Characterization and catalytic activity of dispersed transition metal oxide particles on carbon substrates


Material Information

Characterization and catalytic activity of dispersed transition metal oxide particles on carbon substrates
Physical Description:
vi, 139 leaves : ill. ; 28 cm.
Davis, Jack G., 1950-
Publication Date:


Subjects / Keywords:
Photoelectron spectroscopy   ( lcsh )
X-ray spectroscopy   ( lcsh )
Manganese oxides   ( lcsh )
Iron oxides   ( lcsh )
Chemistry thesis Ph. D
Dissertations, Academic -- Chemistry -- UF
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1988.
Includes bibliographical references.
Additional Physical Form:
Also available online.
Statement of Responsibility:
by Jack G. Davis, Jr.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 024929552
oclc - 20117437
System ID:

Full Text







fl -'r LF'


I would like to express my appreciation to Dr. Vaneica Young for
her guidance and very helpful suggestions during the course of my
research project. I would also like to thank members of my
research committee who were also of help.
Paul McCaslin, Linda Volk, Mike Clay, and Peter ten Berge all
made my stay at the University of Florida much more enjoyable by
their friendship and advice.


ACKNOWLEDGMENTS .......................................... ii
ABSTRACT ................................................... v
1 INTRO DUCTION ..................................... 1
Fundamental Principles of X-Ray Photoelectron
Spectroscopy ................................... 1
Instrum entation .................................. 3

Overview of Molecular Orbital Theory .............. 17
Approximations to Molecular Orbital Theory ......... 31
VALENCE BAND DENSITY OF STATES .................. 38
Background ................................... 38
Modification of CNDO2/U Algorithm ............... 43
Construction of VBDOS curves ................... 50

Introduction .................................. 57
Variable Angle XPS (VAXPS) .................... 59
Quantitative Analysis by XPS .................... 61
Preparation of Samples ........................ 64
R results ....................................... 70


Introduction .................................. 104
Kinetics of Reactions .......................... 106
Experim ental ................................. 110
R results ...................................... 118
Correlation of Rate Law Expressions
with Experimental Data ......................... 125
6 CONCLUSIONS AND FUTURE WORK................... 129
REFERENCES .............................................. 135
BIOGRAPHICAL SKETCH ...................................... 139

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




December 1988

Chairperson: Vaneica Y. Young
Major Department: Chemistry
After modification of a CNDO2/U algorithm, a valence band
density of states curve (VBDOS) is constructed from the resulting
eigenvectors and eigenvalues for a distribution of MnO dimer
structures. The resulting curve shows a remarkable similarity to the
X-ray photoelectron spectroscopy (XPS) spectrum of a thick
continuous film of MnO.
In order to investigate the properties of small particles using
XPS, they must be isolated on supports. Different methods of
preparing dispersed oxides on carbon foil supports are investigated.
In particular, vapor and solution deposition are used to fabricate
various transition metal oxides on carbon foil. Vapor deposition is
used to disperse MnO and Mn203 on the carbon substrate and solution
deposition is used to disperse MnO2 and Fe203 on the substrate. Data

acquired by XPS from the samples of both methods indicate that two
completely different surfaces result. Vapor deposition yields a
surface in which the particles are confined to the surface. Solution
deposition yields a surface whereby the particles have been
incorporated into the surface. The electronic structure of these
particles are investigated using valence band photoemission.
When the dispersed Mn oxides on carbon foil are placed in
permanganate solution, it is found that the dispersed MnO has a
greater effect on the autocatalytic decomposition of permanganate
than does dispersed Mn203. It is also found that dispersed Mn203 has
a greater effect on this reaction than either the bulk or continuous
film of the Mn(lll) oxide. Dispersed MnO2 on carbon foil are judged to
have too high an effect on the reaction, possibly because of the
particles which are incorporated into the surface.


Fundamental Principles of X-Ray Photoelectron Spectroscopy

The analytical methodology of X-Ray Photoelectron Spectroscopy
(XPS) is an effective means for the chemical analysis of surfaces. In
XPS, a solid is bombarded with X-Ray radiation, typically of energies
of either 1487 or 1254 eV. Photoemission of electrons from the
surface region of the solid occurs. The ejected photoelectrons are
dispersed according to their kinetic energy (if an electrostatic
analyzer is used) and then counted. The intensity of the signal is a
function of the number of counts at a given kinetic energy.
Conservation of energy requires that

hv= B.E. + K.E. + aspect, (1)

where hv is the energy of the incident X-Ray radiation, B.E. is the
binding energy and K.E. is the kinetic energy of the photoelectron and
aspect is the spectrometer work function. The value of aspect. will
vary from instrument to instrument, thus each instrument must be
calibrated for this quantity. From the measured kinetic energy of the
photoelectron its binding energy can be calculated by

B.E. = hv K.E. aspect. (2)

From the binding energies of photoelectrons, chemical information

can be deduced about the surface.

Electron Detector / .
X-Rays e
Anod / \Sample e \
ee e e \
ee eI-
Filament e e e e e Vacuum

' \^^ Photoelectrons >

Figure 1-1. Schematic Representing Basic XPS Principles.

X-Ray bombardment of a surface can produce another type of

electron other than a photoelectron. Upon the photoemission of an
electron, a vacancy will develop after ejection. An electron with
lower binding energy from another subshell can move in and fill the

vacancy. This movement results in the dissipation of energy which

can cause a third electron, known as an Auger electron, to be

ejected. Auger electrons are denoted by three upper case letters. The
first letter represents the subshell containing the electron which is

ionized by the incident X-Ray radiation, the second represents the

subshell containing the electron which fills the vacancy, and the
third the subshell from which the Auger electron is ejected. The
signal intensity or peak height is affected by parameters which do
not affect the binding energies. One such parameter is the

differential cross section for photoionization [1], which is given by

do(E)/dQ= GT(E)/4 [1 + BP2(cos(0)]. (3)
It is seen that the intensity is a function of the total cross section
for photoionization, which is the probability of observing an
electron of a given energy for ionization, and the angle 0 between
the incident photon beam and the direction of ejection of the
photoelectron. The B term is known as the asymmetry parameter
and is a term which is characteristic of a given molecular orbital.
The value of B indicates what the preferred direction of the
photoelectrons will be with respect to the incident photon beam. If
B=+2 as is the case for a spherically symmetric distribution of
charge (an atomic s orbital), then the phrLtoelectrons will be
preferentially ejected at angles of 900 to the photon beam [1]. For
orbitals having angular momentum (i.e. p, d or f), B values will be
less than +2, which will cause the photoelectron to be preferentially
ejected at different angles. If the intensity is plotted against 0 for
many different values of B it is found that there is a "magic angle"
of 54.70 where the intensity is independent of 0. When
spectrometers are operated at this magic angle, total cross sections
can be used directly in quantitative analysis.


General Principles

An X-ray photoelectron spectrometer is composed of a x-ray
source, a sample analysis chamber (SAC), an electron or energy
analyzer, and a detector system, which is usually interfaced with a
computer. X-rays are generated by bombarding a material (target)

with high energy electrons. When these electrons impinge upon the
target they knock out its electrons, which creates vacancies. The
photons are generated as a result of the higher energy electrons of
the target filling the vacancies created by the ejection of the lower
energy electrons. This process creates photons of various energies
which then pass through either a Be or an Al window. This acts to
partially filter the bremsstrahlung or x-ray continuum from the
desired Kx rays. Even after filtration, however, only about 50% of

the photons are of the desired energy. The contribution to the
photoelectron spectrum by the bremsstrah ung is not important
because it is distributed over 2 KeV while the Ka rays are

concentrated in a peak of 1eV FWHM [1]. In addition to the Kco 1,2

line, other lines are present due the to difference in energy between
the LII and LIII levels. This difference is important because Ka 3,4

gives satellites in the spectra. With monochromatization it is
possible to reduce the width of Al Kx 1,2 radiation to as little as 0.2

eV [2]. X-ray radiation can be monochromatized by allowing it to
impinge upon a crystal which will cause it to be dispersed. After
dispersion, radiation of a particular energy can be selected by means
of a slit. This method is known as slit filtering. Another method
involves the matching the dispersion of the crystal with that of the
spectrometer. In this technique the X-rays are dispersed by the
crystal before they reach the sample. Their dispersion will cause
the resulting photoelectrons to come out at slightly different
energies, depending on their position along the target. This
dispersion is compensated for by the spectrometer so all electrons
will be ejected as if they came from atomic orbitals having the

same energy [1]. Still another means to reduce the inherent line

width of the radiation is the fine focusing method, which uses a

rotating anode to concentrate the beam before it impinges upon the

crystal. Its advantages are that it gives more photons after

emergence from the monochromator and does not put any

restrictions on the sample. Monochromatization removes those lines

responsible for source satellites in the spectra.







Figure 1-2.

Schematics showing different types of
a) Slit Filtering
b) Dispersion Compensation

Rotating Sample

Figure 1-2.--continued
c) Schematic showing the fine focusing method of

It also removes all of the high energy brehmsstrahlung radiation,
which is responsible for the decomposition of organic samples and
some inorganic salts. The one disadvantage is that there is a large
intensity loss for those instruments which employ a large sample
area like the retarding grid instruments [2]. Figure 1-2 illustrates
the above methods of monochromatization.
To avoid collisions between the X-rays and photoelectrons with
the surrounding gas molecules, it is necessary to contain the source
and sample in a high vacuum sample analysis chamber (SAC). In our
instrument typically pressures as low as 10-9 torr are achieved.
With some degree of effort, namely by placing liquid nitrogen in the
cold trap of the SAC, pressures as low as 5 X 10-11 torr are
possible. The next major component of an X-Ray Photoelectron
Spectrometer is the electron or energy analyzer. Photoelectrons are
generated with a very broad spectrum of energies. Before an electron

with a particular kinetic energy can be counted by the detector
system it must be separated from photoelectrons which have kinetic
energies different from itself. The function of the electron analyzer
is to perform this separation. The spectrum of photoelectron kinetic
energies relative to the sample is not identical to the spectrum of
photoelectron kinetic energies relative to the electron analyzer,
because the sample and the spectrometer share a common ground, as
seen by Figure 1-3. However, they are in one-to-one correspondence,
since they differ by a constant factor, as shown by the following

Ekin = E'kin + (0sam-0spect)
hv = EbinE + E'kin + 0sam
hv = EbinE + Ekin + aspect.

Sample Electron Analyzer Det.


vacuum ki Ek
level Ekin
Ii vYBcuum level
$am_--_n_ specl Fermi level

Figure 1-3. Diagram showing principles for the calculation of
binding energies.

EbinF is the Fermi level referenced binding energy of the electron in
the sample. Thus, binding energies will be measured correctly only
when the Fermi level of the sample is pinned to the Fermi level of
the spectrometer. This is not possible for insulators, and the
problem is further aggravated by sample charging which results
from the photoemission process. The work function of the
spectrometer is determined by calibrating it with a known standard,
for example, the binding energy of the Au4f electron, which has a
literature value of 83.8 eV. Auxiliary referencing must be employed
for insulators. Most common are the gold decoration technique or
referencing to the Cis contamination peak.
There are three main types of analyzers which perform these
functions. The two most important are the retarding grid and
dispersion types. The retarding grid analyzer forces the electron to
traverse a potential difference between two grids. This analyzer has
a poor resolution and is not employed in any commercial x-ray
photoelectron spectrometer. The dispersive type analyzer, which is
most commonly used today, separates photoelectrons either
according to their momentum or energy by making them traverse
either a magnetic or electrostatic field respectively. The earliest
dispersion analyzers were of the magnetic type. In this type
analyzer, the photoelectrons are sorted according to their
momentum. The equation which relates the magnetic field, with the
path and momentum of the photoelectron is given by


The magnetic field is B, m and i) are the mass and velocity of the
electron and Po is the radius of the orbit of the electron. Double

focusing can be understood by looking at Figure 1-4. If the
photoelectron enters the analyzer; at an angle 0 to po (the optic

circle), in the xy plane or at an angle g. with respect to the z-axis,
then it will return to the optic circle after it has traversed an angle
of ,'2 or 255 degrees. In other words, the analyzer has the ability
to redirect the deviation of the photoelectron whether it deviates in
or out of the plane of po. The major advantage of the magnetic

dispersion analyzers is that a greater field can be supplied for the
study of high energy (>5000eV) photoelectrons. At this energy
relativistic effects become significant and the optics needed to
study such electrons are better understood for magnetic

1 ~rmage
OpticeCirde x

^^ Source

Figure 1-4. Principle of Magnetic Double Focusing Electron Analyzer.

One disadvantage of this type analyzer is that of cost. A magnetic
analyzer comparable to an electrostatic one would cost
approximately ten times as much. Another disadvantage of the
magnetic analyzer instruments is their sensitivity to stray
magnetic fields. If a resolution of 0.01% is desired the stray field
must be reduced to 0.1 mG over a very large volume. To achieve this
reduction of stray magnetism, very large compensating coils must
be used. This makes the instrument as a whole very space consuming.
Also, the stray fields must be monitored continuously [2].
Besides magnetic dispersive analyzers, there are also
electrostatic type analyzers, which are more commonly used today
because they cost less to construct and are less cumbersome. Their
principle disadvantage is that they are unsuitable for studies
involving high energy photoelectrons. If one is analyzing electrons
with an eV greater than 2000, relativistic effects become
noticeable. The optics for dealing with these effects are better
understood for magnetic than for electrostatic analyzers. In an
electrostatic analyzer, the photoelectron is dispersed according to
its kinetic energy along a predescribed path. This is done by forcing
the photoelectron to traverse an electrostatic field instead of a
magnetic one. Such a field is achieved by placing a potential between
two plates. The geometry of these plates determines the type of
electrostatic analyzer. Several of the possible geometries of the
plates are hemispherical and cylindrical. In our instrument
hemispherical plates are employed. If we look at Figure 1-5 we can
see a schematic diagram of a hemispherical electron analyzer.
Equation (5) shows the relationship between the voltage between the

plates, the kinetic energy of the photoelectron, and the radius of its

orbit in the analyzer.

Electron Anal Uzer

e.rd..L n--'- /--i Multichennel
Retardeti Dn
Secti on__ __ Detectar

X -Reus

Figure 1-5. Schematic illustration of electron analyzer with
retardation section.

V=E/e(R2/R 1 -R 1/R2) (5)

This equation states that when a particular voltage is being applied

between the plates, only photoelectrons of a particular kinetic

energy will be able to completely traverse space between the plates
and reach the detector. To measure the kinetic energies of all
photoelectrons which are generated, the voltage is continuously
varied over some specified range, usually with the aid of a computer
which has been interfaced with the system. To reduce or lessen the
tolerances of many of the mechanical components of the
spectrometer, the photoelectron passes through a retardation
section prior to its entry into the analyzer. If the kinetic energy of
the electron is reduced from Ekin to some final EO, the relative

resolution required from the analyzer is reduced from AEkin/Ekin to
AEkin/EO. If, for example, a spectrometer is required to achieve a
resolution of 0.1eV for 10O00eV electrons or 0.01% as the result of
retardation, it would be required to achieve a resolution of 0.1ev for
100eV electrons. Retardation does reduce the intensity of the signal,
however, so a trade-off between intensity and resolution results.
After the photoelectrons have been dispersed according to their
kinetic energies, they must be detected and counted. Almost all
detectors in XPS utilize continuous-dynode electron multipliers of
the "channeltron" type [3-5]. These devices consist of glass tubes
which have been doped with lead and then treated in such a way so
as to leave the surface coated with a semiconducting material with
a very high secondary electron emissive power [3]. A voltage of a
few kV is placed between the ends of these tubes and electron
multiplications of the order 106-108 are achieved by repeated wall
collisions as electrons travel down the inside of the tube [6]. As the
voltage between the plates of the analyzer is swept, the electron
counts at different kinetic energies is usually stored with the aid of
a computer. Computer control is advantageous because it is
desirable to make repeated scans over a spectral region to average
out instrument drifts and to eliminate certain types of noise [6]. In
many cases repeated scans are mandatory. Weak signals can result
due to a small amount of the analyte or if the take-off angle (angle
between the sample and the analyzer about which more will be
covered in a later chapter) is at or near zero. The output from the
electron multiplier can be linked directly to a plotter or printer for
a single continuous sweep. The data system also allows one to

perform peak fits whereby .spectral data can be resolved into
Gaussian or Lorentzian distributions. For example the Oi s spectra
can be resolved into a given number of these distributions which can
then be compared to literature values. By doing so, various chemical
species on the surface which contain oxygen can be deduced. Besides
performing peak fits, it is possible to deconvolute XPS spectra so
as to mathematically remove instrumental linewidth contributions.
This term is not to be confused with peak fit.

Instrument Employed

The instrument utilized for this work is the KRATOS XSAM 800.
This instrument has a dual anode (either Mg or Al can be selected) x-
ray source, a sample analysis chamber which can be pumped on by
either a roughing, turbomolecular, ion, or titanium sublimation type
pump, a hemispherical electron analyzer which includes a
retardation section to reduce the kinetic energy of the
photoelectrons before entering the analyzer, a detector consisting of
a electron multiplier and a data system (A Digital Micro PDP-11
with 256 K bytes of RAM) to control the scanning of the
photoelectrons and to collect and store the data. Peak fits are also
possible at the convenience of the operator.
The X-ray source consists of a filament assembly which is
essentially a tube with tungsten filaments which have been coated
with thorium on either side. The thorium has a lower work function
than the tungsten which makes the emission of electrons more
efficient. The anode is a hollow metal rod with one end open so as

to allow cooling water to pass through and the other end capped
with copper. The copper end resembles a roof with one side plated
with Al and the other side with Mg. To generate x-rays, the
emission stabilizer circuitry is first activated. This is a new circuit
design by KRATOS which is supposed to generate a more stable
emission of electrons from the filaments. After the emission
stabilizer is activated, the high voltage power supply is switched
on. Before this can occur, three safety interlocks must be satisfied.
First, the pressure of the cooling water to the anode must be
sufficient; second, the ion pump to the x-ray source must be on; and
third, the ion gauge must be switched on. If these interlocks are
satisfied, the power can be turned on. The desired voltage of the x-
ray radiation is dialed up. Usually a value of 15kV is selected. The
emission current is selected, and usually a value of 15mA is chosen.
At this point the x-rays can be generated either by computer
command or local command. When the Al anode is selected, the x-
rays have an energy of 1487 eV, and when the Mg anode is chosen,
the radiation is 1254 eV.
The sample analysis chamber (SAC) normally is kept at pressures
between 10-9 to 10-11 torr. To achieve this extremely low
pressure, it must be pumped in stages. First a roughing pump is used
to evacuate the chamber to approximately 0.5 torr. After this
pressure is obtained, a turbomolecular pump is used to bring the
vacuum down to the 10-3 torr region. The ion pump is switched on
and in conjunction with the turbomolecular pump evacuates the
chamber down to 10-6 torr. The valve between the SAC and the
turbomolecular pump is closed and the ion pump brings the vacuum

down to the operating range of 10-9 torr. The specifications of the
machine state that it is capable of reaching a pressure as low as
5X10-11torr. To reach this pressure, however, it is necessary to add
liquid nitrogen to the cold trap in the SAC. This procedure is very
time consuming and is usually not necessary. To improve the
efficiency of the ion pump, it may be used in conjunction with a
titanium sublimation pump.
After the SAC has been pumped down to the 10-9 torr range and
the X-ray source has been activated, data can be acquired. This can
be done with or without computer control. Almost always computer
control is the option chosen. As stated previously the computer is a
Digital Micro PDP-11 with the RT-11 operating system and 256K
bytes of RAM. DS800, the software written by KRATOS, allows the
user to acquire and process data off line. To acquire data, one simply
chooses that option from the master menu which first appears after
the system is booted. After the data acquisition menu appears one
selects the regions) which is (are) to be scanned, the number of
sweeps to be performed in that region and the time allowed for each
sweep. In addition, the operator has the choice of excitation source
(either Al or Mg), low or high magnification, low or high resolution,
and analyzer mode (either FRR or FAT). After the parameters have
been selected, data acquisition can begin. One chooses the run option
and assigns a file name to that run. The file name can be anything
with six or fewer characters. One parameter that the computer
cannot control on our instrument is the angle between the analyzer
and the sample (sometimes referred to as the take off angle). This
has to be adjusted manually. After the data has been collected and

stored, it can be viewed by choosing the off line processing option.
After this option is chosen the data file name is typed in and the
data appears in the form of spectra. The y-axis represents the
intensity of the signal (number of photoelectron counts for a given
kinetic energy) and the x-axis can represent either the binding or
kinetic energy of the photoelectrons. It is also possible to perform
depth profiling, which is done in conjunction with the ion gun. The
ion gun creates argon ions which are directed as a beam to the
sample. The beam strips away successive layers of the sample and
the composition of each layer is determined. One of the
disadvantages of this technique is that it destroys the sample. If one
is using the sample in a catalytic experiment it might be preferable
to use the sample after it has been analyzed. Also, matrix effect
data is not attainable. Very rarely is the analyte going to be
unaffected by the matrix in which it resides. Depth profiling works
from the premise that all of the molecules are going to be sputtered
at an equal rate, which is not the case. For example, in my work with
dispersed manganese dioxide on carbon substrates, one factor which
causes the molecules to be sputtered at different rates is the
difference in the weights between the manganese and carbon atoms.


Overview of Molecular Orbital Theory

Fundamental Principles

One aspect of the research that is undertaken here involves the
effect of variation of the molecular geometry of clusters of the
formula (MnO)2 on the XPS valence band density of states (VBDOS).
To show this effect, results of molecular orbital theory are
compared with XPS valence band spectra. This comparison is
possible as a result of the theory of Koopman [7-8], which states
that the negative of an eigenvalue or molecular orbital energy (-6) is
equal to the binding energy (B.E.) of the photoelectron as given by

- = B.E.

By making comparisons between eigenvalues and XPS valence band
data, information can be deduced about the electronic structure of
discontinuous clusters of deposited material on the surface of a
substrate. Using the postulates of quantum mechanics, the
eigenvalues can be obtained from the wavefunction of a system, as
discussed briefly below.

The particle, detected by XPS, can be regarded as a wave, with
the wavelength given by the de Broglie relation given by
or X=h/p, (6)

where p is the particle momentum and h is Planck's constant [6].
This wave behavior of the electron can be characterized by a wave
function P(r,t) which contains all the information possible about it
[9]. The Schrodinger equation, which allows eigenvalues to be
extracted from the wavefunction, can be derived from the de Broglie
relationship and the classical time-independent wave equation. It is
given by

ih/4n 2(a/at) I(r,t) = -(h2/8n2m) AP(r,t) + V(r,t)T'(r,t), (7)

where A is the Laplacian operator o2/ax2 + a2/ay2 + a2/az2. This
equation describes the motion of a particle when it is under the
influence of a potential V(r,t). The Laplacian operator and the
potential acting together on W(r,t) form the Hamiltonian operator,
which represents the total energy of the system. The wavefunction
T(r,t) must meet certain conditions, however, in order for the
equation to be valid. These conditions derive from the postulate of
quantum mechanics which states that a system of particles must be
described by a square-integrable function. Thus '(r,t) = T (ql,q2,q3,-
.... W* wi, w2, w3, .*, t), where the q's are the space coordinates, the

w's are the spin coordinates, and t is the time coordinate. T*T is the
probability that the space spin coordinates lie in the volume element

dr (=dti, dr2, -) at time t, if is normalized. To be acceptable the
function must be single valued, nowhere infinite, continuous, with a
piecewise continuous first derivative [8] as seen in Figure 2-1.

a) Not Single Valued

b) Not Continuous


c) Has Infinite


d) Acceptable


Figure 2-1.

Illustrations of unacceptable and acceptable




The more familiar form of the' Schrodinger equation (7) is

HT = ET, (8)
where H is the Hamiltonian operator and E is the energy. Equation (8)
is an example of a class of equations called eigenvalue equations as
shown by

Opf=cf, (9)

where Op is an operator, f is a function called an eigenfunction and c
is a constant called an eigenvalue. Therefore T is an eigenfunction.
In equation (7) V is the potential energy, and the second derivatives
of the wavefunction are related to the kinetic energy. This is so
because the second derivative of P with respect to a given direction
of measure is the rate of change of slope (i.e. the curvature) of P in

that direction. A wave function with more curvature will yield a
greater kinetic energy. This is in agreement with the de Broglie
relationship which states that a wave with a shorter wavelength
will have a greater kinetic energy. Since we have a constant E, the
wave must have more curvature in regions where the potential
energy is low and visa versa [8]. The wavefunctions which are
associated with a particle are related to its momentum by equation
(6). In addition, the wavefunctions are eigenfunctions of the
Schrodinger equation (8) and must meet the conditions which are
illustrated in Figure 2-2. Also, the absolute square of the
wavefunctions (i.e. Il12) is proportional to the probability density
for finding a particle. If is an eigenfunction of equation (8), then
k' is also an eigenfunction, where k is a constant. Since k can be any

number, a problem arises as to which value k should be. Since the
absolute square IT'12 is proportional to the probability density for
finding an electron,

AT HigherE

-=- Lower E



Figure 2-2. Illustration showing relationship between wave
curvature and energy.
(a) When the potential energy V=0, the higher energy
has more curvature (more wiggly). (b) As V increases
the wavefunction becomes less wiggly [8].

then the probability of finding a particle between x=-oo and x=-o
must be 1. The following equation is given by

k*k f I*(x)T(x)dx =1. (10)

If selection of the k multiplier is made such that (10) is satisfied,
then the wave function T' = kT is normalized. If two different

eigenfunctions 'ga, 'b are integrated over x, they must give zero as
a result, as shown by equation (11)

J 'Pa'Pbdx=0. ab (11)

Such wavefunctions are orthogonal. If wavefunctions are normalized
and orthogonal, they are said to be orthonormal.
A second postulate states that when a dynamical variable of an
operator is measured, that the measurement is one of the
eigenvalues of that operator. If a large number of identical systems
have the same function T then the average number of measurements
on the variable M is given by

Mav = J T*MP dr. (if T is normalized) (12)

By linking the postulates together, and knowing the Hamiltonian
operator represents the total energy of the system, the average
value for the energy of a large number of identical systems can be
obtained. This is given by equations (13) and (13a).

E = J W*HT de (if T is normalized) (13)

The bra-ket form of (13) is
E = I H I (13a)

In molecular orbital calculations, the energy of the system is
minimized according to equation (13). Before this can be done
however, it is necessary to formulate the Hamiltonian operator for
the system and the form of the wavefunction T'. To formulate the

Hamiltonian operator, one must account for the total energy of the
system. Terms must be formulated for the kinetic energy of the
electrons, the potential between the electron and the nucleus and
the electrostatic repulsion between electrons. Formulation of these
terms results in an equation given by

n N n n-1 n
H = -(1/2) Ai2 I I ( Zu/r.ti) + I 1/rij. (14)
i=1 I=1 i=1 i=1 j=i+1

The letters i and j are indices for the n electrons and is an index
for the N nuclei. In principle wavefunctions can be symmetric or
antisymmetric. An antisymmetric wavefunction has equal amounts
of area represented by (+) and (-) regions. Symmetric wavefunctions
do not. This principle can be seen in Figure 2-3.



Figure 2-3. Figure showing principle of antisymmetric and
symmetric wavefunctions.
a) antisymmetric; b) symmetric
The Heisenberg uncertainty principle states that the ability to "see"
electrons in an atom would perturb it so strongly that it could not be

assumed to be in the same state after measurement. Therefore there
is no way of distinguishing electron (1) at position rli, from electron
(2) at position r2. If we want to know r, we can only average
together rl and r2. Since we cannot distinguish electron (1) from
electron (2), the wavefunction can not be written as simply the
product of one electron functions 0 of the form

V = 01(1)02(2)03(3)... (15)

Since electrons are fermions, which are particles with half integral
spin, the wavefunctions are required to be antisymmetric with
respect to electron exchange. This behavior is accounted for if the
wavefunctions are written in the form of a Slater determinant [6].
For example, for a two electron system, the wavefunction would be
given by the following equation:

T 1 0 (1) 42 (1)
V2 *(2) 0 (2)

The general form for the Slater determinant is

0(1)02(1).... n (1)
01(2)02(2) Wn(2)
01(n)2(v) nn)

In molecular quantum mechanics it is very important to calculate
eigenvectors and eigenvalues which represent electrons moving in a

"self consistent field" or SCF. The reason for this importance is seen
by looking at the last term in the equation for the Hamiltonian
operator (14), which is the interelectronic repulsion operator.
Because electrons repel each other, the electron density is more
diffuse than it would otherwise be. Electron (2) "sees" electron (1)
as a smeared out, time averaged cloud. Electron (2) "sees" electron
(1) as a smeared out, time averaged cloud. Electron (2) is thus
screening the positive nucleus from electron (1). Since the nucleus
is being screened, electron (1) will occupy a less constricted orbital
than it otherwise would. If electron (1) is in a ls orbital, its orbital
as a result of this screening is represented by

Is'(1)= qt3/ir exp(-rl ). (electron (1)) (16)

A numerical value for C, which is related to the screened nuclear
charge seen by electron (1), can be determined. Likewise electron (2)
is being screened by electron (1) in its expanded orbital C. A value
for ' can therefore be determined for electron (2). C' will be
different from C because the shielding of the nucleus by electron (1)
is different from the previous step. Each change in C for electron (1)
necessitates a change in '. This process is continued until the two
values (i.e. C and C') converge. When this happens electrons (1) and (2)
are being screened by the same amount. The potential due to the
nucleus and charge cloud of each electron causes the orbital for each
electron to be self consistent. The electrons move in a self
consistent field [8].

The one electron wavefunction Oi for a molecular orbital must be
expressed in some mathematical form. The manner in which this is
done is to express it as a linear combination of atomic orbitals
(LCAO). If Oi is the molecular orbital it can be expressed by
Oi = I cjixj, (17)
where the Xj's are the atomic orbitals and the cji's are the
coefficients. The coefficients of the atomic orbitals are known as an
eigenvector. The atomic orbitals Xji can be written as a function of
the following variables:
X(r,O,S) =Rnl(r)Ylm(O,O). (18)
The variables r, 8, and 0 are expressed in terms of spherical polar
coordinates as illustrated by Figure 2-4. The Ylm(0,0) part has
angular dependence and is explained by spherical harmonics.
x=r sin8 cos
y=rsin 8 sin r /
z=r cos 8

0 v


Figure 2-4. The relationship between spherical polar (r,e,p) and
cartesian coordinates (x,y,z).
It can further be broken down as exemplified by the equation given

YIm(9,0) = Olm(e)Dm(O). (19)

To be sure that the wavefunction will be unchanged if 0 or 0 is
replaced by (0+2n) or (0+2n)the spherical harmonics depend upon the
angular-momentum quantum numbers I and m, which arise in the
solution of differential equations involving angular coordinates 0
and 0. The radial part of the atomic orbital Rnl(r) is a function of
exponential decay function (exp) and can take either the Slater-type
[10] form
rn-1 exp(-Cr) (20)

or the Gaussian type form
rn-1 exp(-r2). (21)

The Slater type orbital is used in the research undertaken in this
dissertation. Slater functions behave better in the region of r=0 and
do not fall off as sharply as do the Gaussian type orbitals [11].
The orbital exponent is a function of how "spread out" the orbital
is. The formula for is given by

S= (Z-s)/n*, where n* is the effective principal number,

s is the screening constant and Z is the atomic number. The greater
the screening by the other electrons, the smaller will be the value of
and the more diffuse the orbital will be. The effect of the value of
CL on the orbital is shown by Figure 2-5.

/ Larger "inner" STO

R(r) /

-Smaller "outer" STO

0 r

Figure 2-5. Schematic showing the effect of the Slater exponent
on the radial portion of an atomic orbital.

Self-Consistent Field Theory

As was mentioned in the previous section, it is necessary to
determine eigenvalues and eigenvectors which are the result of
electrons moving in a self-consistent field. From equation (13), the
energy is obtained by allowing the Hamiltonian operator H to operate
on the probability density {*-'. H can be broken down into a one
electron part Hi and a two electron part H2 as illustrated by

H = Hi + H2. (22)
The one electron part is a function of the kinetic energy of the ith
electron and the potential between that electron and the nucleus. If
summed over all electrons we have
HI = Hcore (p), (23)
where Hcore (p) = (-1/2) Ap2 ZArpA-1. (24)

The two electron part of equation (22) is H2 = ,. rpq-1 (25)
Equation (13a) is given by
E = <'| H IT>,
where E = <'1 Hi IT> + <'1 H21'>. (26)

The wavefunction must be written as a Slater determinant so that
electron exchange can be incorporated into it. Allowing a
permutation operator P to act on the wavefunction is the equivalent
to writing the wavefunction as a Slater determinant. The
expectation value of the one electron operator is

Hii = Jf 'i(1) Hcore "i(1) drli. (27)

The two electron Hamiltonian is a function of the two electron
operator 1/rpg (equation 25). This operator gives the electrostatic
coulomb repulsion energy between two charge clouds [8]. A matrix
element of this electrostatic coulomb repulsion is defined as

Jij = fJf 1i*(1)Wj*(2) (1/rpq) 'k(1)W 1(2) d'rldT2. (28)

The value of this integral represents the repulsion between electron
(1) on orbitals Ti and Wk and electron (2) on orbital Tj and 'F1. Since
the charge clouds are everywhere negative, their product causes J to
be everywhere positive. The entire matrix would represent the
electrostatic repulsions between all orbitals in the molecule
including differential overlap where i*()TWk(1). Another integral

that results from the evaluation of the two electron Hamiltonian is
the exchange integral denoted by K. This integral gives the
interaction between an electron "distribution" and another electron
in the same distribution [8]. The exchange integral is given by

Kij = JJf i*(1) Tj*(2) (1/rpq) 'j(1) Ti(2) dti d'r2. (29)

By collecting terms the formula for the total electronic energy is
given by
n n n n
E = 2 Hii + ,Jii+ I I (2Jij- Kij) (30)
i i i j(i)
and the orbital energies are given by

e= H + ,{ 2Jij Kij}. (31)

The derivation of the Fock operator is very complex and it is not
necessary for it to be presented here. If interested in its derivation
consult reference 8 appendix 7. By utilization of the previously
mentioned terms, the Fock operator is given by

F = [Hcore + X(2Jj (1) Kj(1))]. (32)
(1) j
which leads to the following equation in the eigenvector form.

Foi = 4i (33)

Self consistency is achieved by making an initial guess at the
molecular orbitals ij. These MOs are used to construct a Fock

operator, which is used to solve for the new MOs (i.e.()'). These are
then used to construct a new F' and so on until no significant change
is detected. The solutions are said to be self-consistent.

Approximations to Molecular Orbital Theory

Basic Principles of Complete Neglect of Differential Overlap (CNDO)

The SCF principles that were outlined above involve very lengthy
algorithms (some methods have more than 80,000 lines of code) and
as a result, require a considerable amount of computer memory and
CPU time in order to execute. As a result, approximations have been
applied to SCF principles. Thus considerably shorter codes (typically
between 1200-1600 lines) requiring less computer memory are
obtained. One of the best known examples of approximate molecular
orbital theory is complete neglect of differential overlap (CNDO)
written by John A. Pople and associates in 1965 [12-13]. Such an
approximate method is also referred to as semiempirical because
the eigenvectors and eigenvalues no longer result solely from the
principles of quantum mechanics. Experimental data is used in the
formulation of the Fock matrix.
The first approximation in CNDO which is applied to SCF theory
applies to the formulation of the overlap integral matrix. This
matrix is composed of values which show the degree of overlap
between the various atomic orbitals in the molecule. The
approximation consists of replacing the overlap matrix by a unit

matrix whereby all elements are zero except the diagonal elements
which are 1. In the normalization of Roothaan's equations J.(FIO-
eSgi)C'Ui = 0 [14]. By ignoring differential overlap, S -u =0 for Iu In
other words, the atomic orbitals are treated as if they were
orthogonal and as a result, the Roothaan equations reduce to
X FgCu = eiCgi where Fg is the Fock operator, C-ui is the
eigenvector or coefficients to the same atomic orbital as the Fock
operator and the Ei's are the eigenvalues or molecular orbital
energies. This approximation becomes more severe as the
internuclear distance decreases, however, because it causes larger
and larger electron populations to be ignored. The second
approximation results in a simplification of equation (28), which
computes the matrix elements of the electrostatic coulomb
repulsion between charge clouds. The approximation neglects all
differential overlaps in two electrons integrals. Differential overlap
occurs when i*(l1)Pj(1)0, where probability density is coming
from electron (1) over orbitals i and j. Such electron densities are
exceedingly numerous and also exceedingly small. Ignoring
differential overlap means than equation (28) vanishes unless i=k
and j=l. This has the obvious benefit of reducing the number of
integrals that need to be evaluated. The third approximation, which
results from the second, is to reduce the number of coulomb
repulsions to one value per atom pair. Differential overlap can be
monoatomic, where iYk is on the same atom or diatomic where
'PiFk is on different atoms. For the monoatomic case neglect of
differential overlap causes invariance of rotation to be negated. This
means that rotation of an atom with respect to another atom will

result in a different set of eigenvalues and eigenvectors. To restore
invariance, there is an additional approximation made. The remaining
two electrons integrals will not be dependent upon the nature of the
atomic orbitals, but on the atoms to which Tj and k belong [11].
This can be shown as

(iilkk) = FAB for all i on atom A and for all k on atom B.

r'AB is the average electrostatic repulsion between any electron on
atom A and any on atom B [11]. The value of FAB is given by

FAB = ifJ SA2(1) (1/r12) SB2(2) dtldt2. (34)

As equation (34) shows all orbitals are taken to be of the "s" type.
The fourth approximation is to neglect differential overlap in the
interaction integrals involving the cores of other atoms where

(i|VBjk) = VAB if i=k. If ick the integral vanishes.

VAB is the interaction between any electron on atom A with the
core of atom B. Therefore, any differential overlap between two
atomic orbitals on atom A will be ignored in the calculation of this
The last approximation made in CNDO is to allow off diagonal
matrix elements in the Hamiltonian to be proportional to the
overlap integrals. This is shown by

Hik = P3ABOSik,


where PABo is the bonding parameter, which is characteristic of a
particular atom. As the overlap increases, the bonding capacity of
the overlap will increase [11]. With all these approximations, the
Fock matrix elements can be computed and are given by

Fgg = Ugt + ( PAA 1/2 Pg) FAA + (PBB FAB VAB) and (36)
Fgu = P3ABOS u 1/2 Pp) FAB. I,u (37)

Equation (36) can be rearranged into
Fjj = Utg + (PAA 1/2 Pgg) FAA + [-QB FAB + (ZB FAB VAB)]
and the total energy can then be derived. This is shown by

E Total = (1/2) Pg (Hg + Fl) + _,ZAZBRAB-1. (38a)
Ilv A To achieve self-consistency an initial guess is made of the
molecular orbital coefficients. The diagonal elements of the Fock
matrix (i.e. Fg) come from experimental values for the ionization
potentials ( i.e. Ug in equation (38)). The off diagonal elements (i.e.
FgI)) are replaced by PABOSgu. The electrons are then assigned to
M.O.s with the lowest energy (i.e. lowest eigenvalues). The density
matrix, which is given by
PU = c icJ)i, (39)
is calculated from the coefficients of the occupied atomic orbitals.
This matrix is used to formulate a new Fock matrix Fg). When the

Fock matrix is diagonalized a new set of eigenvectors and

eigenvalues are produced. They are then used to reassign the
electrons in pairs to the molecular orbitals with the lowest energy
and to construct a new density matrix. These steps are repeated
until self-consistency is achieved [11]. Figure 2-6 shows the effect
of self consistency upon the radial part of a wavefunction. This
program utilizes the modifications made in the second
parameterization of CNDO. These modifications include the
incorporation of the "zero penetration effect" which equates the last
term in parenthesis in equation (38) to zero and the replacement of
the ionization potentials in the UlI term with the average of the
ionization potential and the electron affinity (i.e. -1/2 (lg + A )).


A relatively new CNDO algorithm was selected for this project.
Unlike the version written by Pople et al., [12,13] this program is
parameterized for the first 81 elements of the periodic table. This
is possible by the utilization of the concept of "fictitious atoms,"
whereby those elements which have their valence electrons
distributed over two or more subshells with different principal
quantum numbers, are treated as two or more atoms which are
centered at the same coordinate. Figure 2-7 illustrates this

(before SCF determination)

4n2 y2

Figure 2-6.

Effect of SCF calculation upon the electron distribution.

Although this program retains the concepts outlined in the
previous section, one of the major differences between this
program and the earlier version is in the formulation of the
coulomb repulsion matrix.

(F~kl 3d

Figure 2-7.

First Row Transition Element
Illustration of how a first row transition element is
treated in CNDO2/U.

From the last section it was stated that there is one matrix element
calculated per atom pair. However for a first row transition element
there are ns and (n-1) d valence electrons. It therefore would be
necessary to calculate the following r values:

F (n-1)d (n-1)d ; r(n-1)d ns ; Fns ns.

In addition to this modification in the formulation of the coulomb
integral matrix, there is also a modification in the calculation of
the atomic energies of transition type elements. Since the transition
element is considered to be two atoms it is necessary to calculate
two atomic energies for each atom. The equation used is given by

AE = s*ENEG(s) + p*ENEG(p) + (TE2 r(l,l)/2.) + d*ENEG(d). (40)

where AE equals the atomic energy, ENEG is the average of the
ionization potential and the electron affinity for the respective
subshell. TE equals the total number of electrons and s, p, and d
equal the number of s, p, and d electrons, respectively.



Dispersed particles or clusters, which usually have catalytic
properties on inert substrates such as carbon, silica, or aluminum
oxide, have been the focus of much investigation [15-26].
Investigations have focused primarily on their electronic structure-
experimentally, through the use of electron spectroscopy and
theoretically, through various molecular orbital algorithms. The
electronic structure of these deposited clusters can be investigated
by observing the shifts in binding energy of the main photoelectron
peaks in the relevant core regions and through acquisition of valence
band spectra. It has been reported in a study on the electronic
structure of catalytic metal clusters (i.e. Pd and Pt) that the valence
band undergoes a narrowing and a shift away from the Fermi level
relative to the bulk metal as the metal clusters become more highly
dispersed [27]. Unlike dispersed metals on inert substrates,
relatively little attention has been paid to the electronic structure
of dispersed metal oxides on inert substrates. As a result, the
electronic structure of dispersed and bulk like MnO on carbon foil
has been performed by Zhao and Young [28]. It was determined that

Figure 3-1.

Comparison of spectra of highly dispersed and
continuous film of MnO on carbon foil.
a) The valence band spectrum of highly dispersed MnO on
carbon foil.
b) The valence band spectrum of a thick continuous film
of MnO on carbon foil.




I Ii

15 10 5 0 -5

B.E. (eV)

for highly dispersed MnO on carbon foil (a coverage of 0.22) the
valence band undergoes a narrowing and a shift in its binding energy
away from the Fermi level relative to bulk material, as can be seen
in Figure 3-1. In the case of supported metal particles, this behavior
led to investigations which sought to correlate XPS valence band
behavior with results from molecular orbital (M.O.) algorithms to
determine the minimum or threshold number of atoms required for a
cluster to exhibit properties of the bulk material. For example, R.C.
Baetzold et al. performed an investigation into the determination of
the particle size required for bulk metallic properties [29]. Extended
Huckel calculations were performed on clusters with face-centered
cubic (fcc) geometry of sizes ranging from 13 to 79 atoms. The
results from the calculations were used to construct valence band
density of state curves and the width of the d band was determined.
For the largest cluster size (i.e. 79) it was determined that the d
bandwidth was 86% of that for the bulk material and for a cluster of
13 atoms the d bandwidth was 50% of that of the bulk material.
Table 3-1 shows the results.

Table 3-1
Comparison of d bandwidths with cluster size

cluster Pd2 Pdl 3 Pd3l Pd55 Pd79 Bulk Pd

d band-
width 0.80 1.54 2.26 2.57 2.65 3.08

What is done can be summed .up by the following statement. For (M)x,
where X=1,2,3,4 *** and where M is some monomer unit, as X is
increased the electronic structure evolves to that of the bulk
material. In Baetzold's study the selection of one geometry (i.e. fcc)
for all cluster sizes is arbitrary. It seems reasonable that when a
substance is deposited either fractionally or partially on a substrate
that it will not form the same geometry for all possible values of X.
The goal of this research project is to see what effect the variation
of the molecular geometry for a given cluster size would have on the
valence band density of states (VBDOS). Could a "bulk like" VBDOS
curve be constructed by variation of the smallest possible cluster
size for MnO, namely (MnO)2 ?
As was mentioned briefly at the beginning of chapter 2, the
binding energy of the photoelectron is linked to eigenvalues derived
from molecular orbital theory by the rule of Koopman [7-8], which
can be represented by

-C (eigenvalue)= binding energy of the photoelectron. (41)

This rule is not entirely correct, however, because it is based on an
incorrect assumption-that the orbitals remain frozen orbitals
during photoemission. This is a "static" approximation. In actuality
the remaining electrons "relax" towards the site of photoemission
because of reduced screening of the nuclear charge. This relaxation
imparts a certain amount of kinetic energy to the ejected electron,
thus reducing its binding energy. If comparisons are being made
between clusters of different size, where the extent of relaxation is

different, this limitation of Koopman's rule could be a problem. In
addition, the rule of Koopman also neglects correlation energy,
which is due to electron repulsion. Since we are interested in the
relative eigenvalues for clusters where the number of atoms is held
constant, this limitation should not be a problem.

Modification of CNDO/2U Algorithm

In this work, the semiempirical molecular orbital method of
CNDO/2U [30] is utilized to determine the eigenvectors and
eigenvalues. This method treats the valence electrons as Slater
orbitals and used parameters which are fitted empirically [1]. This
new version of CNDO can be utilized on any element in the periodic
table whose atomic number is less than or equal to 81. One
fundamental difference between this method and the Pople method is
the way it treats elements whose valence electrons are distributed
over different subshells with different principal quantum numbers,
which is illustrated by Figure 2-7. The molecule TO for example,
where T is a transition element, would be treated as three atoms
with T accounting for two of them at the same coordinate. If this
treatment were carried over to the Lanthanides, where the valence
electrons are dispersed over three subshells, we would have three
atoms at the same coordinate (i.e. T(n-2)f, T(n-1)d, and Tn) [30].
Before the eigenvalues and eigenvectors of the various geometries
of (MnO)2 could be determined, it was necessary to check the
accuracy of the program by determining its ability to calculate
dissociation energies of various diatomic molecules. The author

performed such calculations, which are listed in Tables 3-2 and 3-3.
For those diatomic molecules which contain no transition elements,
most of the calculated dissociation energies are in good agreement
with experimental values, as seen in Table 3-2. When the
dissociation energies of diatomics which contained either one or
two transition elements are determined, there is very poor
agreement between the calculated results and experimental values,
as shown in Table 3-3. These results indicated that there is a
problem in the way transition elements are treated. It was decided
to modify those areas of the program which are a manifestation of
treating the transition element as two atoms centered at one
coordinate. One such area which needs to be modified is the
computation of the atomic energies.
In M.O. theory the atomic energy is related to the dissociation
energy by
Edissoc = Ebond = Etot Eatomic. (42)

Table 3-2
Bond Energies and Lengths of Selected Diatomic Molecules.

AB Bond Length Exp.a Dissociation Energy Exp.a
(A) (eV)
HF 1.00 0.92 -6.01 -5.90
LiH 1.54 1.60 -6.71 -2.50
IH 1.63 1.61 -4.47 -3.09
CO 1.22 1.13 -21.96 -11.09
SnO 2.25 1.84 -1.46 -5.46

a) Ref [30] -

Table 3-3
Bond Energies and Lengths of Selected Diatomic Molecules which
Contain one or more Transition Elements.

AB Bond Length Exp. Dissociation Energy Exp.a
(A) (eV)

Mn2 2.8 3.4b -286.18 -0.23
FeO 1.58 1.57 a -142.17 -4.20
FeS 1.90 -149.5 -3.31
MnBr 2.3 -108. -3.22
Fe2 2.7 -179.24 -1.06

a) Ref. [31] b) Ref. [32]

Since the transition element is being treated as two atoms it is
necessary to compute two atomic energies, as shown by

AE= s*ENEG(s) + p*ENEG(p) + (TE2 *F(l,l)/2) + d*ENEG(d). (43)

The ENEG terms are equal to the average of the ionization potential
and the electron affinity of the respective subshell. TE is the total
number of electrons and s,p, and d are the number of s,p, and d
electrons respectively. The r(l,l) term is the monocenter coulomb
repulsion for either the s or d subshell. The total atomic energy is
then computed by adding the "atomic energy" of the s and d shells
together. The program is modified to compute one atomic element
per transition atom and this is done by determining the probability

of a valence electron being either an s (i.e. s/TE) or a d (i.e. d/TE).
The equation utilized is

AE = s*ENEG(s) +p*ENEG(p) +((s/TE)*TE2 F(l,l)/2.) + ((d/TE)*TE2*
F(I,I)/2.) + d*ENEG(d). (44)

The effect of this modification on the diatomic molecules containing
transition elements is shown in Table 3-4. The reason that these
diatomics are chosen to test the accuracy of the program is that
they represent the two types of bonds that are encountered when the
eigenvectors and eigenvalues of the (MnO)2 cluster are being
determined (i.e. the T-T and T-O types where T is a transition
Table 3-4
Results of atomic energy modification to CNDO/2U on selected
diatomics which contain one or more transition elements.

AB Bond Length(A) Dissociation Energy (eV) Exp.(eV)

FeO 1.58 40.9 -4.20
Mn2 2.80 1.0 -0.23
MnO 1.70 38.0 -3.70
Fe2 3.0 -4.52 -1.06
FeS 1.90 33.02 -3.31
MnS 2.00 30.9 -2.85

For molecules representing the T-T bond, (i.e. Mn2 and Fe2) there is
good agreement between data and calculated results. For molecules

representing the T-O type bond however, there is poor agreement.
This makes it necessary to modify the program so that it calculates
a lower total energy for T-0 and T-S bonds.
Another part of the algorithm which can be modified is that part
which computes the coulomb repulsion matrix. This matrix is a set
of values which represent the electrostatic repulsion between the
charge clouds occupied by the electrons in the system. Equation (28)
is the formula for the electrostatic repulsion between electrons (1)
and (2) distributed over orbitals i,j,k and I. When the approximations
of CNDO are applied to (28), equation (34) results. CNDO calculates
the average electrostatic repulsion between any electron on atom A
and any electron on atom B instead of the electrostatic repulsion
between orbitals. Therefore, one matrix element is calculated per
atom pair. For a T-O type molecule the following 3X3 coulomb
matrix is formulated.

0 4s 3d
0 Foo F04s F03d
4s FO4s F4s4s F3d4s
3d FO03d r4s3d F3d3d

and for a T-T type atom the following 4X4 matrix is formulated.

4s 3d 4s 3d
4s F4s4s F3d4s F4s4s F4s3d
3d F4s3d F3d3d F3d4s F3d3d
4s F4s4s F3d4s [4s4s F4s3d
3d r4s3d F3d3d F4s3d F3d3d

The relationship between' the electronic energy and the coulomb
integral matrix elements are seen by the equations derived from
CNDO approximations to SCF theory (i.e. equations 38 and 38a). The
term (ZBFAB-VAB), which represents the potential difference
between the core ( the nucleus and non-valence electrons) and the
valence electrons of atom B, is set to zero as a result of CNDO
approximations. A valence electron on atom A experiences no
potential difference between these components of atom B, thus the
possibility of its penetration into B is eliminated. Such a
penetration would lead to a net attraction yielding a lower bond and
dissociation energy between A and B. To modify the program so that
a greater net attraction between A and B is realized, the term-
QBFAB is modified. This term represents the effect of the potential
due to the total charge on atom B [11]. If FAB is increased, the
potential due to atom B (i.e. Mn or Fe) will be more negative. This
increase would have the same effect as increasing VAB which would
make the potential of the core of B greater than the valence
electrons of B. This should then cause a greater net attraction
between A and B to develop, which is what is desired. By looking at
the first matrix, it can be seen that FAB are represented by 1O3d
and rO4s, both of which are bicenter. The problem is to determine
by how much these elements should be increased. Possibilities exist
that the best results might be obtained by multiplying the FO4s and
f03d by the same or different amounts. By the process of trial and
error it has been determined that if each r value is multiplied by
1.1, the best fit between dissociation energy, ionization potential
and equilibrium bond distance results. After applying the above

corrections to the original program, the values shown in Table 3-5
result. Increasing the bicenter matrix elements of a T-O molecule
should have the effect of increasing the electron population in the
overlap between the oxygen and the d orbitals of the transition
element. This is substantiated by looking at the population matrix,
calculated from the equation
PgU = 2 ,c*gi c-i, (45)

where the c's are the coefficients of the atomic orbitals. This
matrix does indeed show an increased electron population in the
overlap between the oxygen and the d orbitals.

Table 3-5
Final values for the bond lengths, dissociation energies and
ionization potentials of selected diatomics after the atomic energy
and coulomb integral modifications were made to CNDO/2U.

Bond Ionization Dissociation
AB Length Exp. Potential Exp. Energy Exp.
(A) (eV) (eV)

FeO 1.40 1.57b -11.69 -8.71 a -6.77 -4.20

FeS 1.90 -5.02 -0.50 -3.70

Mn2 2.80 3.4e -7.08 -6.9c 1.00 -0.23

MnO 1.45 1.77b -8.06 -5.68 -3.70

Fe2 3.00 -16.09 -6.30e -4.52 -1.06

MnS 1.80 -11.53 -10.16 -2.85

C) Ref [34] d) Ref [35] e) Ref [32]

a) Ref [33] b) Ref [31]

Construction of VBDOS curves

After the modifications were made to the program it was then
possible to begin the acquisition of the eigenvalues and eigenvectors
of various geometries of the formula (MnO)2. In order for the
eigenvalues and eigenvectors of a given geometry to qualify for
comparison with XPS valence band spectra, it is necessary for that
geometry to pass the self consistency test-i.e. two successive
iterations of the total electronic energy must agree to within 0.01
eV. To pass this test it was necessary that the program calculated a
total electronic energy that was within 0.01eV of the previous
value. The determination of suitable geometries was like the
coulomb matrix modification, a trial and error process, but after a
long and arduous process, six suitable geometries were found, which
are illustrated by Figure 3-2.

Mn- 0O--- Mn-0 -
Linear Rectangular

0 \n

Mn Mn M Tetrahedral
Semicircular 1 r
0 n

Mn 0 Mn-

Cross Mn Angular Mn--- 0

Suitable geometries for cluster (MnO)2.

Figure 3-2.

After these geometries were determined, it was necessary to
construct valence band density of states (VBDOS) curves from the
eigenvalues and eigenvectors of each geometry. VBDOS curves are
constructed by centering a Gaussian distribution function about each
eigenvalue which has a FWHM of 0.85 eV (the natural line with for
AIKa( radiation is this value) and then summing them over some
specified energy range. The relative heights of the Gaussian
distribution functions are determined by the following formula.

Ht. of Gaussian = Z c2ij aj (46)

The cij 's are the coefficients of the atomic orbitals and the aj's are
the Scofield cross sections [36]. A VBDOS curve has been
constructed for each of the geometries shown in Figure 3-2. Figure
3-3 illustrates these VBDOS curves. Weighted distributions of these
geometries can be used to construct composites. These simulate the
effect of having active sites on the substrate surface which
stabilize different geometrical configurations of the clusters. The
weights simulate the relative fraction of each type of site on the
surface. One such composite, which can be seen by Figure 3-4,
shows a remarkable agreement with the lineshape obtained for very
thick MnO supported on carbon foil. Although there is no evidence
for "geometry' selective active sites on the supports normally used
for heterogeneous catalysts, there is also no evidence against such
sites. Based on Figure 3-4, the variation in the molecular geometry
for a given value of X does have an effect on the VBDOS. Valence band
broadening is not only due to an increase in the number of atoms in a


cluster, but can also be due to the formation of a distribution of
geometries of different stability on the surface. Thus a meaningful
interpretation of the experimental VBDOS curves of supported metal
oxide particles cannot be made without supplementary information
on the particle size distribution.












0 0
(D -
(o 0



U 0

I. .t".

0 *. O

"- "" 0 0-
i- ) (q 07




L 4J

4i-- 0

E 40D


'7 C,1 at(9 "
0000 ?

c I V

9^ 00

c (D



u 8. cr-^.u 0"



10 C9 in 10' 0i -

co . .07
(M -


c E
v C8
Lnr^ ul 'n us U) 'n o

a ;

Figure 3-4.

Valence band density of states of a calculated (MnO)2
composite compared to spectrum of a continuous film
of MnO.


-18 -1G -14 -12 -10 -8 -G -4

b) XPS

bulk MnO

-14 -12 -b -6 -4 -2 0



A large number of heterogeneous catalysts are composed of a
material (e.g. a metal like Pt or Pd) which has been dispersed upon
an inert support like silica, alumina, or carbon. These dispersed
substances can be characterized by x-ray photoelectron
spectroscopy. More specifically, XPS can be used to determine the
relative concentrations of the various constituents and the relative
particle size distributions of the supported material. The relative
concentrations of the various constituents on a surface are
determined by determining the relative peak areas of the
constituents concerned. This then allows a value for the number
atom ratio of one constituent atom to another to be computed. The
effect of differences in the relative particle size distributions of
the supported material can be qualitatively determined from shifts
in binding energy within a given region. Such effects may be
observed in both the core level regions and in the valence band
regions. Valence band distribution narrowing is one of the most
dramatic changes which can occur. Zhao and Young [28] prepared
carbon surfaces containing submonolayers of MnO, as confirmed by

the relative concentrations' of the constituents on the surface.
Changes in the electronic structure which accompany reduction in
the particle size distributions are determined for the highly
dispersed MnO by comparing its valence band region with that of the
bulk oxide. In that investigation, the oxide was prepared by vapor
deposition. The number atom ratios calculated from XPS data
indicated a low coverage ( represented by D) for the dispersed oxide
(the assumption is made that the lower the coverage value, the more
highly dispersed is the oxide). It is also shown that it is possible to
form a thick continuous film of the oxide on the carbon foil when a
number atom ratio (i.e. NMn/(NC+NMn)) of 1.0 is achieved at take off
angles between 0 and 50 degrees. The effect of changes in the
particle size distribution consisted of shifts in the position of the
maximum of core, Auger and valence level spectra and a narrowing
of the valence band level. This chapter will report results of similar
experiments on other dispersed oxides of manganese and on
dispersed ferric oxide. These characterizations are based on data
acquired from samples prepared by vapor deposition and solution
deposition. Because in solution deposition it might be possible for
the oxide particles to diffuse in from the edge of the carbon foil, it
is necessary to also perform quantitative analyses of samples
which have had 0 edges, 1 edge and 2 edges directly exposed to the
solution which is used in the solution deposition method in addition
to the normal sample which has 4 edges directly exposed.

Variable Angle XPS (VAXPS)

In the introduction of this chapter the term take off angle is
mentioned. The take off angle is the angle between the electron
analyzer and the sample, which is commonly denoted by the Greek
letter 0. The angle 0 is varied in almost any quantitative analysis
undertaken, because its variation causes a variation in the effective
mean electron escape depth, which is more commonly referred to as
the probe depth. The mean electron escape depth is the average
distance a photoelectron will travel before it undergoes an inelastic
collision, a collision whereby it looses energy from the initial
energy acquired from the photoemission process. Unfortunately
inelastic collision is the fate that awaits most photoelectrons
which are generated. This is what severely limits the depth that one
can investigate in XPS.

-, -

Figure 4-1. Schematic showing the attenuation of the incident x-ray
radiation. The distance the photoelectron can travel
before it suffers inelastic collision is quite short
compared to the distance the x-ray can travel.

This is illustrated schematically in Figure 4-1. The distance the x-
rays can travel into the solid is much greater than the
photoelectrons can travel. The relationship between the mean and
the effective mean electron escape depth is illustrated in Figure 4-
2. Even though the mean electron escape depth is constant, the probe
depth, or the effective mean electron escape depth becomes smaller
as the take off angle becomes smaller until at a take off angle of 0
degrees, the photoelectrons generated are originating from the top
monolayers of atoms only. Conversely at a take off angle of 90
degrees, the spectra that are obtained are the most "bulklike". The
value of the effective mean electron escape depth depends upon the
substance that is being analyzed. For metals this depth can be as
small as 10 Angstroms. To change the take off angle in the KRATOS
XSAM 800, one simply rotates the sample probe.

To analyzer
X-Rays \ /

mean electron / effective mean electron
escape depth A8 -escape depth


Figure 4-2. Diagram showing the relationship between the take off
angle and the effective mean electron escape depth.

Quantitative Analysis by XPS

XPS investigations often require that the relative concentrations
of the various components of a surface be determined. If one
averages over depth, then these relative concentrations can be
expressed as number atom ratios (i.e. Nx/Ny) which can be
calculated from the following equation.

Nx,i/Ny,j = (lx,i/ly,j) (oy,j/ax,i) (Xy(Ej)/Xx(Ei)) (F(Ej,Ea)/F(Ei,Ea))-
(T(Ej/Ea)/T(Ei,Ea)) (47)

The I's are the time normalized intensity ratios, the a's are the
asymmetry corrected photoionization cross sections, Ei and Ej are
the kinetic energies of photoelectrons i and j respectively and Ea is
the analyzer pass energy. The subscripts x and y are the specific
elements and subscripts i and j represent specific levels in x and y
respectively [37]. X(Ei) is the inelastic mean escape depth, F(Ei,Ea)
is the electron optical factor and T(Ei,Ea) is the analyzer
transmission function. Fortunately experimentation has given
results which allows equation 47 to be simplified. It has been shown
that the product of the electron optical factor, the analyzer
transmission function and the inelastic mean escape depth cancel
each other [38]. The simplified form of equation 47 is



If the elements x and y have not been recorded with identical
window widths, then the window width must be incorporated into
equation 48.

Nx,i/Ny,i = (Ix,i/lyi) (oy,i/ax,i) (Wx/Wy) (49)

The number atom ratio evaluated at 0=0 degrees is an estimate of
the coverage (0). The determination of the relative concentrations
of the various constituents of a sample is made relatively simple
with the DS800 software. After the spectra are acquired, the
operator constructs the desired window that needs to be quantified
and then simply types either "Q/A" or "Q/I" at the prompt, where
"Q/A" means quantify area and "Q/I" means quantify intensity. To
quantify area, "The peak area is calculated by summing the counts at
all energies within the appropriate window, multiplying by the
energy step between the channels and dividing by the total time
spent acquiring each channel (i.e. the dwell time)" [39]. The
quantification of peak intensity is determined by computing the
difference between the most intense and the least intense channels
within the spectral window, which is divided by the dwell time. The
dwell time is another name for time normalization that was
mentioned at the beginning of this section. Time normalization
ensures that the quantities obtained are not affected by the
operators choice of acquisition parameters. The atomic
concentration % is then calculated from equation (50).

[(xi/qi)/ Y (xi/qi)] X 100. (50)

The variable xi is the ratio .of the raw area/intensity of the ith peak
and qi is the corresponding quantification factor (i.e. related to a
which was discussed earlier). The quantification factor or a is a
term which is a manifestation of how sensitive the instrument is to
an electron in a particular orbital. The quantification factor is a
function of the size of the atomic orbital.
Not only is the relative concentration of elemental constituents
which comprise a surface desired, but also the relative
concentrations of different chemical species which are contributing
to a given spectral window (i.e. Ols, Cls). Analysis of a given
photoelectron peak is achieved by peak fitting a number of different
symmetrical line shape functions (either Gaussian or Lorentzian) to
the peak. With the DS 800 software it is possible to express a
distribution as a combination of a Gaussian and Lorentzian function.
The distribution can be made to have as much as 50% Lorentzian
character. To fit these distribution functions to a photoelectron
peak, it is first necessary to construct a synthetic window (a
different type of window than needed to determine the relative
atomic concentrations). The synthetic window is first constructed
by typing "syn" at the prompt. After the operator states the name of
the window (i.e. Clis, Ols etc.) the cursor appears and the window
width is determined by pressing the space bar at the desired
positions (in eV) at the beginning and the end of the window. After
the window is formed a table appears which requires the operator to
state what type(s) of functions) are to be fit, (i.e. either Gaussian
or Lorentzian or combinations thereof), the element IDs, the
positions of maximum intensity of the component distributions and

the beginning and ending of a given distribution. Some degree of
estimation and guesswork is required here and it takes practice for
the operator to become proficient. In addition the operator can
require the computer to fit a given distribution to within a given
FWHM. After this is done, the interactive synthesis option is
selected and the spectral window that the peak fit is being
performed upon appears in the upper right portion of the display that
appears. The background is subtracted out and the distributions that
the operator selects appear under the photoelectron peak. The degree
of fit is displayed, which at this point is usually very high. Next,
the autofit command is utilized by typing "au" at the prompt. The
computer fits the distributions selected to the peak, calculates the
final fit value and displays it. The lower the value of the fit, the
better it is. Figure 4-3 is an example of four distributions being fit
to the 01 s region of four year old carbon foil, for which the 01 s
contamination can not be completely eliminated.

Preparation of Samples

As is mentioned in the introduction section of this chapter,
vapor and solution deposition are utilized to prepare dispersed
manganese oxides on carbon foil, while the ferric oxide samples are
prepared by solution deposition only. To vapor deposit a manganese
oxide onto carbon foil, it is first necessary to vaporize the
manganese. Vaporization of the manganese is achieved by heating it
resistively in a tungsten filament basket. The vapor produced is
allowed to impinge upon a piece of carbon foil which is cut into a

0 -


C 0

(0 0

%e C

j) 0)
o ct )

0) C

.-- c 70
a. c

o CL

0 M.C
0 C-) 0)





/ /

if If)

c u '. T. LO
-- r O

OLD I I T al
/. O. "*- ,,o[ CO

o ? I

a ,. P) -.--' O

0-- a- c,. In . .

.LO ," J--) ,- .-

LL (n "- C#.
w- ID Iy P

3) C: LO .,O U- . .
0 a -. ..


0~~~I "l 0 d,'

L 0 0 0 0 0
u-. ,,- m L .C
-J _n li m L C : ,1 z
C) c0 # -4 o m m
_j -n -1 r-4 mU cr <
ii. ~~ ~ Sn S- **i oc 'i
co Icu '-j.C c '-4in in I
2w i ni r-^ LL
0) 0 *-r
CC, .o. c *
o cui ~ r '-4 C') tf '4) 0
0l If) 3)i IJ ) 0 f 0 00
o; ** ***
1x1 I3 nN t-C **;** il> C') C') C') Cf
C'Ct i )J Ll 0 C S lb3 3
0)- LL- 0 r- 'A -< | < CT Ci

rectangle with dimensions of, approximately 1 cm X 1.2 cm. This is
done in a sample analysis chamber of a Hewlett Packard X-ray
photoelectron spectrometer at a base pressure of ~ 10-7 torr. Figure
4-4 illustrates that after the Mn is vaporized, it reacts with oxygen
to form either the Mn(ll) or Mn(lll) oxide, depending upon the
reaction conditions. These samples are outgassed by heating them
to 235 C for 20 minutes before recording spectra. If there is any
MnOOH on the surface, this procedure will convert it into Mn203.
Dispersed manganese dioxide on carbon foil is prepared by solution
deposition since is has been shown that MnO2 can be deposited by
means of the KMnO4 decomposition [40]. Absence of KMnO4 itself is
confirmed by the absence of K core level peaks in the survey scans.
The reaction for this decomposition is given by equation (51), where
the fate of the negative charge has not been established.

Mn(s) Mn(v) MnO/C
W filament Oxidation
basket (poor vacuum)

2500 C


Figure 4-4. Schematic showing method of vapor deposition of
manganese oxides on carbon foil.

Equation (51a) symbolizes deposition onto the carbon foil.
MnO4- MnO2 + 02 (51)
MnO2 Mn0O2/C (51 a)
C foil

Solution deposition is accomplished by placing carbon foil
rectangles into permanganate solution (= 3.3 X 10-4 F) and allowing
the dioxide to deposit. The samples are kept in solution for various
lengths of times. Some were left in for 2 hours, some for 20 hours,
and the remainder for 100 hours. To prepare the samples which
contained either 0, 1, or 2 edges, a relatively large piece of carbon
foil ( 5cm X 5 cm) was placed in the permanganate solution for
100 hours. Figure 4-5 shows how the XPS samples with various
edges exposed to the permanganate solution were selected after the
100 hours of deposition time. The rectangles were cut from the
large rectangle with the use of a razor knife.

/ Double edge

No edge

Single edge sample
Figure 4-5. Diagram showing how samples containing the number of
edges indicated were harvested after they were allowed
to have Mn02 deposit on them for 100 hours.

Solution deposition of ferric oxide onto carbon foil necessitated the
precipitation of the ferric ion out of solution before the carbon foil
could be added. The precipitation was accomplished by adding a 1M

NaOH solution dropwise to a. 0.1 N ferric nitrate solution. The NaOH
was added until there was no further change in the appearance of the
brown precipitate that was formed. The pH of the solution containing
the precipitate at this point was 14. The carbon foil was then placed
into the precipitate containing solution and the precipitate was
allowed to deposit for either 18 or 100 hours. After the time
allowed for the deposition to occur had expired the samples were
washed alternately with de-ionized water and acetone. The samples
resulting from manganese oxide deposition were also washed after
deposition. The acetone facilitated the removal of any organic on
the surface and aided in drying. Prior to letting either the Fe(lll) or
Mn(IV) oxide deposit onto the carbon foil, it is necessary to remove
contaminant oxygen from the surface of the foil. According to the
literature, this can be accomplished by heating the foil to 5310 C in
ultrapure nitrogen [41] or by heating it to 2100C in a vacuum [28].
Since the carbon foil is four years old, the oxygen peak could not
completely be removed. Quantitative analysis of the above samples
was performed after the acquisition of data by VAXPS. Regions in
which data are acquired are the 01s, Cls, Mn2p, Fe2p, and the
valence band (VB). It is also necessary to acquire data in the the
regions just mentioned for the carbon foil not being subject to
deposition of manganese dioxide. This allows the determination of
the quantity of contaminant oxygen on the surface and comparison
with carbon foil which has been subject to deposition. The valence
band region is scanned from 15-0 eV.


After the carbon foil rectangles are placed in the permanganate
solution for either 2, 20, or 100 hours, XPS data is acquired in the
regions mentioned in the previous section. The 01s region of the
manganese free carbon foil is compared with the 01 s regions of the
carbon foil samples which are subject to deposition to see if any
deposition had occurred. Figure 4-3 shows a peak fit of the 01 s

region of manganese free carbon foil acquired at a take off angle of
0 degrees. The manganese free carbon foil contains oxygens species
in four different chemical environments. Several of the oxygen
species have been identified in a paper written by Young [42]. The
additional species are probably due to the fact that the carbon foil
has aged. The carbon foil used by Young in the investigation to
identify inherent surface oxygen species was new. The carbon foil
used for the deposition of manganese dioxide is approximately 4
years old. Figure 4-6, which illustrates the Mn2p peaks of a sample
of carbon foil which has been allowed to soak in potassium
permanganate solution 100 hours, confirms that manganese
deposition has indeed occurred. Figure 4-7, which shows a peak fit
for the 01 s region at a 0 degree take off angle for a sample prepared
with 100 hours of deposition time, shows that there is an additional
peak at 529.0 eV. This value compares very favorably with the
literature value of 529.3 eV for the 01s peak of bulk Mn02 [43]. The
0/Mn ratio is 1.72, within experimental error of the expected ratio
2.0. Figure 4-8 illustrates a peak fit for the 01s region of the same
sample at a 85 degree take off angle. In this case, the 0/Mn ratio is

.-= 00
0 0

L- 4-
Q 0
C" ,_ -

o "-j

E o CZ
(0 _0 _0


w o
t<- -:: n

02 a
a) cO
-!d > a)

U _

0 ctscp),
4- a)

cz -Ci a)

C C o 01
2 3: r- co





*. So -

!,1 ,1]. ..rj
I-~~, .,,-q

,In %0


,' u i .,;:
a' -" C

L.U -4 "o -.4
LU0 co

=I .L'
m W

11 CO . .5.

LcU ?' C

- i-


h i.. I" t)
* 1)i no

I- lu
c ^-^^0

-* -, -^ 0 (C

0 ^0
Ll. -4

'^ cX(~ro) ~seu

4-c 0



C 0


z 0

0 0
0) ^


.4- U )

C. 7C) -0














ml Vv - vD 0 CD
aI 0 .~0 cO
Iii C03 0* 0 I
CO co -' Cf toi

-T7 -4 '
,cc CO - U W- T-



1T f-

oA 0 4 $'I U)

0 0 0 0 0

lJ, -1


'j C






V.- i"

( U



s ni

0 0
0 0

CL 0t

o 0



a) 0
70 CZ

0 ui

a) 0
0. 2 o

0 00
f.- o

U ) (D

0 o .-


m0 "=
, $. -





r_ ^_^ /'

co~~A coC"-^^ r
............. ^~^ .. " [ I


w Ll
I-- -'- O

hm ... 0D 'a
x '""- ""y'"/

"" llj lj[ r. in

OCL C" N1 CJ "m

ou cu cl, I

5- .


w u 7 C% Ch tD 0 m
0- 77 ) ,J .- <


I.D m. E
LU a c

L: Cj uNc'J ('

fo cu co * o o 0 -
'.D 0'

C'J COt 0 C 0 10 r i-
_ *m4 n coin ( -, t

U -
fLU C. nI V) 'i : -
X * -- ,,J -- c
-- '-r*- *. .. c ,o ro c \J
..J.C U) Jf -'-" LU D 5, L-
.-1 '.o in- In; ..a ILl
0 0.-- 0 'f"
" .. .. (_i fri --" ;'.J "-* < '--' "-i " "- "
,Q. il C, S- UJ O O O O O
Li" -- -- -/
_ - - E:i ED 'i V/ ,ft ,.t cn
rE C ** *3*l V ilil t
LU 3ET Z ** **"." 1'h VI1 fl*QI *t0 U
-J C;U ",W -.-" 0 0 3 .,3 _,
(. ii *.- *'-. -0 0 *- u fl m '
I CO -,'-, .--I '3 CT5) s 07;C

1.90. By comparison of the- percent area values of the 529.0 eV
distribution of Figures 4-7 and 4-8, it can be seen that deposited
oxygen and therefore MnO2 is being incorporated into the layers of
the carbon foil (i.e. MnO2 increases with increasing take off angle).
This conclusion is substantiated by the number atom ratios NMn/NC
calculated at increasing values of the take off angle. As the angle is
increased, the amount of Mn increased, as shown in Table 4-1. This
is quite surprising since we expected to be able to form a continuous
layer of the oxide at the surface like it was possible to do for vapor
deposited MnO and Mn203. At 0=0 it is possible to form a continuous
film of the vapor deposited manganese oxide [28], as shown in
Table 4-2. It should be pointed out that the value of 1.00 is
determined not by NMn/NC but by NMn/(NMn+NC) in Table 4-2.

Table 4-1
Values of NMn/NC for solution deposited MnO2 on carbon foil.

e(degrees.) 2 hrs. 20 hrs. 100 hrs.

0 0.055 (0) 0.13 (0) 0.073(D)
15 0.13 0.20
35 0.26 0.83
55 0.046 0.22 1.24
85 0.041 0.18 1.07

Table 4-2
Coverage values for vapor deposited MnO and Mn203.


MnO 0.18
MnO 1.00 (no discernible Cl s)
Mn203 0.3
Mn203 0.65
Mn203 1.00 (no discernible Cl s)

Solution deposition of ferric oxide on carbon foil gave similar
quantitative results to that of the solution deposited manganese
dioxide. Figure 4-9, which illustrates the Fe2p peaks of a carbon
foil sample which was allowed to remain in the precipitated Fe
solution (see preparation of samples section) for 100 hours,
confirms that iron has indeed deposited onto the surface. In Figures
4-10 and 4-11, which are peaks fits of the 01is region of solution

deposited ferric oxide on carbon foil at take off angles of 0 and 85
degrees respectively, a peak can be seen at 529.1 eV at 0 degrees
and a peak at 529.2 eV at 85 degrees. This compares very favorably
with a literature value of 529.3 eV for the 01 s peak for the bulk
Fe203. The 0/Fe ratios for the sample analyzed at 0 and 85 degree
take off angles are 1.8 and 1.6 respectively, within experimental
error of the expected 1.5 for Fe203. There is incorporation of the
oxygen from the iron oxide into the layers of the surface of the
carbon, but the incorporation does not occur to the same extent as
for Mn02. This conclusion is made from the 100% area values in

Figures 4-10 and 4-11.



0 0
cu 0

U) a

o -r-

Cc o


0Q-C 0

0 C



cu f.
E ._ o

o CO.-

0 0
C- -C (0

0 U: 0_
S. "o


'6--= _o^

0 1-

.I- oc n

0 C

%- I-
0 -o o

.e 0 0 (

Cz- M- Cl- V;)

CZ a)t aU )
C .) 7 0 70


m C .J - OJ

-j ,- *- .-

Ix 01 ru- rJ- ,J
r: csoI)i
LL *
~? *< - *^- *-


I: I


5 in


E o co

0) a)CO

..- u- 03

;- ) --
o c 0

a -
3: a, )

x: 0 1--

= CZ


*-. '*- ^ Vi
*-" c' C

0.. o o
0 0 a)
Qz (0 a (




---.. /i
.- .." j" /

. -.j ....... ..'-.:1'/
o -- '-

-4 LU
bJ *" "" / .' ~'2
'p, "\. "- -y o

I -F
k ) de .' '*'
,:n s*. -... / L.

!ij Clf) '., T.. *' v
' ( %o %, r

LD~' ur 7

W \1 0'* f' --*' If)
II o., 1A

,. C -r v Lb
" **....- Y ,

7) 0 D.D LL.O

ii tO^ **-.. 'i

ID C'. ( (I C

rr atij

-IJ 17" 0 10 LO tr u' *.'
,,C, *. *

0 0
U- m
cz 115 (0 0. r-o mn
0 u "

CC) *... -4" 0 0 .-,
S"f) LL 0 0 cO Jo a-
I- r.O r.u j u -r .

(0 "- ^ '.D II
n_ In? .'q- 0:' "r .' "--
LiJ ,, ,, i-. "-> "

.n r 1/ CO C OJ T 0".
",J C. 0J ' LL oJ 0TJ 0 <*'o
*..J a-'' *5). .' ."
fKr, .. OJ 1CU (I <}*. i
,-ij co o.=j .> c r 1 o
..- oj .. *_0. .'
uj~" ... .. w
*D " " -. U? @ 0". 0. 0r ,r

c' ", 0".f " -,- Q-- 0- < I -:X
^ '. CJ LL "-' -.I 0" 0..' 09 0

At 0=0, it is 196991 and at .0=85 it is 354465. If the second value

for 100 % area is divided by the first a ratio of = 1.8 is obtained. The
same ratio for the MnO2 deposition yields a ratio 8.4, an
approximate 4 fold increase. The conclusion that ferric oxide is not
being incorporated to as great a degree into the layers of the carbon
foil as manganese dioxide can also be drawn from the area % of the
529 eV peak. At 85 degrees the 529 eV peak constitutes 52.9% of the
area under the curve for the sample prepared by the solution
deposition of MnO2 while it constitutes only 19.3% of the area under
the curve for the sample prepared by the solution deposition of
Fe203. The percentage of area contributed by the 529 eV peak is
greater at e=0 degrees than at 85 degrees for dispersed ferric oxide.
This further substantiates the differences in the degree of
incorporation between the two oxides. The number atom ratios also
indicate that less of the oxide is being incorporated into the carbon
foil. Table 4-3 gives the number atom ratios for the Fe and C atoms
as resulting from solution deposition.

Table 4-3
N Fe/NC values for samples prepared from the solution deposition of
Fe203 on carbon foil.

take off angle 18 hours 100 hours

0 0.057 0.044
10 0.12 0.095
25 0.098 0.089
55 0.075 0.092
85 0.063 0.083

The photoemission results clearly show that there is a difference
between the way that particles deposit from the vapor onto carbon
foil and the way that particles deposit from solution onto carbon
foil. The foil itself contains numerous gross defects, which are
evident as microscopic tears from SEM photomicrographs. For a
defect free foil, the surface monolayer would be expected to act as a
barrier to deposition material, thus all particles would be confined
to the top monolayer. It is obvious that particles can penetrate a
defect-laden barrier. The extent of penetration might be expected to
depend upon the particle deposition rate, the deposition time, and
the defect density. Our results on vapor deposited manganese oxide
particles show that the particles are confined mainly to the surface,
and thus indicate that defect penetration is a minor process. This is
expected because the particle dose (deposition rate multiplied by
deposition time) is low and the defect density is small, as can be
judged qualitatively from the fact that x-ray diffraction results
show that the foil is semicrystalline. This means that it has more
defects than a crystalline solid, and less defects than a
polycrystalline solid. Thus, the probability that a particle deposits
in a defect or that it deposits close enough to a defect to fall in
from a random walk is expected to be small. Besides the possibility
of a particle depositing through a defect in the surface, there is also
the possibility that a particle can leave the surface of the carbon
once it has been deposited. For the solution deposition of Mn02, at
least 4 scenarios can be envisaged, as follows:
1) Small particles of oxide produced by photodecomposition in the
solution phase deposit on the surface and penetrate through defects.

2) Small particles of oxide produced by photoemission in the
solution phase deposit on the surface and intercalate between layers
of carbon by penetrating along exposed edges. They then can leave or
"fall off" the carbon foil edges at a later time.
3) Permanganate anions intercalate the carbon foil and particles of
MnO2 are produced by in-situ decomposition, while small particles
produced by photodecomposition in solution phase deposit only on
the surface.
4) Some combination of all three.
Photoemission results can be used to investigate these scenarios.
Based on an earlier study of fresh carbon foil [42], the Ols peaks at
532.1 eV and 533.2 eV may be associated with graphite oxide, Cx+
(OH-)y (H20)2, where the 532.1 eV peak is due to OH- and the 533.2
eV peak is due to H20. For the four year old carbon foil, there is no
significant difference between the ratios of the areas of Ols peaks
at 529.8 eV and 530.9 eV to the area of the 0 s peak at 532.1 eV for
take off angles of 35 and 85 degrees, as shown in Figures 4-12 and
4-13. Thus, the subsurface has an almost homogeneous distribution
of carbon oxidation species (the relative amount of the peak at 533.2
eV increases slightly with the take off angle) and segregating
species in the depth explored. However after exposure of 4-year old
carbon foil to dilute, neutral permanganate for 100 hours, the
situation is that shown in Table 4-4. There are significant changes
(>> 5%) in the peaks at 533.2 eV and 529.7 eV. Similar results are
obtained in the case of the deposition of iron (111) as the hydrous
oxide, which can be seen in Table 4-5. The drastic reduction in the

c in

o c


U) o


4- a)

0 .
C1 3





7 <

-, .

c -

C) C
,- -.

m" CU ---'4 CT.
." / ." / '"

ft f t< ........., i

7w r-a O

""-. .. ", i -^

.....--. .-,.. .,* -j-

f-i 4*. ^ ^'

~ "\., "\

,=- L -r-. LL:
Lr'j ,-,--- i - i i i -^ i - i l l - l -- I-

I- . .
i- r t r-

0 .<; 0 ':_ J .-T

1--4 OJ O aD 0 a .
0 -
( ,J .- -f -4 a a
CL Q fa C J '-4'C

-#-; ib 1- 1- --
o l ') -u 0 C L t
"") L, ,s,<-j 0. C -1 .E-
0) u- r--- "" I"11 c- ci I '1 *:I
c .o cJu '.j -) ti in) i- ( "
.- ,.n rO I) -* LU

I.iljW J L W 0 0
a. o -'I
-. J_, C ILW 0 C C '

-J. 3 u. n) C f_*** .,-I I .:. :
Of; IL 'DJ *- o c Ct Lr



o c
..Q ,-

o c


Q- (1)



^ a-a)-

T --- -/o

14 .1

0)- 0

0 D

|. j C

C' D

0 CD 3) j^ "' 1'7**

-w P. 7- 5 J ,

III [- -
LU ^:--. '1 n
-": "% ...-....A" '

0 a C" C) ".

0 C

-1 0 C 0
o V-, (Y) Coo -4 :*I'
-" L If) U-) '

CD U< -4 fi f-
qz o '- c:'~
S.,-,- . . .*

'-4 ;5:iJI O
0 ~ ; U C 3) CU '-' C'J
_5 10 eJ
.J -- o -4 1 -. -. -'4

w- Z) m C (Y) Ce) cd 'I-j
-i Ix I->) U> w e: =
LC, L.-' Co I "- I

Ck (j *- 13" L3 t

00 $>0 .I S ., 1:1:r
LL r r,- "" a )i c0 Lc':i :)

o *J-" *.,-
C 4' 75 CO (0() .f
nt: 0 Il 0 00 C
7": .o ,,. ".. C' < V V O
"" ;." tC L LUt 0 00 ,
W a Z ** ** 4(0 (ft'f
hz c. n a ,- .-
--I I u n? ILU 0 -J J::: = :::
oj co *- *"t o "' o

533.2 eV peak is probably due to the dehydration of the surface
caused by the evaporating acetone wash. The increase in the peak at
529.7 eV cannot be directly correlated to the amount of the deposit.
There is a greater amount of deposited MnO2 than of deposited Fe203
after 100 hours. The ratio of NMn/NC to NFe/NC is 12.9 at a take off
angle of 85 degrees. The corresponding ratio of the change on Ols at
529.7 eV, column 4- column 3, is 1.7. Thus, the increase in the
species at 529.7 eV cannot be due to attack on carbon by MnO4- or
Fe+3, since it should be much larger in the case of MnO4- attack. If
we regard each pair of planes of carbon as a "two wall cuvette", then
they could entrain permanganate solution between them by capillary
action provided the interplane separation is large enough. One would
then expect to observe potassium peaks in the XPS spectra, but none
are observed. It seems unlikely that MnO4- would be selectively
drawn in, because of the charging problem. If the interplane
separation is the same as that of the graphite, then both K+ and
MnO4- are too large to be drawn in.

Table 4-4
Ols levels for carbon foil-- new [42], 4 year old (take off angle=85o)
and after deposition of MnO2 for 100 hours on the latter ( take off

Peak Relative % of Total Non Metal Oxygen

529.7 14.3 51.3 ?
530.9 24.4 26.9 ?
532.1 60.0 28.6 16.9 graphite oxide
533.2 40.0 31.8 4.9 bound water

Table 4-5
Ols levels for carbon foil-- new [42], 4 year old (take off angle=850)
and after deposition of Fe203 for 100 hours on the latter (take off

Peak Relative % of Total Non Metal Oxygen

529.9 14.3 36.5 ?
530.8 24.4 30.2 ?
532.1 60.0 28.6 33.3 graphite oxide
533.2 40.0 31.8 0 bound water

Thus scenario 3 is at best a minor process. Resolving the importance
of this scenario would allow the catalysis results to be better
addressed. Figure 4-5 illustrates how samples were prepared which
have either 0, 1, or 2 edges exposed to the solution during the
deposition process. It was stated that the samples were cut from
the large rectangle of carbon foil with a razor knife. It is possible
that the knife edge may have dislodged some of the manganese
dioxide particles in the area of the cut. This is probably not a
serious problem however since the x-ray beam is confined to the
center of the carbon foil. The results of the "edge" experiment are
shown in Table 4-6. The table indicates that an increase in the
number of edges does not cause an increase in the amount of
manganese oxide that is being incorporated in to the carbon foil. In
fact, the opposite effect seems to be occurring. If scenario 1 is the
major method for interlayer incorporation of MnO2, one would expect

the three samples to show the same distributions, because the
defects are uniformly distributed. Thus, the results do not support
the initial expectations for either scenario 1 or scenario 2. However
reflection shows that these results could be obtained for either
scenario 1 or scenario 2. Because the particles are more mobile at
edges, in the 0 edge region, a particle must perform a sequence of
directed "kicks" on a nearest neighbor in order to escape. However
particles near an edge can easily fall off into the dilute colloid
medium. If scenario 2 is occurring, then a decrease in the number of
particles which have intercalated into the surface should be seen if
the number of edges exposed to the permanganate solution is
increased. Likewise, if particles are being uniformly incorporated
into depths, they will be depleted more rapidly where 2 edges are
available for escape and least rapidly where 0 edges are available
for escape. Table 4-6 shows the largest number atom ratios for the
0 edge sample and the smallest number atoms ratios for the 2 edge
sample. Thus, these results cannot distinguish between scenario 1
and scenario 2; many more experiments will be needed before the
mechanism of particle incorporation can be established.
Besides being able to determine the relative concentrations of
various constituents on a surface, XPS can also be used to study the
electronic structure of particles on inert supports. The change in the
electronic structure of these particles can be monitored by
variations in the valence band density of states (VBDOS), which was
discussed in the previous chapter. One of the variations which
VBDOS undergoes when there is a change in the electronic structure