaces.
It is felt that the full potential of the ELS
technique in characterizing tin oxide has not yet been
realized. It is recommended that future development
of this technique proceed though the study of derivative
loss spectra, i.e. N(E)' or N(E)". A significant amount
of detail is believed to be lost in the background of
N(E) measurements such as those presented here. Also,
it may be possible to obtain a more complete understanding
of the atomic nature of the conduction bands by studying
core level losses associated with 0 Is, Sn 3d and Sn
3p levels using dipole selection rules.
119
It has been demonstrated that the experimental tools
developed for ESD measurements are in place and function
properly. Preliminary data has been presented which
suggests that ESD will prove very useful in studying
the interaction of water (and other adsorbates) with
tin oxide. Further work in this area should attempt
to apply the full capabilities of the ESD technique to
the identification of adsorbed species and surface binding
sites. It is felt that combining ESD and UPS measurements
would be a powerful approach for studying adsorption
(particularly of water) on tin oxide surfaces.
Zirconium
The chemisorption properties of polycrystalline
zirconium have been found to vary dramatically depending
on the thermal history of the sample. Chemisorption
on this surface is found to be suppressed by heating
for prolonged periods of time above the HCP-to-BCC phase
transition temperature at 1135K. The chemisorption
behavior can be correlated roughly with the appearance
or disappearance of a zirconium MW Auger peak. A slow
phase transition at the surface has been postulated as
the cause of the variation in chemisorption properties.
It is felt that future work on zirconium may benefit
from ESD measurements. 'Variations in surface geometry
or in the types of adsorption sites available on "state
120
1" and "state 2" zirconium may be detectable in ESD as
changes in the desorbing ion kinetic energy distribution.
APPENDIX A
A BRIEF DESCRIPTION OF
THE EXPERIMENTAL TECHNIQUES
X-Ray Photoelectron Spectroscopy (XPS)
Surface analysis by XPS involves irradiating a sample
with nearly monoenergetic soft x-rays and energy analyzing
the emitted electrons. The photons, in this case Mg
Ka x-rays at 1253.6 eV, interact with atoms in the solid
causing electrons to be emitted by the photoelectric
effect. The photoemission process is illustrated in
Figure A-l. The emitted electrons have kinetic energies
given by
KE = hv BE s
where h v is the photon energy, BE is the binding energy
of the orbital from which the electron originates and
s is the spectrometer work function. By analyzing the
kinetic energy distribution of the emitted electrons
a spectrum is obtained which corresponds (roughly) to
the number denisty of electrons per binding energy
interval. In general, electrons from narrow, well-defined
energy states (i.e. core levels) are those of interest.
121
122
VALENCE
LEVELS
CORE
LEVELS
Figure A-l.
Photoemission process.
123
The binding energy of a given peak in the spectrum
may be regarded as the ionization energy of that particular
shell (orbital) in the emitting atom. Because each element
has a unique elemental spectrum, the observed peaks from
a mixture are approximately the sum of the elemental
peaks of the constituents. However, the binding energy
(ionization energy) of a given orbital in an emitting
atom may be effected by the chemical environment. Changes
in the valence electronic structure of an atom due to
chemical bonding may be reflected as a "chemical shift"
in the measured binding energies of the core level
electrons. Core level binding energies may therefore
be used as an indicator of the "valence (oxidation) state"
of an atom. For example, the reported binding energies
of iron 2p 3/2 peaks for the metal and oxides are 706.8
eV, 709.3 eV and 711.0 eV for Fe, FeO and Fe2C>3,
respectively [31].
The surface sensitivity of the XPS measurement is
determined by the mean free path of the emitted electrons.
The mean free path is (roughly) the average distance
an electron is expected to travel in a solid without
undergoing an inelastic collision with other electrons
(see Section V for a more accurate definition). The
mean free path generally decreases with decreasing electron
kinetic energy. Emitted electrons undergoing inelastic
collisions lose some fraction of their kinetic energy
and appear in the spectrum at an apparent binding energy
124
greater than that of the parent photoelectron peak. These
"loss" features generally show up as a broad background
from which the elastically emitted electrons are easily
distinguished. The mean free path of electrons emitted
, O
in XPS generally ranges from 5 to 20 A. This short mean
free path results in a parent photoelectron line which
originates from atoms in the "top few" atomic layers
of the solid.
Auger Electron Spectroscopy (AES)
Surface analysis by AES is based on the radiationless
decay process discovered by P. Auger in 1925. Figure
A-2 illustrates the process. The process is initiated
by the creation of a core hole (i.e. ejection of a core
level electron) typically excited by an impinging electron
beam. The creation of the core hole leaves an ion in
an excited state. The atom subsequently decays to a
doubly-ionized lower energy state when an electron from
a higher energy level drops into the core hole and
simultaneously releases its energy to another (Auger)
electron which is emitted from the atom. The energy
given up in the transition from the singly to doubly
ionized state is absorbed by the Auger electron and
determines its kinetic energy. As in XPS, an energy
analysis of the emitted electrons is performed to determine
the positions (kinetic energies) of the Auger electron
peaks. Because each atom gives a characteristic spectrum,
125
O
Lj C2S)
K CIS)
Figure A-2, KL]_L2 Auger decay process.
126
the kinetic energies of the Auger electrons can be used
to identify the composition of the solid surface.
The core hole which is created to initiate the process
is like that which is left after the ionization of a
core level in XPS. Indeed, Auger peaks may be observed
in XPS spectra as a result of the core-hole (Auger) decay
process. Unlike the photoelectron in XPS however, the
kinetic energy of the Auger electron is set by the decay
process and is independent of the energy of the ionizing
radiation. As in XPS, the surface sensitivity of AES
is set by the kinetic energies (short mean free paths)
of the emitted electrons.
Electron Energy-Loss Spectroscopy (ELS)
The ELS measurements presented in Section III and
IV are for energy losses of the order of electronic
transitions (see Section III for a more complete
discussion). In ELS an impinging (primary) beam of
electrons at energy Ep strikes the solid and excites
various electronic processes such as valence and core
level ionizations, Auger processes and plasmon excitations.
The kinetic energy distribution of backscattered electrons
near the primary beam energy is measured. The primary
electrons which strike the surface and are backscattered
to the detector may interact either elastically (i.e.
with no loss of energy) or inelastically giving up some
fraction of their energy to the "loss" processes. Energy
127
UNOCCUPIED
LEVEL
OCCUPIED
LEVEL
Ep -AE
Figure A-3. Electron energy-loss process.
128
differences between the elastically and inelastically
scattered electrons are reported as energy losses and
(hopefully) related to the characteristic processes that
occur within the solid. An idealized illustration is
given in Figure A-3. The power of this technique is
that it probes not only the filled electronic levels
of the solid (like AES and XPS) but also the higher lying
unoccupied levels. The result is an often complex spectrum
which is a convoluted picture of both the filled and
unfilled states.
The ELS measurement probes the inelastic processes
which are generally responsible for the surface sensitivity
of XPS and AES. The analysis depth of the technique
is mainly dependent on the primary beam energy. From
mean free path considerations it can be seen that the
higher the incident beam energy the deeper the analysis
region within the sample.
Electron-Stimulated Desorption (ESP)
When an electron beam is directed at a sample it
can result in the desorption of cations, anions and neutral
species. For the results presented here a mass analysis
of desorbing positive ions is performed (see Section
V). The ESD technique possesses an extreme surface
sensitivity because of the very short mean free path
of low energy cations in a solid. The high probability
of reneutralization results in the escape and detection
129
of cations from the outer layer of the material only.
Because the ESD technique involves a mass analysis of
desorbing species it is directly sensitive to surface
hydrogen unlike XPS, AES and ELS. This makes ESD a
particularly useful technique for studying H2 and H2O
adsorption.
The ESD of positive ions from metal oxide surfaces
is generally thought to occur as the result of an Auger
decay process [30], This model proposed by Knotek and
Feibelman (KF) has been successful in correlating large
amounts of experimental data for metal oxide systems.
In the KF model the process is begun (as in AES) by the
creation of a core hole. During the subsequent Auger
decay process the resulting positive ion may desorb if
it has a kinetic energy in excess of the surface binding
energy.
While normal (intra-atomic) Auger processes may
cause desorption, significantly more information may
be obtained from inter-atomic processes. In the
inter-atomic decay mechanism the core hole created on
a particular atom is filled by a decaying electron from
a nearest neighbor atom. The ionized nearest neighbor
then desorbs and is detected. By correlating the
desorption of a particular species with the threshold
ionization energies of surface atom core levels the ESD
measurement becomes a specific probe of the surface binding
site (i.e. nearest neighbor) of the desorbing species.
130
To date, only one conflicting example has been found
where second nearest neighbor desorption results from
a core level ionization process in an oxide [79].
APPENDIX B
COMPUTER-INTERFACED
DIGITAL PULSE COUNTING CIRCUIT
Introduction
Pulse counting is an important means of detecting
signals when they are particularly small as in x-ray
photoelectron spectroscopy (XPS), ultraviolet photoemission
spectroscopy (UPS), electron-stimulated desorption (ESD),
electron energy-loss spectroscopy (ELS) and numerous
others. Pulse counting can be performed using a
commercially available analog ratemeter, but it is more
convenient to use a computer-interfaced, digital pulse
counter because the timing of the data collection process
can be controlled precisely thereby allowing the use
of an on-off heater circuit [80], multiple scans can
be added together readily and the original data is stored
easily so that it can be recalled or digitally filtered
[81], as necessary. A digital pulse counter which
interfaces to a laboratory computer through a general
purpose 16-bit parallel interface is described here.
To start a count the computer sends a control word
containing the desired counting time to the parallel
interface. Pulses are sent to the counter after
131
132
amplification and discrimination. When counting is
finished the device sends a signal back to the parallel
interface. The computer can either wait for this signal
or use it as an interrupt. The pulse count is read then
by the computer as a 16-bit word.
Cricuit Description
The counter consists of the following four parts:
(1) an onboard timer circuit, (2) a timing counter, (3)
an event counter, and (4) a control logic section. The
on-board timer (Figure B-l) is a 1.0000-MHz crystal
followed by four stages of divide-by-ten logic (Type
74390). Thus base frequencies of 100, 10, and 1 KHz
and 100 Hz are available for timing. These frequencies
are jumper selectable. Since the device counts pulses
over any period from 1 to 65,535 (177777 octal) clock
cycles, timing intervals are available from 10 psec to
655 sec. Of course, all of the above numbers may be
changed easily to suit the user, and 1.0 MHz is available
directly off the clock. Higher-frequency clocks can
be used and may be necessary if higher time resolution
is required as in the time-of-flight modification discussed
later in this appendix. However, the clock frequency
must not exceed the response time of the integrated
circuits. It should be .noted that the use of a higher
frequency clock will alter the base frequencies and timing
intervals accordingly.
133
74390
74390
SYS
CLK
= Ref. o r- o
Figure B-l. On-board timer schematic showing jumper
selectable system clock rates.
100 KHz
134
Figure B-2 shows the control logic section as used
with a Heath Model H-ll-2 parallel interface module and
the LSI 11 (Digital Equipment Corporation) computer.
When the command word is sent from the computer to the
parallel
interface
the
latter
generates a
"take
data"
signal (TD). When
this
signal
goes low it
clocks
a HI
into the
first of
three D-type
flip-flops
(Type
7474).
The next
pulse from
the
on-board
timer clocks
this
signal
synchronously into the second flip-flop causing Q2 to
go LO. Q2 clears the first flip-flop and thereby restores
itself to the HI state at the next clock pulse. Thus
Q2 is a single HI-LO-HI pulse of width equal to one timing
cycle. It is used to load the timing counter, to clear
the event counter, to start the counting and (if necessary)
to report back to the parallel interface that the command
word has been received.
Counting is started by simultaneously enabling the
event counter and gating timing pulses into the timing
counter. This is done through the third flip-flop which
is clocked HI by Q2.
Counting is stopped in two ways. When the timing
counter reaches zero, its final borrow signal goes LO.
When the event counter fills (overflow condition), its
final carry signal goes HI. The carry signal is inverted
and then ANDed with the borrow signal so that when either
event occurs the third flip-flop is cleared, a "data
135
7474 Vi 7474
Figure B-2. Schematic of the control logic section
136
sent" (DS) signal is returned to the parallel interface
and counting stops. Computer software can check the
count for the value 177777 (octal) which would indicate
an overflow condition.
Figure B-3 shows the timing and event counters.
The former is a series of four synchronous 4-bit binary
up-down counters (Type 74193) wired to count down. An
initial value is applied to this counter through the
C inputs and loaded by the pulse at Q2 (see Figure B-2).
This value is the command word sent from the computer
and is available at the parallel interface when the TD
signal goes LO. Timing pulses are gated into the least
significant bit (LSB) of this counter. The borrow signal
from the most significant bit (MSB) goes LO when the
counter has reached zero.
The event counter is a series of four asynchronous
4-bit binary upcounters (Type 74161). These counters
are cleared when the pulse at Q2 goes LO, and counting
is enabled when that pulse returns to HI. Should the
counter saturate, the carry from the MSB will go HI and
stop the process.
Figure B-4 shows the actual wiring diagram. The
integrated circuits (ICs) as
laid
out will
fit
on a
4"
x 8"
printed circuit board.
The
7 400 and
7408
ICs
are
NAND
and AND gates. There
are
37 connections
to
the
parallel interface (C0-C15,
D0-D15, TD, DT
, DS
, +5
V,
137
C15C14C13C12
C11 C10C9C8
C7 CSCS C4
C3 C2C1 CO
C = INITIAL VALUE FOR TIMING COUNTDOWN
(A) TIMING COUNTER
D15D14D13D12
D11 D10 09D8
D7D6D5D4
D3 D2 D1 DO
(B) EVENT COUNTER
Figure B-3. (A) Timing and (B) event counter schematics.
138
LOAO (L)
Figure B-4. Layout and wiring
(pull up) resistors are 10K ohms.
diagram.
All unmarked
139
and REF.) and one EVENT connection. This last connection
should be shielded properly to avoid counting spurious
pulses.
Time-of-Flight Modification
This circuit may be modified easily to perform another
important class of experiments such as electron-stimulated
desorption (ESD) in which an electrostatic analyzer can
be used as a time-of-flight (TOF) mass spectrometer [73].
In this type of experiment an initial signal causes ions
to desorb off a sample. The different masses require
different flight times to reach the counting circuitry
so a delay is required before counting for a set period
of time.
In the circuit as described in the previous section,
both the timing counter and event counter are started
as a result of the TD signal. For TOF measurements the
TD signal is activated simultaneously with the ionizing
event at the surface at time zero. The "TD signal is
used to start the timing counter as before but not the
event counter. The delay period has been loaded previously
into the timing counter so that after the timing counter
has counted down its final borrow signal is used to enable
the event counter which counts for a selected period
of time.
A brief description of the ESD TOF circuit
modification is given here. The output of the third
140
flip-flop, Q3 (Figure B-2) is used as a trigger for
the Bl input of a 74LS123 one-shot. A potentiometer
is connected across the external timing circuit of the
one-shot to allow the output pulse width to be varied.
The one-shot pulse (typically 300 nsec) activates a MRF427A
power transistor which is used to switch 50 V off a
deflection plate in order to pulse an electron beam onto
the sample. At the end of the delay period, the final
borrow from the timing counter triggers the A2 input
of the 74LS123 one-shot. The output pulse of this one-shot
is used to enable the event counter, and pulses are
collected for a
time
period equal
to the
pulse
width
of the electron
beam.
The negative
edge of
the
output
of this one-shot
is
used to trigger another one-shot
which clears the
third
flip-flop and
sends a
"data
sent"
(DS) signal to the parallel board. A time resolution
of 0.1 psec can be obtained if a 10-MHz clock is used
with no divide-by-ten circuitry.
Acknowledgments
The original version of the device described here
was conceived by G.B. Hoflund. The first working model
was designed and constructed by R.E. Gilbert with helpful
discussions from Sonny Johnson. A version incorporating
the TOF modification was conceived by G.B. Hoflund and
D.F. Cox using ideas contributed by Mort Traum. This
141
second version was designed and constructed by D.F. Cox
with helpful discussions from R.E. Gilbert.
REFERENCES
1.
M.M.P. Janssen
21(1976)861.
and J.
Moolhuysen,
Electrochim.
Acta
2.
M.M.P.
Janssen
and J.
Moolhuysen,
Electrochim.
Acta
21(1976)869.
3. M.M.P. Janssen and J. Moolhuysen, J. Catal.
46(1977)289.
4. M.R. Andrew, J.S. Drury, B.D. McNicol, C. Pinnington
and R.T. Short, J. Appl. Electrochem. 6(1976)99.
5. A. Katayama, J. Phys. Chem. 84(1980)376.
6. M. Watanabe, S. Venkatesen and H.A. Laitinen, J.
Electrochem. Soc. 130(1983)59.
7. A.C.C. Tseung and S.C. Dhara, Electrochim. Acta
19(1974)845.
8. G.B. Hoflund, D.A. Asbury and R.E. Gilbert, to be
published.
9. R.R. Cecil, W.S. Kmak, J.H. Sinfelt and L.W. Chambers,
Oil Gas J. 7(1972)50.
10. E.M. Blue, Hydrocarbon Processes 48(1969)141.
11. M.J. Fuller and M.E. Warwick, J. Catal. 29(1973)441.
12. B. Hori, N. Takezawa and H. Kobayashi, J. Catal.
80(1983)437.
13. F. Solymosi and J. Kiss, J. Catal. 41(1976)202.
14. N. Niwa, T. Minami, H. Kodama, T. Hattori and Y.
Murakami, J. Catal. 53(1978)198.
15. K. Wakabayashi, Y. Kamiya and N. Ohta, Bull. Chem.
Soc. Japan 40(1967)2172.
142
and R. Reid, J. Catal.
143
16. H.H. Herniman, D.R. Pyke
58(1979)68.
17. G.B. Hoflund, D.F. Cox, G.L. Woodson and H.A.
Laitinen, Thin Solid Films 78(1981)357.
18. G.B. Hoflund, D.F. Cox and H.A. Laitinen, Thin Solid
Films 83(1981)261.
19. G.B. Hoflund, D.F. Cox, F. Ohuchi, P.H. Holloway
and H.A. Laitinen, Appl. Surface Sci. 14(1982-83)281.
20. C.E. Smith, J.P. Biberian and G.A. Somorjai, J.
Catal. 57(1979)426.
21. R. Bacaud, P. Bussiere and F. Figueras, J. Catal.
69(1981)399.
22. Z. Zarpinsky and J.K.A. Clarke, J. Chem. Soc. Faraday
Trans. I 71(1975)893.
23. H. Verbeek and W.M.H. Sachtler, J. Catal. 46(1976)257.
24. N. A. Pakhomov, R.A. Buyanov, E.M. Moroz, E.N.
Yurchenko, A.P. Chernyshey, N.A. Zaitseva and G.R.
Kotelnikov, React. Kinet. Catal. Lett. 14(1980)329.
25.
A.
Katayama, Chem. Lett. (1978)1263.
26.
D. Elliot, D.L. Zellmer and
Electrochem. Soc. 117(1970)1343.
H.A.
Laitinen, J.
27.
Yu.
Kul
E. Roginskaya, D.A. Dulin,
'kova and A.I. Gel'bshtein,
S.S. Stroeva, N.V.
Kinetika i Kataliz
9(1968)1143.
28. H.A. Laitinen and T.M. Hseu, Anal. Chem. 51(1979)1550.
29. M. Yokoyama and H.A. Laitinen, unpublished results.
30. M.L. Knotek and P.J. Feibelman, Phys. Rev. Letters
40(1978)964.
31. C.D. Wagner, W.M. Riggs, L.E. Davis, J.F. Moulder
and G.E. Muilenburg, Handbook of X-Ray Photoelectron
Spectroscopy, Perkin-Elmer, 1979.
32. C.L. Lau and G.K. Wertheim, J. Vac. Sci. Technol.
15(1978)622.
33. T.W. Capehart and S.C. Chang, J. Vac.
18(1981)393.
Sci.
Technol.
144
34. W.E. Morgan and J.R. VanWazer, J. Phys. Chem.
77(1973)964.
35. J. Rosenfarb, H. A. Laitinen, J.T. Sanders and H.A.
VanRinsvelt, Anal. Chim. Acta 108(1979)119.
36. M.G. Mason, Phys. Rev. B 27(1983)748.
37. P.H. Citrin and G.K. Wertheim, Phys. Rev. B
27(1983)3176.
38. R. Bouwman and P. Biloen, Anal. Chem. 46(1974)136.
39. M. Komiyama, Y. Ogino, Y. Akai and M. Goto, J. Chem.
Soc. Faraday Trans. 79(1983)1719.
40. D.F. Cox, G.B. Hoflund, W.H. Hocking and H.A.
Laitinen, to be published.
41. R.A. Powell, Appl. Surface Sci. 2(1979)397.
42. R.A. Powell and W.E. Spicer, Surface Sci. 55(1976)681.
43. P.A. Cox, R.G. Egdell, C. Harding, A.F. Orchard,
W.R. Patterson and P.J. Tavener, Solid State Commun.
44(1982)837.
44. J. Robertson, J. Phys. C: Solid State Phys.
12(1979)4767.
45. M.L. Knotek, Surface Sci. 101(1980)334.
46. M.L. Knotek, V.O. Jones and V. Rehn, Phys. Rev.
Letters 43(1979)300.
47. P.L. Gobby, Ph.D. Dissertation (Montana State
University, May 1977), University Microfilms
International, Ann Arbor.
48. E. Rudberg and J.C. Slater, Phys. Rev. 50(1936)150.
49. J.E. Rowe, Solid State Commun. 15(1974)1505.
50. R. Ludeke and A. Koma, Phys. Rev. Letters 34(1975)817.
51. E. Colavita, M. DeCrescenzi, L. Papagno, R.
Scarmozzino, L.S. Caputi, R. Rosei and E. Tosatti,
Phys. Rev. B 25(1982)2490.
52. A. Messiah, Quantum' Mechanics, Vol. II, John Wiley
& Sons, New York.
145
53.R. Ludeke and L. Esaki, Phys. Rev. Letters
33(1974)653.
54. W. Spence, J. Appl. Phys. 38(1967)3767.
55. E.W. Giesekke, H.S. Gutowsky, P. Kirkov and H.A.
Laitinen, Inorganic Chem. 6(1967)1294.
56.
R. Summit, J.A. Marley and N.F.
Chem. Solids 25(1964)1465.
Borelli, J.
Phys.
57.
R.D. McRoberts, C.G. Fonstad and
Rev. B 10(1974)5213.
D.
Hubert,
Phys.
00
LO
E. deFre^art, J. Darville and J
.M.
Giles,
Solid
State Commun. 37(1980)13.
59. N. Yamazoe, J. Fuchigami, M. Kishikawa and T. Seiyama,
Surface Sci. 86(1979)335.
60. S. Munnix and M. Schmeits, Phys. Rev. B 27(1983)7624.
61. A.J. Bevolo, J.D. Verhoeven and M. Noack, J. Vac.
Sci. Technol. 20(1982)943.
62. P. Bayat-Mokhtari, S.M. Barlow and T.E. Gallon,
Surface Sci. 83(1979)131.
63. H.A. Bethe and E.E. Salpeter, Quantum Mechanics
of One- and Two-Electron Atoms, Springer-Verlag,
Berlin, 1957.
64. P.W. Palmberg, J. Vac. Sci. Technol. 12(1975)379.
65. T. Kaji, H. Yamazoe and T. Seiyama, Nippon Kagaku
Zassi 7(1980)1088.
66. H.A. Laitinen and N.H. Watkins, Anal. Chem.
47(1975)1352.
67. H.A. Laitinen and N.H. Watkins, J. Electrochem.
Soc. 123(1976)804.
68. H. Yoneyama and H.A. Laitinen, J. Electroanal. Chem.
Interfacial Electrochem. 75(1977)647.
69. H. Yoneyama and H.A. Laitinen, J. Electroanal. Chem.
Interfacial Electrochem. 79(1977)129.
70. See for example: R.L. Kurtz and V.E. Henrich, Phys.
Rev. B 26(1982)6682, for results on T2O3; V.E.
Henrich, G. Dresselhaus and H.J. Zieger, Solid State
Commun. 24(1977)623, for results on Ti02-
146
71. D. Menzel, in: Photoemission and the Electronic
Properties of Surfaces, Eds. B. Feuerbacher, B.
Fitton and R.F. Willis, John Wiley & Sons, New York,
1978.
72. J.A. Knapp, G.J. Lapeyre, N.V. Smith and M.M. Traum,
Rev. Sci. Instrum. 53(1982)781.
73. M.M. Traum and D.P. Woodruff, J. Vac. Sci. Technol.
17(1980)1202.
74. L.E. Davis, N.C. McDonald, P.W. Palmberg, G. Riach
and R. Weber, Handbook of Auger Electron Spectroscopy,
Perkin-Elmer, 1978.
75. J.S. Foord, P.J. Goddard and R.M. Lambert, Surface
Sci. 94(1980)339.
76. L.R. Danielson, J. Vac. Sci. Technol. 20(1982)86.
77. H. Poppa and F. Soria, Phys. Rev. B 27(1983)5166.
78. J.-M. Lin and R.E. Gilbert, to appear in Appl. Surface
Sci.
79. Y.X. Wang, F. Ohuchi and P.H. Holloway, J. Vac.
Sci. Technol. A 2(1984)732.
80. G.B. Hoflund and R.E. Gilbert, J. Vac. Sci. Technol.
18(1981)56.
81. A. Savitzky and M.J.E. Golay, Anal. Chem.
36(1964)1627.
BIOGRAPHICAL SKETCH
David Fullen
Cox
was born in
Marietta,
Georgia,
July 4, 1956.
He
is the son of
Marjorie
and Loyd
and has one
older
sister, Nancy.
David was
; married
to the former Teresa Louise Ludovici on August 11, 1984.
They have no children.
David was raised in Marietta, Georgia, a town in
the northern metropolitan Atlanta area, where he attended
school and graduated from Marietta High School in 1974.
He entered the University of Tennessee in the Fall of
1974 and participated in the Engineering Cooperative
Education Program before receiving his Bachelor of Science
in Chemical Engineering in 1979. David entered graduate
school in chemical engineering at the University of Florida
in the Fall of 1979 and received his Master of Science
degree in 1980. He continued in the Ph.D. program at
the University of Florida where he has been until the
present.
147
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Gar B. Hoflnd, Chariman
Associate Professor of Chemical
Engineering
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
j pCLJ J~f
Paul H. Holloway
Professor of Materials Science
and Engineering
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
t.-.-. iPTt
Herbert A. Laitinen
Graduate Research Professor
of Chemistry
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
U
Joh^' P. O'Connell
Chairman and Professor of
Chemical Engineering
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Dinesh 0. Shah
Professor of Chemical
Engineering
This dissertation was submitted to the Graduate
Faculty of the College of Engineering and to the Graduate
School, and was accepted as partial fulfillment of the
requirements for the Degree of Doctor of Philosophy.
/ LsJlu/ Ct
Dean, College of Engineering
Dean for Graduate Studies
and Research
December, 1984
UFDissertations, 08:38 AM 6/10/2008, UF Libraries:Digital Dissertation Project
Internet Distribution Consent Agreement
In reference to the following dissertation:
AUTHOR:
Cox, David
TITLE:
Surface characterization and chemisorption properties of polycrystalline systems : Sn02 Pt/Sn02 and Zr /
(record number: 496891)
PUBLICATION DATE:
1984
, as copyright holder for the aforementioned
dissertation, hereby grant specific and limited archive and distribution rights to the Board of Trustees of
the University of Florida and its agents. I authorize the University of Florida to digitize and distribute the
dissertation described above for nonprofit, educational purposes via the Internet or successive
technologies.
This is a non-exclusive grant of permissions for specific off-line and on-line uses for an indefinite term.
Off-line uses shall be limited to those specifically allowed by "Fair Use" as prescribed by the terms of
United States copyright legislation (cf, Title 17, U.S. Code) as well as to the maintenance and preservation
of a digital archive copy. Digitization allows the University of Florida to generate image- and text-based
versions as appropriate and to provide and enhance access using search software.
This gi ant of permissions prohibits use of the digitized versions for commercial use or profit.
Signature of Copyright Holder
7c5
Printed or Typed Name of Copyright Holder/Licensee
Personal information blurred
>//ofe R
Date of Signature
Printed for "David F. Cox"
2
SURFACE CHARACTERIZATION AND CHEMISORPTION PROPERTIES
OF POLYCRYSTALLINE SYSTEMS:
Sn02, Pt/Sn02 and Zr
By
DAVID FULLEN COX
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1984
Copyright 1984
by
David Fullen Cox
ACKNOWLEDGMENTS
The author would like to thank Gar Hoflund for the
guidance and encouragement he furnished in his role as
research advisor. Special thanks go to Gar also for
many hours spent in the discussion of academic and purely
nonacademic matters and for invaluable assistance rendered
in the author's quest to master the French (not to mention
the English) language, pate7 de foie gras! Thanks go
also to Herb Laitinen for the benefit of his expertise
on tin oxide in all its forms and for providing the
laboratory facilities used for sample preparation. Thanks
go to Paul Holloway for his helpfulness in allowing the
use of his SIMS apparatus and for his patience in enduring
the associated visits. A special thank you also goes
to Dick Gilbert at the University of Nebraska for a
multitude of software and hardware contributions which
played such a major role in all the results presented
here.
The author thanks G.B Hoflund and H.A. Laitinen
for financial support supplied through sponsored research
grants. Thanks go also to the Department of Chemical
Engineering and the State of Florida for financial support.
Lastly, the author acknowledges the Florida Guaranteed
Student Loan Program for unilateral support during the
final six months of his degree program.
iii
TABLE OF CONTENTS
PAGE
ACKNOWLEDGMENTS iii
LIST OF TABLES vi
LIST OF FIGURES vii
ABSTRACT xi
SECTION
ONE GENERAL INTRODUCTION 1
Motivation 1
Format 4
TWO AN XPS INVESTIGATION OF TIN
OXIDE SUPPORTED PLATINUM 8
Introduction 8
Experimental 10
Results and Discussion 13
Conclusions 21
THREE AN ELECTRONIC AND STRUCTURAL
INTERPRETATION OF TIN OXIDE
ELS SPECTRA 23
Introduction 23
Experimental 25
Background 26
Results and Discussion 29
Conclusions 56
FOUR A STUDY OF THE DEHYDRATION OF
TIN OXIDE SURFACE LAYERS 58
Introduction 58
Experimental 61
Results and Discussion 62
Conclusions 71
IV
PAGE
FIVE AN OBSERRVATION OF WATER ADSORPTION
ON TIN OXIDE USING ESD AND
GRAZING-EXIT-ANGLE XPS AND AES 73
Introduction 73
Experimental 74
Results and Discussion 81
Conclusions 90
SIX THE INTERACTION OF POLYCRYSTALLINE
ZIRCONIUM WITH 02, N2, CO AND N20 91
Introduction 91
Experimental 92
Results and Discussion 93
Conclusions 115
SEVEN GENERAL CONCLUSIONS AND
RECOMMENDATIONS FOR FUTURE
RESEARCH 117
Pt Sn Oxide 117
Zirconium 119
APPENDICES
A A BRIEF DESCRIPTION OF THE
EXPERIMENTAL TECHNIQUES 121
X-Ray Photoelectron Spectroscopy (XPS) 121
Auger Electron Spectroscopy (AES) 124
Electron Energy-Loss Spectroscopy (ELS) 126
Electron-Stimulated Desorption (ESD) 128
B COMPUTER-INTERFACED DIGITAL PULSE
COUNTING CIRCUIT 131
Introduction 131
Circuit Description 132
Time-of-Flight Modification 139
Acknowledgments 140
REFERENCES 142
BIOGRAPHICAL SKETCH 147
v
LIST OF TABLES
TABLE
3-1. Electronic Configurations of Atomic
Tin and Oxygen and the Stoichiometric
Oxides.
5-1. Variation in O/Sn Ratio With Emission
Angle.
PAGE
40
83
vi
LIST OF FIGURES
FIGURE
2-1.
2-2.
3-1.
PAGE
Pt 4f XPS spectrum of an electro-
chemically platinized substrate.
The lower binding energy doublet
is characteristic of Pt metal,
and the higher binding energy
doublet is characteristic of
a Pt species chemically bonded
to be the tin oxide substrate. 14
Pt 4f XPS spectra of a sample
prepared by platinum chemisorp
tion. Spectrum (A) is obtained
immediately after pumpdown,
(B) after high temperature oxida
tion, (C) after high temperature
reduction and (D) after a high
temperature anneal _in vacuo. 17
Energy level diagram representing
the Sn02 band structure. The
locations of the major occupied
(unoccupied) valence and core
(conduction) states involved
in the energy loss spectrum
are shown. The approximate
locations of the SnO VBM and
CBM are indicated by dashed
lines. 30
N(E) ELS spectrum of tin oxide
after a high temperature vacuum
anneal. The high primary beam
energy (Ep = 1500 eV) at normal
incidence results in primarily
a bulk sensitivity. The indicated
features are characteristic
of well-annealed SnC>2 The
low energy features are due
to VB > CB transitions,
and the high energy loss features
are core > CB transitions. 33
vii
PAGE
3-3. Variation in the N(E) ELS spectrum
with primary beam energy, Ep.
The set of spectra represent
a depth profile of the annealed
material. The growth of the
27 eV feature is due to an
increasing oxygen deficiency
as the spectra become more surface
sensitive.
3-4. EN(E) ELS spectrum for Ep = 50
eV. The main loss feature at
13 eV shows that the annealed
material is essentially SnO
at the surface.
3-5. ELS spectra for a sample annealed
at 600C. The angle of incidence
for the primary beam is 45.
With Ep = 1500 eV it is seen
that the bulk is primarily SnC>2.
The Ep = 300 eV spectrum shows
SnO at the surface and evidence
of a defect structure between
bulk Sn02 and the surface.
3-6. Valence band XPS spectra after
a (A) 600C anneal and (B) 2
KeV argon-ion bombardment. The
spectra are more bulk than surface
sensitive.
3-7. ELS spectra following 2 KeV argon-
ion bombardment. The valence
band features show the bulk
to be Sn02like while the core
features reveal a significant
concentration of defects. The
VB features in the more surface
sensitive spectrum illustrate
an amorphous structure at the
surface due to sputtering.
3-8. Valence band XPS spectra following
(A) a short 500C anneal after
sputtering and (B) subsequent
oxygen exposure. The presence
of a mixture of SnO and Sn02
is indicated in (A).
36
37
44
46
49
52
vm
PAGE
3-9.
ELS spectra corresponding to Figures
3-8(A) and (B) respectively.
54
4-1.
Valence band XPS spectra after
(A) hydration by exposure to
atmospheric humidity, (B) a
500C vacuum anneal for 45 minutes
and (C) a 600C vacuum anneal
for 30 minutes.
64
4-2 .
ELS spectra of (A) subsurface
and (B) surface regions after
hydration due to atmospheric
humidity.
65
4-3 .
ELS spectra of (A) subsurface
and (B) surface regions after
a 500C vacuum anneal.
68
4-4.
ELS spectra of (A) subsurface
and (B) surface regions after
a 600C vacuum anneal.
70
5-1.
Variation in path length with
emission angle.
76
5-2.
Deflection circuit for desorption
event initiation.
79
5-3.
Time-of-flight spectrum for mass
analysis.
86
5-4.
Ion kinetic energy distribution
after sputtering.
88
LO
1
in
Ion kinetic energy distribution
following water exposure.
89
6-1.
AES spectra taken after (A)
hours of heating and (B) 1'
hours of heating below the HCP-to-
BCC transition temperature.
2
4
94
6-2.
AES spectra of state 1 zirconium
after room temperature exposure
to (A) nitrogen and (B) nitrous
oxide.
98
IX
PAGE
6-3 .
6-4 .
6-5.
6-6.
6-7 .
6-8.
6-9.
6-10.
XPS spectra showing the zirconium
3d peaks for (A) clean zirconium,
(B) N2 exposure, (C) N2O exposure
and (D) oxygen exposure
XPS spectra showing the zirconium
3s peak for (A) clean zirconium,
(B) nitrogen exposure and (C)
N2O exposure. The nitrogen
Is peak appears at 396 eV in
(B) and (C)
XPS spectra showing the oxygen
Is peak after (A) O2 adsorption
on zirconium and (B) N2O adsorp
tion on zirconium
AES spectrum for clean zirconium
after heating near the melting
temperature for 3 hours. The
AES 17 5 eV peak is greatly dimi
nished which is characterisitic
of state 2 zirconium
AES spectrum taken after clean
state 2 zirconium is exposed
to CO contamination from the
electron beam for 8 hours. The
carbon peak shows characteris
tics of both graphite and carbidic
carbon
XPS spectrum of the carbon Is
peak corresponding to the AES
spectrum shown in Figure 6-7.
Both graphitic and carbidic
carbon are present
AES spectra after exposing state
2 zirconium to CO at (A) room
temperature and (B) high tempera
ture but allowing the sample
to cool during the exposure
XPS spectra of the carbon Is peak
corresponding to the AES spectra
shown in Figure 6-9. The room
temperature adsorption produces
approximately equal amounts
of graphitic and carbidic carbon
as shown in spectrum (A) while
the high temperature adsorption
results in predominantly carbidic
carbon as shown in spectrum
(B)
99
100
102
104
106
107
108
109
PAGE
6-11.
A-l.
A-2 .
A-3 .
B-l.
B-2 .
B-3 .
B-4 .
(A) AES spectrum taken after expos
ing state 2 zirconium to nitrogen
at 5x10 Torr for 15 minutes
at room temperature. A small
amount of carbon and oxygen
contamination accumulated during
the long exposure and subsequent
AES run. (B) AES spectrum taken
after exposing state 2 zirconium
to nitrogen initially at high
temperature and then allowing
the sample to cool during the
exposure. (C) AES spectrum
taken after allowing the sample
to remain in vacuum for 3 days
at room temperature. State
2 zirconium has transformed
into state 1 zirconium. (D)
AES spectrum taken after exposing
the state 1 zirconium of spectrum
(c) to nitgrogen at 5xl0-^ Torr
for 5 minutes at room temperature. Ill
Photoemission process.
122
KLqL2 Auger decay process.
125
Electron energy-loss process.
127
On-board timer schematic showing
jumper selectable
rates .
system
clock
133
Schematic of
section.
the
control
logic
135
(A) Timing and
schematics.
(B)
event
counter
137
Layout and wiring
unmakred (pull
diagram. All
up) resistors
are 10K ohms. 138
xi
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
SURFACE CHARACTERIZATION AND CHEMISORPTION PROPERTIES
OF POLYCRYSTALLINE SYSTEMS:
Sn02f Pt/Sn02 and Zr
By
DAVID FULLEN COX
December, 1984
Chairman: Gar B. Hoflund
Major Department: Chemical Engineering
X-ray photoelectron spectroscopy (XPS) is used to
characterize platinum supported on tin oxide. A feature
in the platinum 4f XPS spectrum associated with the bond
formed between supported platinum and the tin oxide
substrate is identified. The bond is believed to form
with surface lattice oxygen resulting in a Pt-O-Sn surface
species. This substrate-bonded species appears to act
as a nucleation site for cyrstallite growth in both the
electrochemical desposition of platinum and in the
sintering of supported platinum.
It is demonstrated that electron energy-loss
spectroscopy (ELS) is an acceptable technique for
xix
distinguishing between the different oxides of tin. The
major features in the N(E) loss spectrum are interpreted
as due to collections of optically allowed interband
transitions. It is shown that depth profile information
about tin oxide may be obtained by varying the primary
electron beam energy. Combined ELS and valence band
XPS results indicate that a significant amount of
structural information may be inferred from the size,
shape
and/or
position
of
the
N(E) ELS features.
Core
level
features are found
to
be quite sensitive to
the
presence of
defects
in
an
SnC>2 lattice with
some
specificity as to the type of defect.
The chemisorption properties of polycrystalline
zirconium have been found to vary dramatically depending
on the thermal history of the sample. Chemisorption
on this surface is found to be suppressed by heating
for prolonged periods of time above the HCP-to-BCC phase
transition temperature at 1135K. The chemisorption
behavior can be correlated roughly with the appearance
or disapperance of a zirconium MW Auger peak. A slow
phase transition at the surface is postulated as the
cause of the variation in chemisorption properties.
xm
SECTION I
GENERAL INTRODUCTION
Motivation
The primary motivation for the work presented here
is an interest in the catalytic properties exhibited
by tin oxide supported platinum. It has been demonstrated
that platinized tin oxide surfaces display higher catalytic
activities than platinum electrodes in the electrochemical
oxidation of methanol [1-5] and for the reduction of
oxygen in alkaline [6] and 85% phosphoric acid solutions
[7]. A similarity between the electronic properties
of platinum in these supported systems and in the
industrial platinum-tin bimetallic reforming catalyst
supported on alumina has been demonstrated recently [8].
The bimetallic catalyst is known to exhibit improved
stability and higher average catalytic activity than
platinum supported on alumina [9,10].
Tin oxide and its modified forms also exhibit a
significant catalytic activity. Tin oxide has been shown
to be active for the catalytic oxidation of CO [11,12]
and the reduction of NO [12-14]. Chromia-doped tin oxide
is very active for the reduction of NO with CO, H2 and
C2H4 [13]. Antimony-doped tin oxide is known to be active
1
2
toward the selective (partial) oxidation of propylene
to acrolein [15] and the oxidative dehydration of butene
to butadiene [16]. Much of this behavior is believed
to be linked to the oxidation and reduction of active
sites on the catalyst surface.
The catalytic properties of the tin oxide support
are believed to play an important role in platinized
tin oxide catalysts. For example, Tseung and Dhara [7]
have postulated that a spillover mechanism may be involved
in the electrochemical reduction of oxygen. Their results
suggest that adsorbed oxygen migrates from the supported
platinum to the tin oxide surfaces before undergoing
reduction. The possible importance of spillover mechanisms
in platinum/tin oxide catalyzed reactions and the apparent
redox behavior of tin oxide surfaces both indicate that
a successful surface characterization of the supported
system must include a determination of the tin oxidation
state.
The main focus of the work presented here is the
application of ultrahigh vacuum (UHV) surface probes
to a fundamental characterization of tin oxide and
platinized tin oxide catalysts. It is hoped that this
characterization will aid in understanding the
physiochemical properties affecting the preparation and
the catalytic behavior of these systems. To this end,
a realistic system is studied which utilizes
3
polycrystalline tin oxide as the support material. The
polycrystalline nature of the support presents immediate
problems in terms of structural characterization because
the usual surface-structure-sensitive technique, low-energy
electron diffraction (LEED), is not applicable. For
this reason the approach primarily has been to use electron
spectroscopies as probes of the electronic properties
of the materials. The direction taken in the work
presented here has been influenced by some earlier studies
which have already been reported in the literature [17-19].
The earlier results will not be repeated here but will
be referenced when appropriate to the discussion of present
results.
One unforeseen result of the study of tin oxide
has been a broadening of interests to include structural
effects in electronic spectra (see Sections III and IV).
Out of this interest has grown a study of the chemisorption
properties of clean polycrystalline zirconium.
Observations in the literature of anomalous effects on
the Auger electron spectrum of zirconium due to gas
adsorption combined with conflicting results on the uptake
rate of adsorbates are responsible for the selection
of zirconium for study. Though not directly related
to the characterization of platinized tin oxide, these
zirconium results are also presented here.
4
Format
The results presented here are divided into
independent sections, each of which is complete within
itself. Each section deals with one aspect of the surface
characterization of a platinum-tin oxide catalyst with
the exception of Section VI which deals with the apparent
structural dependence of the chemisorption properties
of polycrystalline zirconium.
Section II presents results on the characterization
of supported platinum on tin oxide. X-ray photoelectron
spectroscopy (XPS) has been used in this characterization
to investigate the valence (oxidation) states of the
supported platinum. In situ thermal and chemical
treatments are used to help identify the various platinum
oxidation states and to investigate the nature of the
chemisorption bond formed between platinum and the tin
oxide substrate.
While XPS has proven useful in determining the valence
states of supported platinum, it has been of limited
use in characterizing the oxidation state of tin in the
support. Though core level XPS can distinguish between
metallic tin and tin oxide, it can not distinguish between
the different oxides of .tin, SnO and SnC>2. In view of
the apparent importance of the determination of the tin
valence state, the majority of effort has been devoted
to finding a suitable characterization technique for
5
the tin oxide support. The chosen technique is electron
energy-loss spectroscopy (ELS).
Section III is a discussion of the use of ELS in
the characterization of tin oxide. An interpretation
of the spectra is given in terms of the excitation of
interband transitions. Using ELS in conjunction with
valence band XPS, it is demonstrated that a significant
amount of structural information about polycrystalline
tin oxide may be inferred from changes in the electronic
structure as probed by ELS.
An understanding of the interaction of water with
tin oxide surfaces is believed to be of primary importance
in elucidating the chemisorption properties of platinum
on tin oxide. This belief is supported by considerable
evidence that the active chemisorption sites are surface
hydroxyl groups [6,19]. In Section IV ELS is applied
to the study of the dehydration of a tin oxide sample
after exposure to atmospheric humidity. The flexibility
of analysis depth provided by this technique affords
a particularly worthwhile characterization of the
subsurface region.
Because of a phenomenon known as electron-stimulated
desorption (ESD), ELS is not useful in studying the
interaction (i.e. adsorption) of water with the surface
of tin oxide. The incident electron beam used for ELS
actually removes the adsorbed species of interest from
6
the surface. This phenomenon has been observed in the
earlier studies [17-19] which show that significant surface
modification can result from an impinging electron beam.
However, the ESD phenomenon can itself be used to probe
the interaction between surface and adsorbate.
Section V presents some preliminary observations
of the adsorption of water on tin oxide.
Grazing-exit-angle XPS provides a measure of the water
adsorbed from the background vacuum in the UHV system.
Preliminary results are also presented which demonstrate
the potential of ESD in characterizing the interaction
of water with tin oxide surfaces.
Section VI presents a study of the chemisorption
properties of polycrystalline zirconium. The study
focusses on the chemisorption behavior as a function
of the thermal history of the sample. A slow phase
transition in the surface region is postulated as the
cause of a suppression in chemisorption after high
temperature annealing. A zirconium feature in the Auger
electron spectrum is shown to be an indicator of the
chemisorption properties of the surface.
Section VII presents a summary of conclusions for
the entire study along with recommendations for future
research. Two appendices are included after Section
VII. Appendix A contains a brief description of the
physical processes involved in the experimental techniques
7
used in this work. A brief introduction to XPS, AES,
ELS
and ESD is
given.
Appendix
B describes
the
computer-interfaced
digital
pulse
counting circuit
used
for
data collection
in XPS
, ELS
and
time-of-flight
ESD
measurements.
SECTION II
AN XPS INVESTIGATION OF TIN OXIDE
SUPPORTED PLATINUM
Introduction
The study of tin oxide supported platinum is motivated
by the interesting catalytic properties displayed by
mixed Pt-Sn systems. It has been shown that platinized
tin oxide exhibits a catalytic activity 50 to 100 times
greater than that of platinum electrodes for the
electrochemical oxidation of methanol [1-5]. Watanabe
et al. [6]. have studied the effects of platinum loading
on the catalytic activity for oxygen reduction in alkaline
solution. Their results show that the catalytic activity
of highly dispersed platinum on tin oxide may exceed
that of platinum electrodes by a factor of four or more
[6]. An interest in impure H2 fuel cells using 85%
phosphoric acid at 150C has prompted Tseung and Dhara
[7] to study supported Pt on antimony-doped tin oxide
because of the corrosion resistance and electrical
conductivity exhibited by this system. Their results
show a significant increase in the catalytic activity
for oxygen reduction over that of platinum black. In
addition, the similarity between platinized tin oxide
8
9
and the Pt-Sn bimetallic reforming catalyst has been
demonstrated recently [8].
The catalytic effects of low coverages of strongly
bound oxygen on Pt single-crystal surfaces has been
demonstrated by Smith, Biberian and Somorjai [20]. They
interpret dramatic, oxygen-coverage-dependent changes
in activity and selectivity for hydrogenation and
dehydrogenation reforming reactions as being due to the
formation of surface Pt oxides. Interestingly, it has
been shown that the formation of Pt-Sn alloys generally
results in lower catalytic activities [21-24]. These
observations suggest that oxygen may be partially
responsible for the catalytic properties exhibited by
Pt/tin oxide systems.
Several X-ray photoelectron spectroscopy (XPS) studies
have been performed on Pt-Sn systems [4,5,8,25]. These
studies all agree that tin is largely present as tin
oxide while platinum is present in the metallic form
and as Pt^+ and Pt4+ in the form of oxides and hydroxides.
These previous XPS studies have examined intimately mixed
systems of tin oxide and platinum oxides and metal. In
the present work a supported platinum system is studied.
The use of this system in conjunction with in situ chemical
and thermal treatments has allowed the assignment of
a Pt "oxidation state" characteristic of the chemical
bond formed between the metal and the tin oxide substrate.
10
Experimental
The tin oxide substrates are prepared by the thermal
hydrolysis of Sn (IV) from a solution containing 3 M
SnCl4 5H20, 1.5 M HC1 and 0.03 M SbCl3. The solution
is sprayed onto the hot surface of a titanium foil held
at 500C in air. The formation of tin oxide occurs
according to:
SnCl4 + 2H20 > Sn02 + 4HC1
The resulting planar film is a polycrystalline, n-type,
Sn02 semiconductor with the rutile structure. The antimony
is incorporated into the film at a concentration
approximately twice that of the spray, i.e. 2% [26].
This dopant is known to be in solid solution with the
tin oxide [27], and it acts as a donor which raised the
conductivity of the film to a level adequate for
electrochemical studies [26]. Spraying is continued
O
until a tin oxide layer approximately 6000 A to 7000
O
A thick is obtained as determined by the colors of the
interference fringes of the layer. After cooling in
air, the samples are polished with 0.25 pm diamond paste.
The use of alumina as a polishing compound is avoided
because of the overlap of platinum 4f and aluminum 2p
peaks in XPS [8].
A previous Auger electron spectroscopy (AES) and
XPS investigation has shown that the surface of an
11
antimony-doped tin oxide film prepared in the described
manner may contain a number of surface contaminants [17].
Among these surface contaminants are carbon, chlorine,
potassium, sodium, calcium and sulfur in varying amounts.
Argon ion bombardment and high temperature oxygen
treatments have proven to be effective in removing this
surface contamination, but the effects on the tin oxide
surface (see Section III) and platinum oxidation state
(as shown below) are substantial. Hence, an understanding
of surfaces such as those being used in electrochemical
studies [6,28] may require analysis in the presence of
several types of surface contamination.
Two different techniques are used in the present
study to prepare the supported Pt. The first of these
techniques is electrochemical in nature. In essence,
the platinum is plated from a 5x10^ m solution of H2PtClg
buffered at a pH of 6.8. The process is carried out
for varying amounts of time at -0.5 V versus SCE. The
second technique utilizes a chemisorption mechanism.
The substrate is pretreated by exposure to a 10 M NaOH
solution at 90C for 30 minutes. The pretreated substrate
is washed in distilled water and then exposed to an 80C
solution of 0.01 M KOH and 500 ppm Pt (IV) from Na2Pt(OH)g.
The Pt loading is dependent on the exposure time for
both preparation techniques. Regardless of the procedure
used, the samples are washed in distilled water after
12
platinization and solvent cleaned before mounting in
the vacuum system.
An important consideration in the preparation of
these supported platinum catalysts is the relative rate
of crystallite nucleation to growth. The indications
are that the growth of crystallites is favored over
nucleation in the electrochemical preparation [29]. The
chemisorption technique, however, has been shown to be
capable of producing highly dispersed (>90%) platinum
at low loadings [6]. The alkaline pretreatment is believed
to hydroxylate the surface, thereby increasing the number
of active sites available for Pt chemisorption [6,19].
Electron beam effects on these surfaces can be
dramatic. The removal of carbon, chlorine, oxygen,
hydrogen and sodium by electon-stimulated desorption
(ESD) has been observed previously [17-19], These beam
effects somewhat limit the usefulness of AES as an
analytical tool on these surfaces making XPS the preferred
technique because of its less destructive nature. However,
an understanding of the ESD phenomenon in terms of an
interatomic Auger decay model [30] shows promise in helping
to unravel the chemistry of these surfaces (see Section
V and ref. 19).
All XPS spectra were collected with a Physical
Electronics double-pass CMA using Mg K a X-rays as an
excitation source. A pass energy of 50 eV (AE/E = 0.014)
13
was used throughout. All binding energies are referencecd
to the tin 3d 5/2 peak at an assumed energy of 486.4
eV [31]. It has been shown that there is no change in
this core level binding energy for the different oxides
of tin [32-34] making this peak an excellent reference.
The base pressure in the vacuum system for this study
was lxl0-9 Torr. Details of the vacuum system have been
given previously [17].
Results and Discussion
Figure 2-1 shows the Pt 4f XPS spectrum of an
electrochemically platinized sample. The plating process
was carried out for 40 minutes at -0.5 V versus SCE.
A fairly high platinum loading of about 40 \ig/cm2 is
obtained by this process as estimated from a previous
Rutherford backscattering (RBS) study [35]. Deconvolution
of the spectrum reveals the presence of two platinum
species. In the assignment of these features platinum
chloride species are neglected. A check for surface
chlorine contamination using standard sensitivity factors
[31] showed the concentration to be low (Cl/Pt < 0.1).
The lowest binding energy doublet in Figure 2-1
has the 4f 7/2 peak at 71.2 eV and the 4f 5/2 peak at
74.5 eV. These features are assigned to Pt metal in
agreement with the work of Katayama [5,25]. The binding
energy reported here is about 0.4 eV higher than that
14
Figure 2-1. Pt XPS spectrum of an electrochemically plati
nized substrate. The lower binding energy doublet is
characteristic of Pt metal, and the higher binding energy
doublet is characteristic of a Pt species chemically bonded
to the tin oxide substrate.
15
generally reported for the bulk Pt metal [31]. This
observation of a higher binding energy for supported
clusters (crystallites) over that of bulk metal is in
agreement with general expectations [36]. This shift
is most likely the result of differences in the reference
levels (work functions) of the bulk metal and the tin
oxide support and/or a decrease in the final-state
extra-atomic relaxtion energy as a result of the change
from bulk metal to small cluster [37].
The doublet at higher binding energies in Figure
2-1 has a 4f 7/2 peak at 72.3 and a 4f 5/2 peak at 75.5
eV. This doublet is shifted about 1.1 eV above Pt metal
and about 0.5 eV below the position expected for a Pt(OH)2
species [5,25,31]. The higher oxides of platinum all
fall to significantly larger binding energies which removes
them from consideration (see below). Examination of
a second substrate electrochemically plated for only
1/4 the time (i.e. 10 minutes) gives a Pt 4f spectrum
(not shown) characterized primarily by this high binding
energy species observed in Figure 2-1. These results
suggest that the higher binding energy doublet in Figure
2-1 may be associated with a platinum species directly
bonded to the tin oxide substrate. Further, the appearance
of Pt metal at longer plating times demonstrates that
this substrate-bonded Pt species acts as a nucleation
site for the growth of metallic crystallites by
16
electrochemical deposition. These results are consistent
with a model of nucleation and crystallite growth suggested
by earlier work on the electrodeposition of platinum
on tin oxide [29].
Figure 2-2 shows the Pt 4f peaks for a sample prepared
by the chemisorption technique. The pretreated tin oxide
substrate was exposed to the 80C Na2Pt(OH)5 solution
for one hour. The platinum loading is approximately
2 pg/cm^. Because the Pt loading is small, the
signal-to-noise ratio does not justify a spectrum
deconvolution. However, the use of in_ situ chemical
and thermal treatments allows a manipulation of the Pt
valence state for a more complete determination of the
supported species. Because only a trace of chlorine
was detected, the possibility of platinum chloride species
was again discounted.
Figure 2-2a shows the spectrum obtained immediately
after pumpdown. The position of the doublet indicates
that the substrate-bonded species is the predominant
form of platinum obtained by the chemisorption procedure.
However, the peak widths also suggest the presence of
small amounts of Pt metal at lower binding energies and
Pt(0H)2 at slightly higher binding energies. The
observation of the substrate-bonded species as the primary
form of platinum is consistent with earlier work showing
nucleation is preferred over crystallite growth during
chemisorptive platinization [6].
NCE>
17
Figure 2-2. Pt XPS spectra of a sample prepared by plati
num chemisorption. Spectrum (A) is obtained immediately
after pumpdown, (B) after high temperature oxidation,
(C) after high temperature reduction and (D) after a
high temperature anneal in_ vacuo.
18
Figure 2-2b shows the effect of a high temperature
(600C) in. situ oxidation in 11 Torr of O2 for 30 minutes.
The oxygen treatment shifts the Pt 4f XPS peaks to higher
binding energies. The presence of the higher oxides,
PtO and PtC>2f is clearly indicated by structure on the
high binding energy side of the specturm. The PtO features
are shifted approximately 2.9 eV higher with respect
to Pt metal while the Pt02 features are shifted about
3.9 eV [31]. Though there is no evidence of Pt metal
in Figure 2-2b, the shoulder at 72.3 eV is clear evidence
of the persistence of the substrate-bonded Pt species.
The observation of this species in conjunction with PtO
and Pt02 confirms that the substrate-bonded species is
not simply a stoichiometric platinum oxide.
A 500C in_ situ reduction in lxlO-^ Torr of H2 for
30 minutes results in the spectrum shown in Figure 2-2c.
An XPS inspection of the Sn 3d core levels shows no sign
of a reduction of the substrate to bulk Sn metal. However,
the PtO and Pt02 species observed in Figure 2-2b have
undergone a complete reduction leaving primarily Pt metal.
The loss of the higher oxides coupled with the appearance
of a Pt 4f 7/2 peak at 71.2 eV confirms the earlier
assignment of this binding energy to Pt metal. Evidence
of the substrate-bonded species is also found in Figure
2-2c in the form of a shoulder on the high binding energy
side of the 4f 5/2 peak.
19
A subsequent 800 anneal in. vacuo has little effect
on the XPS peak positions as shown by Figure 2-2d. The
platinum remains primarily in the metallic state with
a small contribution due to the substrate-bonded species.
The presence of this substrate-bonded species after high
temperature annealing confirms that these features are
not due simply to a platinum hydroxide or hydrate species.
Decomposition or dehydration of such species would be
expected at significantly lower temperatures.
Before the oxidation-reduction cycle the platinum
from the chemisorption preparation is largely present
in the substrate-bonded form. This observation suggests
a high dispersion as found by Watanabe et al. for samples
prepared in a similar fashion [6]. The high temperature
oxidation-reduction cycle results in a sintering of the
supported species as shown by the large fraction of Pt
metal in Figure 2-2c. The remaining presence of the
substrate-bonded species suggests that a fraction of
these species acts as nucleation sites for the crystallite
growth as was observed in the electrochemical platinization
process.
The constancy of the XPS peak positions for the
substrate-bonded species obtained by either the
electrochemical or chemisorption process indicates a
similarity in the species formed regardless of the
20
procedure used. It has been shown above that this species
may not be identified as simply a PtOx or Pt(OH)y species.
Likewise, the binding energy shift of this species with
respect to Pt metal is not consistent with that observed
in the formation of Pt-Sn alloys [38]. These observations
suggest that the bond formed with the surface occurs
through surface lattice oxygen. The formation of a Pt-O-Sn
substrate-bonded Pt species is postulated. The tenacity
displayed by this species in resisting complete reduction
by chemical and thermal treatments is characteristic
of a species exhibiting such a strong interaction with
the substrate. Komiyama et al. have observed a similar
resistance to reduction by ion bombardment of strongly
interacting rhenium species on iron oxide [39].
Previous work on samples prepared by the chemisorption
technique offers insight into the mechanism of formation
of the substrate-bonded species. Watanabe et al. Have
shown that an alkaline pretreatment of the substrate
prior to platinization results in an increased Pt uptake
[6]. It is believed that the pretreatment hydroxylates
the surface and provides an increased number of active
chemisorption sites for the platinum species in solution.
Earlier studies using secondary-ion mass spectrometry
(SIMS) [40] and ESD [19] lend support to the surface
hydroxylation model by showing significant increases
in surface hydrogen and oxygen after the alkaline
21
pretreatment. Platinum chemisorption is believed to
occur by replacement of the proton on the surface hydroxyl
group with the loss of a coordinated ligand from the
platinum solution species. Under the pH conditions used
for chemisorption from a H2PtCl6 solution, the
chloroplatinate undergoes hydrolysis resulting in the
replacement of two chlorines by hydroxyl groups.
Chemisorption should occur via
Sn-OH + Pt(OH)2Cl42- > Sn-O-Pt(OH)CI42- + H20
with a subsequent dehydroxylation and loss of chlorine
from the
surface
complex.
For
chemisorption
from an
alkaline
solution
of Na2Pt(OH)5
the surface
hydroxyl
group is
ionized
through the
loss
of the acidic
proton.
Chemisorption is expected to occur via
Sn-O" + Pt(OH) g2- > Sn-O-Pt(OH)52- + OH-
leaving the substrate-bonded species after dehydration
of the surface complex.
Conclusions
XPS has been used to study tin oxide supported
platinum prepared by electrochemical and chemisorption
techniques. Features in the Pt 4f spectrum have been
22
assigned to a species chemically bonded to the substrate.
The position of these features is independent of the
preparation technique used. Tn situ chemical and thermal
treatments confirm that this substrate-bonded platinum
is not simply a PtOx or Pt(OH)y species. The platinum
is believed to bond through surface lattice oxygen giving
a Sn-O-Pt surface species. High temperature reduction
results in a sintering of these species, but the inability
to completely reduce the platinum is indicative of the
strong chemical interaction between the platinum and
tin oxide.
A model for the chemisorption of platinum on tin
oxide is proposed. Surface hydroxyl groups are believed
to be the active chemisorption sites for platinum species
in solution. The chemisorption process is believed to
occur through the replacement of the hydroxyl group proton
with the loss of a coordinated ligand from the platinum
species.
SECTION III
AN ELECTRONIC AND STRUCTURAL INTERPRETATION
OF TIN OXIDE ELS SPECTRA
Introduction
The spectroscopic study of tin oxide surfaces is
complicated by the difficulty in distinguishing between
the two oxides of tin, SnO and Sn02. Several x-ray
photoelectron spectroscopy (XPS) studies have failed
to detect any changes in core level binding energies
between SnO and Sn02 [32-34]. Similar problems are
encountered using Auger electron spectroscopy (AES) where
no significant differences in kinetic energies or line
shapes are found [41]. As expected, however, the valence
band spectra of the two oxides do differ. An ultraviolet
photoelectron spectroscopy (UPS) study of tin oxidation
by Powell and Spicer [42] and a valence-band XPS study
by Lau and Wertheim [32] have shown these differences,
but interpretation difficulties associated with analysis
depth have proven to be substantial.
Electron energy-loss spectroscopy (ELS) is a technique
which offers flexibility of analysis depth and is sensitive
to changes in the valence band density of states. Powell
[41] has shown that ELS may be used to distinguish between
23
24
the two oxides of tin and has given a preliminary
interpretation of the spectra in terms of differences
in plasmon frequencies. For SnC>2 a main loss feature
at 19.5 eV was identified while for SnO a main loss feature
was found at approximately 13.5 eV. A combined UPS
and high-resolution electron energy-loss spectroscopy
(HREELS) study of 3% Sb doped and undoped SnC>2 has shown
the room temperature occupied conduction bands to be
very f ree-electron like [43] in agreement with a bulk
tight-binding band structure calculation [44]. For the
heavily doped sample an HREELS loss feature at 0.55 eV
was found. Based on the experimentally determined carrier
concentration and effective mass ratio, the 0.55 eV loss
feature was identified as a surface plasmon loss associated
with conduction band electrons from Sb donors. Since
valence band and core level electrons in SnC>2 are not
free-electron like, the higher energy ELS losses in the
present study are not assigned to plasmon losses.
While an interpretation of the ELS spectrum would
be useful for distinguishing between the two oxides of
tin, an additional benefit may be derived due to the
usefulness of ELS measurements in the interpretation
of electron-stimulated desorption (ESD) threshold studies.
It has been shown that core level transitions can be
correlated with desorption thresholds and may specify
adsorbate binding sites [30,45,46]. In particular, the
25
ability to distinguish between transitions from Sn 4d
and 0 2s core levels which cannot be resolved using XPS
could be most useful in understanding the chemistry of
tin oxide surfaces.
Experimental
The polycrystalline tin oxide films used in this
study were prepared by spraying a solution of 3 M SnCl4
and 1.5 M HC1 onto a titanium foil maintained at 500C
in air. Unlike the samples used in Section II and in
previous studies [17-19], a high purity (99.998%) anhydrous
SnCl4 reagent was used. The resulting samples were found
to have significantly less surface contamination. Trace
chlorine and carbon contamination was found to be removed
quickly in. situ by heating at 500C in 10 Torr of oxygen
for about 5 minutes. This procedure gave a clean oxide
surface as determined by AES.
The samples were annealed ini vacuo initially and
were heated briefly and allowed to cool before each
individual measurement. Using angle-resolved ultraviolet
photoelectron spectroscopy (ARUPS) on an ion-sputtered
SnC>2 (001) single crystal surface, Gobby [47] has shown
that the annealing process (550C to 835C) strengthens
the primary emission from the valence bands and increases
the sharpness and magnitude of the anisotropic emission
indicating a well ordered crystal. Similar annealing
26
effects in the sharpness and magnitude of ELS spectra
and on the magnitude of core level emission in XPS have
been observed for polycrystalline tin oxide samples in
this study.
All spectra were collected with a double-pass CMA.
Details of the vacuum system have been published previously
[17]. The ELS data were taken in the retarding (N(E))
mode to allow a comparison with the data of Powell [41].
All ELS spectra were collected with a pass energy of
25 eV ( A E/E = 0.014) with the exception of the 50 eV
primary beam measurement. This spectrum was recorded
in a nonretarding (EN(E)) mode to suppress the large
signal from secondary electrons at near zero kinetic
energies. All
ELS
spectra
were collected using
100
nA
beam currents
and
pulse
counting
detection.
The
XPS
spectra were
taken
using
a Mg K a
source and
a 50
eV
analyzer pass energy. The base pressure in the vacuum
system for this study was 1 x 10^- Torr.
Background
For energy losses of the magnitude of electronic
excitations, the inelastic scattering event may be
described in terms of optical (dipole) selection rules
in cases where the primary electron energy is high enough
to justify the Born approximation. It is generally thought
that primary electron beam energies above 100 eV to 200
eV satisfy this criterion [48-51].
27
Consider a primary electron of momentum h K scattered
inelastically into a state h K' resulting in an interband
transition between one-electron states, |k,l> >
V 1
V
Momentum
>
conservation requires AK = k'
->
- k
->
-> ->
+ G where AK = K
- K' and
G is a reciprocal lattice
vector.
Energy conservation
> ~y
requires h^( | K | 2- j k'
I2)
= 2m( £]< i
->*
,l'e k,l>
where ^ q,l
V-.
is the eigenenergy of
the
one-electron state
-7"
1q,i>.
Not only must energy
and
momentum be conserved, but the matrix element
- -v
T =
must be nonzero [48,51,52]. Expansion in powers of
(AKr) yields the selection rules. It has been shown [48,51]
that the monopole term vanishes due to orthogonality
and that retaining only the dipole (linear) term gives
T = i AK < k + AK, 1' | r | k, 1>
For small | AK | Rudberg and Slater [48] have shown a fair
approximation at small energy losses or large | K | may
be obtained by considering only direct transitions, k
= k'. Hence, in the regime where the Born approximation
applies the selection rules are essentially optical in
nature. To a first approximation, the energy dependence
of the loss spectrum should be similar to that measured
28
in optical absorption [49]. Since the momentum transfer
>
in the ELS transition, h A k, may be different than in
the optical process, it is expected that a broadening
in the energy dependence of the ELS features will occur
with respect to the optical features [48].
Because the results from this study are for
polycrystalline samples with a random grain size which
is small compared to the excitation volume, the orientation
of K with respect to the crystal axes may be assumed
to be random. Therefore, the ELS spectra presented here
represent an average over the entire Brillouin zone.
The present results should be most comparable to optical
absorption studies of polycrystalline samples.
It should be mentioned that a breakdown in dipole
selection rules is possible for low beam energies and
large energy losses. In this case the expansion of the
41
phase factor, exp(i A k *r), must be carried to the
quadratic (quadrupole) term to obtain an accurate
description. Ludeke and Koma [50] and Colavita et al.
[51] have taken advantage of this effect to identify
loss features due to quadrupole-allowed transitions between
dipole-unallowed states. No such identifications have
been made in the present work.
Using a generalization of the joint density-of-states
function for optical interband transitions which includes
finite momentum changes, Ludeke and Esaki [53] have shown
that the energy-loss distribution due to transitions
29
from narrow, filled initial states to empty conduction-band
final states may be proportional to the conduction-band
(CB) density of states. This density-of-states
interpretation requires the initial state to be isolated
with no additional scattering channel existing near the
same energy loss. An additional complication may arise
if there is a significant modulation of the scattering
cross section due to a partial filling of the conduction
bands from a competing scattering channel originating
from a different initial state. In spite of these problems
it should be possible to obtain some picture of the CB
density-of-states in
tin
oxide
if the
Sn
4d
and 0 2s
core levels couple
to
final
states
of
significantly
different energy.
Results
and Discussion
Figure 3-1 is
an
energy
level
diagram
depicting
the band structure of SnC>2. The character of the
electronic states in the valence and lower conduction
bands is due to Robertson [44]. The assignment of Sn
4f character to high lying conduction band states is
due to Gobby [47]. The width of the valence bands and
the location of the three major features therein are
from the available photoemission data [32,47]. The
position of the 0 2s and, Sn 4d core levels are from XPS
measurements made in this laboratory with no attempt
25
SN 4F
SN5P 02P
SN 5S
02P LONE PAIR
MIN. BONDING 02P
02P SN5S BONDING
-22
- 18
12
9.5
3.6
a
-2
-4
-7.5
-9.5
Figure 3-1. Energy level diagram representing the SnC>2
band structure. The locations of the major occupied
(unoccupied) valence and core (conduction) states involved
in the energy loss spectrum are shown. The approximate
locations of the SnO VBM and CBM are indicated by dashed
lines.
31
at deconvolution. Photoemission results [47] were used
to locate the states in the conduction bands which couple
strongly to various valence and core states as discussed
below. The cut-off position at the top of the conduction
bands was determined form the ELS spectrum in Figure
3-2 based on the interpretation of the high-energy loss
features given below. While all the states are represent
by single horizontal lines, some are quite broad and
may extend over 5 eV or more. The dashed lines in the
band gap and lower conduction bands represent the
approximate location of the SnO valence-band maximum
(VBM) and conduction-band minimum (CBM) respectively.
These assignments are due to photoemission results for
SnO [32] and optical absorption on highly defect laden
tin oxide films [54].
Annealing Effects
Figures 3-2 to 3-4 show the ELS data for a sample
annealed at 7 50C in_ vacuo. Each of these spectra were
recorded for a normal incidence primary beam of specified
energy, Ep. The annealing process was carried out until
the background chamber pressure went through a clear
maximum (about 45 minutes). Giesekke et al. [55] have
shown using thermogravimetric analysis and electron
diffraction that the decomposition of tin (IV) hydroxide
proceeds through four distinct crystalline hydrogen
containing compounds before yielding SnC>2 above 600C.
32
The observed pressure maximum during the annealing process
is indicative, in part, of this dehydration. The
hygroscopic nature of tin oxide and the study of hydrated
surfaces is discussed in Sections IV and V.
Figure 3-2 is the loss spectrum for a 1500 eV primary
beam. This spectrum may be divided into two parts; the
higher energy loss features above
about
28
eV
and
the
features
at lower energy losses.
The lower
half
of
the
spectrum
consists of two major features
at
19.5
eV
and
13 eV in agreement with the SnC>2 spectrum reported by
Powell [41]. Additionally, extrapolation of the linear
portion of the leading edge of the loss spectrum to the
baseline gives a minimum energy loss of 3.6 eV. This
value is equal to the best available optically determined
band-gap energy for SnC>2 single crystals [56,57 ] and
the calculated lowest energy direct-allowed one-electron
transition ( Ft > ft ) found by Robertson [44]. Using
constant-intial-states (CIS) ARUPS measurements and
angle-integrated UPS for SnC>2 (001), Gobby [47] has shown
that VB-to-CB transitions are dominated by excitations
form an initial state about 1.5 eV below the VBM to final
states near 10 eV, 13 eV and 19 eV to 22 eV higher in
energy as shown in Figure 3-1. Inspection of Figure
3-2 reveals a shoulder in the loss spectrum near 10 eV
as well as the two higher energy features. This 10 eV
loss feature also corresponds to a collection of VB-to-CB
NCE>
33
Figure 3-2. N(E) ELS spectrum of tin oxide after a high
temperature vacuum anneal. The high primary beam energy
(Ep = 1500 eV) at normal incidence results in primarily
a bulk sensitivity. The indicated features are
characteristic of well-annealed SnC>2. The low energy
features are due to VB > CB transitions, and the high
energy loss features are core > CB transitions.
34
dipole-allowed transtions at the r point in the Brillouin
zone for bulk SnC>2 as found by Robertson. It is concluded
that the lower energy loss features in Figure 3-2 are
due to collections of optically (dipole) allowed interband
(VB -> CB) transitions.
The loss features above 30 eV are strongly dependent
on the thermal history of the sample and are dominated
by core-to-conduction-band transitions from tin 4d and
oxygen 2s levels. Gobby [47] has shown that these core
levels couple to final states in two energy regimes.
Coupling to CBs which are 3 2 eV to 36 eV above the core
level
is
observed easily in
UPS
while coupling
to
the
lower
CBs
(the
CB minimum
lies
approximately
26.6
eV
above
the
core
levels) is
not
observable due
to
the
photoemission threshold and large background of secondary
electrons. At higher photon energies coupling to CBs
37 eV and higher relative to the core levels is observed.
This coupling begins to strengthen at 40 eV above the
core level, but higher energies were not used because
of a lack of photon intensity. However, a higher energy
CB final state was identified for an initial state feature
in the lower VBs. This final state falls about 45 eV
above the core level, and Gobby suggests that it is a
Sn 4f derived state (see Figure 3-1). In Figure 3-2
a range of energy-loss features from about 29 eV to 48
eV are visible. The strongest features fall near 36
35
eV and 46 eV in excellent agreement with the photoemission
results of Gobby.
On the basis of the similarities between the
photoemission results for single crystal Sn02 and the
energy-loss spectrum, Figure 3-2 is interpreted as being
characteristic of a well-annealed (though polcrystalline)
Sn02 material. Additionally, these similarities support
the conclusion that the main features observed in the
ELS spectrum are due to single inelastic events possibly
in conjunction with elastic scattering events. Because
of the long mean free path of electrons near 1500 eV,
the spectrum in Figure 3-2 (Ep = 1500 eV) is primarily
due to contributions from the bulk of the material.
Figure 3-3 shows the effect on the loss spectrum
of varying the primary beam energy from 1500 eV to 200
eV. Figure 3-4 shows the EN(E) loss spectrum for a 50
eV primary beam. Decreasing the beam energy decreases
the analyses depth due to a reduction in the electron
mean free path with kinetic energy. The set of spectra
in Figures
3-3
and 3-4, therefore
, represent a depth
profile
of
the
vacuum-annealed
tin
oxide material. In
Figure
3-3
the
main change in
the
valence band region
is seen
to
be
a growth of the
12
eV to 13 eV feature
relative to the 19 eV feature with decreasing beam energy.
This change is most apparent in Figure 3-4 where a feature
near 13 eV dominates the spectrum. Changes in the core
NCE>
36
Figure 3-3. Variation in the N(E) ELS spectrum with
primary beam energy, Ep. The set of spectra represent
a depth profile of the annealed material. The growth
of the 27 eV feature is due to an increasing oxygen
deficiency as the spectra become more surface sensitive.
ENCE)
37
Figure 3-4. EN(E) ELS spectrum for Ep = 50 eV. The
main loss feature at 13 eV shows that the annealed material
is essentially SnO at the surface.
38
level region are more dramatic. The core level losses
may be resolved into two features. The large loss feature
at 46 eV is seen to decrease rapidly with beam energy
leaving a separate feature near 36 eV. Concurrent with
the loss of the 46 eV feature, the growth of a feature
at 27 eV is observed.
By comparison to the work of Powell [41], the changing
valence-band derived features in Figures 3-3 and 3-4
may be loosely interpreted as a change in the tin oxide
from a SnC>2 compound in the bulk to a more SnO-like
material at the surface. Because the SnO-like feature
near 13 eV dominates the spectrum only for a 50 eV primary
beam, it appears that such a material exists in the near
surface region, possibly in the top few atomic layers.
This interpretation is reasonable in view of the well
documented oxygen loss from tin oxide surfaces during
high temperature annealing [47,58,59]. Such oxygen losses
have been observed frequently in this laboratory.
Decreases in surface O/Sn ratios from near 2 down to
1 on annealing have been monitored with AES and XPS.
The interpretation of changes in the core-to-CB
region of the spectrum leads to the same conclusion as
derived from the VB-to-CB features, but some discussion
of the symmetry of the initial and final states involved
is required. The band structure calculations of Robertson
[44] and Munnix and Schmeits [60] as well as the ARUPS
39
measurements of Gobby [47] show that the SnC>2 valence
bands are mostly 0 2p like with only a small admixture
of Sn derived states. The lower conduction bands are
primarily Sn 5s and 5p like, and within 3 eV to 4 eV
of the CBM these states are 90% Sn 5s like [44], To
a first approximation the atomic character of these states
suggests that the electronic structure of SnC>2 may be
considered to be ionic. Within this ionic approximation
the electronic configurations of the atomic and
stoichiometric oxide systems are those given Table 3-1.
For SnC>2 the highest occupied states are oxygen
2p like, and the lowest unoccupied states are tin 5s
like in basic agreement with the band structure
calculations. Reduction of Sn02 to SnO populates the
Sn 5s states leaving the lowest unoccupied states more
Sn 5p like. Likewise, the removal of oxygen from SnC>2
to form a nonstoichiometric oxide should result in a
mixing of Sn 5s states into the valence bands (possibly
as defect states) leaving a more Sn 5p like CBM. Such
a variation in symmetry near the conduction band minimum
should be apparent in the energy-loss spectrum. In
particular, a Sn 4d core-to-CBM transition will be dipole
unallowed for a Sn 5s dominated CBM, but dipole allowed
(Al=l) for a Sn 5p like CBM. Hence, the loss of oxygen
from SnC>2 should result in a change in the Sn 4d
core-to-CBM transition from unallowed to allowed. Notice
40
Table 3-1. Electronic Configurations of Atomic
Tin and Oxygen and the Stoichiometric Oxides
Atomic Tin
Atomic Oxygen
Stannous Oxide, SnO
Stannic Oxide, SnO?
gn : [Kr] 4d10 5s2 5p2
0 : Is2 2s2 2p4
Sn2+ : [Kr] 4d-*-0 5s2 5p
02- : Is2 2s2 2p6
Sn4+ : [Kr] 4d10 5s 5p
O2- : Is2 2s2 2p6
41
in Figure 3-1 that little change is expected in the energy
of the CBM between SnC>2 and SnO.
The
changing
nature
of
states near
the
tin oxide
CBM
may
be seen
clearly
in
Figures 3-3
and
3-4. The
46
eV loss feature may
be
interpreted as
i a
transition
from the Sn 4d core to a high lying Sn 4f-like CB state
[47]. The 27 eV feature may be interpreted as a Sn 4d
core-
-to-CBM transition
[61]. A feature near
2 7 eV
has
been
observed
in the
N(E) loss spectrum
for
both
SnO
and
Sn metal
[41,62]
but not for SnC>2.
For
the
case
of metallic tin, this feature may be viewed as a transition
from the Sn 4d core to empty states above the Fermi level.
The growth of the 27 eV loss feature in conjunction with
the decrease in the 46 eV feature may be interpreted
as a change in the CBs. The growth of the 27 eV loss
feature with decreasing beam energy is characteristic
of the changing nature of the CBM due to a deficiency
of oxygen in the surface region. This interpretation
is supported by the relative strengths of the two
transitions. The d > f transition is expected to
be stronger than the d > p transition [63].
The insensitivity of the 36 eV loss feature to
incident beam energy relative to the Sn 4d core features
discussed above suggests that the atomic origin of this
core derived feature is- significantly different. From
Figure 3-3 the main change in this feature is a gradual
42
decrease in intensity with decreasing beam energy. The
assignment of this 36 eV loss feature to an 0 2s core-to-CB
transition can explain this trend for a material exhibiting
a decreasing oxygen concentration on moving from the
bulk to the surface. This is precisely the situation
encountered in the present case.
Because the 0 2s and Sn 4d core levels couple to
CB final states of significantly different energy, the
energy-loss distribution due to these transitions may
be viewed as approximately proportional to the CB density
of states over a narrow range. There is certainly some
overlap between the 0 2s and Sn 4d transitions in the
neighborhood of the 0 2s feature. At the extremes,
however, near the 4 6 eV or 27 eV feature the
density-of-states interpretation should be valid although
substantial matrix element differences are likely between
these two regions. The observation from Figure 3-3
that changes in the core level features at high beam
energies are more dramatic than in the VB loss features
suggests the Sn 4d core features are more sensitive to
low concentrations of crystal structure defects than
the VB features. The Sn 4d core level losses may be
viewed as an strong indicator of the structural order
of the tin oxide material. Supporting evidence is found
from results on ion-sputtered samples.
43
Sputtering Effects
In order to increase the surface sensitivity of
the ELS measurement, the sample orientation was changed
to give the coaxial electron beam from the CMA an incident
angle of 45 with respect to the sample normal. This
change allowed reasonably surface sensitive measurements
with higher primary beam energies, and it eliminated
the problem of low-energy secondary electrons inherent
in the use of low electron beam energies (50 eV) for
N(E) measurements. Also, the probability of encountering
additional quadrupole-allowed features was minimized.
Assuming a straight line incident and exit path for an
electron scattered through a nominal angle of 137.7
(fixed by the CMA [64]), a very crude estimate of the
ELS analysis depth based on sample orientation and electron
mean free path can be made. Since the main features
observed in the ELS spectrum are due to single inelastic
events possibly in conjuction with elastic events, a
total path length of twice the mean free path of an
electron at the primary beam energy seems appropriate.
O
These
asssumptions lead
to an estimate of 5
to
10 A
(2
to
4
atomic
layers)
at
Ep = 2 00 eV and 15
to
20 A
(5
to
7
atomic
layers)
at
Ep = 1500 eV. These
estimates
should be viewed as qualitative at best.
Figure 3-5 shows the ELS data for a sample annealed
at 600C. The spectrum for Ep = 1500 eV shows the
structure characteristic of a well annealed SnC>2 bulk
NCE>
44
ENERGY LOSS CEV)
Figure 3-5. ELS spectra for a sample annealed at 600C.
The angle of incidence for the primary beam is 45. With
Ep = 1500 eV it is seen that the bulk is primarily SnC>2.
The Ep = 200 eV spectrum shows SnO at the surface and
evidence of a defect structure between bulk Sn02 and
the surface.
45
material. The more surface sensitive spectrum for Ep
= 200 eV has a sharp structure near 13 eV which is
characteristic of SnO [41]. The broad feature centered
at 18 eV is not characteristic of either SnO or Sn02,
and it most likely comes from a subsurface
nonstoichiometric defect structure accompanying the change
in structure from Sn02 in the bulk to SnO at the surface.
The 27 eV feature is also present indicating a structure
which is oxygen deficient in comparison to Sn02.
Figure 3-6a is the valence band XPS spectrum for
the 600C annealed sample. The resolution of the VB
XPS data is seen to be poor. This poor resolution is
due to a combination of very low signal intensity, the
x-ray line width, x-ray satellite emission from Sn 4d
and 0 2s core levels and data smoothing. In spite of
these difficulties, the general shape of the VB emission
is similar to that for SnC>2 as found by Lau and Wertheim
[32]. The obvious lack of surface sensitivity in this
measurement is not unexpected. Because the kinetic energy
of the valence band photoelectrons is large ( > 1200
eV), the mean free path is correspondingly large.
Additionally, the sample orientation is such that the
angle between the surface normal and the cylinder axis
is very nearly equal to the nominal 42.3 acceptance
angle of the CMA [64]. Since photoemission from
polycrystalline materials is expected to peak at the
n
46
BINDING ENERGY CEV)
Fiqure 3-6. Valence band XPS spectra after a (A) 600C
anneal and (B) 2 KeV argon-ion bombardment. The spectra
are more bulk than surface sensitive.
47
surface normal, the VB XPS results shown here have their
largest contribution from high energy electrons at near
normal emission. Hence, Figure 3-6a is primarily due
to the bulk SnC>2 material.
The annealed sample characterized by Figures 3-5
and 3-6a was ion sputtered with 2 KeV argon ions. Figure
3-6b illustrates the change in the VB XPS spectum. AES
and core level XPS show no evidence of a reduction to
metallic tin in this particular case, but the preferential
sputtering of oxygen is demonstrated by a drop in the
O/Sn ratio. Ion sputtering introduces a shoulder on
the VB emission near a binding energy of 2 eV to 3 eV.
A similar feature has been observed in ARUPS and
interpreted as emission from defect states associated
with a deficiency of oxygen [47]. Interestingly, this
sputter-induced feature lies near the same binding energy
as the highest lying SnO VB feature [32]. There is
even a fair correspondence between the SnO VBM as found
by Lau and Wertheim and the low binding energy edge of
the defect emission.
The assignment of the shoulder in Figure 3-6b to
defect states rather than SnO is justified by the ELS
spectra for the sputtered sample in Figure 3-7. The
more surface sensitive spectrum, Ep = 200 eV, shows a
broadening of the characteristic SnO feature at 13 eV.
The entire valence band portion of the spectrum becomes
48
broad and relatively featureless as a result of sputtering.
This broadening may be interpreted as a change from the
SnO structure at the surface to a more amorphous structure
caused by sputtering. For Ep = 1500 eV the loss spectrum
is sensitive to the bulk within the region probed by
the VB XPS measurements. While there is some broadening
and a small shift toward lower energy losses, the VB
features are still very much Sn02 like in approximate
agreement with the VB XPS spectrum shown in Figure 3-6b.
The core level loss features reflect the defect presence
much more strongly than the VB loss features. The absence
of the 46 eV feature and prominence of the 27 eV feature
confirm the change from a well-annealed Sn02 structure
to a more oxygen-deficient defect structure after
sputtering.
The damage induced by ion sputtering is heaviest
in the top few layer's of the solid as illustrated by
Figure 3-7. Thus, the amorphous structure at the surface
implied by the valence band features for Ep = 200 eV
is not unexpected. Sputtering damage in layers deeper
in from the surface may result from ion implantation,
knock-in and other ion-matrix phenomena, but the damage
in these deeper layers should be significantly less than
near the surface. Bearing in mind that ion bombardment
effects become less apparent as the experiment becomes
more bulk sensitive, a comparison of the results between
N
49
ENERGY LOSS CE^)
Figure 3-7. ELS spectra following a 2 KeV argon-ion
bombardment. The valencp band features show the bulk
to be SnC>2 like while the core features reveal a signifi
cant concentration of defects. The VB features in the
more surface sensitive spectrum illustrate an amorphous
structure at the surface due to a sputtering.
50
annealed and sputtered samples suggests that a significant
amount of qualitative structural information may be gained
from the N(E) energy-loss spectrum. The width and
center-of-gravity position of the valence band features
can be used as a gross indicator of the tin oxide
structure. A matrix characteristic of a stoichiometric
form of tin oxide is suggested by sharper, more well
defined valence band loss features near 19.5 eV for SnC>2
and near 13 eV for SnO as was found by Powell [41]. A
broadening and shift in energy between these two
characteristic features suggests an increasing structural
disorder. The radical change in core level features
in comparison to VB features suggest a higher sensitivity
to lattice defects. In particular, the 46 eV feature
appears to be an excellent indicator of the SnC>2 structure.
Even when the valence band features appear to be very
Sn02~like, the presence of defects is indicated by the
loss or decrease of the 46 eV feature relative to the
VB features. This interpretation of structurally related
changes in the ELS spectrum is strongly supported by
the combined LEED and ELS study of de Fresart et al. [58]
on SnC>2 (110 ) .
Oxygen Effects
It is shown above that a growth of the 27 eV feature
and loss of the 46 eV feature reflects a change in tin
oxide away from a well-annealed Sn02 material. Some
51
distinction between the origins of the changes in these
two core level features can be made. This distinction
requires a measure of the oxygen concentration which
is provided by core level XPS using standard sensitivity
factors [31]. To make comparisons with VB XPS useful,
attention is limited to the more bulk sensitive energy-loss
measurements for 1500
eV primary
beam
energies.
The
quantitation of oxygen
levels within
the
matrix by
core
level XPS presents a
problem due
to a
difference in
analysis depth with respect to VB
XPS
and ELS.
This
problem is minimized by using the O/Sn ratios determined
in this manner as only a rough measure of the oxygen
concentration further into the bulk. The O/Sn ratios
are reported within an uncertainty of 0.03 which describes
the reproducibility of the measurements. No uncertainty
in the sensitivity factors is reported. In this regard
trends in the O/Sn ratios are more important than the
absolute values.
For the sputtered sample described by Figures 3-6b
and 3-7, the O/Sn ratio is 0.96. Annealing the sputtered
sample in_ vacuo at 500C for 20 minutes repairs some
of the sputter-induced damage. Figure 3-8a shows the
effect on the VB XPS spectrum. A decrease in the defect
feature at low binding energies is observed, and a
splitting in the VB features at 4.5 eV binding energy
appears. This splitting is characteristic of a mixture
NCE>
52
SINDIH6 EhCRSY
Figure 3-8. Valence band XPS spectra following (A) a
short 500 C anneal after sputtering and (B) subsequent
oxygen exposure. The presence of a mixture of SnO and
Sn02 is indicated in (A). '
53
of SnO and SnC>2 [32]. The short anneal also increases
the O/Sn
ratio
to
1.14 presumably
due
to some oxygen
diffusion
into
the
surface region
from
the bulk. The
energy-loss spectrum in Figure 3-9a also shows more
evidence of structural repair caused by annealing. The
largest valence band feature is sharper and centered
at 19.5 eV, and the presence of the high energy loss
feature near 45 eV is again slightly visible. Both of
these features indicate the presence of an SnC>2 structure.
The presence of the 27 eV feature reveals an oxygen
deficiency relative to SnC>2r and the size and shape of
the feature near 13 eV suggests the possibility that
SnO is present. However, the 13 eV feature is a
convolution of SnO and Sn02 features which yields little
information by casual inspection.
Subsequent treatment in. situ with 11 Torr of O2
at 500C for 15 minutes results in the addition of a
significant amount of oxygen to the matrix, O/Sn = 1.34.
Figure 3-8b shows the effect on the VB XPS spectrum.
The splitting which is apparent in Figure 3-8a is removed,
and the shape of the VB emission is predominantly that
of Sn02* The addition of oxygen also affects the ELS
spectrum as seen in Figure 3-9b. The change in VB features
is minimal. The main loss feature falls at 19.5 eV as
expected for an Sn02 material, and there is a decrease
in the feature near 13 eV relative to the 19.5 eV feature
NCE>
54
EM-R6Y LOSS (EV)
Figure 3-9. ELS spectra corresponding to Figures 3-8(A)
and (B) respectively.
55
suggesting a loss of the SnO contribution to the spectrum.
The most apparent changes occur in the core level features.
A small increase in intensity of the feature near 36
eV is observed. This increase is consistent with the
assignment of this feature to 0 2s-to-CB transitions.
It can be seen that the 27 eV feature in Figure
3-9b is greatly diminished. The loss of this feature
by annealing in oxygen substantiates the earlier
interpretation that it is associated with a loss of oxygen
from the Sn02 structure and be may interpreted as due
to a change in symmetry of the states near the CBM. It
seems that the growth of the 2 7 eV feature reflects a
loss of coordinating oxygen or a lowering of the valency
of the tin. This loss may occur through the formation
of defects such as oxygen vacancies in a nonstoichiometric
or amorphous oxide, through the formation of stoichiometric
SnO or through the formation of metallic Sn.
Figure 3-9b demonstrates that the 46 eV and 27 eV
loss features are not strictly interdependent. The weak
intensity of the high-energy loss feature suggests a
sensitivity to defects other than those associated only
with a deficiency of oxygen in a Sn02 lattice. It is
postulated that the high-lying conduction-band final
states associated with this transition are strongly
dependent on the periodic potential of the SnC>2 lattice
and easily perturbed by the presence of defects. This
56
strong dependence may occur if the states are less atomic
in nature than the valence and lower conduction bands
while still containing a fraction of tin 4f character
as suggested by Gobby [47].
Conclusions
The use of ELS combined with valence band
photoemission and results of band structure calculations
provides a powerful means for studying tin oxide surfaces.
In this study an assignment of the major features in
the tin oxide N(E) energy-loss spectrum is made. The
loss features are assigned to collections of optically
(dipole) allowed interband transitions based on a previous
photoemission study by Gobby [47]. It is found that
the low-energy portion of the spectrum may be associated
with valence-to-conduction-band transitions, and the
higher energy-loss features are due to
core-to-conduction-band transitions. Of these core level
features, it is possible to distinguish between transitions
from Sn 4d and 0 2s levels even though these features
cannot be resolved in XPS.
It is demonstrated for tin oxide surfaces that depth
profile information may be obtained using ELS. By varying
the primary electron beam energy and hence the analysis
depth, it is shown that a high temperature anneal results
in bulk Sn02 under an oxygen deficient structure which
is essentially SnO at the surface.
57
By using ELS in conjunction with valence band XPS,
it is found that a significant amount of structural
information may be inferred from the size, shape and
position of the N(E) ELS features. In particular,
distinctions can be made between SnC>2, SnO and defect
or amorphous structures. The Sn 4d core level features
are found to be much more sensitive to defects in an
SnC>2-like lattice than are the VB features. A loss feature
at 27 eV assigned to transitions from Sn 4d levels to
states near the CBM is associated with atoms in a lower
oxidation state or in a lattice deficient in oxygen
relative to SnC>2. An SnC>2 loss feature near 45 eV is
shown to be very sensitive to defects not necessarily
associated with oxygen vacancies or deficiencies, but
the specific type(s) of structural defect(s) associated
with the behavior of this high-energy-loss (45 eV) feature
has not yet been determined.
SECTION IV
A STUDY OF THE DEHYDRATION OF TIN OXIDE
SURFACE LAYERS
Introduction
The chemisorption properties of tin oxide surfaces
can be significantly influenced by the interaction with
water. Kaji et al. have demonstrated that it is possible
to fixate Cu (II) and Pd (II) complex ions on hydrated
tin oxide surfaces in the preparation of propylene
oxidation catalysts [65]. The modification of tin oxide
electrode surfaces by an alkaline pretreatment has been
shown to give enhanced cell emf responses to changes
in pH [28]. This enhancement is thought to occur through
the hydrolysis of surface Sn=0 bonds to give Sn-OH surface
species. The specific adsorption of Fe (III) and Pb
(II) cations has been shown to occur on these hydrated
surfaces apparently by replacement of the proton on the
surface hydroxyl groups [66,67]. Similarly, the specific
adsorption of bromine and iodine anions on tin oxide
occurs only on hydrated surfaces [68,69]. Most recently
it has been shown that an increase in Pt uptake rates
during chemisorption frbm solution occurs on hydrated
tin oxide surfaces [6]. Pt dispersions can exceed 90%
58
59
on these surfaces, and the resulting catalytic activity
per surface Pt atom exceeds that of metallic Pt electrodes
for the electrochemical reduction of C>2.
Secondary-ion mass spectrometry (SIMS) and
electron-stimulated desorption (ESD) were used in a
previous study to examine alkaline-pretreated tin oxide
for evidence of surface hydroxylation [19], ESD
demonstrated higher yields of both H+ and 0+ after the
alkaline pretreatment suggestive of significant surface
hydroxylation. A small signal due to 0H+ desorption
was also observed. Results using dynamic SIMS showed
no apparent differences in the bulk regardless of
pretreatment. It has become apparent, however, that
hydrogen is a major constituent in most tin oxide films,
and it appears that the actual film composition may be
best described as Snx0yH2. SIMS depth profiles of H+,
0+, 0H+ and SnH+ species indicate an excess of hydrogen
and/or hydroxide or hydrated species at the surface of
tin oxide films [40]. A steep concentration gradient
O
within approximately the outer 30 A of the material
indicates that hydration is not limited strictly to the
outer atomic layer. While the degree of hydration is
greatest for the alkaline-treated samples, significant
hydration occurs over the same depth for samples exposed
to atmospheric humidity only. This observation is
indicative of the hygroscopic nature of tin oxide surfaces.
60
The complexity of the interaction of water with
tin oxide is demonstrated by the work Giesekke et al.
on the decomposition of bulk tin (IV) hydroxide [55].
Using thermogravimetric analysis, it was determined that
the decomposition of SnC>3H2 leads to the formation of
Sn2C>5H2 above 250C, Sn^gf^ between 325C and 360C,
Sn816H2 at 500C and SnC>2 above 600C. Electron
diffraction clearly shows that each dehydration product
is a different crystalline substance. Though an accurate
determination of the structures was not possible, a study
of proton magnetic resonance line shapes shows the
structures to be complex. None of the substances can
be described as simple hydrates or hydroxides.
In Section III it was shown that electron energy-loss
spectroscopy (ELS)
is
sensitive
to electronic changes
in tin oxide and
is
useful as
either a
surface or
subsurface probe.
With
the aid
of valence
band x-ray
photoelectron spectroscopy (XPS), it was also shown that
certain changes in the ELS spectrum may be related to
structural changes in the material. ELS and valence
band XPS are used in the present study of the hydrated
layer formed on tin oxide by exposure to atmospheric
humidity. A preliminary observation of water adsorption
using grazing-exit-angle XPS and ESD is given in Section
V.
61
Experimental
The preparation of the polycrystalline tin oxide
film used in this study has been described in Section
III. Once prepared the sample was exposed to atmospheric
humidity for several months to allow hydration of the
near surface region. All spectra were collected with
a double-pass CMA. The ELS data were taken in the N(E)
mode to allow for a direct comparison with the data of
Powell [41]. A 100 nA, 0.1 mm diameter primary electron
beam was used. All ELS spectra were recorded with a
25 eV pass energy ( AE/E = 0.014) using pulse counting
detection. The XPS spectra were taken using a Mg Ka x-ray
source and a 50 eV analyzer pass energy. The base pressure
in the vacuum system for this study was lxl0~l Torr.
Details of the vacuum system have been given previously
[17] .
The ELS spectra were taken using the coaxial electron
gun in the CMA at an incident angle of 45 with respect
to
the sample
normal. As shown
in
Section
III,
ELS
measurements sensitive to the top
few
atomic
layers
can
be
obtained in
this configuration
using
a 200
eV primary
beam energy (Ep). Using a 1500 eV primary beam energy
significantly decreases the surface sensitivity of the
ELS measurement. This higher beam energy makes ELS more
sensitive to the subsurface region with an estimated
O
analysis depth of approximately 20 A.
The valence band
62
(VB) XPS spectra obtained with this sample orientation
are also more sensitive to the subsurface region, and
the VB XPS analysis depth is expected to be similar to
the high energy (Ep = 1500 eV) ELS measurements. Both
the subsurface VB XPS and ELS measurements are sensitive
to the same region in which previous SIMS results [40]
suggest that hydration occurs.
Results and Discussion
Prior to analysis, the sample was cleaned in_ situ
by heating to 500C in 10 Torr of O2 for 5 minutes. This
procedure removed carbon and chlorine contamination while
leaving a trace amount of K on an otherwise clean oxide
surface as determined by Auger electron spectroscopy
(AES). This contamination is known to be segregated
at the surface [17], and it may be removed easily by
Ar+ bombardment. However, in order to preserve the
hydrated layer of the sample, no ion bombardment was
used. While the O2 treatment may have affected the very
near-surface region of the sample, the following data
reveal that the subsurface layers were not dehydrated.
Figure 4-1 shows the valence band XPS data for the
sample after various treatments. Figure 4-la is the
spectrum recorded after the in. situ cleaning. Figure
4-2a illustrates the effect of a 500C vacuum anneal
on the spectrum. The annealing process was continued
63
until the background chamber pressure went through a
distinct maximum (about 45 minutes). Figure 4-lc shows
the result of a similar (30 minute) 600C anneal. Figures
4-2, 4-3 and 4-4 show the ELS spectra corresponding to
Figures 4-la, 4-lb and 4-lc respectively.
The valence band spectrum in Figure 4-la is
characteristic of a sample hydrated due to exposure to
atmospheric humidity. The large feature near 10 eV binding
energy (7 eV below the valence band maximum (VBM)) is
largely due to this hydration though the form of the
incorporated water is unknown. The 10 eV feature is
similar to that found for water adsorption on other oxides
[70].
Figure 4-2a shows the ELS spectrum (Ep = 1500 eV)
corresponding to Figure 4-la. The features at energy
losses less than about 30 eV are due to
valence-to-conduction-band transitions as discussed in
Section III. The main VB loss feature in Figure 4-2a
falls at 20 eV which is characteristic of a Sn02~like
material (see Section III and ref. 41). Additionally,
a large shoulder associated with the VB loss features
is observed. This feature has not been previously
reportad, and
it
appears
to
be composed
of two
loss
features near
24.5
eV and
27
eV. This shoulder is
not
characteristic
of
SnC>2. '
These additional
features
may
be interpreted as transitions from the hydrate-induced
nce:>
64
Figure 4-1. Valence band XPS spectra after (A) hydration
by exposure to atmospheric humidity, (B) a 500C vacuum
anneal for 45 minutes and (C) a 600C vacuum anneal for
30 minutes.
ncet:>
65
i n 111 i 11111111 ji'i h m n | u i i f 1111 j 1111 i 11 j 11111 i u i n 111 i i n 11
un 11111111111111111111111 in 11111111111111111111111111 m 11111
68 50 46 30 20 10
ENERGY LOSS CEV)
Figure 4-2. ELS spectra of (A) subsurface and (B) surface
regions after hydration due to atmospheric humidity.
66
lower valence band feature observed in Figure 4-la to
conduction band, states. This interpretation is in
agreement with the ultraviolet photoelectron spectroscopy
(UPS) measurements of Gobby [47]. The UPS results show
that the lower valence band feature couples strongly
to conduction band states in a range from 2 5 eV to 3 0
eV higher in energy. Features at energy losses greater
than about 3 0 eV are due to core-to-conduction-band
transitions. In particular, the broad feature centered
near 36 eV is due to a set of 0 2s-to-CB transitions
(see Section III).
Figure 4-2b (Ep = 200 eV) shows the more surface
sensitive ELS spectrum of the hydrated sample. The main
VB loss feature falls near 19 eV suggestive of an Sn02-like
material. The lack of higher energy-loss VB features
in this spectrum indicates a surface which is dehydrated
relative to the subsurface layers. This dehydration
is most likely the result of ESD from the surface under
the influence of the primary electron beam and/or some
superficial dehydration due to the elevated temperature
used during the oxygen cleaning procedure.
The main VB loss features in Figures 4-2a and 4-2b
suggest that the sample is fully oxidized in both the
surface and subsurface regions. This is confirmed for
the surface region in Figure 4-2b by the lack of a 27
eV Sn 4d core level loss feature which would be
67
characteristic of a deficiency of oxygen relative to
SnC>2 (see Section III). Although the presence of such
a 27 eV feature would be obscured in Figure 2a by the
hydrate-induced VB loss feature, the absence of any low
binding energy structure (2 eV to 3 eV) in Figure 4-la
reveals that no oxygen deficiency exists (see Section
III). It is apparent, however, that the structure of
the material is perturbed relative to a well-annealed
SnC>2 rutile structure. This perturbation is evidenced
by the lack of a core level loss feature near 4 5 eV in
Figure 4-2 (see Section III). For the hydrated subsurface
layers, the perturbation is easily understood as due
to the addition of excess oxygen and hydrogen from the
water of hydration while at the surface beam damage is
the most likely cause.
Figure 4-lb shows the effect of a 500C vacuum anneal
on the valence bands. The large feature near 10 eV in
Figure 4-la has been greatly reduced suggestive of a
dehydration of the subsurface region, and a spectrum
very similar to that characteristic of SnC>2 remains (see
Section III and ref. 32). This change is also reflected
in the ELS spectrum in Figure 4-3a. While the main VB
loss feature remains near 20 eV, the features in the
25 eV to 27 eV region associated with the hydrated oxide
are substantially decreased. The valence band ELS features
in Figure 4-3a are quite Sn02 like in agreement with
NCE>
68
t i i j 111111 i i q i n i n i Fi ] 1111111111111111 j i j j 11111 j 111111111111
iiiiilmim nil limn ill miimlm n i mlm mm In in m
ee
S8
46 38 20
QCR6Y LOSS CEV:
16
Figure 4-3. ELS spectra of (A) subsurface and (B) surface
regions after a 500C vacuum anneal.
69
the VB XPS spectrum of Figure 4-lb. Concurrent with
the change in VB features, the appearance of a small
core level loss feature near 45 eV is observed in Figure
4-3a. The weak presence of the 45 eV loss feature
represents the beginning of a change in the subsurface
oxide to a true SnC>2 structure (see Section III). However,
a significant perturbation of this structure is still
apparent. At the surface the annealing has resulted
in an oxygen deficiency as illustrated by Figure 4-3b
(Ep = 200 eV). The broad valence band features with
increased intensity at 13 eV show that the surface has
changed from Sn02~like to a more SnO-like material. The
oxygen deficiency in the surface region is confirmed
by the appearance of the small feature near 27 eV (see
Section III).
Further annealing at 600C has only a small effect
on the VB XPS spectrum shown in Figure 4-lc. The feature
near 10 eV binding energy is completely removed leaving
a spectrum characteristic of Sn02 (see Section III and
ref. 32). Similarly, the ELS spectrum in Figure 4-4a
shows the core level loss features characteristic of
a well-annealed Sn02 material indicating the nearly
complete dehydration of the subsurface region. The
annealing process has, however, effected a reduction
at the surface. The sharp sturcture near 13 eV in Figure
4-4b is characteristic of SnO as found by Powell [41].
n < e;>
70
ENERGY LOSS CEV)
Figure 4-4. ELS spectra of (A) subsurface
regions after a 600C vacuum anneal.
and (B) surface
71
The broad feature centered near 18 eV is not characteristic
of SnO or SnC>21 and it may be interpreted as due to a
nonstoichiometric defect structure accompanying the change
from subsurface (bulk) SnC>2 to surface SnO (see Section
III) .
The changes observed in the subsurface layers using
ELS clearly indicate a temperature dependence in the
decomposition of the hydrated near-surface region. The
dehydration product observed at 500C is a forerunner
to the formation of a true Sn02 compound near 600C in
the subsurface region. These observations are in agreement
with the work of Giesekke et al. [55] on the thermal
decomposition of bulk tin (IV) hydroxide. The formation
of SngOggl^ could account for the apparent perturbation
of the subsurface crystal structure evidenced by Figure
4-3a while causing only a small variation in the VB density
of states from that expected for SnC>2
Conclusions
The near-surface region of a hydrated polycrystalline
tin oxide film has been studied. A large increase in
the lower VB density of states has been observed for
hydrated subsurface layers using VB XPS and ELS. These
observations are in agreement with SIMS data [40] which
suggests that hydration -due to exposure to atmospheric
O
humidity occurs to depths of at least 30 A.
72
The thermal decomposition appears to proceed in
a stepwise fashion. The subsurface hydrated layers yield
SnC>2 near 600C, but the surface undergoes a reduction
to SnO. A comparison with existing data on bulk tin
(IV) hydroxide decomposition leads to an interpretation
consistent with the formation of an intermediate
hydrogen-containing compound in the subsurface region
near 500C.
SECTION V
AN OBSERVATION OF WATER ADSORPTION ON TIN OXIDE
USING ESD AND GRAZING-EXIT-ANGLE
XPS AND AES
Introduction
As discussed in Sections I and IV, the chemisorption
properties of tin oxide surfaces can be significantly-
affected through the interaction with water. In
particular, the chemisorption of platinum on tin oxide
surfaces is believed to occur at hydroxylated surface
sites (see Section II and ref. 6). In Section IV, the
dehydration of tin oxide surfaces has been studied using
valence band x-ray photoelectron spectroscopy (XPS) and
electron energy-loss spectroscopy (ELS). These techniques
have proven particularly useful in studying the subsurface
layers of the material. Though ELS may be made quite
surface sensitive by lowering the primary beam energy,
the study of adsorbates on tin oxide with this technique
is made difficult by the phenomenon of electron-stimulated
desorption (ESD).
Some preliminary observations of water adsorption
on tin oxide are given in this section. ESD experiments
and grazing-exit-angle XPS and Auger electron spectroscopy
(AES) measurements provide this observation. Though
73
74
the data presented here is incomplete, it provides an
interesting comparison with the dehydration study in
Section IV, with previous ESD data on alkaline and
non-alkaline treated tin oxide surfaces [19] and with
the work of Giesekke et al. [55] on the dehydration of
bulk Sn (IV) hydroxide. The incomplete nature of the
ESD experiments is due to a prolonged (16 months and
counting) failure of the Physical Electronics double-pass
CMA. Grazing-exit-angle XPS and AES are used because
of the unavailability of a preferred technique, ultraviolet
photoelectron spectroscopy (UPS).
Experimental
The preparation of the polycrystalline tin oxide
film used in this study has been described in Section
III. After preparation, the sample is rinsed in distilled
water and solvent cleaned prior to insertion into the
vacuum system. Before analysis the sample is cleaned
in situ by heating to 500C in 10 Torr of O2 for 5 minutes.
This procedure removes trace chlorine and carbon surface
contamination leaving a clean oxide surface as determined
by AES. The AES, XPS and ESD data were taken with a
PHI double-pass CMA. The AES spectra are collected in
the nonretarding (EN(E) ) mode using a 3 KeV, 200 mA/cm2
electron beam. For XPS the analyzer was run in the
retarding (N(E)) mode with a pass energy of 5 0 eV ( AE/E
75
= 0.014). Details of the vacuum system have been published
previously [17]. The base pressure for this study was
5xl0-10 Torr.
The surface sensitivity of the electron spectroscopies
(AES and XPS) can be improved by collecting the emission
at angles away from the sample normal, i.e. at a more
grazing exit angle. The path length, L, that an escaping
electron (photoelectron or Auger electron) must travel
through a solid is related to the signal attenuation
due to inelastic collisions. Signal attenuation as a
function of path length can be described by an exponential
decay law with a uniform attenuation length. The
attenuation length, X, is known as the mean free path.
Hence,
I exp(-L/X )
where I is the signal intensity. Figure 5-1 illustrates
the increased surface sensitivity obtained at grazing
exit angles for a perfectly flat surface. If an emitting
source (atom) is a fixed distance, D, below the surface,
the path length, L, traversed within the solid increases
with increasing exit angle, 0, as
L' = D/cos 0
76
Figure 5-1. Variation in path length with emission angle.
77
Therefore, at fixed D, the signal intensity decreases
with increasing 0 making the measurement more surface
sensitive. In general, experimentally observed increases
in surface sensitivity are not as large as expected from
the above analysis. Two possible reasons for the deviation
are surface roughness and a decay of total signal with
increasing 0 causing a reduction in the signal-to-noise
ratio [71].
For all measurements the sample was mounted with
approximately a 4 5 angle between the CMA axis and the
surface normal. This orientation directs the sample
normal into the 42.3 6 acceptance cone of the CMA
[64]. Grazing exit angles are chosen with the 12 angular
acceptance aperture on the angle resolving drum mounted
coaxially within the inner cylinder of the second stage
of the CMA [72]. Using the relationship derived by Gobby
[47], the exit angle for a given drum setting may be
found.
ESD experiments are performed by using the CMA in
a time-of-flight (TOF) mode which allows for a simultaneous
determination of the mass and energy of desorbing ions.
For these measurements the analyzer is operated at a
constant pass energy of about 80 eV. This pass energy
(kinetic energy of the analyzed ions) sets the flight
time of the ions through the analyzer (about 4 nsec for
H+). Because the CMA only passes charged particles of
78
the proper kinetic energy, species of different masses
(but same charge) have different axial velocities through
the CMA. Hence, the flight time of an ion through the
analyzer is directly proportional to the square root
of the mass-to-charge ratio. Traum and Woodruff [73]
have discussed in-depth the analyzer characteristics
which effect the flight time and mass resolution. Unity
mass resolution is possible for mass-to-charge ratios
(m/e) of at least 20.
To operate the CMA in a TOF mode for ESD experiments,
a computer-interfaced digital pulse counting circuit
is used (see Appendix B). The TOF modification to the
pulse counter allows it to perform three functions:
(1) it initiates the desorption event,
(2) it delays for a programmed flight time
(3) and it measures (counts) the signal pulses.
Before beginning the TOF analysis, the coaxial
electron gun in the CMA is configured with a +50 V charge
on the lower deflection plate. This voltage deflects
the electron beam (typically below 200 eV) downward out
of the analysis area (focal region) of the analyzer.
The pulse counter circuit initiates the desorption event
by supplying a 300 nsec TTL pulge to the base of a n-p-n
power transistor in series with the deflection plate
and ground (see Figure 6-2). This pulse "grounds" the
deflection plate and swings the electron beam into the
79
TTL PULSE INPUT
+50 V
5K OHMS
1/
MRF427A
DEFLECTION
PLATE
Figure 5-2. Deflection circuit for desorption
initiation.
event
80
analysis region for 300 nsec. The circuit delays for
a programmed flight time before enabling an event counter
which records signal pulses for a similar 300 nsec period.
The count is subsequently read into the computer where
it is stored, and the process is repeated. By scanning
the programmed delay time a TOF (i.e. m/e) spectrum is
obtained. The only real-time constraint is that a total
time span be observed between desorption events at least
equal to the flight time of the most massive species
in the spectrum. This delay clears the analyzer of ions
before the initiation of a new desorption event.
To analyze low energy positive ions like those
obtained in the ESD experiment, the CMA is operated in
an accelerating mode. The inner cylinder and accelerating
grid which are connected internally are set initially
at -70 V, and the sample is biased at +10 V. This
potential difference between the sample and accelerating
grid raises ions initially at zero kinetic energy up
to the 80 eV analyzer pass energy thereby allowing their
detection. By ramping the accelerating grid to more
positive potentials, ions of higher initial energy
(typically 10 to 20 eV) can be measured. If the TOF
analyzer is operated at a fixed flight time, an energy
distribution spectrum of a single desorbing species may
be obtained. In this experiment, part of the accelerating
potential is imposed on the sample to provide a voltage
81
difference with an outer grid which is grounded to the
magnetic shield of the CMA. In this way any spurious
signal due to ESD from this grid is shifted to apparent
negative kinetic energies and is easily recognizable
[73] .
Results and Discussion
XPS and AES
Measurements have been made following three different
in situ treatments. These treatments include a 500C
and a 600C annealing step as was studied in Section
IV. Measurements have also been made following a 2 KeV
argon ion bombardment. Three types of AES and XPS
measurements are reported. Angle integrated results
are obtained with the angle resolving aperture retracted.
Angle dependent results have been obtained at normal
emission (0 4) and at a 70 2.5 grazing exit angle.
The results are presented in terms of O/Sn ratios. For
XPS this determination is made using the area under the
0 Is and Sn 3d 5/2 peaks corrected with standard
sensitivity factors [31], The AES measurements are made
in a similar fashion using the peak-to-peak heights of
the O KLL (512 eV) and Sn MNN (437 eV) transitions [74],
A problem encountered when using the angle resolving
aperture for these measurements is a drop in total signal
and in signal-to-noise ratio. In AES this drop is not
82
a significant problem because of the magnitude of the
signal, but in XPS the drop is so large that a minimum
seven hour period is required to accumulate enough signal
for a single O/Sn ratio determination. The measurements
were undertaken originally to demonstrate the depletion
of oxygen near the surface due to annealing (as observed
in Section IV) but water adsorption from the background
vacuum is observed instead because of the extended period
of time required to collect the data. Assuming a
background of water at the base pressure for this study
(5x10~5 Torr), the surface receives a 12.5 Langmuir (1L
= lxl06 Torr-sec) dose over a seven hour period. For
a unity sticking coefficient this dose represents about
12 monolayers of water. The water adsorption from the
background vacuum observed here is manifested by an
increased O/Sn ratio in the most surface sensitive (70)
XPS measurement. The adsorption of CO is believed not
to be a contributing factor because no carbon or CO
desorption signal is observed in the subsequent ESD
experiments.
The XPS and AES results are given in Table 5-1.
Angle integrated measurements (with the aperture retracted)
are made quickly after a given treatment before any
significant H2O adsorption occurs. These results indicate
a drop in the O/Sn ratio with increased annealing
temperatures and argon ion bombardment. The angle
83
Table 5-1. Variation in O/Sn Ratio With Emission Angle.
XPS
AES
508C
ANNEAL
ANGLE INTEGRATED 1.4
NORMAL EMISSION 1.5
78 DEGREE 2.1
1.3
600C
ANNEAL
ANGLE INTEGRATED 1.3
NORMAL EMISSION 1.4 -1.5
70 DEGREE 1.8-2.3
l.l 1.2
2K EV
SPUTTER
ANGLE INTEGRATED 1.0
NORMAL EMISSION 1.1
70 DEGREE 1.5
1.0
84
integrated results are consistently similar to, but
slightly lower than, those obtained at normal emission.
This observation illustrates that the signal intensity
is highest at the sample normal as expected for a
polycrystalline material. A similar observation has
been made in Section III regarding the lack of surface
sensitivity in valence band XPS spectra. It is worth
noting that angle integrated measurements taken several
hours after a given treatment show a small increase in
O/Sn ratio like that observed with the seven hour normal
emission measurements.
The increase in O/Sn ratio observed for 70 emission
illustrates a significant uptake of H2O at the surface
from the background vacuum. Indeed, the large O/Sn ratio
of 2.3 observed in one case, suggests the formation of
a hydrated surface. Regardless of the order in which
the data is taken (i.e. the total exposure), the 70
emission always shows a substantially higher O/Sn ratio
indicating H2O adsorption at the surface. Water adsorption
during the normal emission measurements also explains
the small increase observed relative to the angle
integrated measurements.
The AES results given in Table 5-1 show no variation
with exit angle, and the O/Sn ratio is generally lower
than that obtained by angle integrated XPS. The lower
O/Sn value relative to XPS is probably due, in part,
to the increased surface sensitivity of AES. The kinetic
energies of the AES peaks are more than 200 eV less than
85
are observed in AES even after several hours of exposure
to the background vacuum is due to the ESD phenomenon.
As observed in a previous study [19], the surface
concentration of desorbing species can be rapidly depleted
under an electron beam of high current density. It is
believed that water adsorbed on the surface is quickly
removed by the incident beam used for the AES analysis
and is therefore undetected.
ESD
The first TOF ESD measurements made in this laboratory
are reported here. The data was acquired during the
process of tuning the instrument for the first time.
Unfortunately, an electron gun failure ended this
familiarization procedure before a good rapport could
be developed with the experimental set-up. Therefore,
the results shown here do not represent the full
capabilities of the equipment.
Figure 5-3 is the TOF spectrum of a tin oxide surface
sputtered with 2 KeV argon ions. The desorption of H+
and 0+ is clearly visible at 4.4 usee and 17.6 psec,
respectively. The feature at 18.6 nsec is possibly due
to H20+ desorption, but it is most likely due to 0+
desorption with different initial conditions than in
the 17.6 usee peak. These possibilities can be checked
by increasing the accelerating potential and compressing
the flight times in the entrance region of the analyzer.
ION YIELD
86
Figure 5-3. Time-of-flight spectrum for mass analysis.
87
The 12 angular aperture has been used to select species
desorbing at near normal angles. Traum and Woodruff
[73] have shown that a significant increase in mass
resolution is possible by using the 4 angular acceptance
aperture. With the smaller aperture it should be possible
to resolve 0+ and 0H+ species.
Figure 5-4 shows the ion kinetic energy distribution
for the sputtered sample after exposure to a high current
density electron beam for 30 minutes. Figure 5-4a shows
the total ion yield, 5-4b the time-gated (mass resolved)
H+ ion yield and 5-4c the time-gated 0+ ion yield. It
is seen that the CMA may be used for a simultaneous mass
and energy determination.
Figure 5-5 shows the total ion kinetic energy
distribution for the same sample after exposure to the
background vacuum for two hours. No time-gated
distributions were obtained in this case. Though a power
supply problem encountered during the analysis prevents
an accurate determination of the true kinetic energy
scale, the total ion energy distribution is seen to be
very different after H2O adsorption. This variation
suggest that ESD will prove useful in distinguishing
between different forms of hydrogen and oxygen on the
tin oxide surface.
NCE}
88
0 5 10
ION KINETIC ENERGY CEV)
Figure 5-4. Ion kinetic energy distribution after sputter-
ing.
NCE>
89
Figure 5-5. Ion kinetic energy distribution following
water adsorption.
90
Conclusions
Grazing-exit-angle XPS demonstrates that a significant
uptake of H2O occurs on tin oxide surfaces. This uptake
appears to proceed to the point of surface hydration
as expected from similar observations in Section IV.
Grazing-exit-angle AES demonstrates that this hydration
formed under UHV conditions is limited to the outer layers
of the material because it is easily removed by electron
bombardment. The ESD results demonstrate that the
experimental set-up used here is viable. Observed
differences in ion kinetic energy distributions for
sputtered surfaces before and after H2O adsorption
illustrates the potential of the ESD technique as a tool
for studying chemisorption on tin oxide.
SECTION VI
THE INTERACTION OF POLYCRYSTALLINE ZIRCONIUM
WITH 02, N2, CO AND N20
Introduction
Very few surface studies have dealt with zirconium.
Foord, Goddard, and Lambert [75] have studied the
interaction of zirconium with CO, NO, N2, 02 and D2 using
Auger electron spectroscopy (AES), work function
measurements and thermal programmed desorption (TPD).
In a later study Danielson [76] attempted to quantify
the zirconium AES peak heights as a function of amount
of carbon and oxygen contamination. These studies both
demonstrate that, quite unlike most metals, heating causes
adsorbates to migrate from the surface into the bulk
rather than to desorb. This is very convenient with
regard to producing a clean surface, but much information
is lost about the gas-solid interaction because TPD cannot
be used. However, Foord et al. were able to determine
the diffusion coefficients for surface to bulk transport
of C, N and 0 over a range of temperature by monitoring
the surface composition as a function of time with AES.
Only deuterium shows desorption behavior as the sample
is heated.
91
92
The two previous studies disagree on an important
point. Foord et al. claim that the zirconium 175 eV
AES peak height is the most sensitive to chemisorbed
species, but Danielson's study shows the 175 eV peak
height to be almost completely insensitive to surface
carbon and oxygen. This present study relates the
previously observed phenomena to the past history of
the sample, particularly with respect to heating the
sample above the HCP-to-BCC phase transition at 1135K.
The results suggest that the phase transition in the
surface region occurs slowly at room temperature over
a time period of several days and that the chemisorption
properties of the zirconium toward nitrogen change
dramatically over the same period. This appears to be
time-dependent chemisorption on a clean metal surface
due to electronic changes in the valence band caused
by alterations in geometric structure. Poppa and Soria
[77] have recently reported similar reductions in the
amount of CO and H2 adsorbed on (111)-type palladium
layers after annealing at 600K. Low dose, argon ion
bombardment restored the high adsorption probability.
Experimental
The experiments were carried out in an ultrahigh
vacuum chamber which has a base pressure of 3xlO""H Torr.
A PHI double-pass cylindrical mirror analyzer was used
to perform AES and XPS.
93
The zirconium
sample
was a foil
of
approximate
dimensions 12x3x0.5
mm and
99.9% pure.
The
sample was
cleaned in a hydrofluoric acid solution in order to remove
most of the accumulated oxide layer. Next it was solvent
cleaned in ethanol and then mechanically supported between
stainless steel rods. The sample was heated resistively.
Results and Discussion
An AES spectrum showed the sample to be contaminated
initially with carbon, oxygen and nitrogen. These
contaminates were removed easily by heating below the
HCP-BCC phase transition to drive them into the bulk
as discussed in the earlier papers [75,76]. Figure 6-la
shows an AES spectrum taken after 2 hours of heating.
The small amount of nitrogen and oxygen which are on
the surface initially have disappeared, and the carbon
peak shape has changed from graphitic to carbidic after
heating [74]. The oxygen diffuses into the bulk faster
than the carbon which is consistent with the conclusion
of Foord et al. The carbon peak decreases so slowly
that it requires heating overnight to be reduced to the
height shown in Figure 6-lb.. The sample could be cleaned
in a few minutes by heating above the transition
temperature. Cleaning in this manner does not
significantly alter the 'sample because even as much as
20 layers of contamination adds only 1 ppm of bulk
PNCE^/DE ARBITRARY UNITS
94
100 200 300 400 500 600
KINETIC ENERGY CEV)
Figure 6-1. AES spectra taken after (A) 2 hours of heating
and (B) 14 hours of heating below the HCP-to-BCC transition
temperature.
95
contamination after migration from the surface into the
bulk.
The zirconium AES transitions lie below 200 eV.
Foord et al. state that the 92 eV peak is of the type
(MMN + MNN) thus involving only core level electrons,
and the 120 eV signal has three contributions (one from
an MNN transition, one from an MMN transition and one
from an MNV transition). The 150 eV signal arises from
an MNV transition exclusively, and the 175 eV transition
arises exclusively from an MW transition [75]. Thus,
it is reasonable to base relative peak heights on the
92 eV peak which involves only core level electrons and
is relatively insensitive to changes in the valence band.
It is important to notice that the heights of the 150
and 175 eV peaks relative to the 92 eV peak change after
the cleaning process. Foord et al. attribute much of
this variation to contamination by sulfur and chlorine
which would produce AES peaks at 150 and 181 eV,
respectively. Although a similar variation in the AES
peaks is found in this study, XPS demonstrates that both
sulfur and chlorine are absent. Therefore, it is likely
that sulfur or chlorine were not present in the earlier
study. Foord et al. present another explanation later
in their paper in which they claim that the peak heights
are sensitive to changes in the valence band caused by
chemisorption. This explanation is consistent with the
96
present study and with the work of Danielson and will
be discussed in more detail below.
Adsorption of N2, N2O, O2 and CO were performed
by cleaning the sample and bringing the system pressure
up to 5xl0-6 Torr for some period of time (typically
1 to 10 minutes). The sample either was allowed to cool
before gas exposure, allowed to cool during exposure
to the gas, or held at an elevated temperature during
the adsorption process. In general the higher sample
temperatures result in increased amounts of adsorbate
at the surface unless the temperature is so high that
the adsorbed gas diffuses into the bulk more rapidly
resulting in a low surface concentration of adsorbate.
It is observed qualitatively that the room temperature
sticking coefficients are very low. It is estimated
that about 0.1 monolayers of contaminant accumulate on
the surface over a 24 h period at a pressure of 1.0x10^
Torr. This crude observation suggests a sticking
coefficient which is less than 0.01 for 02/ CO, N2 and
N2O. A more rigorous determination of the sticking
coefficients of these gases would be difficult. Since
the gases do not desorb thermally, TPD cannot be used
to determine the amount of gas adsorbed. Similarly,
calibrated uptake measurements would be difficult to
perform because the sticking coefficients are so small.
Pumping and/or outgassing due to the chamber walls, ion
97
gauge, etc., would interfere with the measurements. One
known exception is hydrogen which does thermally desorb.
Lin and Gilbert [78] have measured its sticking coefficient
on zirconium finding that it exhibits a maximum of 6.5xl0-^
near 700K and falls to about 4x10^ below 350K and
above 1000K. These results are similar to the observation
presented here for CO, N2, N2O and O2.
Figure 6-2 shows AES spectra after room temperature
exposure to (a) nitrogen at 5xl0-^ Torr for 5 minutes
and (b) nitrous oxide at 1x10"^ Torr for 2 minutes.
Spectra are not shown for O2 and CO exposures because
CO gives a spectrum with a carbon peak identical to Figure
6-la and a typical oxygen peak as does O2. The influences
of the adsorbates on the zirconium XPS 3d peaks are shown
in Figure 6-3. The clean spectrum is shown in Figure
6-3a for comparison. The spectrum due to oxygen exposure
is shown in Figure 6-3d. It can be deconvoluted into
two spectra due to zirconium oxide and zirconium metal.
The spectrum in Figure 6-4b due to N2 exposure shows
a very small shift of about 0.2 eV with a slight broadening
of the peaks while Figure 6-4c due to N2O exposure shows
the combined effects of both nitrogen and oxygen exposure.
The changes caused by adsorption are even more apparent
in the zirconium 3s region shown in Figure 6-4. The
clean zirconium spectrum is shown in Figure 6-4a for
comparison purposes. Figure 6-4b is due to N2 exposure
DNCE2/DE ARBITRARY UNITS
98
Figure 6-2. AES spectra of state 1 zirconium after room
temperature exposure to (A) nitrogen and (B) nitrous
oxide.
NCE} ARBITRARY UNITS
99
BINDING ENERGY CEV)
Figure 6-3. XPS spectra showing the zirconium 3d peaks
for (A) clean zirconium, (B) N2 exposure, (C) N2O exposure
and (D) oxygen exposure.
NCET} ARBITRARY UNITS
100
Figure 6-4. XPS spectra showing the zirconium 3s peak
for (A) clean zirconium, (B) nitrogen exposure and (C)
N2O exposure. The' nitrogen Is peak appears at 396 eV
in (B) and (C).
101
and Figure 6-4c is due to exposure of clean zirconium
to nitrous oxide. Figure 6-4b shows a splitting of the
zirconium 3s peak into two peaks; one due to metallic
zirconium and a second due to nitride formation which
is shifted by 1.5 eV. Also, a nitrogen Is peak appears
at 396 eV which is characteristic of a nitride [33].
Attempts to produce a nitrogen peak characteristic of
molecular N2 or N2O were unsuccessful at room temperature
or
above.
Adsorption of N2O
(Figure
6-4c) shows a
similar
splitting
in the zirconium
3s peak
as Figure 6
-4b
due
to
nitride
formation and a
similar
nitrogen Is
peak
at
396
eV.
In addition, a
shoulder
has formed
on
the
zirconium 3s peak about 4.5 eV greater in binding energy.
This shoulder is attributed to oxide formation. The
shift is comparable to the shift in the zirconium 3d
peaks due to oxide formation [31]. Figure 6-5 compares
the XPS oxygen Is peak at 530.6 eV due to oxygen adsorption
(Figure 6-5a) with the oxygen Is peak due to N2O adsorption
shown in Figure 6-5b. They are essentially identical
indicating dissociative adsorption of N2O.
The sample temperature only affects the amount of
adsorbate at the surface which depends on both the
adsorption process and the surface-to-bulk diffusion
process. These results indicate that CO, N2, N2O and
O2 adsorb dissociatively on polycrystalline zirconium
at room temperature and above. This observation is in
NCEJ) ARBITRARY UNITS
102
11111111111111111 m jTm 1111111111111111
i 11 1 I 1 1 1 I 1 I 1 I I [ 1 I 1 I I II I 1 I 1 1 1 1 I 1 1 1 I I f I
550 540 530 520 510
BINDING ENERGY CEVO
Figure 6-5. XPS spectra showing the oxygen Is peak after
(A) O2 adsorption on zirconium and (B) N2O adsorption
on zirconium.
103
agreement with the conclusion of Foord et al. but provides
more direct evidence of dissociative adsorption.
Unusual adsorption phenomena is observed after heating
the sample at high temperature for prolonged periods.
An AES spectrum taken after cooling the clean sample
to room temperature is shown in Figure 6-6. This spectrum
results from heating near the melting point for about
3 hours. The largest change occurs in the 17 5 eV peak.
It now has become a very small feature whereas in the
AES spectra presented earlier the 175 eV peak is one
of the more prominent features. When the 17 5 eV peak
appears as in Figure 6-6 (i.e., the two peaks in the
170-185 eV region are of comparable size), the zirconium
will be said to be in "state 2," otherwise zirconium
will be referred to as "state 1."
Both CO and N2 adsorption experiments have been
performed after the high temperature heating period (i.e
on state 2 zirconium). The CO adsorption experiments
are performed in three different ways: (1) exposing the
sample to CO contamination from the electron beam for
long periods, (2) dosing the sample at room temperature
and (3) dosing the sample while it is initially hot and
allowing it to cool. It is observed that exposing a
clean sample to the electron beam causes a very slow
growth of carbon and oxygen peaks probably due to CO
cracking by the hot filament. Figure 6-7 shows an AES
DNCE)/DE ARBITRARY UNITS
104
KINETIC ENERGY CV)
Figure 6-6. AES spectrum for clean zirconium after heating
near the melting temperature for 3 hours. The AES 175
eV peak is greatly diminished which is characteristic
of state 2 zirconium.
105
spectrum of a clean sample after about 8 hours of beam
exposure. The carbon peak appears to contain both carbidic
and graphitic features. The corresponding XPS carbon
Is
peak
is
shown in Figure 6-8.
It
shows
two peaks;
one
at
284
eV
due to graphitic
carbon and
another at
281
eV
due
to
carbidic carbon.
Most
of the
carbon is
in the graphitic form.
CO exposures to clean state 2 zirconium at room
temperature and at high temperature are shown in Figures
6-9a and 6-9b, respectively. The exposure is for 8 minutes
at a pressure of 5x10^ Torr. In the high temperature
exposure, the sample was allowed to cool from above the
transition temperature during the exposure. The two
spectra are very different. At room temperature only
a small amount of CO adsorbs compared to the same CO
exposure for state 1 zirconium. However, much more CO
adsorbs during the high temperature exposure as evidenced
by the size of the carbon peak. The oxygen peak is very
small in Figure 6-9b because oxygen diffuses rapidly
into the bulk at elevated temperature. The corresponding
XPS carbon Is spectra are shown for room temperature
adsorption in Figure 6-10a and for high temperature
adsorption in Figure 6-10b. Both spectra contain peaks
due to graphitic and carbidic carbon. However, the
relative amounts are different. The room temperature
adsorption results in about equal amounts of the two
DNCO/DE ARBITRARY UNITS
106
Figure 6-7. AES spectrum taken after clean state 2 zircon
ium is exposed to CO contamination from the electron
beam for 8 hours. The carbon peak shows characteristics
of both graphitic and carbidic carbon.
NCE> ARBITRARY UNITS
107
Figure 6-8. XPS spectrum of the carbon Is pe§k correspond
ing to the AES spectrum shown in Figure 6-7. Both graphi
tic and carbidic carbon are present.
DNCE^/DE ARBITRARY UNITS
108
KINETIC ENERGY CEV)
Figure 6-9. AES spectra after exposing state 2 zirconiiim
to CO at (A) room temperature and (B) high temperature
but allowing the sample to cool during the exposure.
NCE) ARBITRARY UNITS
109
29S 290 28S 280 27S 270
BINDING ENERGY CEV¡>
Figure 6-10. XPS spectra of the carbon Is peak correspond
ing to the AES spectra shown in Figure 6-9. The room
temperature adsorption produces approximately equal amounts
of graphitic and carbidic carbon as shown in spectrum
(A) "while the high temperature adsorption results in
predominantly carbidic carbon as shown in spectrum (B).
110
types of carbon while the elevated temperature adsorption
results in mostly the carbidic form. These data suggest
that CO adsorbs dissociatively forming an oxide and a
graphitic layer of carbon followed by transformation
of the graphitic carbon into carbidic carbon. The
transformation occurs more rapidly at elevated temperatures
as does diffusion of the oxide and carbidic form of the
carbon into the bulk.
Nitrogen adsorption shows very unusual behavior
for state 2 zirconium compared to state 1 zirconium.
With state 1 zirconium it is easy to adsorb large amounts
of nitrogen at room temperature as evidenced by the large
AES nitrogen peak shown in Figure 6-2a. Figure 6-2a
results from an exposure of 5x10" Torr of nitrogen for
5 minutes. With state 2 zirconium, nitrogen does not
adsorb appreciably at room temperature. This can be
seen in Figure 6-lla which results from a room temperature
exposure to nitrogen at a pressure of 5x10^ Torr for
15 minutes. A small amount of carbon and oxygen
contamination has accumulated during this long exposure
and the subsequent AES run. Figure 6-llb shows a state
2 surface in which the nitrogen adsorption was carried
out on a hot surface which cooled during the exposure.
The exposure was 5xl0-6 Torr of nitrogen for 5 minutes.
More nitrogen adsorbs than during the state 2 room
temperature adsorption, but much more adsorbs during
a state 1 room temperature adsorption.
Ill
IS)
I-
H
z
3
>-
oc
<
O'
t-
H
CD
O'
<
UJ
Q
\
A
LlJ
v
Z
a
Figure 6-11. ('A) AES spectrum taken after exposing state
2 zirconium to nitrogen at 5xl0-^ Torr for 15 minutes
at room temperature. A small amount of carbon and oxygen
contamination accumulated during the long exposure and
subsequent AES run. (B) AES spectrum taken after exposing
state 2 zirconium to nitrogen initially at high temperature
and then allowing the sample to cool during the exposure.
(C) AES spectrum taken after allowing the sample to remain
in vacuum for 3 days at room temperature. State 2 zircon
ium has transformed into state 1 zirconium. (D) AES
spectrum taken after exposing the state 1 zirconium of
spectrum (C) to nitrogen at 5xl0-^ Torr for 5 minutes
at room temperature.
iee 200 300 m 500 600
KINETIC ENERGY CEV)
112
After allowing the state 2 zirconium sample to remain
in vacuum at room temperature for 3 days an unexpected
phenomenon occurs. The AES zirconium 175 eV peak increases
in size as shown in Figure 6-llc. Not only does it
resemble the state 1 17 5 eV peak of earlier scans, but
the sample again adsorbs nitrogen at room temperature
as shown in Figure 6-lld for a 5 minute exposure at 5x10
Torr of N2
Two possible explanations of this time-dependent
chemisorption phenomena are presented here. The first
is that the HCP-BCC transition occurs slowly at least
in the surface region. The change in the atomic positions
appears to cause major changes in the valence electronic
struture as reflected in the AES levels which involve
valence electrons. The fact that nitrogen adsorption
essentially does not occur when the AES zirconium 175
eV peak is absent indicates that the valence electrons
which participate in this AES transition are the valence
electrons which are primarily responsible for bonding
zirconium with nitrogen. An attempt was made to observe
the valence band using XPS, but the cross sections of
the valence electrons for photoemission using x-rays
were to small to yield meaningful information. A similar
statement applies to CO adsorption on zirconium but to
a lesser extent because some CO adsorbs at room temperature
when the AES 175 eV peak is diminished.
113
A second possible explanation is that adsorption
of hydrogen (i.e., the presence of zirconium hydride)
is responsible for the presence of the valence structure
which participates in the AES 175 eV peak. Hydrogen
desorbs from zirconium giving TPD peaks at about 700,
1000 and 1135K [78]. The third peak occurs at the phase
transition temperature and is the largest. This behavior
would explain why the AES 17 5 eV peak diminishes after
heating above the transition temperature and why the
175 eV peak grew back after 3 days of exposure to
background hydrogen in the system. Although the presence
of hydrogen generally cannot be detected with XPS or
AES, it should have been possible to detect the desorbing
hydrogen. However, none was observed. This explanation
also suggests that CO
and N2 adsorb
more
rapidly
on a
hydrided surface
than
on clean zirconium metal
which
seems unlikely.
Also,
heating state
1
zirconium
near
the melting point
for
short periods
of
time (about 1
minute) does not convert it to state 2 even though most
hydrogen initially present would be desorbed. Therefore,
the second possibility does not seem very probable but
will not be ruled out completely without further tests.
As stated previously, Foord et al. report that the
17 5 eV peak height is the most sensitive AES peak to
adsorption while Danielsn states that the 150 eV peak
height is the most sensitive and that the 17 5 eV peak
114
height is relatively insensitive to adsorption. This
present study actually agrees with both Foord et al.
and Danielson by showing the 175 eV peak to be the most
sensitive peak if the sample has not been heated above
the HCP-BCC transition temperature for prolonged periods
while the same 175 eV peak is relatively insensitive
to adsorption after extensive heating above the transition
temperature. This is consistent with the earlier papers
in that Foord et al. argon ion sputtered and typically
heated below the transition temperature while Danielson
routinely heated above the transition temperature at
1200 to 1800K during the adsorption experiments. In
essence Foord et al. studied state 1 zirconium while
Danielson studied state 2 zirconium. Danielson also
used very large carbon monoxide exposures of 1x10^ Torr
for 400 minutes at 1200K for adsorption which overcame
the low sticking coefficient problem for state 2
adsorption. The results of this present study indicate
that the adsorption properties of zirconium and the AES
relative peak heights depend upon the previous history
of the sample. These obervations imply that it is not
particularly useful to plot AES relative peak heights
of. zirconium as a function of adsorbate concentration
unless the zirconium sample is in a well-characterized
state.
115
Conclusions
Zirconium is reactive toward CO, O2, N2 and N2O,
but the room temperature sticking coefficient of these
gases on zirconium is low (C0.01). This low value is
probably due to the fact that the adsorption is
dissociative and requires activation energy to proceed.
No evidence has been found which indicates that these
molecules bond molecularly to zirconium at room temperature
or above although they may bond molecularly below room
temperature.
A clean zirconium surface can be in two different
"states" as reflected by the chemisorption properties
and the electronic structure given by AES. State 1 occurs
when the sample has not been heated for prolonged periods
(many hours depending upon the temperature) above the
transition temperature, and state 2 occurs after prolonged
heating above the transition temperature. State 1 is
characterized by a large AES zirconium 17 5 eV peak while
this peak is very small for state 2. This peak is due
to an MW transition thus indicating that the valence
electronic strutures of the two states are different.
State 1 adsorbs CO and N2 much more readily than state
2. It can be concluded that the valence electrons
responsible for the AES 175 eV peak also participate
in the adsorption process. The transformation between
state 1 and state 2 is reversible. The rate of
116
transformation of state 1 to state 2 is much faster than
for the state 2-to-state 1 transformation because it
occurs at high temperature. This behavior is also
characteristic of a phase change. The bulk phase
transformation from HCP to BCC occurs rapidly [78], but
apparently the transformation in the surface region occurs
slowly.
SECTION VII
GENERAL CONCLUSIONS AND
RECOMMENDATIONS FOR FUTURE RESEARCH
Pt Tin Oxide
A feature has been identified in the Pt 4f XPS
spectrum associated with the bond formed between supported
platinum and the tin oxide substrate. The bond is believed
to form with surface lattice oxygen resulting in a Pt-O-Sn
surface species. This substrate-bonded species appears
to act as a nucleation site for crystallite growth in
both the electrochemical deposition of platinum and in
the sintering of supported platinum.
Future attempts to characterize the metal-substrate
interaction may be aided by the study of a more
geometrically ideal system. Low-energy electron
diffraction (LEED) studies of the SnC>2 (110) face have
shown that several specific types of oxygen-deficient
defect structures may be obtained by annealing at different
temperatures [58]. By vapor depositing Pt in. situ on
a tin oxide single crystal the interaction with specific
defect surfaces and tin atoms of varying coordination
could be studied.
It has been demonstrated that the oxidation state
of supported platinum may be manipulated in. situ and
118
characterized by XPS. This observation should prove
useful in future molecular beam studies relating catalytic
activity to surface characterization. It is recommended
that the first such attempts should be directed at tin
oxide surfaces with high platinum loadings. High loadings
will benefit the XPS analysis by increasing the
signal-to-noise ratio and should benefit the molecular
beam studies by providing more reaction products.
An acceptable technique for characterizing tin oxide
has been developed. It has been demonstrated that ELS
is useful in characterizing both the surface and subsurface
regions of the tin oxide support. The interpretation
of the ELS data presented here also provides a basis
for the interpretation of ESD threshold measurements
in the study of chemisorption on tin oxide surfaces.
It is felt that the full potential of the ELS
technique in characterizing tin oxide has not yet been
realized. It is recommended that future development
of this technique proceed though the study of derivative
loss spectra, i.e. N(E)' or N(E)". A significant amount
of detail is believed to be lost in the background of
N(E) measurements such as those presented here. Also,
it may be possible to obtain a more complete understanding
of the atomic nature of the conduction bands by studying
core level losses associated with 0 Is, Sn 3d and Sn
3p levels using dipole selection rules.
119
It has been demonstrated that the experimental tools
developed for ESD measurements are in place and function
properly. Preliminary data has been presented which
suggests that ESD will prove very useful in studying
the interaction of water (and other adsorbates) with
tin oxide. Further work in this area should attempt
to apply the full capabilities of the ESD technique to
the identification of adsorbed species and surface binding
sites. It is felt that combining ESD and UPS measurements
would be a powerful approach for studying adsorption
(particularly of water) on tin oxide surfaces.
Zirconium
The chemisorption properties of polycrystalline
zirconium have been found to vary dramatically depending
on the thermal history of the sample. Chemisorption
on this surface is found to be suppressed by heating
for prolonged periods of time above the HCP-to-BCC phase
transition temperature at 1135K. The chemisorption
behavior can be correlated roughly with the appearance
or disappearance of a zirconium MW Auger peak. A slow
phase transition at the surface has been postulated as
the cause of the variation in chemisorption properties.
It is felt that future work on zirconium may benefit
from ESD measurements. 'Variations in surface geometry
or in the types of adsorption sites available on "state
120
1" and "state 2" zirconium may be detectable in ESD as
changes in the desorbing ion kinetic energy distribution.
APPENDIX A
A BRIEF DESCRIPTION OF
THE EXPERIMENTAL TECHNIQUES
X-Ray Photoelectron Spectroscopy (XPS)
Surface analysis by XPS involves irradiating a sample
with nearly monoenergetic soft x-rays and energy analyzing
the emitted electrons. The photons, in this case Mg
Ka x-rays at 1253.6 eV, interact with atoms in the solid
causing electrons to be emitted by the photoelectric
effect. The photoemission process is illustrated in
Figure A-l. The emitted electrons have kinetic energies
given by
KE = hv BE s
where h v is the photon energy, BE is the binding energy
of the orbital from which the electron originates and
s is the spectrometer work function. By analyzing the
kinetic energy distribution of the emitted electrons
a spectrum is obtained which corresponds (roughly) to
the number denisty of electrons per binding energy
interval. In general, electrons from narrow, well-defined
energy states (i.e. core levels) are those of interest.
121
122
VALENCE
LEVELS
CORE
LEVELS
Figure A-l.
Photoemission process.
123
The binding energy of a given peak in the spectrum
may be regarded as the ionization energy of that particular
shell (orbital) in the emitting atom. Because each element
has a unique elemental spectrum, the observed peaks from
a mixture are approximately the sum of the elemental
peaks of the constituents. However, the binding energy
(ionization energy) of a given orbital in an emitting
atom may be effected by the chemical environment. Changes
in the valence electronic structure of an atom due to
chemical bonding may be reflected as a "chemical shift"
in the measured binding energies of the core level
electrons. Core level binding energies may therefore
be used as an indicator of the "valence (oxidation) state"
of an atom. For example, the reported binding energies
of iron 2p 3/2 peaks for the metal and oxides are 706.8
eV, 709.3 eV and 711.0 eV for Fe, FeO and Fe2C>3,
respectively [31].
The surface sensitivity of the XPS measurement is
determined by the mean free path of the emitted electrons.
The mean free path is (roughly) the average distance
an electron is expected to travel in a solid without
undergoing an inelastic collision with other electrons
(see Section V for a more accurate definition). The
mean free path generally decreases with decreasing electron
kinetic energy. Emitted electrons undergoing inelastic
collisions lose some fraction of their kinetic energy
and appear in the spectrum at an apparent binding energy
124
greater than that of the parent photoelectron peak. These
"loss" features generally show up as a broad background
from which the elastically emitted electrons are easily
distinguished. The mean free path of electrons emitted
, O
in XPS generally ranges from 5 to 20 A. This short mean
free path results in a parent photoelectron line which
originates from atoms in the "top few" atomic layers
of the solid.
Auger Electron Spectroscopy (AES)
Surface analysis by AES is based on the radiationless
decay process discovered by P. Auger in 1925. Figure
A-2 illustrates the process. The process is initiated
by the creation of a core hole (i.e. ejection of a core
level electron) typically excited by an impinging electron
beam. The creation of the core hole leaves an ion in
an excited state. The atom subsequently decays to a
doubly-ionized lower energy state when an electron from
a higher energy level drops into the core hole and
simultaneously releases its energy to another (Auger)
electron which is emitted from the atom. The energy
given up in the transition from the singly to doubly
ionized state is absorbed by the Auger electron and
determines its kinetic energy. As in XPS, an energy
analysis of the emitted electrons is performed to determine
the positions (kinetic energies) of the Auger electron
peaks. Because each atom gives a characteristic spectrum,
125
O
Lj C2S)
K CIS)
Figure A-2, KL]_L2 Auger decay process.
126
the kinetic energies of the Auger electrons can be used
to identify the composition of the solid surface.
The core hole which is created to initiate the process
is like that which is left after the ionization of a
core level in XPS. Indeed, Auger peaks may be observed
in XPS spectra as a result of the core-hole (Auger) decay
process. Unlike the photoelectron in XPS however, the
kinetic energy of the Auger electron is set by the decay
process and is independent of the energy of the ionizing
radiation. As in XPS, the surface sensitivity of AES
is set by the kinetic energies (short mean free paths)
of the emitted electrons.
Electron Energy-Loss Spectroscopy (ELS)
The ELS measurements presented in Section III and
IV are for energy losses of the order of electronic
transitions (see Section III for a more complete
discussion). In ELS an impinging (primary) beam of
electrons at energy Ep strikes the solid and excites
various electronic processes such as valence and core
level ionizations, Auger processes and plasmon excitations.
The kinetic energy distribution of backscattered electrons
near the primary beam energy is measured. The primary
electrons which strike the surface and are backscattered
to the detector may interact either elastically (i.e.
with no loss of energy) or inelastically giving up some
fraction of their energy to the "loss" processes. Energy
127
UNOCCUPIED
LEVEL
OCCUPIED
LEVEL
Ep -AE
Figure A-3. Electron energy-loss process.
128
differences between the elastically and inelastically
scattered electrons are reported as energy losses and
(hopefully) related to the characteristic processes that
occur within the solid. An idealized illustration is
given in Figure A-3. The power of this technique is
that it probes not only the filled electronic levels
of the solid (like AES and XPS) but also the higher lying
unoccupied levels. The result is an often complex spectrum
which is a convoluted picture of both the filled and
unfilled states.
The ELS measurement probes the inelastic processes
which are generally responsible for the surface sensitivity
of XPS and AES. The analysis depth of the technique
is mainly dependent on the primary beam energy. From
mean free path considerations it can be seen that the
higher the incident beam energy the deeper the analysis
region within the sample.
Electron-Stimulated Desorption (ESP)
When an electron beam is directed at a sample it
can result in the desorption of cations, anions and neutral
species. For the results presented here a mass analysis
of desorbing positive ions is performed (see Section
V). The ESD technique possesses an extreme surface
sensitivity because of the very short mean free path
of low energy cations in a solid. The high probability
of reneutralization results in the escape and detection
129
of cations from the outer layer of the material only.
Because the ESD technique involves a mass analysis of
desorbing species it is directly sensitive to surface
hydrogen unlike XPS, AES and ELS. This makes ESD a
particularly useful technique for studying H2 and H2O
adsorption.
The ESD of positive ions from metal oxide surfaces
is generally thought to occur as the result of an Auger
decay process [30], This model proposed by Knotek and
Feibelman (KF) has been successful in correlating large
amounts of experimental data for metal oxide systems.
In the KF model the process is begun (as in AES) by the
creation of a core hole. During the subsequent Auger
decay process the resulting positive ion may desorb if
it has a kinetic energy in excess of the surface binding
energy.
While normal (intra-atomic) Auger processes may
cause desorption, significantly more information may
be obtained from inter-atomic processes. In the
inter-atomic decay mechanism the core hole created on
a particular atom is filled by a decaying electron from
a nearest neighbor atom. The ionized nearest neighbor
then desorbs and is detected. By correlating the
desorption of a particular species with the threshold
ionization energies of surface atom core levels the ESD
measurement becomes a specific probe of the surface binding
site (i.e. nearest neighbor) of the desorbing species.
130
To date, only one conflicting example has been found
where second nearest neighbor desorption results from
a core level ionization process in an oxide [79].
APPENDIX B
COMPUTER-INTERFACED
DIGITAL PULSE COUNTING CIRCUIT
Introduction
Pulse counting is an important means of detecting
signals when they are particularly small as in x-ray
photoelectron spectroscopy (XPS), ultraviolet photoemission
spectroscopy (UPS), electron-stimulated desorption (ESD),
electron energy-loss spectroscopy (ELS) and numerous
others. Pulse counting can be performed using a
commercially available analog ratemeter, but it is more
convenient to use a computer-interfaced, digital pulse
counter because the timing of the data collection process
can be controlled precisely thereby allowing the use
of an on-off heater circuit [80], multiple scans can
be added together readily and the original data is stored
easily so that it can be recalled or digitally filtered
[81], as necessary. A digital pulse counter which
interfaces to a laboratory computer through a general
purpose 16-bit parallel interface is described here.
To start a count the computer sends a control word
containing the desired counting time to the parallel
interface. Pulses are sent to the counter after
131
132
amplification and discrimination. When counting is
finished the device sends a signal back to the parallel
interface. The computer can either wait for this signal
or use it as an interrupt. The pulse count is read then
by the computer as a 16-bit word.
Cricuit Description
The counter consists of the following four parts:
(1) an onboard timer circuit, (2) a timing counter, (3)
an event counter, and (4) a control logic section. The
on-board timer (Figure B-l) is a 1.0000-MHz crystal
followed by four stages of divide-by-ten logic (Type
74390). Thus base frequencies of 100, 10, and 1 KHz
and 100 Hz are available for timing. These frequencies
are jumper selectable. Since the device counts pulses
over any period from 1 to 65,535 (177777 octal) clock
cycles, timing intervals are available from 10 psec to
655 sec. Of course, all of the above numbers may be
changed easily to suit the user, and 1.0 MHz is available
directly off the clock. Higher-frequency clocks can
be used and may be necessary if higher time resolution
is required as in the time-of-flight modification discussed
later in this appendix. However, the clock frequency
must not exceed the response time of the integrated
circuits. It should be .noted that the use of a higher
frequency clock will alter the base frequencies and timing
intervals accordingly.
133
74390
74390
SYS
CLK
= Ref. o r- o
Figure B-l. On-board timer schematic showing jumper
selectable system clock rates.
100 KHz
134
Figure B-2 shows the control logic section as used
with a Heath Model H-ll-2 parallel interface module and
the LSI 11 (Digital Equipment Corporation) computer.
When the command word is sent from the computer to the
parallel
interface
the
latter
generates a
"take
data"
signal (TD). When
this
signal
goes low it
clocks
a HI
into the
first of
three D-type
flip-flops
(Type
7474).
The next
pulse from
the
on-board
timer clocks
this
signal
synchronously into the second flip-flop causing Q2 to
go LO. Q2 clears the first flip-flop and thereby restores
itself to the HI state at the next clock pulse. Thus
Q2 is a single HI-LO-HI pulse of width equal to one timing
cycle. It is used to load the timing counter, to clear
the event counter, to start the counting and (if necessary)
to report back to the parallel interface that the command
word has been received.
Counting is started by simultaneously enabling the
event counter and gating timing pulses into the timing
counter. This is done through the third flip-flop which
is clocked HI by Q2.
Counting is stopped in two ways. When the timing
counter reaches zero, its final borrow signal goes LO.
When the event counter fills (overflow condition), its
final carry signal goes HI. The carry signal is inverted
and then ANDed with the borrow signal so that when either
event occurs the third flip-flop is cleared, a "data
135
7474 Vi 7474
Figure B-2. Schematic of the control logic section
136
sent" (DS) signal is returned to the parallel interface
and counting stops. Computer software can check the
count for the value 177777 (octal) which would indicate
an overflow condition.
Figure B-3 shows the timing and event counters.
The former is a series of four synchronous 4-bit binary
up-down counters (Type 74193) wired to count down. An
initial value is applied to this counter through the
C inputs and loaded by the pulse at Q2 (see Figure B-2).
This value is the command word sent from the computer
and is available at the parallel interface when the TD
signal goes LO. Timing pulses are gated into the least
significant bit (LSB) of this counter. The borrow signal
from the most significant bit (MSB) goes LO when the
counter has reached zero.
The event counter is a series of four asynchronous
4-bit binary upcounters (Type 74161). These counters
are cleared when the pulse at Q2 goes LO, and counting
is enabled when that pulse returns to HI. Should the
counter saturate, the carry from the MSB will go HI and
stop the process.
Figure B-4 shows the actual wiring diagram. The
integrated circuits (ICs) as
laid
out will
fit
on a
4"
x 8"
printed circuit board.
The
7 400 and
7408
ICs
are
NAND
and AND gates. There
are
37 connections
to
the
parallel interface (C0-C15,
D0-D15, TD, DT
, DS
, +5
V,
137
C15C14C13C12
C11 C10C9C8
C7 CSCS C4
C3 C2C1 CO
C = INITIAL VALUE FOR TIMING COUNTDOWN
(A) TIMING COUNTER
D15D14D13D12
D11 D10 09D8
D7D6D5D4
D3 D2 D1 DO
(B) EVENT COUNTER
Figure B-3. (A) Timing and (B) event counter schematics.
138
LOAO (L)
Figure B-4. Layout and wiring
(pull up) resistors are 10K ohms.
diagram.
All unmarked
139
and REF.) and one EVENT connection. This last connection
should be shielded properly to avoid counting spurious
pulses.
Time-of-Flight Modification
This circuit may be modified easily to perform another
important class of experiments such as electron-stimulated
desorption (ESD) in which an electrostatic analyzer can
be used as a time-of-flight (TOF) mass spectrometer [73].
In this type of experiment an initial signal causes ions
to desorb off a sample. The different masses require
different flight times to reach the counting circuitry
so a delay is required before counting for a set period
of time.
In the circuit as described in the previous section,
both the timing counter and event counter are started
as a result of the TD signal. For TOF measurements the
TD signal is activated simultaneously with the ionizing
event at the surface at time zero. The "TD signal is
used to start the timing counter as before but not the
event counter. The delay period has been loaded previously
into the timing counter so that after the timing counter
has counted down its final borrow signal is used to enable
the event counter which counts for a selected period
of time.
A brief description of the ESD TOF circuit
modification is given here. The output of the third
140
flip-flop, Q3 (Figure B-2) is used as a trigger for
the Bl input of a 74LS123 one-shot. A potentiometer
is connected across the external timing circuit of the
one-shot to allow the output pulse width to be varied.
The one-shot pulse (typically 300 nsec) activates a MRF427A
power transistor which is used to switch 50 V off a
deflection plate in order to pulse an electron beam onto
the sample. At the end of the delay period, the final
borrow from the timing counter triggers the A2 input
of the 74LS123 one-shot. The output pulse of this one-shot
is used to enable the event counter, and pulses are
collected for a
time
period equal
to the
pulse
width
of the electron
beam.
The negative
edge of
the
output
of this one-shot
is
used to trigger another one-shot
which clears the
third
flip-flop and
sends a
"data
sent"
(DS) signal to the parallel board. A time resolution
of 0.1 psec can be obtained if a 10-MHz clock is used
with no divide-by-ten circuitry.
Acknowledgments
The original version of the device described here
was conceived by G.B. Hoflund. The first working model
was designed and constructed by R.E. Gilbert with helpful
discussions from Sonny Johnson. A version incorporating
the TOF modification was conceived by G.B. Hoflund and
D.F. Cox using ideas contributed by Mort Traum. This
141
second version was designed and constructed by D.F. Cox
with helpful discussions from R.E. Gilbert.
REFERENCES
1.
M.M.P. Janssen
21(1976)861.
and J.
Moolhuysen,
Electrochim.
Acta
2.
M.M.P.
Janssen
and J.
Moolhuysen,
Electrochim.
Acta
21(1976)869.
3. M.M.P. Janssen and J. Moolhuysen, J. Catal.
46(1977)289.
4. M.R. Andrew, J.S. Drury, B.D. McNicol, C. Pinnington
and R.T. Short, J. Appl. Electrochem. 6(1976)99.
5. A. Katayama, J. Phys. Chem. 84(1980)376.
6. M. Watanabe, S. Venkatesen and H.A. Laitinen, J.
Electrochem. Soc. 130(1983)59.
7. A.C.C. Tseung and S.C. Dhara, Electrochim. Acta
19(1974)845.
8. G.B. Hoflund, D.A. Asbury and R.E. Gilbert, to be
published.
9. R.R. Cecil, W.S. Kmak, J.H. Sinfelt and L.W. Chambers,
Oil Gas J. 7(1972)50.
10. E.M. Blue, Hydrocarbon Processes 48(1969)141.
11. M.J. Fuller and M.E. Warwick, J. Catal. 29(1973)441.
12. B. Hori, N. Takezawa and H. Kobayashi, J. Catal.
80(1983)437.
13. F. Solymosi and J. Kiss, J. Catal. 41(1976)202.
14. N. Niwa, T. Minami, H. Kodama, T. Hattori and Y.
Murakami, J. Catal. 53(1978)198.
15. K. Wakabayashi, Y. Kamiya and N. Ohta, Bull. Chem.
Soc. Japan 40(1967)2172.
142
and R. Reid, J. Catal.
143
16. H.H. Herniman, D.R. Pyke
58(1979)68.
17. G.B. Hoflund, D.F. Cox, G.L. Woodson and H.A.
Laitinen, Thin Solid Films 78(1981)357.
18. G.B. Hoflund, D.F. Cox and H.A. Laitinen, Thin Solid
Films 83(1981)261.
19. G.B. Hoflund, D.F. Cox, F. Ohuchi, P.H. Holloway
and H.A. Laitinen, Appl. Surface Sci. 14(1982-83)281.
20. C.E. Smith, J.P. Biberian and G.A. Somorjai, J.
Catal. 57(1979)426.
21. R. Bacaud, P. Bussiere and F. Figueras, J. Catal.
69(1981)399.
22. Z. Zarpinsky and J.K.A. Clarke, J. Chem. Soc. Faraday
Trans. I 71(1975)893.
23. H. Verbeek and W.M.H. Sachtler, J. Catal. 46(1976)257.
24. N. A. Pakhomov, R.A. Buyanov, E.M. Moroz, E.N.
Yurchenko, A.P. Chernyshey, N.A. Zaitseva and G.R.
Kotelnikov, React. Kinet. Catal. Lett. 14(1980)329.
25.
A.
Katayama, Chem. Lett. (1978)1263.
26.
D. Elliot, D.L. Zellmer and
Electrochem. Soc. 117(1970)1343.
H.A.
Laitinen, J.
27.
Yu.
Kul
E. Roginskaya, D.A. Dulin,
'kova and A.I. Gel'bshtein,
S.S. Stroeva, N.V.
Kinetika i Kataliz
9(1968)1143.
28. H.A. Laitinen and T.M. Hseu, Anal. Chem. 51(1979)1550.
29. M. Yokoyama and H.A. Laitinen, unpublished results.
30. M.L. Knotek and P.J. Feibelman, Phys. Rev. Letters
40(1978)964.
31. C.D. Wagner, W.M. Riggs, L.E. Davis, J.F. Moulder
and G.E. Muilenburg, Handbook of X-Ray Photoelectron
Spectroscopy, Perkin-Elmer, 1979.
32. C.L. Lau and G.K. Wertheim, J. Vac. Sci. Technol.
15(1978)622.
33. T.W. Capehart and S.C. Chang, J. Vac.
18(1981)393.
Sci.
Technol.
144
34. W.E. Morgan and J.R. VanWazer, J. Phys. Chem.
77(1973)964.
35. J. Rosenfarb, H. A. Laitinen, J.T. Sanders and H.A.
VanRinsvelt, Anal. Chim. Acta 108(1979)119.
36. M.G. Mason, Phys. Rev. B 27(1983)748.
37. P.H. Citrin and G.K. Wertheim, Phys. Rev. B
27(1983)3176.
38. R. Bouwman and P. Biloen, Anal. Chem. 46(1974)136.
39. M. Komiyama, Y. Ogino, Y. Akai and M. Goto, J. Chem.
Soc. Faraday Trans. 79(1983)1719.
40. D.F. Cox, G.B. Hoflund, W.H. Hocking and H.A.
Laitinen, to be published.
41. R.A. Powell, Appl. Surface Sci. 2(1979)397.
42. R.A. Powell and W.E. Spicer, Surface Sci. 55(1976)681.
43. P.A. Cox, R.G. Egdell, C. Harding, A.F. Orchard,
W.R. Patterson and P.J. Tavener, Solid State Commun.
44(1982)837.
44. J. Robertson, J. Phys. C: Solid State Phys.
12(1979)4767.
45. M.L. Knotek, Surface Sci. 101(1980)334.
46. M.L. Knotek, V.O. Jones and V. Rehn, Phys. Rev.
Letters 43(1979)300.
47. P.L. Gobby, Ph.D. Dissertation (Montana State
University, May 1977), University Microfilms
International, Ann Arbor.
48. E. Rudberg and J.C. Slater, Phys. Rev. 50(1936)150.
49. J.E. Rowe, Solid State Commun. 15(1974)1505.
50. R. Ludeke and A. Koma, Phys. Rev. Letters 34(1975)817.
51. E. Colavita, M. DeCrescenzi, L. Papagno, R.
Scarmozzino, L.S. Caputi, R. Rosei and E. Tosatti,
Phys. Rev. B 25(1982)2490.
52. A. Messiah, Quantum' Mechanics, Vol. II, John Wiley
& Sons, New York.
145
53.R. Ludeke and L. Esaki, Phys. Rev. Letters
33(1974)653.
54. W. Spence, J. Appl. Phys. 38(1967)3767.
55. E.W. Giesekke, H.S. Gutowsky, P. Kirkov and H.A.
Laitinen, Inorganic Chem. 6(1967)1294.
56.
R. Summit, J.A. Marley and N.F.
Chem. Solids 25(1964)1465.
Borelli, J.
Phys.
57.
R.D. McRoberts, C.G. Fonstad and
Rev. B 10(1974)5213.
D.
Hubert,
Phys.
00
LO
E. deFre^art, J. Darville and J
.M.
Giles,
Solid
State Commun. 37(1980)13.
59. N. Yamazoe, J. Fuchigami, M. Kishikawa and T. Seiyama,
Surface Sci. 86(1979)335.
60. S. Munnix and M. Schmeits, Phys. Rev. B 27(1983)7624.
61. A.J. Bevolo, J.D. Verhoeven and M. Noack, J. Vac.
Sci. Technol. 20(1982)943.
62. P. Bayat-Mokhtari, S.M. Barlow and T.E. Gallon,
Surface Sci. 83(1979)131.
63. H.A. Bethe and E.E. Salpeter, Quantum Mechanics
of One- and Two-Electron Atoms, Springer-Verlag,
Berlin, 1957.
64. P.W. Palmberg, J. Vac. Sci. Technol. 12(1975)379.
65. T. Kaji, H. Yamazoe and T. Seiyama, Nippon Kagaku
Zassi 7(1980)1088.
66. H.A. Laitinen and N.H. Watkins, Anal. Chem.
47(1975)1352.
67. H.A. Laitinen and N.H. Watkins, J. Electrochem.
Soc. 123(1976)804.
68. H. Yoneyama and H.A. Laitinen, J. Electroanal. Chem.
Interfacial Electrochem. 75(1977)647.
69. H. Yoneyama and H.A. Laitinen, J. Electroanal. Chem.
Interfacial Electrochem. 79(1977)129.
70. See for example: R.L. Kurtz and V.E. Henrich, Phys.
Rev. B 26(1982)6682, for results on T2O3; V.E.
Henrich, G. Dresselhaus and H.J. Zieger, Solid State
Commun. 24(1977)623, for results on Ti02-
146
71. D. Menzel, in: Photoemission and the Electronic
Properties of Surfaces, Eds. B. Feuerbacher, B.
Fitton and R.F. Willis, John Wiley & Sons, New York,
1978.
72. J.A. Knapp, G.J. Lapeyre, N.V. Smith and M.M. Traum,
Rev. Sci. Instrum. 53(1982)781.
73. M.M. Traum and D.P. Woodruff, J. Vac. Sci. Technol.
17(1980)1202.
74. L.E. Davis, N.C. McDonald, P.W. Palmberg, G. Riach
and R. Weber, Handbook of Auger Electron Spectroscopy,
Perkin-Elmer, 1978.
75. J.S. Foord, P.J. Goddard and R.M. Lambert, Surface
Sci. 94(1980)339.
76. L.R. Danielson, J. Vac. Sci. Technol. 20(1982)86.
77. H. Poppa and F. Soria, Phys. Rev. B 27(1983)5166.
78. J.-M. Lin and R.E. Gilbert, to appear in Appl. Surface
Sci.
79. Y.X. Wang, F. Ohuchi and P.H. Holloway, J. Vac.
Sci. Technol. A 2(1984)732.
80. G.B. Hoflund and R.E. Gilbert, J. Vac. Sci. Technol.
18(1981)56.
81. A. Savitzky and M.J.E. Golay, Anal. Chem.
36(1964)1627.
BIOGRAPHICAL SKETCH
David Fullen
Cox
was born in
Marietta,
Georgia,
July 4, 1956.
He
is the son of
Marjorie
and Loyd
and has one
older
sister, Nancy.
David was
; married
to the former Teresa Louise Ludovici on August 11, 1984.
They have no children.
David was raised in Marietta, Georgia, a town in
the northern metropolitan Atlanta area, where he attended
school and graduated from Marietta High School in 1974.
He entered the University of Tennessee in the Fall of
1974 and participated in the Engineering Cooperative
Education Program before receiving his Bachelor of Science
in Chemical Engineering in 1979. David entered graduate
school in chemical engineering at the University of Florida
in the Fall of 1979 and received his Master of Science
degree in 1980. He continued in the Ph.D. program at
the University of Florida where he has been until the
present.
147
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Gar B. Hoflnd, Chariman
Associate Professor of Chemical
Engineering
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
j pCLJ J~f
Paul H. Holloway
Professor of Materials Science
and Engineering
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
t.-.-. iPTt
Herbert A. Laitinen
Graduate Research Professor
of Chemistry
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
U
Joh^' P. O'Connell
Chairman and Professor of
Chemical Engineering
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Dinesh 0. Shah
Professor of Chemical
Engineering
This dissertation was submitted to the Graduate
Faculty of the College of Engineering and to the Graduate
School, and was accepted as partial fulfillment of the
requirements for the Degree of Doctor of Philosophy.
/ LsJlu/ Ct
Dean, College of Engineering
Dean for Graduate Studies
and Research
December, 1984
UFDissertations, 08:38 AM 6/10/2008, UF Libraries:Digital Dissertation Project
Internet Distribution Consent Agreement
In reference to the following dissertation:
AUTHOR:
Cox, David
TITLE:
Surface characterization and chemisorption properties of polycrystalline systems : Sn02 Pt/Sn02 and Zr /
(record number: 496891)
PUBLICATION DATE:
1984
, as copyright holder for the aforementioned
dissertation, hereby grant specific and limited archive and distribution rights to the Board of Trustees of
the University of Florida and its agents. I authorize the University of Florida to digitize and distribute the
dissertation described above for nonprofit, educational purposes via the Internet or successive
technologies.
This is a non-exclusive grant of permissions for specific off-line and on-line uses for an indefinite term.
Off-line uses shall be limited to those specifically allowed by "Fair Use" as prescribed by the terms of
United States copyright legislation (cf, Title 17, U.S. Code) as well as to the maintenance and preservation
of a digital archive copy. Digitization allows the University of Florida to generate image- and text-based
versions as appropriate and to provide and enhance access using search software.
This gi ant of permissions prohibits use of the digitized versions for commercial use or profit.
Signature of Copyright Holder
7c5
Printed or Typed Name of Copyright Holder/Licensee
Personal information blurred
>//ofe R
Date of Signature
Printed for "David F. Cox"
2
|
