Desorption due to recoil induced by neutrino emission and auger relaxation of 37Cl following the
electron capture decay of 37Ar
by Lin Zhu
A thesis submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in
Physics
Montana State University
© Copyright by Lin Zhu (1995)
Abstract:
A novel experiment was developed in this work to study the desorption and the Auger relaxation
processes of 37Cl following the 37Ar electron capture decay. For the first time, the desorption of 37Cl
ions due to recoil induced by neutrino emission in this decay process was observed. The kinetic energy
distribution of the desorbing 37Cl ions was accurately measured by using coincidence techniques. The
resulting 37Cl ion energy ranges from 5 eV to 13 eV with a maximum at around 9 eV and an FWHM
about 3 eV. The charge state distribution of the desorbing 37Cl ions was also measured. The resulting
charge state distribution is: 53% of the total ions have charge +e, 21% have charge +2e and 26% have
charge +ne, where n ≥ 3. The desorbing probability of 37Cl ions was measured by two independent
experiments which gave the result of 9.4±1.2%. The energy distribution, the charge state distribution
and the desorbing probability of 37Cl ions are all quite different as compared with the expected values
for an isolated Cl atom. These differences are explained in the desorption model involving charge
exchange and Coulomb repulsion between 37Cl ions and their surrounding atoms.
The electron capture decay also creates a highly unusual initial state in the 37Cl atom which allows
direct observation of some novel relaxation processes which are. amenable to many body theory, but
essentially impossible to probe experimentally with conventional techniques. For the first time, direct
evidence of the double Auget decay of a K-hole and the large shift in energy (22 eV) of an LMM
Auger line was reported. The double Auger decay probability and energy distribution of the two double
Auger electrons were measured by using coincidence techniques. The resulting double Auger decay
probability ranges from 12±0.3% to 15±0.4% of the total Auger decay, The preferred energy
distribution of the double Auger emission is for one of the electrons to take most of the energy, with
the second receiving the small remaining balance. A model explaining the large shift in energy of an
LMM Auger line is also provided. 
DESORPTION DUE TO RECOIL INDUCED BY NEUTRINO EMISSION 
AND AUGER RELAXATION OF 37Cl FOLLOWING 
THE ELECTRON CAPTURE DECAY OF 37Ar
by
' LINZHU
A thesis submitted in partial fulfillment 
of the requirements for the degree
of
Doctor of Philosophy 
in
Physics
MONTANA STATE UNIVERSITY—BOZEMAN 
Bozeman, Montana
May 1995
J ) 3 7 8
"ZL rD i
APPROVAL
of a thesis submitted by
LIN ZHU
This thesis has been read by each member of the thesis committee and has been found to 
be satisfactory regarding content, English usage, format, citations, bibliographic style, and 
consistency, and is ready for submission to the College of Graduate Studies.
Dr. Gerald Lapeyre ^
(S ta tu r e )  y  f (Date) I
Approved for the Department of Physics
Dr. John Hermanson
(Signature)
L 1M r
(Date)
Approved for the College of Graduate Studies
Dr. Robert Brown
(Signature) (Date)
iii
STATEMENT OF PERMISSION TO USE
In presenting this thesis in partial fulfillment of the requirements for a doctoral 
degree at Montana State University—Bozeman, I agree that the Library shall make it 
available to borrowers under the rules of the Library. I further agree that copying of this 
thesis is allowable only for scholarly purposes, consistent with “fair use” as prescribed in 
the U.S. Copyright Law. Requests for extensive copying or reproduction of this thesis 
should be referred to University Microfilms International, 300 North Zeeb Road, Ann 
Arbor, Michigan 48106, to whom I have granted “the exclusive right to reproduce and 
distribute my dissertation in and from microform along with the non-exclusive right to 
reproduce and distribute my abstract in any format in whole or part.’’
Signature
f  f7Date
iv
DEDICATED TO MY GRANDMOTHER 
AND MY PARENTS
VVITA
Lin Zhu was born on October 29, 1962 in Beijing, China to Shijun Zhu and 
Wenjuan Xiang. He has a younger sister, Hong Zhu.
Lin Zhu was raised in the beautiful campus of Tsinghua University, Beijing China 
where he attended the university-attached primary school in 1969, junior high school in 
1974 and high school in 1977. In 1980, he was admitted to Tsinghua University, one of 
the best schools in China. He completed his undergraduate education with a B.S. degree 
in electrical engineering in 1985. In the same year, he started his graduate study in the 
Graduate College of Tsinghua University. From 1987 to 1989, he studied in the Physics 
Department of the Technical University of Munich, Munich, West Germany. In 1989, he 
came to the wonderful small town called Bozeman. He got his M.S. degree in physics 
from Montana State University in 1992.
Xvi
PREFACE
<
Eight years ago, when Dr. Avci, a surface scientist met Dr. Hindi, a nuclear 
physicist half a world away in Saudi Arabia, they came up with a brilliant idea: using 
surface analysis techniques to study nuclear reactions. Six years ago. Dr. Menzel invited 
me to his office, “I just talked to Prof. Lapeyre in Montana. You should not disappoint 
them when you get there,” he told me in German, “because I have just put up a lot of 
good words for you.” Four years ago, there began this research project. With the help of a
I!
German professor, I got the opportunity to work under the guidance of professors with 
Turkish, Lebanese and American origin. I would like to dedicate this report to these 
mentors who have shown me how small this world could be for physicists.
My sincere thanks to those who made this possible. To Dr. Lapeyre for his help in 
my coming to this great country, for his continuous support, supervision and confidence 
in me. To Dr. Avci for his scientific inspiration, encouragement and patient guidance 
throughout this thesis work. To Dr. Hindi who has helped me throughout this project with 
his sharp instincts in physics and tireless working spirits. To Dr. Anderson for his 
important criticisms and advice that have been and will remain as a beneficial experience 
for me. To Dr. Smith, Dr. Tuthill and Department Chairman, Dr. Hermanson for their 
great support in my graduate study. To the wonderful staff, Alice, Margaret and Rose for 
their cheerful help. To Erik and Norm for their excellent technical support.
vii
TABLE OF CONTENTS
Page
VITA.................................................................................................................................... ...
PREFACE................ !..........................................................................................................vi
TABLE OF CONTENTS..................................................................................................vii
LIST OE TABLES............................................................................................................. ...
LIST OF FIGURES........................................................................................................... xi
ABSTRACT......................................................................................... ;.................... , ......xv
1. INTRODUCTION...........................................................................................;.............I
Concepts of the Work.......................................................................................... 2
Desorption due to Recoil Induced by Neutrino Emission.................. 2
Surface Interactions of 37Cl Ion............................................................. 5
Auger Relaxation of 37Cl Atom............................................................. 8
Historical Review of Related Works...........................................   10
2. THEORY..........................................................................................................................14
Recoil Induced by Neutrino Emission..................................... '......................... 14
Models on Stimulated Desorption........................................................................17
The Menzel-Gomer-Redhead (MGR) Model 18
viii
TABLE OF CONTENTS—Continued
The Knotek-Feibelman (KF) Model...................................................... Tl
Desorption Due to “Coulomb Explosion” ............................................25
Theory of Auger Transition.........................  29
Auger Effect..................................................................................... ;.......29
Auger Energies and Intensities........................ !.....................................33
Double Auger Transition......................................................................... 38
3. EXPERIMENT............................................................................................................... 45
Experimental Setups............................   45
The UHV System..............................................  ....46
The Substrates..............................   49
The Detectors......................... .................................................................. 51
The Position Sensitive Detector..............................................................56
The Radioactive 37Ar Sources.......... ;...............i.....................................61
Coincidence Techniques and Electronic Systems............................................. 66
Experimental Procedure......................................................................■................ 73
4. RESULTS AND DISCUSSIONS......................  77
Desorption due to Recoil Induced by Neutrino Emission................................77
Energy Distribution of 37Cl Ions.............................................................77
Charge-State Distribution of 37Cl Ions...................   83
Page
Results on Multiple and Mixed Layers............ .....................................91
Desorption Probability of 37Cl Ions........................................................95
Discussion of the Desorption Results.................................................... 99
Auger Relaxation of a 37Cl atom......................................................................... 108
37Cl LLM and LMM Auger Peaks.........................................................109
Discussion of 37Cl LMM Auger Results............. :................................ 113
Double Auger Decay of a 37Cl atom...................................................................117
37Cl Double Auger Probability................................................................ 118
37Cl Double Auger Energy Distribution  ........................ .................127
Discussion Of37Cl Double Auger Decay Results...............  133
5. CONCLUSIONS...........................................................................................................140.
Conclusions............................................................................................................. 140
Perspective on Neutrino Mass Limit Study.......................   143
REFERENCES.............................................................................   146
APPENDICES............................................................. ..1...................................................156
Appendix A: Introduction to Neutrino Physics
.
and Neutrino Mass Limit Study................................................. 156
Appendix B: Designing Diagrams of the Detectors........................................168
Appendix C: Plasmon Decay into Multi-Electron-Hole Pairs in Si(l 11)... 176
ix
TABLE OF CONTENTS—Continued
Page
XTable . Page
2 .1 Table of the Energies, Symbols and Number of
Electrons for Ar and Cl Core Levels...................................................................29
2.2 Table of Final States of Ar KLL Auger Series.................................................. 32
2.3 Table of Ar LMM Auger Energies and Relative Intensities............................ 36
2.4(a) Table of Auger Transition Probabilities for a K-hole Vacancy..................... 36
2.4(b) Table of Auger Transition Probabilities for L-hole Vacancies...................... 37
2.5 Table of Final Charge State Distribution of Ar after
Auger Transition of an Is Initial Hole............................................................... 38
4.1 Table of Charge State Distribution Results
from Experiments and Simulations....................................................................84
A l.I  T able of the Three Generations of Quarks and Leptons................................. 161
A l.2 Table of Neutrino Mass Fitting Results............................................................. 165
LIST OF TABLES
xi
I . I Electron capture decay and 37Cl ions desorption in 37Ar ->37C1 + v ..............3
1.2 The original 9.5 eV recoil energy and the additional energy due to
charge exchange and Coulomb repulsion of the 37Cl ions.............................. 6
1.3 The unusual initial state of 37Cl atom from EC decay......................................9
1.4 Normal Auger transition and double Auger transition.....................................10
2 .1 The energy level scheme of MGR model...........................................................19
2.2 The energy level scheme of Antoniewicz model.............................................21
2.3 The KF model for O+ ion desorption.................................... ........................... 23
2.4 Schematic displaying terms of a Hubbard Hamiltonian with intrasite
Coulomb term Ueff............................ ................................................................... 27
2.5 KLL Auger transition and LLM Coster-Koing Auger transition...................30
2.6 Energy changes of an Auger electron entering analyzer
with work function................................................................................................ 34
2.7 Relative intensities of Auger peaks of Ar with Is hole calculated
by using RACS................................................ ....................................................37
2.8 Brueckner-Goldstone diagrams for the partial amplitudes
of double Auger decay...................................................................................... . 40
2.9 Diagrams for double Auger decay models........................................................42
2.10 Double Auger energy distribution calculated for atomic Ne..........................43
3 .1 UHV system schematics....................:.................................................................. 47
LIST OF FIGURES
Figure Page
LIST OF FIGURES—Continued
Figure Page
3.2 Aiiger peaks of Ar and Carbon from the substrates
at different temperatures...................................................................................... 48
3.3 TPD spectrum of Ar from the graphite substrate............................................ 50
?
3.4 The mounting bridge between the graphite substrate
and the cold finger............................................... .................................................51
3.5 Diagram of the home-made channeltron based detector.................................. 53
3.6 Diagram of the home-made microchannel plates based detector.................. 55
3.7 MCP pulse high distribution and detecting efficiencies for
electron and ion.... ............................................................................................... 55
3.8 Geometry of a circular arc terminated resistive anode encoder..................... 59
3.9 Diagram of the home-made MCP-RAE based PSD detector.........................60
3.10 The gamma ray spectrum of the radioactive gas source taken
by an HP-Ge detector............................ ...............................................................62
3.11 ■ The gamma ray spectrum of the 37Ar gas source..............................................63
3.12 Double sealing 40Ca in quartz ampoules before nuclear irradiation..............65
3.13 Schematic diagram of the first coincidence measurement system.................67
3.14 True coincidence counts and accidental coincidence counts..........................68
3.15 Schematic diagram of the second coincidence measurement system...... . 70
3.16 The logic block diagram of the second coincidence
measurement system........................................................................................... 71
4 .1 The Monte-Carlo simulation of the 37Cl ion recoil
velocity distribution.................................................... .........................................79
■
xii
LIST OF FIGURES—Continued
Figure Page
4.2 Time-of-flight spectrum obtained by using coincidence measurement
and the corresponding energy spectrum of the desorbing 37Cl ion................ 81
4.3 Comparison of ToF spectra between the Monte-Carlo
simulations and the experimental results...........................................................82
4.4 Using retarding field energy analyzer to measure
charge state distribution....................................................................................... 85
4.5 Retarding field energy spectrum of the desorbing 37Cl ion.............................86
4.6 Calculating charge state distribution from the RFE spectrum........................87
4.7 Comparison of charge state distribution and the REF spectra
between the Monte-Carlo simulation and the experimental results............... 89
4.8 Comparison of the maximum kinetic energy obtained by
ToF and RFE spectra............................................................................................ 90
4.9 RFE spectra of desorbing 37Cl ion at different coverage.................................92
4.10 Comparison of RFE spectra, total ion counts and maximum
kinetic energies among different mixtures of radioactive and stable Ar...... 93
4.11 Comparison of energy distribution of desorbing
37Cl ion from different coverage.............................................................. .......... 95
4.12 KLL and LMM Auger electron-electron coincidence ToF spectrum............98
4.13 Local environment of 37Cl ion............................................................................ 100
4.14 The simulation results of Cl+ ion exchange charge with metal surface.........104
4.15 The model explaining charge exchange and Coulomb
. repulsion during 37Cl ion desorption...................................................................106
4.16 The RAE spectrum of 37Cl LMM Auger............................. .............................110
xiii
LIST OF FIGURES—Continued
Figure Page
4.17 Comparison of LMM Auger spectra among 37Cl following
EC decay, Cl from CsCl and physisorbed 40Ar.................................................I l l
4.18 37Cl LMM Auger spectrum measured with new radioactive source.............. 113
4.19 Comparison between 37Cl LMM Auger spectrum and Cl
conventional Auger lines..................................................................................... 114
4.20 The Auger cascade model explaining the LMMh peak...................................116
4.21 Spectrum of coincidence measurement between
KLL Auger and LMM Auger.......................... ................. .................................. 119
4.22 Spectrum of coincidence measurement between two .
Double Auger electrons...................................................................:................... 120
4.23 The RFE spectrum of 37Cl LMM Auger by the MCP detector.......................123
4.24 Spectra of coincidence measurement for KLL and LMM
Auger electrons with different retarding screen voltages........:...................... 126
4.25 Spectra of coincidence measurement for double Auger
electrons with different retarding screen voltages........................... ................128
4.26 Double Auger energy distribution......................................................................130
4.27 Differential energy dependence of the double Auger probability as
compared with theoretical prediction........................................................... .....131
4.28 Double Auger energy distribution measured with the new source..........;.....132
4.29 Schematic diagram of a triple coincidence spectrum....................................... 134
4.30 . Comparison of double Auger coincidence spectra between
experimental and simulation results......................... .........................................138
xiv
ABSTRACT
A novel experiment was developed in this work to study the desorption and the 
Auger relaxation processes of 37Cl following the 37Ar electron capture decay. For the first 
time, the desorption of 37Cl ions due to recoil induced by neutrino emission in this decay 
process was observed. The kinetic energy distribution of the desorbing 37Cl ions was 
accurately measured by using coincidence techniques. The resulting 37Cl ion energy 
ranges from 5 eV to 13 eV with a maximum at around 9 eV and an FWHM about 3 eV. 
The charge state distribution of the desorbing 37Cl ions was also measured. The resulting 
charge state distribution is: 53% of the total ions have charge +e, 21% have charge +2e 
and 26% have charge +ne, where n > 3. The desorbing probability of 37Cl ions was 
measured by two independent experiments which gave the result of 9.4+1.2%. The energy 
distribution, the charge state distribution and the desorbing probability of 37Cl ions are all 
quite different as compared with the expected values for an isolated Cl atom. These 
differences are explained in the desorption model involving charge exchange and 
Coulomb repulsion between 37Cl ions and their surrounding atoms.
The electron capture decay also creates a highly unusual initial state in the 37Cl 
atom which allows direct observation of some novel relaxation processes which are. 
amenable to many body theory, but essentially impossible to probe experimentally with 
conventional techniques. For the first time, direct evidence of the double Auger decay of 
a K-hole and the large shift in energy (22 eV) of an LMM Auger line was reported. The 
double Auger decay probability and energy distribution of the two double Auger electrons 
were measured by using coincidence techniques. The resulting double Auger decay 
probability ranges from 12+0.3% to 15+0.4% of the total Auger decay, The preferred 
energy distribution of the double Auger emission is for one of the electrons to take most 
of the energy, with the second receiving the small remaining balance! A model explaining 
the large shift in energy of an LMM Auger line is also provided.
ICHAPTER I
INTRODUCTION
On February 23, 1987, a supernova burst in the Large Magellanic Cloud 
(SN1987a) and sent rays of neutrinos to the earth. The sudden release of the IO53 erg of 
energy from the dying star triggered merely 100 or so “neutrino” counts in huge 
underground detectors around the world. This, however, caused enough excitement for 
several groups of physicists— the burst of SN1987a gave them a golden opportunity to 
study the mysteries of neutrinos.1,2 Among these mysteries, the biggest one is whether 
neutrino has a rest mass or not, and what that mass might be. The neutrino mass problem 
is one of the most fundamental questions in physics. The answer to this question could 
have profound implications for particle physics, astrophysics and cosmology.3
The mass information of a neutrino can be obtained from a nuclear reaction 
involving the creation or annihilation of a neutrino.1'3 One of these kinds of nuclear 
reactions is the electron capture (EC) decay. In this thesis, the desorption due to recoil 
induced by neutrino emission and the Auger relaxation processes of 37Cl following the 
EC decay of 37Ar is reported. It is demonstrated in this work that the EC decay of 37Ar 
provides not only a possible approach to the neutrino mass problem but also a rare
2opportunity for some novel studies in surface and atomic physics.4"7 The focal point of 
this thesis is to report the results of these novel studies.
The problems addressed in this thesis work fall into two categories: The first one 
is the study of the desorption due to recoil induced by neutrino emission in the EC decay 
process, 37Ar —> 37Cl + v, where V is a neutrino. In this EC decay process, could the 
emission of the neutrino cause enough recoil o f the adsorbed daughter atom 37Cl to break 
the absorption bond? If the desorption does occur, what factors would affect the energy 
distribution of those desorbing daughter atoms? The second category is the study of the 
Auger relaxation processes of the 37Cl daughter atom. The EC decay creates a deep 
vacancy in the energy levels of the 37Cl atom while the whole atom is still in a neutral 
charge state. As discussed later, this initial state is highly unusual as compared with those 
commonly encountered initial excitations. What are the relaxation processes or the decay 
channels associated with this particular initial state? To better understand these questions, 
the concepts of these studies will be introduced first in the following paragraphs.
Concepts of the Work 
Desorption due to Recoil Induced by Neutrino Emission
The first part of the study is to measure the spectrum of recoil velocities of 37Cl 
ions following the EC decay of 37Ar physisorbed on a well defined surface. In this EC
3decay process which converts 37Ar into 37Cl, a proton in the nuclei of 37Ar captures a Is 
or 2s electron and becomes a neutron, while a neutrino is emitted and the 37Cl daughter 
atom recoils in the direction opposite to that in which the neutrino moves. This recoil may 
break the bond between the 37Cl atom and the substrate and cause desorption of the 37Cl 
ions from the surface (Fig. LI). The initial Is or 2s vacancy created by the EC process 
will decay through Auger cascades or x-ray emission. Therefore the daughter 37Cl atom is 
usually multiple ionized after the EC decay.
Figure 1.1. (a) Electron is captured from a low (Is) orbital: 37Ar —» 37Cl + v. (b) 37Cl ion 
recoils and desorbs from surface due to neutrino emission.
For an isolated 37Ar atom, the recoil process is a two-body problem with a total 
reaction energy, <2 = 814 keV (Ref. 8). Energy and momentum conservation dictates that
4the recoil energy must be 9.54 eV, if a massless neutrino with about 814 keV of energy is 
emitted. If a neutrino with mass mvis also emitted in a certain fraction of the EC decays, 
then the corresponding recoiling ions will have lower kinetic energy. The fractional 
change in recoil kinetic energy is AE/E = (mJQ)2 (Ref. 3), where Q, the total reaction 
energy is the rest mass (me2) difference between 37Ar and 37Cl. Thus the heaviest massive 
neutrino that could be emitted in this EC decay would have a rest mass of 814 keV. If a 
250-keV neutrino were to be emitted, for example, the fractional change in recoil energy 
of the 37Cl would be 9.4%. The lowest neutrino mass to which this method could be 
sensitive depends on the ability to resolve the peak associated with the emission of the 
massive neutrino from the dominant peak associated with the emission of the massless 
neutrino in the recoil spectrum.
There are well known and calculable effects that would contribute to the 
broadening in the recoil energy spectrum for an isolated 37Ar atom. Some of these effects 
can be studied by computer simulations. However, for an 37Ar atom physisorbed on a 
cold surface, and possibly surrounded by other Ar atoms, it is not obvious whether the 
recoil due to neutrino emission could break the bond between 37Cl atom and the substrate 
and cause desorption. Even if the desorption does occur, it is still much more difficult, as 
compared to the isolated 37Ar case, to determine what effects would contribute to the 
broadening in the recoil energy spectrum. To approach these questions, it is necessary to 
examine the surface interactions during the 37Cl ion desorption process.
5Surface Interactions of 37Cl Ion
Adsorption and stimulated desorption have been very useful tools to study 
surfaces. Since the 1960s, several models have been developed to describe the 
interactions between adsorbates and the substrates. Current models involve detailed 
knowledge of surface binding energies, adsorption site symmetries and potentials, 
substrate phonon spectra, and surface electronic structures.9"12 The neutrino recoil in this 
work provides an entirely new way to study the desorption dynamics. A simple model of 
the desorption in this case is described as follows: First, the adsorbed 37Ar atoms undergo 
EC decay and become 37Cl ions with 9.54 eV recoil energy to move away from the 
surface. Before these 37Cl ions can leave the surface, the interactions between the 37Cl 
ions and the substrate will result in additional kinetic energies on the top of the 9.54 eV 
recoil energy. These additional energies will contribute to the broadening of the recoil 
energy spectrum. One of the major goals in this work is to understand these interactions.
The first interaction to be discussed is the bonding between 37Ar and the cold 
substrates. The 37Ar atom is physisorbed on gold or graphite substrates. At the critical 
temperature of 65 K Ar begins to form a monolayer on these substrates. By about 30 K, 
multiple adsorbed layers (Ar ice) are formed.13'15 It has been found that these multiple 
adsorbed layers are likely to form a face-centered cubic (fee) polycrystalline layer at the 
experimental temperature of 16 K.16 Therefore the adsorbed Ar could actually form a 
band structure and solid state physics has to be taken into account. The bonding between
6an 37Ar atom and the substrate or the neighboring Ar atoms is established through the van 
der Waals interaction which gives rise to a binding energy of about 100 meV.15 This 
interaction cannot play a major role in the broadening of the recoil energy spectrum 
because the 100 meV interaction energy is much less than the 9.54 eV recoil energy.
(a) *   ^ (b)
Figure 1.2 (a) 37Cl ion obtains 9.5 eV recoil energy from neutrino emission, (b) Charge is 
transferred and Coulomb repulsion gives additional energy 8E to the outgoing 37Cl ion.
The second interaction to be discussed is the charge exchange between the 37Cl 
ions and their local surrounding atoms.17"21 Theoretical calculations and experiments have 
shown that in the gas phase the EC decay of 37Ar will give rise to dominantly 37CI3+, the 
positively threefold-charged ions.22' 23 It is energetically more favorable for the 37CI3+ ion 
to pick up an electron from the neighboring particle, such as an Ar atom, during the initial 
phase of desorption.24 These charge exchanges, however, could significantly broaden the
7recoil energy spectrum. The positively charged 37Cl ions and their positively charged 
neighboring atoms will interact via the Coulomb repulsion giving additional contributions 
to the kinetic energy of the outgoing 37Cl ions (Fig. 1.2). In some cases, the additional 
energy due to this “Coulomb explosion”25 could reach as high as 6 eV. The charge 
exchange process and Coulomb repulsion are the main causes of the broadening of the 
37Cl recoil energy spectrum.
The charge exchange process can also cause an outgoing 37Cl ion to lose all of its 
positive charges and become a neutral atom while desorbing from the surface.26"28 The 
experimental setup in this work can only detect charged particles. An estimation of the 
neutralization cross section and ion desorption yield per electron capture decay will be 
given from the experimental and computer simulation results.
There are other interactions that affect the energy distribution of the 37Cl ions 
desorbing from the substrate. Two examples of them are the thermal vibration of the 37Ar 
atoms on the substrate and the recoil of the 37Cl ions due to Auger electron emission. The 
physical and chemical characteristics of the substrates can also play an important role in 
affecting the energy distribution of the 37Cl ions. Some of these effects will be calculated 
using computer simulation to compare with the experimental results.
A full treatment of each of the interactions discussed above is beyond the scope of 
this work, rather, the emphasis is put on the study of the “Coulomb explosion”
8phenomenon which is of great significance for the understanding of this particular 
desorption process. Because this interaction is the dominant factor in the broadening of 
the recoil energy spectrum, it also becomes a major obstacle to the neutrino mass limit 
measurements. The broadening of the recoil spectrum reduces the sensitivity of the 
neutrino mass limit measurements in this experiment. However, through the present 
studies the major factors which are responsible for the broadening of the recoil spectrum 
are understood. Future experiments will be designed with reduced charge exchange 
effects.
Auger Relaxation of 37Cl atom
The electron capture decay of the 37Ar atom creates a highly unusual electronic 
configuration in the daughter 37Cl atom which is not encountered in ordinary non-nuclear 
physical process. Conventional excitations utilizing electron or photon bombardments 
generally leave the atom in an ionized initial state. However, in the EC decay a proton 
and an electron become a neutron, hence the total net charge of the atom is unchanged. 
The daughter 37Cl atom remains neutral, even though there is a vacancy in its inner shell 
— the hole resulting from the capture of a electron. As shown in Fig. 1.3, this initial state 
is unusual in three ways. First, the electronic configuration of the initial state in 37Cl atom 
is like Cl in energy but resembles Ar in structure, Le., each energy level of the 37Cl atom 
is relaxed from the Ar level to Cl level within the decay time but all levels except Is are 
still fully occupied with electrons, as in the Ar case. Second, this initial state is a neutral
9state with a hole in the K-shell as compared to the ionized initial state obtained by 
conventional means. Third, with 90% of the electron capture occurring from Is orbital, 
the initial state in the 37Cl atom is almost a pure Is vacancy state, while the initial states 
obtained by conventional means are mostly 2s and 2p vacancy states. This unique initial 
state provides valuable opportunities to study some atomic relaxation processes that were 
feasible in theoretical studies but impossible to probe with conventional experiments.
ionized Ar+
0  8  Q Q 0  O 
3s__m_________is____
2p-e@@
3s q
Is ----------------e —
(a)
ionized Cl+
(b)
n e u t r a l  Cl
S 0 Q 0 0 O
—8 -------------0 -----
0 O O O O C
— © -----------0—
- —• ----------e —
(c)
Figure 1.3 (a) and (b) Ionized initial state of Ar+ and Cl+ obtained by conventional means, 
(c) The unusual 37Cl initial state obtained from electron capture decay.
There are two interesting relaxation processes for the unusual initial state studied 
in this work. The first one is the LMM and LLM Auger relaxation of this initial state. In 
an Auger process, the Auger peak energies and structures (splitting, satellite, etc.) are 
governed by the electronic configuration of the element studied. Therefore, the LMM and 
LLM Auger peaks of the 37Cl atom are expected to be like Cl Auger in energy but
10
resemble Ar Auger in structure. One purpose of this study is to look for possible shifts 
and new features in the Auger peaks and to compare them with known Cl and Ar Auger 
peaks.29"39 The second relaxation process studied in this work is a particular phenomena 
called double Auger decay where the filling of the Is hole causes two Auger electrons to 
be emitted simultaneously (Fig. 1.4). While having been predicted by theory, this 
phenomena has never been directly observed.4,40 43 Therefore another goal of this study is 
to measure the probability of the double Auger process and the energy distribution 
between the two double Auger electrons.
Figure 1.4 (a) Normal Auger transition, (b) Double Auger transition.
Historical Review of Related Works
Because this work straddles several disciplines including nuclear physics, surface 
physics and atomic physics, there is little that has been published to serve as a precedent.
11
Therefore, the review of the references is divided into each aspect of this research work: 
the 37Cl ion recoil study, the desorption model study and the Auger relaxation study.
The first measurement of the recoil velocities of 37CI ions in the gas phase was 
reported in 1952 44 Results of experimental22’23 and theoretical45 studies have since been 
reported on the recoil momentum and charge distribution of 37Cl ions. The measurements 
of orbital electron capture ratios in this system have also been reported.46,47 However, not 
much experimental work on the recoil velocities of 37Cl ions has been reported since the 
1950s. Previous experiments in the gas phase lacked the necessary resolution in the recoil 
spectra to search for a massive neutrino because of the effect of thermal velocities of 37Ar 
atoms and the large size of the gas region seen by detectors. In this work 37Ar atoms are 
localized by physisorbing them onto a well defined surface at a low temperature under 
ultrahigh vacuum conditions, where well known surface and nuclear physics techniques 
can be utilized to measure the recoil energies of the 37Cl ions as well as other atomic and 
electronic properties of the surface species. The resolution of this work is expected to be 
much better than those in the gas phase experiments.
When an electron or a photon beam is incident on a surface with adsorbed atoms, 
electronic excitations will be generated by the interactions between the electrons or 
photons and the surface bonded complex, Ie., the absorbates and the substrate surface. 
The decay of these electronic excitations may sometimes cause desorption. This is called 
stimulated desorption}1 The studies on stimulated desorptions have become the focus of
12
increased activities over the past three decades. The results and ideas generated in this 
rather specialized area of surface science have had an important effect on surface science 
and other disciplines.9'12 Most of the existing work focuses on the study of desorption 
stimulated by low energy electrons (ESD) and photons (PSD). In 1964, Menzel, Corner48 
and Redhead49 developed the MGR theory to explain ESD on metal surface. ICnotek and 
Feibelman50 developed the K-F model in 1978 for the desorption in ionic systems. In 
1979, Menzel and Franchy25 explained the desorption due to Auger-like final states in 
covalent materials where the Auger decay of a core hole results in a final state with two 
or three valence holes. “Hence, we can envisage a ‘Coulomb explosion’ as observed in 
gas phase91 molecules”, they reported. Their report is very helpful in establishing the 
model for the desorption process in this research work. A comprehensive review of the 
stimulated desorption was given by Knotek in 1984.11 There have been many papers 
reporting PSD and ESD studies of rare gases on metal or graphite surfaces by MenzeTs 
group 51'60 and other research groups.61"69 The reason for studying these desorption 
models is to develop a new model to understand the broadening of the recoil energy 
spectrum of the 37Cl ions in the desorption process.
The Auger process was first observed seventy years ago. However, until the 
application of the lock-in amplifier by Harris in 1967, Auger signals were very hard to 
detect due to the small signal to noise ratio. Since the 1970s, Auger electron spectroscopy 
(AES) has become one of the most important surface analysis techniques.70"75 The 
theoretical and experimental aspects of Auger emission have been comprehensively
13
reviewed by Sevier75 and Bambynek.70 The K shell, L shell, M shell and Coster-Konig 
Auger transition rates for most atoms have also been calculated.70,74,76 There have been 
many reports on Ar and Cl related Auger studies.29"39,77,78 In 1973, Siegbahn reported the 
LMM Auger study of Ar in the gas phase where they observed more than eighty peaks in 
their high resolution experiment.29 Among these peaks, they attributed more than thirty to 
known electronic transitions. Their results were used as the reference energies for this 
research work. There are also results of the KLL Auger energies of Ar.79"81 For the Auger 
decay of the initial state in this work, another good reference is the theoretical studies by 
Omar and Hahn, where they calculated the Auger lines produced by the decay of Ar ions 
with a deep Is hole.29'34 The first evidence of the double Auger decay process was 
reported by Krause’s group in 1984.40 Some other indirect evidence of this process was 
reported by Becker et al in 1989.41 One of the few theoretical studies on the double Auger 
decay process was reported by Amusia et al. in 1992.4 The experimental studies reported 
were concentrated on measuring the charge state distribution of the ions in the gas phase 
following a double Auger decay.82,83 For example, the ordinary Auger decay of a neon 
atom in the gas phase will result in a doubly charged Ne2"1" ion, therefore, the observation 
of triply charged Ne3+ ions would provide an indirect evidence of the double Auger decay 
process. There has not been much work on the direct measurements of the electrons 
emitted in the double Auger process.
14
CHAPTER 2 
THEORY
Recoil Induced by Neutrino Emission
In today’s theory of elementary particles, neutrinos belong to the family of 
leptons. There are three kinds of neutrinos, electron-type neutrino, muon-type neutrino 
and tau-type neutrino. They are involved in the nuclear reactions where electrons, muons 
or taus are created or annihilated.84"86 A brief induction of neutrino physics is provided in 
Appendix A. In this section, the emphasis is put on calculating the recoil energy of the 
37Cl ions due to the emission of neutrinos in the electron capture reactions.
The object of this study is the 37Ar isotope, which undergoes spontaneous electron 
capture decay (37Ar -> 37Cl + Ve ) with 814 keV total reaction energy.8 The calculation in 
this section assumes that the electron capture decay is an ideal two-body problem which 
can be found in many physics text books. The conservation of energy and momentum 
dictates the recoil energy of the 37Cl ions to be 9.54 eV if a massless neutrino is emitted 
in the EC decay. If neutrinos with different masses were emitted in this reaction, the 37Cl
15
ions would recoil with different kinetic energies. The recoil energy of the 37Cl ion as the 
function of the neutrino mass is calculated below using relativistic kinematics.
The first step is to calculate the 37Cl ion recoil energy, E  assuming neutrino is 
massless (mv = 0). Using natural units, i.e., assuming speed of light, c = I and the Plank 
constant, h =  2tz, the neutrino total energy Ev has the form
Ev = Pv (2.1)
where Pv is the momentum of neutrino. The 37Cl ion recoil energy, E  can be written as 
E  = P2/2M  (2.2)
where P and M  are 37Cl momentum and mass, respectively. Conservation of momentum
and energy requires that
E+  Ev = Q, P = Pv (2.3)
where Q = 814 keV is the total reaction energy. Therefore,
E + P = Q, P 2 = E 2 - 2 E  Q + Q2 (2.4)
From Eq. (2.2) and Eq. (2.4) one can get
E 2 -  2(M + Q)E + Q2 =0 (2.5)
Since E «  Q « M ,  from Eq. (2.5) one can easily get
E s  Q2/2M  = 9.54 eV (2.6)
The next step is to calculate the 37Cl ion recoil energy, E', assuming that neutrino
has non-zero rest mass. The total energy of the neutrino, Ev', now has the form
16
'  ' 2 ,
Ev = ^P v +mv ■ (2.7)
where Pv and mv are the neutrino momentum and rest mass, respectively. Let P 'be the 
momentum of the 37Cl ion; then conservation of energy and momentum requires
E /+  J?'== f / =  f"  (2.8)
From Eq. (2.8) and Eq. (2.9) one can get
E'=-2(M  -H(Z)JE'-H Cf = m," (2.9)
The condition, E '«  Q « M  still holds; therefore, Eq. (2.9) becomes
E ' = (Q 1 - mv ) / 2  M  (2.10)
Comparing Eq. (2.6) with Eq. (2.10), the change of recoil energy, AE, due to the neutrino 
mass is
AE = E - E '=  /2  M  , (2.11)
Finally, from Eq. (2.6) and Eq. (2.11), one can get the percentage change of the recoil 
kinetic energy of 37Cl ion, AE /E  due to the neutrino rest mass mv
AE/E =( mv/Q f  (2.12)
From Eq. (2.12), it is easy to see that a neutrino with 250 keV mass (E=mc2) would result 
in a percentage change, AE /E  =9.4% in the recoil kinetic energy of the 37Cl ion. This 
change corresponds to the difference of recoil energy from 9.5 eV to 8.6 eV. Such a 
change should be easily detectable in this experimental setup. Unfortunately, all of the 
above calculations are based upon 37Ar atom being isolated. In reality the 37Ar atom is 
physisorbed on a cold substrate where charge exchange and other interactions will take
17
place. These interactions will broaden the recoil energy spectrum so that the recoil energy 
peak overlaps the energy difference, AE due to the neutrino mass. The next section is 
devoted to the discussion of these interactions.
Models on Stimulated Desorption
One of the primary goals of surface science is to understand the surface bond. The 
wide array of spectroscopic tools employed in this pursuit include the detection of 
electrons leaving the surface after excitation by radiation in the form of electrons, photons 
or ions. The surface bond or the local environment of an atom is deduced from the energy 
and angular distributions of these emitted electrons.71'75’87 A completely different 
perspective can be obtained from the analysis of the atomic or molecular species leaving 
the surface as ions or neutrals. Techniques based on the detection of desorbing ions 
include electron stimulated desorption and photon stimulated desorption. These process 
are also called desorption induced by electronic transitions (DIET).9'12’48-69
The electronic excitation leading to ion desorption can be a one-electron 
excitation, or much more complex and interesting, a multiple ionized and excited state. 
The dominant mechanism for creating these multiple excited states is the excitation of a 
core hole in the surface-bonded complex. The decay of these core-hole states results in 
the creation of multiple valence-hole final states which are much more efficient at 
inducing desorption than the one-electron excitations.25 The multiple valence-hole states
18
are intrinsically localized, contain a large amount of energy and have much longer life 
time than the normal electronic life times on surfaces.25 In this experiment, the EC decay 
of 37Ar produces not only a 9.54 eV recoil energy but also a highly excited core-hole state 
in the 37Cl atoms. Just as in the DIET case, the decay of this initial state will provide 
additional energy to the 37Cl ions. Therefore, in order to understand the complete recoil 
energy distribution, it is very important to understand these desorption mechanisms.
The Menzel-Gomer-Redhead (MGR) Model48,49
The MGR mechanism for desorption from surface is based on the elementary 
consideration that kinematic processes, i.e.,- direct momentum transfer, are insufficiently 
energetic to cause desorption from surface. Hence, desorption must be due to an 
electronic excitation of the surface bond. The desorption can be envisaged as a two-step 
event, so that the desorption cross section has the form
O=O eP  (2.13)
where Oe is the primary excitation cross section and P is the escape probability. As shown 
in Fig. 2.1, the primary excitation is a vertical, Frank-Condon-Iike excitation from the 
bonding state (M+A) to the repulsive antibonding state (M+A)*, where M stands for the 
metallic substrate and A stands for the absorbed atom. This primary excitation is followed 
by an evolution of both electronic wave functions and nuclear coordinates along a new 
potential energy curve. The transition from (M+A)* to (M+A) is through the infinite set of 
states (M*+A) that differ from the (M+A) only by an excitation of the substrate, which.
19
(M+A),c a p tu re
(M+A),deso rp tion
Figure 2 .1 The energy level scheme of MGR model. The dash curve (M*+A) intercepts 
the antibonding curve (M+A)* at a critical distance of xc.
does not affect the bonding. Hence the curve (M*+A) is identical to the (M+A) but raised 
in energy. The inverse of this de-excitation process, i.e. (M*+A) -> (M +A /->  desorption, 
while energetically feasible, is highly improbable because the excess energy in (M*+A) is 
de-localized in the substrate and its re-localization in (M+A) bond is very unlikely, 
whereas (M+A) —» (M+A)* implies an initial localization of the energy to be dissociated. 
Also as shown in Fig. 2.1, there is a critical distance xc beyond which recapture cannot 
occur because the adsorbate A has sufficient energy to escape in the (M*+A) state. The 
critical distance xc is generally found to be < I A. Menzel and Corner first treated the
escape probability P in Eq. (2.13) as
20
x O v(x) z(x) y
'Xe dx
(2.14)
where t:(x) and v(x) are the lifetime and velocity at position x. Since the reneutralization 
or de-excitation process involves resonance tunneling, it displays this exponential 
dependence on the distance from surface. P  can then be approximated by
where % is assumed to be a constant from x0 to Jtc and At = [2m (xc -X0 ) /  Sr ]m , where Sr 
is the slope of the repulsive curve. This then reduces to
where M  is the molecular weight of the desorbed species and c is a constant. The energy 
spread of the desorbing ions can be approximated as
where Ax is the width of the ground state vibration wave function of the adsorbed species.
hi 1980 Antoniewicz proposed a modification of the MGR model which 
recognized that if a species bonded on a surface is ionized, the attractive potential 
between the positive ion and its image will draw it to the surface. Ionic and neutral 
desorption then can occur by a mechanism illustrated in Fig. 2.2.88 Initial excitations 
occur from (M+A) to some curves lying above the ionic curve (e.g. (M"+A+)*). De­
excitation from (M'+A+)* to (M>A+) can then result in ion desorption (Fig. 2.2a). A 
much more obvious application of this model is to physisorption, where there is no
P = exp ( -At /  t ) (2.15)
P = exp ( - c Mm  ) (2.16)
AE = Sr (xq) A x (2.17)
2 1
bonding structure between A and the surface. Here an excitation from (M+A) to (M>A+) 
again results in the ion being drawn toward the surface. Upon reneutralization and de­
excitation back to (M +A) curve, the adsorbate is far enough up on the repulsive part of 
the curve to desorb. The ion must be drawn to the surface to a distance X<XC for 
desorption to occur (Fig. 2.2b).89,90
&T+A
(M+A),cap tu re
(M+A),desorp tion
M"+A
(a) (b)
Figure 2.2 Energy level schemes of Antoniewicz Model, (a) Positive ion desorption, (b) 
Neutral desorption from a physisorption system.
There are some important points derived from these models. First, there exists a 
threshold for the primary excitations. ESD only happens when the primary electron 
energy is greater than the threshold. Second, there is a much larger ESD yield for neutrals 
than for ions. This is because the cross section, Oe is larger for neutral ESD than ion BSD, 
and the escape probability P, in general, is much smaller for ions than for neutrals, since
22
the processes to neutralize an ion are easier than those to recapture a neutral specie. All of 
the models discussed above involve only one-electron excitations of the surface bond. 
Models discussed in the next section will involve more complicated multielectron excited 
final states.
The Knotek-Feibelman IKFl Model50
There is a second general mechanism for desorption, which involves the 
excitation of a core level followed by the decay to complex desorptive final states rather 
than the one-electron excitations found in the MGR model. While it has since been 
generalized to a wide range of adsorption system, the original idea was developed to 
apply to the desorption of positive oxygen ions (O+) from TiO2, where oxygen is bonded 
in the O2" state. The desorption process implies a large charge exchange where three 
electrons have to be transferred from the O2 state to the O+ state.
TiO2 is called a maximal valency ionic compound. Maximal valency means that 
the cation is positively ionized to the noble-gas configuration (in NaCl , for example, Na 
loses the only one of its valence electron and becomes Na+) and the anions are negatively 
ionized to the noble-gas configuration {e.g. Cl gains one electron and forms Cl"). The 
highest occupied level of the Ti+4 ion is the Ti(Sp) core level at 34 eV below the 
conduction band and the valence band of TiO2 is almost exclusively 0(2p) in character. 
Figure 2.3 shows a simplified schematic of the process leading to desorption of O+ from a
23
ground state of Ti4+O2 2 configuration. As stated above, three electrons must be 
transferred from O2 to yield O+ desorption. If the primary ionization radiation removes 
the electron from the Ti atom’s shallowest core, the predominant core-hole decay will be
Auger e lec trons
Conduction band
Ferm i level
valence band
(a)
Maximal valency
Subm ax im al valency
C+ A" (b)
Figure 2.3 (a) The Knotek-Feibelman model for O+ ion desorption from TiO2. The initial 
hole in Ti(3p) is filled by an interatomic Auger transition in which the O2 loses three 
electrons in its valance band and become the O+ to desorb from the surface, (b) 
Comparison between maximal and submaximal valency compounds. Maximal valency 
results in selective charge transfer from anion A, leading to desorption. In submaximal, 
charge transfer is shared between cation C and anion A, quenching desorption.
24
an “interatomic” Auger process because the higher-laying electrons are lacking on the Ti 
atom. Thus, one valence electron from O2" falls into the core hole and one or two 
electrons will be emitted from oxygen to release the energy of the decay. The loss of two 
electrons due to Auger transition and shake-offs (which can occur up to 10% of the time) 
from the oxygen orbital,50 plus the third electron which fills the Ti(3p) hole transforms 
the oxygen to an O+. The potential in which the O2" is located before the excitation and 
Auger sequence is the sum of an attractive Madelung term and repulsive core overlap 
contribution. In its negative state, the atom is in a stable position defined by the minimum 
in the total potential. However, if  the charge on the oxygen changes sign, the Madelung 
term becomes also repulsive. Thus the oxygen is now in a total repulsive potential and is 
driven to desorb if it is on the surface.
Another way to cause the desorption of O+ from TiO2 is to excite anions (O) 
instead cations (Ti). In fact, by exciting the M22, M1, L23 and L1 levels of titanium or L1 
and K levels of oxygen one can get different yields of O+ depending on the energy 
available from the core-hole decay and the details of the decay process. Energy 
conservation becomes important when the core hole is shallow enough that not enough 
energy is released by decay of an electron from the top of the valence band, or highest 
valence level, to transfer the necessary charge to cause desorption. Now consider the 
ionization of the anion A" to an A0 and subsequently to an A+. The first electron drops 
into the core hole leaving a single valence hole on the anion (A0) and releasing an energy, 
E. This energy is then available to ionize the A0 to A+. If the anion has an ionization
25
energy, /, and the solid has an electron affinity, A, then the available energy, E, must 
satisfy the condition
E>  I  - A + T ' (2.18)
■ where T  is the extra energy required to remove an electron out of the Madelung potential.
" The energy T  is converted into kinetic energy of the ions as it leaves the surface. If the 
anion was originally an A2 , then the loss of a valence electron leaves an A". Removal of a 
second electron requires an energy equal to the gap energy, Eg plus the screened hole-hole 
repulsion energy, U. The subsequent A0 A+ proceeds as before, so the m inimum 
energy condition is
E > Eg + U + (I - A + T) (2.19)
A good approximation to an unscreened value of U for an atomic species is the difference 
between its first and second ionization potential
U = I2 -Ii  (2.20)
These conditions are very important for modeling charge exchange and Coulomb 
repulsion of the desorbing 37Cl ions with their neighbors in this work.
Desorption Due to “Coulomb Explosion” 25
While the original formulation of the Auger model was to explain the desorption 
of ions from ionically bonded surfaces, it was soon demonstrated that ions could be 
desorbed from covalently bonded surface complexes by essentially the same mechanism. 
In the simplest analysis, the product of an Auger decay of a core hole in a covalent system
26
is a two or three valence-hole final state. Hence, one can envisage a “Coulomb explosion” 
as observed in gas-phase91 molecules, where the highly repulsive final state results in the 
production of ionic fragments of the parent species. The presence of multiple valence 
‘holes in a covalent system can result in a repulsive interaction between the unscreened 
nuclei and subsequent production of ion fragments. The multiple holes thus represent 
energy stored in the bond. When two holes exist on an atom or in a bond the most 
obvious way to relieve the large repulsive energy is for one hole to hop away before 
nuclear motion can occur. Typical uncorrelated one-hole hopping times are of the order of 
IO"16 s, whereas desorption times are of the order of IO"13 s. So the life time of hole is too 
short to affect the motion of nuclei. However, one-hole hopping is often slowed by hole- 
hole correlation which can block resonant one-hole hopping process. The importance of 
hole-hole correlation was first pointed out by Cini92-and Sawatsky93 in-explaining the 
existence of atomic-like Auger spectra in marrow d-band metals. This process has also 
been observed in insulators94 and gas-phase molecules.95 This many-body effect can make 
the Auger-induced desorption process effective for covalent materials as well as ionic 
materials.96"100
Now consider a simple Hubbard-like Hamiltonian, as in Eq. (2.21), describing an 
electron in a system with bandwidth W. When two holes are created on an atom or in a 
bond, the correlation energy between the two holes is given by Ueff., which can have 
important contributions due to screening.92,93'95
27
# = ( 2 . 2 1 )
io  ija
Figure 2.4 displays the terms in this Hamiltonian schematically. Typical one-hole hopping 
times in such system are of order o f f  = 1/W (Fig. 2.4 (a)). If Ueff- > W, resonant one-hole 
hopping is essentially blocked and T »  1AV, because it involves either energy transfer 
during the transition or a complicated multiple-hole motion (Fig. 2.4 (b)).92,93 The highly 
repulsive Auger final state is thus given an intrinsically long life time due to this
(c )
Figure 2.4 Schematic displaying terms of a Hubbard Hamiltonian with intrasite Coulomb 
term Ueff. (a) Single valence hole T ~ 1/W. (b) Two valence-hole states with Ueff>  Whave 
one or two hole hopping T >> 1/W. (c ) Presence of unfilled orbital can provide fast 
charge transfer channel hence reduces the chance of desorption.
28
correlation, possibly of the order of IO2 times the normal one-hole lifetime, which allows 
sufficient time for nuclear motion to relieve the energy. Therefore, factors which increase 
Ueff and /or decrease W  will enhance desorption.
Ueff contains both an intrasite Coulomb repulsion term and a screening term.95 The 
intrasite (or intrabond) repulsion term is roughly approximated for atoms by the 
difference between the second and first ionization potential (Eq. (2.20)) and is inversely 
proportional to the ionic radius. The screening term is enhanced by the presence of 
unsaturated bonding, which allows very efficient charge transfer to occur (as shown in 
Fig 2.4(c)), hence reduces the chance of desorption.92"95 The transfer of a charge onto the 
atom from a neighbor is equivalent to the hole being transported from screened site, both 
from the standpoint of energy and total charge.
In this work, the 37Cl ion has an initial state with one core hole which will also 
result in multiple-hole final states. One of the channels to release the trapped energy in 
these multiple-hole states is the hopping of one of the holes from 37Cl ion to its neighbor, 
or equivalently, transferring one electron from the neighbor to 37Cl ion. Further hopping. 
of the holes may be blocked because of energy restriction, e.g., a single charged 37CI+ ion 
cannot exchange charge with a neighboring Ar atom because the first ionization potential 
of Cl is less than that of Ar. Therefore, there is enough time for the 37Cl nucleus to move. 
The ionized 37Cl atom will obtain the kinetic energy through Coulomb interaction with its 
neighbor ions. The details of this model will be developed later.
29
Theory of Auger Transition
Auger Effect
The inner shell vacancy of an atom with multiple electrons can be created by 
photon irradiation,87 energetic electron irradiation70'75 or in this work, an electron capture 
process. Two of the channels the excited initial state can decay through are the radiative 
transition where an electron in a higher energy level jumps down to the vacancy and 
releases energy in the form of a photon, and the non-radiative transition where an electron 
in higher energy level jumps down to the vacancy with the energy transferred to another 
electron to escape from the atom. The latter transition is called Auger effect.70"75
Table 2.1 Number of Electrons, Symbols and Energies (eV) for Ar and Cl Core Levels24
n I 2 . 3
I 0 0 I 0 I 2
j 1/2 1/2 1/2 3/2 1/2 1/2 3/2 3/2 5/2
# of e 2 2 2 4 2 2 4 4 6
Symb K Li L2 L3 Mi M2 . M3 M4 M5
E at 3202.9 320.0 247.3 245.2 25.3 12.4 12.4
E ci 2822.4 270.2 201.6 200.0 17.5 6.8 6.8
30
An atom with multiple electrons can be approximately described in terms of 
single electron wave functions, single electron states and energy levels. These wave 
functions are usually calculated by self-consistent methods. If spin-orbit coupling is 
considered, a single electron state can be described by the four quantum numbers n, I, j  
and nij. as shown in Table 2.1. Where n is the principle quantum number, I is the orbital 
angular momentum quantum number, j  is the total angular momentum quantum number 
and rrij is the quantum number for the z-component of the total angular momentum. An 
Auger transition can be described by three letters as Wi Xp Yq where Wi stands for the 
initial state core hole, Xp and Yq stand for the two holes in the final state.
Figure 2.5 (a) KLL Auger transition, (b) LLM Coster-Koing Auger transition.
31
The Auger transitions involving n = \ and n = 2 levels are called KLL Auger 
series. In fact, there are six possible transitions in this series, KLiL1, KLiL2, KL1L3, 
KL2L2, KL2L3 and KL3L3. If one of the final state holes is in the same shell as the initial 
hole, such as L1L2M1, this kind of Auger process is called Coster-Konig (C-K) transition 
as shown in Fig. 2.5. A C-K transition has a much shorter life time than the regular Auger 
process.70,74,76 Therefore, according to the uncertainty principle, AE-At > h/2jc, the energy 
spreads of the C-K transitions are larger than those of regular Auger transitions. 
Sometimes the energy spread of a C-K transition can reach as high as 10 eV. The C-K 
transitions can also affect the probabilities of other Auger transitions. For example, the 
existence of an LiLpYq C-K transition will reduce the chance for an LiXY  Auger transition 
and increase the chance for an LpXY  Auger transition.76 The effects of the C-K transition 
might play an important role in explaining the features in Ihe37Cl Auger spectrum.
The discussion above agrees quite well with Auger experiments for large Z- 
number atoms. However, it is not quite accurate when applied to middle range or small Z- 
number atoms. For example, there are five KLL peaks for Ar instead of the six predicted 
above. The reason for this is the complexity of the angular momentum coupling of the 
two holes in final states of the Ar atom. Since each subshell in the Ar atom is fully 
occupied with electrons before the transition, the orbital, spin and total angular 
momentum in these subshells are all zero (L = S = J = 0). These subshells are called 
closed subshells. According to quantum mechanics, the coupling of two holes is identical 
to the coupling of two electrons in a closed subshell. For Ar KLL Auger transition, the
32
two holes in the final state are all in the L-subshell. These two holes are coupled with the 
Russell-Saunders coupling scheme or the L-S coupling model.75 In this coupling model, 
the total spin angular momentum, S is formed by adding the spin angular momentum of 
the two holes, S1 and S2, Le., S = S1 + s2. The total orbital angular momentum, L is formed 
by adding the orbital angular momentum of the two holes, I1 and I2, Le., L = I1 + I2. Then 
the total angular momentum, J  is formed by adding the total spin and orbital angular 
momentum, Le., J = S + L. The adding of two angular momentum is accomplished 
according to the angular momentum addition rule in quantum mechanics, e.g., J  = IL -S 1J,
(L:S+1), ..., (L+S). An L-S coupled state can be expressed in the form of the spectral 
term, 2S+1Lj. The KLL series of Ar Auger has six final states described in spectral terms, 
1Sq, 1Pi, ^Pz,1,0, 1So, 3PzjIjO, and 1D2. as shown in Table 2.2, where 3P2jIjO is excluded for 
violating the conservation of parity. An Auger transition with L-S coupling in its final 
state is usually described by the spectral term and the WXY symbol. For example, the 
symbol KLiLi(1So) describes a KLiLi Auger transition whose final state has zero spin 
angular momentum, zero orbital angular momentum and zero total angular momentum.
Table 2.2 Final States of Ar KLL Auger Series81
Final State Holes Spectral Terms
2s2p* 1S0
Zs1Zp5 • iPi, 3PzjIj0*
Zs2Zp* 1So, 3P2,i,o, 1D2
*3P2ji ,o is excluded for violating the conservation of parity.
33
Auger Energies and Intensities
Since the Auger transitions are determined by the electronic configuration of the 
atom studied, each element has its own set of Auger peaks with characteristic energies 
and structures.70 75 The energy of the Auger electrons can be in principle determined by 
the difference of the total energies before and after the transition. Supposing an isolated 
atom with atomic number Z undergoes an Auger transition WXY, the energy available for 
an X  electron to fill the W vacancy is Ew (Z) - Ex (Z), where Ew (Z) and Ex (Z) are the 
binding energies of electrons at the W and X  levels. Therefore the kinetic energy of the 7  
electron obtained in this Auger transition is
Ewxy(Z) =  Ew (Z) - Ex (Z) - Ey (Z) (2.22)
where Ey (Z) is the binding energy of an electron in the Y level. Eq. (2.22) is an 
approximation because Ey (Z) is the binding energy of the 7  level when all the inner 
shells below the 7  level are filled, whereas in Auger transition a hole exists below the 7  
level. Therefore it takes a higher energy, Ey’(Z) to ionize an electron from the 7  level due 
to less screening. In fact, Ey(Z)  is between Ey (Z) and Ey(Z+l) .75
Ey'(Z) = Ey(Z)+ #E y(Z+7)-Ey(Z) 7 (2.23)
where (3 is between zero and one. Hence, the Auger energy of the WXY transition is
EW Z) = Ew (Z) - E% (Z) - Ey (Z) - #  Ey (Z+7) - Ey (Z) )  (2.24)
Similarly, the Auger energy of the WYX transition is
EW Z) = Ew (Z) - Ey (Z) - Ex (Z) - Ex (Z+7) - Ex (Z) 7 (2.25)
34
From a quantum mechanics point of view, the WXY transition is identical to the WYX 
transition. Therefore, Ewxy(X) is equal to Ewyx(Z). A semi-empirical way to calculate the 
Auger energy is to take the average of Ewxy(Z) and Ewyx(Z) and set /3 and /T to be unity101 
Ewxy(Z) = Ew (Z) -1/2 [E x (Z+l) + Ex (Z) ] - 1/2[ Ey (Z+l) + Ey (Z) ] (2.26)
The Auger electrons emitted from solid have to overcome the work function, (ps. The 
Auger electron energy, Ewxy(Z) then becomes
Ewxy(Z) = Ewxy(Z) - (ps (2.27)
By reaching the electron analyzer with work function (pA, the Auger electron energy, 
Ewxy"(Z) then becomes
Ewx^ t(Z) = Ewxy(Z) - (Pa (2.28)
Their relations are shown in Fig. 2.6.
E w x y E ’w x y , E ” w x y
! * AEf=O : * s
Y - e
x  - e
^VACUUM
Ef
w _ e
S am p le  A n a ly z e r
Figure 2.6 Relation between Auger electron energy Ewxy and the kinetic energy when 
Auger electron reaches analyzer, E ttwxy
35
For a conventional Auger process whose initial vacancy is created by electron or 
photon irradiation, both the initial ionization cross section and the Auger transition 
probability contribute to the intensities of Auger peaks.75 The situation is much simpler in 
this work where the initial vacancy is created by electron capture. Thus the Auger peak 
intensities will be proportional to the Auger transition probabilities. By calculating the 
electron-electron interaction the Auger transition probability, Wa can be calculated by the 
Golden Rule as
Wa = f JJpW/M <Pi{h)drA2 (2.29)
where p(k) is the state density and ^ r 1) is the initial hole state, ^ r 1) and ^ r 2) are the 
states electrons come from and Qjir2)= elhr is the emitted Auger electron.75
All of the above discussions are based on the one-electron approximation and the 
assumption that the electromelectron interaction is weak enough so that it does not shift 
too much the core levels, distort the charge distribution or modify the transition rates. 
However, these conditions do not always hold and in those cases the one-electron orbital 
picture will break down. The electron-electron interaction induces shake-up and shake-off 
effects, where the transition of one electron causes another electron to be excited to a 
higher energy level or above the vacuum level. These effects make the Auger spectra 
much more complex. A self-consistent treatment will be needed for these kinds of Auger 
transitions.102 Some theoretical and experimental results on Ar LMM Auger energies, 
intensities and transition probabilities are listed below in Table 2.3 and Table 2.4.
36
Table 2.3 Ar LMM Auger Peak Energies and Relative Intensities.
Transitions Ecai(eV)+ Spectral Terms E.xp(eV)++ Intensity(ab.)++
La M2,3 M2,3 207 207.23 139 ;
1D2 205.62 183
1S0 203.23 55
L3 M2,3 M2,3 205 'Po,U 205.21 272
1D2 203.47 270
1S0 201.09 100
L2 M l M 2,3 193 "Po,I 193.08 37
1Pi 189.49 32
L3 M i M2,3 191 "P2 191.13 87 '
1Pl 187.33 53
L2 M i M i 178 "So 180.06 18
L3 M i M i 176 1S0 177.91 34
+ Calculated from Eq. (2.26) to Eq. (2.28) and taking (pA=5.0 eV ;++ Reference [29]
Table 2.4 (a) Ar Auger Transition Probabilities (%) for a K-hole Initial Vacancy.76
KL1L1 KLiL2 KLiL3 KL1M1 KLiM2 KLiM3 KMM
6.89 7.59 14.54 1.53 0.71 1.36 0.67
KL2L2 KL2L3 KL2Mi KL2M2 KL2M3
1.39 34.44 0.74 0.25 2.87
KL3L3 KL3Mi KL3M2 KL3M3
19.55 • 1.42 2.87 3.30
KLiLi=6.89% KL iL2,3=2 2 .13% KL2^ L2,3=55.38% KLM=I 5.05% KMlVl=O. 67%
37
Table 2.4 (b) Ar Auger Transition Probabilities (%) for L-hole Initial Vacancies.76
L 1L2M 1 L iL2M 2 LiL2M 3 L-1L3M] L 1L3M 2 LjL3M 3 LjM 2M 3 L iM 3M3
17.81 7.42 7.50 34.68 7.33 20.83 .007 .068
L 1M 1M 1 L lM iM 2 L23MiMi L23M i M 2 L2 3M iM3 L2l3M 2M 3 L2 3M 3M3
.769 1 . 2 0 2.39 1.09 0.72 22.83 31.63 45.00
L iL2i3M,=52.50% L1L2i3M2,3=43.07% L,MM=4.43% L2,3M IM=24.64% L2,3M2i3M2i3=76.36%
In this thesis work, the emphasis is put on the study of Auger decay cascades of 
the initial hole in the K-shell of 37Cl atom, especially, the LMM Auger transitions. As 
discussed in the Chapter I , the LMM Augers of 37Cl are expected to be similar to the Ar 
Augers in structure and Cl Augers in energy. The above tables provide a very convenient
30
CD 20
n . .
Energy  (eV)
Figure 2.7 The Auger spectrum predicted by the RACS for Ar with the initial Is hole.
38
reference for the comparison. The similar case like the decay of the 37Cl initial state can 
only be found in the theoretical calculation. In 1992, Omar and Hahn reported the study 
of the decay of an Ar atom with a ls-hole.29"34 All possible Auger transitions with their 
shake-ups and shake-offs during the production and decay of an initial Is hole in Ar are 
calculated by using the model of radiative Auger cascade transitions with the shake-off 
effect (RAGS). The calculated Auger spectrum31 is shown in Fig. 2.7. After Auger decay, 
the Ar atom is ionized in different charge states.29"34 The charge states of the Ar ions are 
determined by the different decay channels and shake-offs. Omar and Hahn also 
calculated the final charge state distribution of Ar ions (FCSD) as shown in Table 2.5.33
Table 2.5 The Final Charge States of Ar after Auger Transitions of an Is Initial Hole.
Charge I 2 3 4 5 ■ 6 7
Y racs(% )31 0.8 6.8 10.0 42.4 28.0 10.3 1.7
Y expi(% )104 2.2 8.8 41.5 33.3 11.5 2.7
Y exp2(% )103 9.9 8.8 43.9 25.8 9.9 1.3
Double Auger Transition
In a normal Auger process, filling of a vacancy in an inner shell by an outer-shell 
electron leads to the emission of a second outer-shell electron. However, if the initial 
vacancy is in a deep inner shell, the transition of the outer-shell electron can sometimes
39
cause the ejection of two electrons simultaneously (Fig. 1.4(b)). This transition is called a 
double Auger decay process.4’40"43 The double Auger decay process was first reported by 
Carlson and Krause from the observation of Ne+3 ions in the gas-phase Ne Auger decay.40 
They concluded that the double Auger decay process accounted for the creation of Ne+3 
ions because in a normal Auger process a Ne atom could not lose three electrons at the 
same time. From the yield of Ne+3 ions they also concluded that double Auger decay 
probability was about 8% 40
hi the one-electron model, the double Auger decay probability has to be zero 
because the initial Xffi and final Xjff  wave functions differ by more than two one-electron
states. In this case, the matrix element (xFf Jy = IZr12JxF j  determining the amplitude of
the double Auger process is zero. Therefore, many-electron correlation has to be taken 
into account to calculate the double Auger decay probability. By using many-body 
perturbation theory 4 the amplitude M of the double Auger decay can be calculated from
9
the electron-electron interactions, Mi as shown in Eq. (2.30) and Fig. 2.8.
i=l
M  _  y
k e . + -  £/, -£*
Jlzf _  V 1 (^ i^ zM /A X ^Z il^ l/i/z )
4 f  ;
, = t
(2.30)
40
M1 +
fiQi
+ exchange(q1,q3)
M, =
M7 = i + exchange(q1,q2)
Figure 2.8 Brueckner-Goldstone diagrams for the partial amplitudes Mi, M4 and M7. 
Lines with arrows to the left and to the right denote holes and particles, respectively.
Where Mi are the partial amplitudes of the process and (pq\u\rs) = (pq\v\rs)-  (Pq\v\sr)
is the difference of the “direct” and the “exchange” matrix elements of the Coulomb 
interaction; i is the initial vacancy and //, f 2, fs  are the final vacancies; qi and q2 are the 
double Auger electrons; and e is the one-particle Hartree-Fock energy. Other amplitudes 
(M2, M3, M5, M6, Mg and Mg) can be obtained out of Mi, M4 and M7 by the cyclic 
replacement of the elements //, f 2 and f 3, respectively. Summation over k in Eq. (2.30) 
implies the summations over hole and discrete electron levels and integration over 
continuous electron spectra. Figure 2.8 shows the Bruckner-Goldstone diagrams of the 
partial amplitudes Mi, M4 and M7. The transition energy in a double Auger decay can be 
expressed as
41
Etot = E f-E i (2.31)
where Ef  and Ei are the energies of initial and final states, respectively. The energy Etot is 
continuously distributed between the two emitted electrons. Therefore, the total 
probability of the double Auger decay process is determined by the expression
rT = J ^ r C E 1) (2.32)
where J(E1) is the probability that the first electron, e.g., qx has the energy Ei . The energy 
E1 is given by
J(E1) = 27t\Mf (2.33)
There are several models which can be derived from the diagram above. The first 
one is the virtual inelastic scattering model as shown in Fig. 2.9(a) where the imaginary 
part of M4 dominates in the total amplitude, i.e.,
\M\ = IhnM4I = (M ^ lA / s X ^ M /A ) )  (2.34)
where k0 is an “intermediate” Auger state that satisfies S0 = k02/2m and E0 =
£ f2 +Bf3 - B i . In this case, a normal Auger decay of the initial hole i occurs and the
intermediate state k0 is created; then the k0 state is inelasticly scattered by electron in /)  
state. The total probability in this case is
•pZM _  T-'A 
1 - ■ 1 /^2/3*0 a o^ (2.35)
where T ^ r3t0 is the probability of normal Auger decay and is the cross section of 
the inelastic scattering of the k0 Auger electron with an electron in a n / ;  state. The
42
energy distribution between the two outgoing electrons is determined by the inelastic 
scattering process of the k0 Auger electron, and if E0 is high enough, £qi »  e t/ .
qi
(a) (b)
Figure 2.9 (a) Virtual inelastic scattering model, (b) Cascade mechanism.
The second model of the double Auger decay process is the cascade mechanism as 
shown in Fig. 2.9(b) The initial hole i can decay through the normal Auger transition, 
producing a vacancy in one of the intermediate shells. This intermediate vacancy, in turn, 
can decay via another Auger transition. The corresponding probability for this model is 
r r  (2.36)
where j  is the intermediate hole and Fj is the total width of this hole. The energy 
spectrum has two lines centered at = ey + e ; -  e, and Eqi =  Efi +Ef -E j .
The third model of double Auger decay is the shake-off model where the initial 
hole, i decays through normal Auger decay, i'1-+ ff 'f i1 + qn Then, the atomic field
43
caused by this transition shakes off another electron q2 from the/? level. The probability 
of the double Auger decay process for this model is
r r = ^ / , , : J ( 9 2 | / 3 ) r  (2.37)
where the second factor is the overlap integral. There are all together six combinations 
out of qj, q2 an d //, / 2 , / 3  states for the overlap term. The energy distribution between the 
two electrons, and £q i, is highly asymmetric, Le., one of them takes most of the Elol
and the other has little share of the energy.
Figure 2.10 Dependence of the double Auger decay probability on the energy of one of 
the outgoing electrons in the Ne ls"'-> 2s'2 2p"‘ + q, + q2 (2S) double Auger transition. 
Calculated by many-body perturbation theory.
44
The results of the many body perturbation theory (MBPT) calculation of the 
energy distribution of the two electrons for Ne double Auger decay4 are shown in Fig. 
2.10. It is clear that the electrons share the Em in the way that one of them takes most of 
the energy and the other takes the small balance of the energy. Since there are very few 
references on the double Auger decay,4’40"44 the experiment results of this research work 
on the double Auger decay probability and the energy distribution will be compared with 
the theoretical calculation of the Ne work4 quoted above.
45
CHAPTER 3
EXPERIMENT
Experimental Setups
"
The main feature of this experiment is to measure the time-of-flight of the 
recoiled 37Cl ions by using coincidence techniques.105'109 The 37Ar atoms were 
physisorbed on a cold flat substrate inside an UHV system. When electron capture occurs, 
the 37Ar nucleus becomes a 37Cl nucleus and a neutrino is emitted. The recoil due to the 
emission of neutrino could cause the 37Cl ion to desorb from the substrate (Fig. 1.1, Fig. 
1.2). Meanwhile, the capture of a deep core-level electron in the 37Ar atom would also 
result in a highly excited initial state in the daughter 37Cl atom. This initial state could 
then decay through radiative (x-ray) or non-radiative (Auger) relaxation and cause the 
ionization of the 37Cl atom. Therefore, if  the desorption were to happen, both the 37Cl ion 
and the Auger electron would escape from the surface simultaneously. By using two 
detectors with one of them detecting ions and the other detecting selected Auger 
electrons, the time-of-flight of the 37Cl ions could be obtained by comparing the time 
difference between ions and electrons arriving at the two detectors. Because the electrons 
are much faster than the ions, this time difference is essentially the time-of-flight of the
46
37Cl ions. The difficult part in this experimental concept is that the detected ions and 
electrons must come from the same EC decay to give a meaningful time difference. This 
problem was solved by using coincidence techniques. The details of this technique will be 
discussed later. The Auger transition study in this work was carried out by using a double 
pass cylindrical mirror analyzer (CMA) to directly measure the emitted Auger electrons 
from the 37Ar atoms absorbed on the back side of the substrates. The double Auger study 
also applied coincidence techniques to identify the correlation among the emitted Auger 
electrons. In the past three years, different 37Ar sources, substrates, detectors and 
electronics have been used in this research work. The following sections will be devoted 
to describe in detail each part of the experimental setups.
The UHV System
The experiments were performed in an UHV system as shown in Fig. 3.1. The 
system was pumped mainly by an ion pump and a Ti-sublimator (TSP). A turbo- 
molecular pump was also provided to the system through a shut-off valve for baking and 
rare gases pumping. The typical base pressure of this system was about 5 x IOt11 Torr. 
The substrates in this experiment were mounted to a cold finger of a closed-cycle helium 
refrigerator. The two parts of the helium refrigerator were connected by a pair of helium 
transfer tubes. The device could deliver about 10 watts of cooling power to the cold 
finger at a temperature of 20 K. With proper shielding the substrates could reach as low 
as 16 K in this work. The temperature of the substrate was monitored by a thermocouple
47
of Au/.07%Fe and chromel. This type of thermocouple has excellent sensitivity at low 
temperatures (17 pV/K at 10 K). During the experiment, the He refrigerator also served 
as a cold trap for
He Cryo 
Pum p
Valve
Turbo
Pum p
Pum p
Figure 3.1 The UHV system of this work. F is the cold finger, S is the substrate and PSD, 
L and R are detectors.
the residual gases because the ion pump was turned off to prevent ions created inside the 
ion pump to affect the measurement. The working pressure in this circumstance was 
about 2 x 10"'" Torr. The refrigerator was mounted on a manipulator on the top of the 
chamber with x, y, z and tilt motions so that substrates could be brought to the focal point
48
of the CMA or other selected points inside the system. In the early stage of the 
experiment, the CMA was also used to determine whether Ar was adsorbed on the 
substrates at certain temperature. At 16 K, the Ar Auger peak was clearly visible only 
when an Ar pressure of 10 7 Torr was maintained in the chamber during measurements in 
order to balance the electron-induced desorption (ESD) effect on the substrates. Figure 
3.2 shows the adsorbed Ar and the substrate Auger peaks at different temperatures.
T=16K
Energy (eV)
150000
124000
"“T 98000
T 16 K
T=SOK
T= 300K
20000
50 150 250 350 450 550
Energy (eV)
Figure 3.2 (a) Ar Auger peak from the AuZSi(Ill) substrate at 16K. (b) Ar and carbon 
Auger peaks from the pyrolytic graphite substrate at 16K, 50K and 300K.
49
The Substrates
There were two sets of substrates prepared in this experiment. The first substrate 
was a I x I cm2 S i( I l l )  wafer. Both polished sides of the wafer were coated with thick 
(0.2-0.3 jim) gold films in another vacuum chamber. Scanning electron microscope 
images showed a smooth film across the entire substrate with slight variations in 
topography visible only at the highest resolution of the microscope (~200A). The last 
several layers of gold were evaporated on the substrate under in situ UHV conditions to 
ensure a clean surface. The CMA was used to monitor the substrate cleanliness. Auger 
spectroscopy showed no measurable contamination of the substrate surface at 16 K.
A major concern of the properties of the substrates was the back-scattering effect 
from the substrates. These back-scattered 37Cl ions would have a . different energy 
compared to those directly desorbed ions, and thus the recoil spectrum would be 
broadened. The back-scattering cross section is proportional to the square of atomic 
number, Z. Hence, the substrates had to use materials withTow Z number.26"28 One of the 
other considerations was that the substrate material had to be conductive so that it would 
not become charged when charged particles left the surface. Considering these conditions, 
carbon became a good candidate. Therefore, the second substrate was a high grade 
pyrolytic graphite (HPG) of 12 x 10 cm2 area and 2 mm thick. The fresh cleaved HPG 
was examined by using a scanning tunneling microscope (STM) in the air, which revealed 
an atomically flat surface. The HPG was cleaved again from both sides right before being
50
put inside the UHV chamber. Auger spectroscopy showed no measurable contamination 
of the substrate surface at 16 K (Fig 3.2 (b)). Argon physisorption and desorption were 
also monitored by a quadrapole mass spectrometer (QMS). It was found that Ar started to 
physisorb (multilayers most likely) at around 30 K and desorbed completely above 55 K 
on this substrate (Fig. 3.3).
10000
S3
r r i
-JS-
D h
m 5000
id(0
CU
0
10 20 30 40 50 60 70
Temperture (K)
Figure 3.3 The temperature programmed desorption (TPD) spectrum of Ar from the HPG 
substrate.
Maintaining a good thermal contact between the substrate and the cold finger was
very critical for obtaining the desired low temperature. Figure 3.4 shows the mounting
bridge between the HPG and the cold finger. The bridge was machined with ultrahigh
purity (99.999%) copper to increase the thermal conductivity at low temperature. (At
A Ar Partia l P ressure
%Xt
%
X\
V
X
51
20K, copper with 99.999% purity is 10 times more thermally conductive than the one 
with 99.9% purity) The sapphire plates with high thermal conductivity and electrical 
resistance at 20 K served to isolate the substrate electrically
I
o . ...........
o :
G rap h ite
(a) (b)
Figure 3.4 The mounting bridge between the graphite substrate and the cold finger. The 
sapphire plates can isolate the graphite substrate from ground. The bridge is soldered to 
the cold finger with indium.
The Detectors
Since the success of this work depends largely on the accuracy of coincidence 
time-of-flight measurements, several special requirements had to be met by the detectors 
which triggered the timing circuits. First, the detectors had to be able to output very sharp
52
pulses. The time resolution of an. electron-electron coincidence measurement must be 
better than a nanosecond. Therefore, the pulse output from the detector had to be in the 
order of nanosecond. Second, most of the measurements in this work required a large 
signal to noise ratio, so the detectors had to have very little background noise. A dark 
count rate in the range of a few counts per second was preferred. Third, the detectors had 
to be able to detect both electrons and positive ions by simply switching the electrical bias 
from outside the UHV chamber. Fourth, the detectors had to have stable gain over the 
time of experiment. Considering all the requirements above, channel electron multipliers 
(channeltron) and microchannel plates (MCP) made by Galileo Electro-Optics were 
adopted to develop the detectors.
There were three generations of detectors developed for this work. The first one 
used channeltron multipliers.110 A channeltron is spiral glass tube two millimeters in 
diameter with materials with high secondary electron emission yield coated inside the 
tube. When a single electron strikes the tube entrance, several secondary electrons will be 
created. These electrons will then create even more electrons like an avalanche. A 
channeltron can detect both positive and negative charged particles by changing the bias 
voltage. Fig. 3.5 shows the schematic of the home made detector with the channeltron 
biased for electron detection. The channeltron was mounted inside a metallic cylinder by 
ceramic rods. Three screens were mounted in front of the channeltron entrance with the 
first and third screen grounded and the middle screen biased by screen voltage, V. The 
middle screen served to stop particles with energy lower than qV, where q is the charge of
53
the particle. A positive voltage was used to retard positive ions and negative voltage to 
retard negative ions and electrons. By scanning the screen voltage, an integrated energy 
spectrum could be measured. The gain of this detector was about IO7, the dark count was 
~ I cps, and the full width at half-maximum (FWHM) of the output pulse was about 4 ns.
Screen  Voltage
Screen s
5 MO
2.7KV
.002MFC han n e Itro n
Signal
C eram ic  Hi Copper
Figure 3.5 Detailed cross section diagram of the first home-made detector with 
channeltron multiplier. Detector is biased for electron detection.
There were some disadvantages for the first generation detectors with channeltron 
multipliers. As shown in Fig. 3.5, the front entrance of the channeltron was a cone shape. 
When electrons or ions reached different parts inside the front cone surface, they could 
have different flight distances. This would bring errors in the time-of-flight
54
measurements. The four-nanosecond FWHM pulse width was also not satisfactory for 
accurate time measurements. Therefore, the next generation detectors employed 
microchannel plates (MCP)111 to substitute channeltron multipliers. A MCP is an array of 
about IO6 miniature electron multipliers. Each one of them works like a tiny channeltron. 
Typical channel diameters are in the range of 10-100 pm and the length to diameter ratios 
are between 40 and 100. Parallel electrical contacts to each channel are provided by the 
deposition of metallic coatings on the front and rear surfaces of MCP, which then serve as 
input and output electrodes. The total resistance between electrodes is about IO9 £2. Such 
MCP, used singly or two of them in a cascade, allow electron multiplication factor of IO4- 
IO7 coupled with ultrahigh time resolution (c l  ns). The spatial resolution of the MCP is 
limited only by the channel dimensions and spacing. The channel diameter of the MCP 
used in this work is 12 pm and the center-to-center spacing is 15 pm. The MCPs have 
direct sensitivity to charged particles like electrons and ions (Fig 3.7 (a)), and energetic 
photons like x-rays and UV light. Therefore, they have been widely used in imaging, 
surface science and nuclear physics.111
Each of the new detectors used two MCPs in cascade. Each MCP was one inch in 
diameter and 400 pm thick. The mounting assemblies and electrode connections for the 
MCP’s were home-constructed as shown schematically in Fig. 3.6. Each detector had 
three screens in front of the MCPs with the middle screen serving as a retarding screen. 
These detectors produced stable and reliable noise-free sharp negative pulses at an
55
average height of 50 mV and with FWHM less than I ns. The dark count rate was about
2-3 cps. A pulse height spectrum of each MCP was measured in order to set the trigger 
thresholds in the amplifiers above electronic noise (Fig.3.7(b)).
*  =Cu
=Channelplate 
^  In su la to r
Figure 3.6 Cross section diagram of MCP based detector. Detector is biased for positive 
ion detection.
Ion E (keV)
0 2 4 6 8 10
IO2
Electron E (eV)
250 500 750 1000
Pulse hight (ch.)
(a) (b)
Figure 3.7 (a) Electron and ion detection efficiency of MCP. (b) A typical pulse high 
distribution spectrum of the MCP used in this work.
J.
In the ion counting mode, a negative high voltage of -2.7 kV was applied between 
the front MCP and the ground to prevent KLL Auger electrons (E<2.6 keV) from entering 
the detector and contributing to the positive ion signal. This voltage also accelerated the 
positive ions to an energy at which their detection efficiency was maximized (Fig. 37(a)). 
In this case, a -2.0 kV voltage was maintained between the front and rear of the MCPs 
which made the voltage across each MCP to be 1.0 kV.
The Position Sensitive Detector
56
Sometimes it is necessary to acquire spatial information of the detected particles. 
For example, the flight distances between sample and MCP detector were different for 
electrons or ions arriving at different points on the front MCP. Compensation could be 
made for the flight path if one could know the exact location where the particle had 
landed. In the past ten years a variety of imaging multichannel detectors have been 
developed.112"1^ 0 The tremendous parallel collection advantage inherent in these devices 
can also be used to greatly reduce data-acquisition times and increase detection 
sensitivities. In this work, a microchannel plate with resistive anode encoder (MCP- 
RAE)112"118 two-dimensional detector was developed.
Instead of a conductive plate behind the rear MCP to collect the multiplied 
electrons and output pulse signal, the MCP-RAE detector uses a plate with evenly coated 
resistive film which outputs signals encoded according to the location where electron hits
57
the front MCP. The theoretical formulation of the encoding mechanism is briefly 
described here.112 Considering a planar resistor whose surface resistivity is denoted r and 
whose electrostatic capacitance per unit area is denoted c, if the potential at a position 
(x,y) is V  and the vector surface current density is S, then Ohm’s law and the law of 
charge conservation require
rS + W  = 0 (3.i)
and
V -S + c V = J  (3.2)
where J  is the incident source of current driving the anode. Combining Eq. (3.1) and Eq. 
(3.2), the two-dimensional time-dependent diffusion equation becomes
V2V -rcV = -rJ  (3.3)
This equation is subject to suitable boundary conditions defined by the edges of the 
anode. An appropriate model for the driving signal J is  a delta function in space and time 
J(x,y,t) = Q d(t) S (x -  x0) S(y -yo) ' (3.4)
with Q is the total charge delivered by the event and (x0,yo) is the event location on the 
anode. An important special case of Eq. (3.3) is the situation where the total terminal 
charge signals are of interest but the detailed diffusion pulse shapes are not. For this case, 
the dc (time average) part of Eq. (3.3) and Eq. (3.4) can change to a Poisson equation 
1V2 V= -rJ  (3.5)
where J  is now the function of location (x,y) only. Equation (3.5) is not analytically 
solvable for arbitrary boundary conditions. It has, however, been solved for several 
special cases of interest for position sensing with solid state detectors, gas counters and
58
MCP’s112. The simplest case is when the resistive surface is infinite in extent. In this case, 
the current, I  extracted at a point I distance away from the source will be inversely 
proportional to I.
I ocW  * (3.6)
This means that far away from the source, the signal output will be lower. Therefore, in 
principle the distance could be obtained by knowing the relative magnitude of the signal.
One practical solution to the boundary problem of Eq. (3.5) leads to the develop­
ment of the MCP-RAE. In 1972 Augustyniak121 et. al. reported their design which 
involved terminating the edges of a uniform resistive surface with resistivity, r by four 
concave circular arcs having radius of curvature, a and resistance per unit length, Rl , as 
shown in Fig. 3.8. The relation is
Rl = r/a (3.7)
By using this design, which followed the suggestion by the Gear theorem, the anode 
could be operated in a distortionless manner. The Gear theorem states that a uniform 
current flow in an infinite sheet having a resistivity r is unaffected by a circular hole of 
radius, a if  the hole is bordered by a line resistor of value Rl = r/a. This theorem holds for 
any arrangements of the holes. One of such arrangement is just the design by Augustyniak 
as shown in Fig. 3.8. In this design the four corners of the plate would share the incident 
charge Q in the way as if the plate is infinite in extent, i.e., the currents, Ia at the four 
corners would be inversely proportional to the distance between the event point (x,y) and 
the corner vertices. Therefore the event position (x,y) can be obtained as
59
A  + A >  + / 3 +  / 4
(3.8)
where d  is the length of the side of the anode and the coordinates are shown in Fig. 3.8.
Figure 3.8 Geometry of a circular arc terminated resistive anode encoder. Electrical 
connections are made at the four corners.
In this work, a IOOkQ resistive anode encoder (RAE) with Ix l inch size made by 
Quanta Inc. was used. Figure 3.9 shows a diagram of this home-constructed MCP-RAE 
detector with biased voltages for ion detection. The detailed design diagrams of each part 
are provided in Appendix B. This MCP-RAE detector provided very high dynamic gain
x
d
60
(107), very low background noise (I cps) and superb time and spatial resolutions. The 
time resolution of a detector is mainly determined by its output pulse width. Because the 
signals from the four corners in this detector were of the order of 100ns, a special 
arrangement was made to extract the timing pulse from the rear side of the second MCP. 
In this way, a sub-nanosecond time resolution was achieved. The spatial resolution was 
checked by putting a mask with tiny holes in front of the detector and looking for the 
output image of these holes. The diameter of these mask holes was 0.2 mm. The resulting 
images of these holes had a FWHM of less than 0.2 mm diameter, indicating that the 
spatial resolution of this detector was better than 1/100, the specification of an MCP-RAE 
detector commercially made by Quanta Inc.. The success of the whole operation relied 
largely on the performance of this detector.
Figure 3.9 Cross section diagram of the home-constructed MCP-RAE detector. Biased for 
positive ion detection.
61
The Radioactive 37Ar Sources
With a half life of 35 days, the 37Ar cannot survive in nature. Argon has an atomic 
number of eighteen and is located between Cl and K in the periodic table. There are 
eleven isotopes ranging from 34Ar to 44Ar. The stable isotopes of Ar include 36Ar 
(0.337%), 38Ar (0.063%) and 40Ar (99.6%). All the unstable isotopes with atomic weight 
lighter than A = 38 decay through electron capture and those heavier than A = 38 decay 
by beta decay.24 Therefore, the isotope 37Ar needed in this experiment had to be prepared 
by designated nuclear reactions. There are several ways to produce 37Ar from its stable 
isotope or neighboring elements. The difficulties involve obtaining enough reaction cross 
section so that the concentration of the produced 37Ar is large enough and the safety 
concern on how to handle these radioactive materials. For this experiment, 37Ar was 
produced by either 36Ar(n,Y)37Ar or 40Ca(n,a)37Ar reactions122 in the Brookhaven 
National Laboratory and extracted locally as described below.
The first 37Ar source prepared for this work was produced from the stable isotope 
36Ar through the nuclear reaction 36Ar + n -> 37Ar + y involving high energy neutrons in 
the center of a nuclear reactor.122 A quartz ampoule filled with 10 ml of 36Ar enriched to 
99.5 atom% was purchased from Isotech Inc.. The neutron capture cross section for this 
reaction is 5 barn;8 irradiation with a flux of IO14 n/cm2-s for seven days would produce 7 
x IO16 atoms of 37Ar corresponding to 0.03% concentration in the remaining 36Ar atoms. 
There were about four weeks waiting time for the short-lived, radioactivity produced in
62
the irradiation to die down to an acceptable level for handling. Therefore, the final source 
used in the experiment would have 4 x IO16 atoms o f 37 Ar corresponding to an activity of 
250 mCi and a 0.015% fraction of the 36Ar gas. Since the density of 36737Ar atoms on the 
surface is 7 x IO14Zcm2 for a monolayer, the initial number of 37Ar atoms on the surface 
would be I x 10" atoms for a monolayer on the I cm2 substrate surface, corresponding to 
an initial decay rate of 2.3 x IO4Zs.
2000 2500 3000 3500 4000
C hann e l
Figure 3.10 The gamma ray spectrum of the radioactive gas source taken by an HP-Ge 
detector. The dominant feature is the 37Ar IB radiation and the background activity mostly 
comes from the quartz ampoule materials excited by the neutron bombardment.
63
The irradiated ampoule was transported from the reactor to the laboratory in a 
heavily lead-shielded container. After the initial check by the nuclear security officer, a 
gamma ray spectrum was taken by an HPGe-detector which revealed the characteristic 
37Ar internal bremsstrahlung (IB) and other radioactive background as labeled in Fig. 
3.10. This assured that the 37Ar was indeed produced by the bombardment of the fast 
neutrons in the center of the nuclear reactor. The ampoule was then introduced into 
another vacuum chamber which was baked at 150° C for 12 hours and then pumped down 
to IO 8 Torn The ampoule was then broken mechanically with a piston attached to a linear 
feed-through. A gamma ray spectrum was taken again just outside the source chamber
Ar 7  ray spectrum
2000
Figure 3.11. The gamma ray spectrum of the 37Ar gas source after the ampoule was 
broken. Noted that most of background radioactivity was blocked and the 37Ar IB 
radiation dominated the spectrum.
64
which confirmed the purity and strength of the radioactive gas. The dominant feature of 
the spectrum was the continuous internal bremsstrahlung radiation which had an end­
point energy at 814 keV (the Q value of 37Ar decay).8 Otherwise, the most prominent 
feature was the 1460-keV 40K line due to the room background (Fig. 3 .I l ) .8 The gas was 
leaked into the main chamber via a Varian leak valve which had an excellent dosage 
control. There was a second leak valve for research grade Ar gas (40Ar) which was also 
used in some parts of the experiment.
The recipe described above could produce only a very small fraction of 37Ar 
among the huge background of 36Ar. The low concentration of 37Ar demanded a higher 
total dosage in the experiment to get the required activity. As discussed in the first 
chapter, this means each 37Cl atom would have more surrounding 36Ar atoms, hence more 
chances to exchange charges. To get a higher concentration of radioactivity, the second 
37Ar source was prepared by a different way. The isotope 40Ca was used to produce 37Ar 
by the 40Ca (n,(X)37Ar reaction. Half of a gram high-purity 40Ca isotope was bombarded by 
fast neutrons for three days near the center of the nuclear reactor. The nuclear reaction 
40Ca + n —> 37Ar + 4He (a) took place and 37Ar and 4He gases were created inside the 
40Ca solid. The cross section of this reaction is about 0.5 barns and the total yield of 37Ar 
was about 20 mCi. Without the background of 36Ar, the impurities now were mainly the 
residual gas inside the source chamber. By this method a tenfold more concentrated 37Ar. 
source was produced.
65
The procedure of handling of the radioactive source was similar to the previous 
one except the solid 40Ca sample had to be melted in order to extract the 37Ar trapped 
inside. It was found that the 40Ca would react with the quartz ampoule when heated above 
the melting point of Ca at 840 °C. To solve this problem, the 40Ca sample was doubly 
vacuum sealed under IO"8 Torr pressure before the irradiation (Fig. 3.12). After the
vacuum  pum p
second seal
f ir s t  seal
Figure 3.12 Double sealing 40Ca within quartz ampoules under high vacuum.
irradiation, the ampoule was heated up to 950°C in air for five minutes. This would 
completely melt the Ca solid and release the trapped 37Ar gas. The inner sealed ampoule 
was broken during the heating because of the chemical reaction between Ca and quartz 
(SiO2) but the outer ampoule would hold until being broken mechanically inside the 
source chamber. The gas obtained this way was ten times more concentrated than before, 
but the residual gas which made up for the impurity background was still too high. A
66
getter pump, which is inactive with Ar but highly reactive with other residual gas will be 
required to purify the gas source in future experiments.
Coincidence Techniques and Electronic Systems
In many physical studies the interesting physical events may be submerged in a 
large background of other events which share the same physical characteristics, such as 
energy, momentum, mass, etc., and the only trace of the events of interest is their 
characteristic time sequence. Utilizing this intrinsic time correlation, coincidence 
technique identifies physical events which happen as the result of the same physical 
process. This technique has been widely used in nuclear physics and some studies in 
surface and atomic physics.105'109 In this research work, coincidence techniques were 
applied together with other surface analysis techniques to study the desorption of 37Cl 
ions and the Auger relaxation.
A typical coincidence measurement setup was employed in the early phase of the 
experiment. The design of the setup is shown in Fig. 3.13. It consists of two MCP 
detectors with retarding screens, two time filter amplifiers (TFA), two constant fraction 
discriminators (CFD), a time-to-amplitude converter (TAC) and a multichannel analyzer 
(MCA) driven by a personal computer. One primary object of this system was to
v
determine whether the signal detected by the second detector was a true coincidence with 
the one detected by the first detector, Le., did the two signals arrive at the two detectors
67
close enough in time to have originated from the same physical process? The two output 
signals from different MCP detectors were amplified, shaped, and fed into the TAC with 
one of them starting and the other stopping the timing circuits. The TAC then output a 
pulse whose height was proportional to the difference of the arrival times of the two input 
pulses. A computer driven MCA would then record the histogram of the TAC pulses.
MCP2
Cryo-
p um p
IBM286
IBM386
CMA
Control
MCP: M ic rochanne lp la te
TFA: T im ing  F il te r  Am p lifie r
CFD: C o n s ta n t F ra c t io n  D isc r im in a to r
TAC: T im e - to -A m p li tu d e  C onve rto r
MCA: M u ltich ann e l A na lyze r 
TD: T ime Delay
CMA: C y lin d rica l M irro r A nalyzer 
V: R e ta rd ing  Screen  Voltage
Figure 3.13 The schematic diagram of the first coincidence measurement system.
A time coincidence of two signals from the same physical process is called a true 
or real coincidence. It is also possible for signals from different processes to trigger the 
timing circuits, this would produce an unwanted chance or accidental coincidence.122 In
6 8
principle, discriminating between true and accidental coincidences is relatively easy. 
After accepting the first signal to start the timing circuit in TAG, only a short period of 
time (time range of TAG) is waited for the second signal to arrive to trigger the stop. The 
longer the waiting period, the greater the possibility of having an accidental count. The 
pulse height spectrum of TAG in the MCA provides a trivial mean to distinguish between 
true and accidental coincidences. The pulse heights, and thus the time differences 
corresponding to the starting and stopping signals of true coincidences have a finite fixed 
time relationship. Accidental coincidences, however, representing random signals, have 
no definite time relationship. Therefore, accidental coincidences produce a uniform range 
of pulse heights whereas true coincidence produced a signal unique pulse, Le., a peak as 
shown in Fig. 3.14.
Figure 3.14 Coincidence peak from (a) ideal detectors and (b) real detectors (Si Auger 
electrons and primary excitation electrons). The area in region I gives true + accidental 
coincidences; region 2 gives accidental coincidence, and their difference gives the true 
coincidence.
69
The electronics of this system is also shown in Fig. 3.13. The two MCP detectors 
were biased with the center one for positive ion detection and the side one for electron 
detection in the 37Cl ion recoil velocity measurements. Electrons with different energies 
could also be selected by the retarding screens. The TFA (model 474, Ortec) supplied 
variable gains and time, integration constants to match the output pulses of the MCP 
detectors. The CFD (model 584, Ortec) transferred the amplified signals from the TFA 
into sharp negative pulses of 500 ps FWHM for better time resolution. The TAC (model 
2143, Canberra) had a time range from 20 ns to I ms. The MCA (Ortec 4000) had 4092 
channels for histograming. Since the electrons are much faster than the ions (typically IO4 
times faster), the time difference obtained from the coincidence peak in the histogram 
was effectively the time-of-flight of the 37Cl ions. The velocity could then be obtained by 
either knowing exactly the distance between the sample and the center detector or by 
shifting the center detector a known distance and measuring the shift of the coincidence 
peak. The integrated recoil energy spectra of the 37Cl ions were also directly measured 
with the center MCP detector alone by scanning the retarding screen voltage. The 37Cl 
Auger energies were directly measured by the CMA facing the other side of the sample. 
The coincidence measurements and the MCP or CMA scanning were both controlled by a 
personal computer with home-made interface and programs.
A much more sophisticated system for coincidence measurement, equipment 
control and data acquisition was developed in the later phase of the experiment. Three 
MCP detectors with one of them position sensitive (MCP-RAE) were used in the new
70
system. The entire chain signal amplifying, shaping, logic operating, timing and recording 
processes were conducted on a NIM bin and a CAMAC crate which could interface 
directly with a personal computer at very high speed. The experimental data could be 
stored in an event-by-event basis on a 2.3 GB magnetic tape with all the timing and 
spatial information in a pre-selected manner. After the experiment, the actual physical 
processes could be revisited by simply replaying the tapes. The schematic diagram and 
the logic block diagram of this system are shown in Fig. 3.15 and Fig. 3.16, respectively.
PSD.
S ta r tl
LogicjOR
IBM-CAMAC
CRATE
Position Signal
Figure 3.15 The schematic diagram of the second coincidence measurement system.
71
4xDAC
RAECl
GATE
4-GATERAEC2
RAEC3
RAEC4
DG4 -
LEFT
-DL2
RIGHT
EVENT
SCALAR INPREG
EVENT
Figure 3.16 The logic block diagram of the second coincidence measurement system. 
Some of the symbols are explained in the text.
The detailed descriptions of each component are omitted here and only the 
operation concepts will be discussed. Each block in Fig. 3.15 and Fig. 3.16 was mounted 
either on a CAMAC crate or on a NEM bin. Three timing signals came from the left MCP 
detector (L), the right MCP detector (R) and the middle MCP-RAE position sensitive 
detector (PSD). Each signal was amplified and shaped by a fast timing amplifier (FTA) 
and a constant fraction discriminator (CFD). The outputs of these CFDs would be 
properly delayed and gated by the delay and gate generators (DG) before fed to the time-
72
amplitude-converters (TAC) to generate coincidence spectra. Coincidences between L 
and R were registered at TACl and ADCI. Coincidences between either L and PSD or R 
and PSD were registered at TAC2 and ADC2. The ADCl and ADC2 are the analog-to- 
digital converters for the first and second TACs, respectively. The PSD also sent out four 
position signals which were amplified by the charge pre-amplifiers (CPA) and linear 
amplifiers (LA) before reaching the ADC-quadruplets (4xADC). The digitized position 
signals from the 4xADC were then compared by the personal computer to calculate the 
positions of the detected particles'.
, Each of the CAMAC crate based devices needs a timing gate, a NIM or TTL logic 
pulse with right time sequence and duration to synchronize the operations. A timing gate 
of a device provides the exact time window within which certain operation will be 
executed. These timing gates were generated by the timing signals through the logic 
operating device (logic AND, logic OR, etc.) and the DCs. For example, the timing gate 
of the TACl coincidence registration was generated by the logic AND of L and R signals, 
i.e., both L and R signals had to be present within the duration of the timing gate to 
trigger the recording of an event in TACL The timing gate of TAC2 was generated by the 
logic AND of the PSD signals and the results of the logic OR of L  and R signals, i.e., 
either L and PSD or R and PSD signals had to be present to trigger the recording of an 
event in TAC2. The identification between the two kinds of coincidence events in TAC2 
was achieved by the input register (ENP.REG) which would record whether PSD was in 
coincidence with L or R. The position information register 4xADC and the single rate
73
counter SCALAR were gated by the event gate. The event gate was generated by the logic 
OR of the TACl timing gate and the TAC2 timing gate. Therefore, the position signals 
and the single counts were saved on the tapes through the personal computer whenever an 
event was registered in either TACl or TAC2. If a higher sampling rate was desired, the 
event gate could also be created by adding (logic OR) every Nth of the PSD single counts 
to the original event gate. This would save the position signals and the single counts 
either on each coincidence count or on each Nth PSD count.
In the block diagram shown in Fig. 3.16, the timing gate of TAC2 was actually the 
event gate. This means the TAC2 would register an event whenever TACl or TAC2 was 
triggered. This would provide a mean of examining a three-event coincidence, i.e., L, R 
and PSD detectors receiving signals simultaneously. The registered data on TAC 1/ADC I, 
TAC2/ADC2, 4xADC, SCALAR and other related parameters could be written to a 
magnetic tape as a word in an event-by-event manner. By replaying the tape, these words 
would be retrieved by the computer in their actual time sequence. The interface between 
computer and the CAMAC crate was controlled by the program developed by Dr. Hindi.
Experimental Procedure
Before any actual measurement, considerable efforts had been made to fine-tune 
the electronic system. The gain of each amplifier was carefully adjusted to meet the 
requirements of time and spatial measurements. The pulse shapes were examined at each
74
of the amplifiers with an oscilloscope to ensure the right amplitudes and widths. Each 
timing gate was logically planned and generated in the right time sequence and for the 
right time duration. Several combinations of the amplifiers were tested and the best 
choice for each measurement was decided. Software was debugged and some applications 
were programmed on the spot. It took at least a month before each round of measurement 
to prepare the electronic system and the radioactive gas source.
The typical experimental procedure involved cooling the substrate from room 
temperature to 16K in about 3 hours, and dosing the sample with radioactive 37Ar. For the 
37Cl ion recoil time-of-flight measurements, the dosage was usually about I L (I L =IO"6 
Torr x I s = 5 xlO"9 Torr x 200 s) of 37z36Ar gases or 37Ar made from 40Ca(n,a)37Ar mixed 
with the residual gases in the source chamber. The sticking coefficient of Ar at 16 K is 
nearly unity; hence I L dosage would produce a monolayer (I ML) coverage (the actual 
coverage might be a little high because the local partial pressure near the dosing nozzle 
could be higher than the ion gauge reading). Two independent measurements were 
conducted simultaneously: The CMA was used to measure the kinetic energy distribution 
of the Auger electrons emitted following the EC decay of 37Ar, and the MCP’s were used 
to measure the time-of-flight of the desorbing ions by recording the coincidences between 
the 37Cl ions and the Auger electrons following the EC decay of 37Ar. The counting 
electronics for this measurement is shown in Fig 3.13 and Fig 3.16. The ion detector was 
biased -2700 V at the front side of the first MCP to prevent fast Auger (KLL) electrons 
from entering. The electron detector had a screen voltage, Vs = -50V to measure both
75
KLL and LMM Auger electrons, or Vs = -250 V to measure only KLL Auger electrons. 
For the recoil energy (near 9.5 eV) and the flight distance (4-6 cm) of interest in this 
measurement, the time-of-fight of the 37Cl ions was about 10 (is. Therefore, the time 
range of the TAG was set at 20 jis full scale to balance the time resolution and accidental 
rate. Because of the low counting rate, all other sources that could contribute to ion or 
electron counts were shut off during the measurement. The ion pump, ion gauge and 
electron gun were turned off and even the view ports were covered to prevent room light 
from getting into the MCPs. The system was maintained under UHV conditions (2 x IO"10 
Torr) by the He cryopump that also cooled the sample. A fresh layer of Ar gas was 
adsorbed every 8 hours for measurements with I L coverage.
For some surface interaction studies multiple layers of Ar gas were adsorbed. The 
energy and charge distributions of 37Cl ions at various coverage were measured directly 
by scanning the retarding screen voltage of the ion MCP detector from 0 to +20 V. The 
time-of-flight of 37Cl ions was measured by coincidence techniques for different 
coverage. The same measurements were also conducted for the 40Ar and 36y37Ar mixture.
hi the 37Cl double Auger study, 6 to 10 L of 37Ar source was dosed every 20 
hours. The MCP detectors were all biased for electron detection. In the EC decay, a KLL 
Auger electron will always be followed by the LMM Auger cascade electrons. 
Coincidence is expected for two detectors receiving KLL and LLM Auger electrons. 
However, if one biases both detectors with screen voltage of -250V, there should be no
76
LMM Auger electrons detected. In this case, the two MCP detectors should not yield any 
true coincidence counts unless there is double Auger decay with both of the electrons 
emitted having kinetic energy higher than 250 eV. The probability of the double Auger 
process and the energy distribution of these double Auger electrons were measured using 
coincidence techniques. The correlation between the KLL Auger electrons and each TTM 
and LMM Auger peaks were also examined by using coincidence techniques. By 
measuring two sets of coincidence simultaneously and recording them in the real time 
sequence, the triple coincidence, Le., all three detectors receiving signals at the same 
time, could then be examined by replaying the recording tape.
Most of the above measurements were repeated by using different substrates and 
different Ar sources. The results reported here for the 37Cl ion time-of-flight and surface 
interaction studies were mostly obtained from the experiments using the first radioactive 
gas source and the first experimental setups. The results of 37Cl Auger relaxation, studies 
were obtained from the experiments using both radioactive gas sources and both 
experimental setups. A triple-event coincidence measurement was attempted in the last 
experiment to study the double Auger decay.
77
CHAPTER 4
RESULTS AND DISCUSSIONS 
Desorption due to Recoil Induced by Neutrino Emission 
Energy Distribution of 37Cl Ions
In this section the energy distribution of the 37Cl ions desorbed due to recoil 
induced by neutrino emission will be presented. In order to gain a better understanding of 
the experimental results, one should first have some idea about what to expect from these 
measurements if the 37Cl ions had no interaction with its surroundings. Ideally, without 
interaction or a massive neutrino, the recoil energy of the 37Cl ions should be well defined 
(9.54 eV) as calculated in Chapter 2. In reality, there are many factors which might 
contribute to the broadening of the energy spectrum even without the substrate surface 
interactions. These “intrinsic factors” which affect the energy spectrum of the isolate 
atoms will be discussed first.
As discussed in the Chapter 2, an isolated 37Ar atom decays 90% of the time by 
electron capture from the K-shell (the K-shell binding energy of the daughter atom 37Cl is
78
2.822 keV), otherwise mainly from the Li-shell (270 eV). This leads the isolated daughter 
37Cl atom (assumed to be initially at rest) to recoil with a kinetic energy of 9.54 eV if the 
neutrino accompanying the decay is massless. This energy will be reduced by mv2/2M if a 
neutrino with mass mv is emitted, where M  is the mass of the 37Ar. In the experiments 
reported here, the 37Ar atoms were physisorbed on a cold substrate with a binding energy 
of less than 100 meV;15 thus the recoil is expected to break the weak absorption bond. 
The K-hole created as a result of the EC reaction decays 90.3% of the time via Auger and 
9.7% of the time via x-ray emission. Note that the 37Cl atom with the K-hole is initially in 
a neutral charge state. When this 37Cl atom decays by emitting Auger electrons and/or 
photons sequentially in a cascade, several new intermediate ion states will be produced 
until a stable final state is reached. The typical lifetime of an Auger decay is about IO"15 
second, during which time the 37Cl ion moves less than 0.1 A.
There are several intrinsic factors which lead to broadening in the recoil energy 
distribution of the Cl ions. First, the 37Ar atom does not start out at rest, but has some 
initial thermal velocity, which at 16 K leads to a spread in the energy of the daughter 37Cl 
of about 3% (or 0.28 eV). The Auger decay of the 37Cl also gives a recoil and hence 
broadening in 37Cl ion energy. A 2.4 keV KLL Auger electron ejected from a 37Cl atom 
moving with 9.54 eV will lead to a broadening of 12.2% (or about 1.15 eV) due to recoil. 
This can be corrected if the direction of the Auger with respect to the Cl ion velocity is 
known. For the time-of-flight (ToF) geometry reported here, the Auger recoil slows down 
the Cl ions by 4.3% (or reduces the energy by 8.3%), with a residual spread of about 4%
79
due to the angular opening of the MCP detector. Similarly, an LMM Auger recoil will 
change the 37Cl ion energy by 3.52% (or 0.34 eV). On the other hand, a 2.6 keV x-ray 
photon will cause only a 0.65% (or 0.06 eV) change in energy of the Cl ions due to recoil. 
There is also broadening due to instrumental effects, such as variation in the flight 
distance over the finite size of the detector and source, and variation in the time ions 
spent between the retarding screens and the detector. A Monte Carlo simulation was 
performed to generate ToF spectra which take all of the above effects into account. The 
result shows a velocity broadening of about 20% if either a KLL or an LMM Auger is 
detected in coincidence with the ion, and a broadening of 14% if only a KLL Auger is 
detected in coincidence with the ion (Fig. 4.1).
ion/KLL+LMM
ion/KLL
£  750
^  500
Time- o f -  f l ig h t  ( f x  s)
Figure 4 .1 The Monte Carlo simulations of the 37Cl ion recoil velocity distributions. Solid 
line: 37Cl ions in coincidence with either KLL or LMM Auger electrons. Dash line: 37Cl 
ions in coincidence with KLL Auger electrons only.
80
The above discussions and simulations dealt with only the isolated 37Ar atom 
case. Comparing with these estimates of what to expect for an isolated 37Ar atom, the real 
case, the results of actual measurements are presented below.
Figure 2(a) shows the ToF spectrum of the recoiled 37Cl ions, taken in coincidence 
with Auger electrons having kinetic energies above 50 eV (-50 V retarding voltage on 
MCP2 screen), i.e., in coincidence with either an LMM Auger electron (E -180 eV) or a 
KLL Auger electron (E -  2.4-2.6 keV). The experiment was conducted by using the first 
experimental setups as shown in Fig. 3.13. The sample was dosed to I ML of the 36/37Ar 
mixture. A +0.5 V bias was also supplied to the MCPl detecting positive ions. The 
spectrum represents a total counting time of 172 hours.
The ToF spectrum shown in Fig. 2(a) was converted to an energy distribution 
spectrum using the measured distance of travel (7.8 cm) between the source and detector. 
Figure 2(b) shows the resulting kinetic energy distribution of the 37Cl ions extracted from 
the ToF spectrum in Fig. 2(a). The absolute velocity scale was also verified by measuring 
the shift in the centroid of the ToF spectrum as the ion detector was moved by a known 
distance. The spectrum does not distinguish the charge state of the ions.
The energies (Fig. 2(b)) of the ions range from 5 eV to 13 eV with a maximum at 
around 9 eV. An important point to note here is that the ToF spectrum is taken in 
coincidence with 37Cl KLL or LMM Augers, hence it is guaranteed that the ions are
8 1
1000
C
m 750
CO 
CO
b 500
A  
in
4 - 1
§ 250
O 
O
0
0 5 10 15 20
T im e of F lig h t (us)
12000
>
w 8000
Sn 
CD 
Oh
in
-g 4000
5
O 
U
0
0 5 10 15 20
K ine tic  E nergy  (eV)
I I I I I I I I I I I I I I I I I I I I I
Ion TOF sp ec trum
Iflj
—
A
I1 I
-  (a) \
\I*t V1;,
Figure 4.2 (a) Time-of-flight spectrum of the desorbing 37Cl ions taken from the 
coincidence between 37Cl ions and 37Cl Auger electrons, (b) Accidentals-subtracted 
energy distribution of the desorbing 37Cl ions, obtained from (a) using the measured time 
and distance of flight.
82
ejected as a result of neutrino emission. Compared with the estimates of an isolated atom, 
there are two features which were not anticipated: (I) The high energy tail of the 
spectrum (>9.4 eV) is too long to be accounted for by neutrino recoil or by the 
uncertainties in the
co in c id en ce  w ith  K  o r  L A ugers
172 h o u rs Monte Carlo 
free  a tom
co in c id en ce  w ith  K  A ugers
42 h o u rs
10 15
Time of flight (yi/,s)
Figure 4.3 Comparison of ToF between Monte Carlo simulations and the results of 
coincidences: (a) ions and KLL or LMM Augers (b) ions and KLL Augers.
83
energy measurement (better than 4%), and (2) the energy spread (FWHM » 3eV) of the 
37Cl ions is larger than the 20% estimate for the isolated atom case. The TbF spectrum in 
coincidence with KLL Augers only was also measured. Figure 3(b) shows the ToF 
spectrum of the desorbing 37Cl ions, taken in coincidence with Auger electrons having 
kinetic energies above 250 eV (-250 V retarding voltage on MCP2 screen), i.e., in 
coincidence with KLL Augers electrons (E ~ 2.4-2.6 keV) only. This measurement used 
the same system as the one in Fig. 4.2(a) except the counting time was 42 hours. The 
FWHM of this spectrum was about 30%, whereas the expected spread was estimated to 
be 14% (superimposed in Fig. 4.3). Therefore there had to be other factors which would 
also contribute to the broadening of the recoil energy spectrum beside the intrinsic 
factors. These “surface interaction factors” will be discussed in later sections.
Charge-State Distribution of 37Cl Ions
Immediately after the EC process, the 37Cl atom is left with a K-hole or an L-hole 
and in a neutral charge state. This is a highly excited state which decays within IO"15 s by 
emitting Auger electrons and/or photons sequentially in a cascade. The 37Cl atoms then 
become positively charged ions with different charge states. This charge-state distribution 
is decided by both the different relaxation channels of the initial K-hole or L-hole and by 
charge exchanges of the 37Cl ions with their neighbors. Like the case of the recoil energy 
distribution study, the charge-state distribution of the isolated 37Ar atom will be examined 
first in the next paragraph.
ZThe charge-state distribution of the isolated Ar (gas phase) ion with an initial Is 
hole has been discussed in Chapter 2. The theoretical and experimental results are listed 
in Table 2.5. Compared to Ar, each charge state of 37Cl ions should have one charge less 
than the corresponding charge state of Ar ions resulting from the same Auger cascade. 
The charge-state distribution of 37Cl ions in the gas phase has been measured by Kofoed- 
Hansen23 and by Snell and Pleasonton22. The two experiments were in rough agreement, 
except for the charge I' state, where Snell and Pleasonton got a much smaller fraction. By 
using the decay probabilities given by Chen, Crasemann and Mark76 and taking into 
account that 90% of the 37Cl start as neutral atoms with a Is hole and 10 % with a 2s hole, 
a theoretical estimate was made by Monte Carlo simulations. The result for the charge I 
■ state is 9.1%, which is much closer to Snell and Pleasonton's value than to Kofed- 
Hansen's. For the other charge states the results are also in reasonable agreement with
Snell and Pleasonton's data. The experiments cited above were not sensitive to neutral
<•
recoils, which can be estimated to constitute < 1% of the decays. The experimental and 
simulation results are listed in Table 4 .1
84
Table 4.1 Experimental and Simulation Results on 37Cl Ion Charge State Distribution.
Charge +1 +2 +3 +4 +5 +6 +7
Exp.l(%)23 26±3 13+4 38±4 18±2 4+1 1+1
Exp.2(%)22 6.2+. I 15.7+.4 39.2+.5 26.7±.4 10.0±.2 1.8+. I .4+1
Monte Carlo 9.1 18.8 48.7 17.7 5.1
85
The charge-state distribution of 37Cl ions which interact with their neighbors is 
now examined in this section. First, it is necessary to explain how to measure the charge- 
state distribution with a MCP based retarding field energy (RFE) analyzer. The RFE 
analyzer is an integrating spectroscopy and is sensitive to the charge states of the ions. 
The counts, C(V) of 37Cl ions at screen voltage, V of the RFE analyzer are related to the 
kinetic energy, E of the 37CEne ions (where «=1,2,3,...) through
= M l
where C(n,V) is the contribution of the n-fold charged ions to the counts at screen voltage 
V and Y,dC (n,V) / dE is the energy distribution of the 37Cl ions as shown in Fig. 4.2(b).
Figure 4.4 shows the effects of charge state and energy distributions to a RFE spectrum.
i3
3
R e ta rd in g  vo ltag e  (V) R e ta rd in g  v o ltag e  (V)
Figure 4.4 The distribution of (a) charge states and (b) ion energy, (c) The RFE spectra of 
q=+l ions(solid line) and q=+2 ions(dash line), (d) The RFE spectrum of all ions.
8 6
A spectrum with charge state and energy distributions as shown in Fig. 4.4(a) and Fig. 
4.4(b) would yield a RFE spectrum as shown in Fig. 4.4(d). Therefore the RFE spectrum 
is a convolution of energy and charge state distributions. If the energy distribution could 
be acquired by other means, the charge states distribution could be in principle obtained 
from the RFE spectrum.
12000 
11000 
10000 
9000
C/3-4-J
§  8000 
o 
U
7000 
6000 
5000 
4000
Figure 4.5 Retarding field energy (RFE) spectrum of the desorbing 37Cl ions from I L 
36737Ar coverage. The vertical dashed line marks the position of the recoil energy of 37Cl 
ions in the gas phase (9.54 eV)
Figure 4.5 shows the retarding field energy (RFE) spectra of the desorbing 37Cl 
ions entering the MCP-RFE detector. The positive voltage applied to the middle screen 
(Fig. 3.13) provided the retardation. Each screen is 90% transparent, with a 100 pm mesh
I 1 I 1 I 1 I ■ I 1 I 1 I i I 1 I
-',IttIHt -
■ —
:
■
•
♦tlHHt
- V t t
— 'tH
♦ ♦
'ttI,:
H
- ' .................................................... ...
i , I ,  I , I I I I I I I I I I I
- 2  0 2 4 6 8 10 12 14
R e ta rd in g  vo ltag e  (V)
87
size. The dosage was IL exposure of 36737Ar mixture with ~50 ppm 37Ar concentration on 
the AuZSi(Ill) substrate. The vertical dashed line in Fig. 4.5 marks the position of the 
recoil energy of 7Cl ions in the gas phase (9.54 eV). There are several interesting 
observations that can be made from this spectrum.
The first one is the charge-state distribution of the desorbing 37Cl ions. Figure 
4.2(b) suggests that more than 95 % of the ion counts fall between 6 eV and 12 eV kinetic 
energy. Therefore, it is safe to assume that 6 eV and 12 eV are the lower and upper 
kinetic energy limits, respectively, for all 37Cl ions desorbing from a monolayer of 36737Ar 
mixture.
12000
10000
<4$- + 3  26%
6 0 0 0
4 0 0 0
R e t a r d i n g  V o lta g e  (V)
Figure 4.6 Charge state distribution obtained from the 37Cl ion RFE spectrum.
88
These limits will be used to quantify the charge state distribution of 37Cl ions desorbing 
from the surface. These kinetic energy limits imply that all ion counts above V = 6 V in 
Fig. 4.6 are due to 37Cl+le ions. Further, the increase in the counts between 3 and 6 eV is 
due to 37Cff2e ions, and the increase in counts between 0 and 3 eV is due to 37Cffne ions, 
with n > 3. The analysis of Fig. 4.6 yields immediately that ~ 53% of the total ions have 
charge +e, ~ 21 % have charge +2e and ~ 26 % have charge +ne, where n > 3 (Fig. 4.6). 
Since the MCPs cannot detect the neutrals, the percentage of the neutral 37Cl atoms 
desorbing can not be measured directly. The indirect measurements suggesting an upper 
limit of 90% for neutralization will be discussed later.
The charge state distribution obtained from the RFE spectrum is substantially 
different from that expected for isolated atom decay (Fig. 4.7(a)). The comparison 
between the simulated RFE spectrum of the isolated 37Ar and the real RFE spectrum is 
shown in Fig. 4.7(b). This discrepancy will be explained in terms of charge exchange and 
subsequent Coulomb repulsion in the discussion section.
The second point that needs to be addressed here is the difference between the 
highest kinetic energy of the 37CI+6 ions measured by RFE and that found by ToF. The 
maximum uncertainty in the energy measurements is less than 4 %, but yet the threshold 
of RFE starts at around 10 eV while the highest energy suggested by ToF is around 13 eV 
(Fig. 4.8). One possibility is that the energy distribution depends on the charge states of
89
100
^  80
C 60 
O
u  40 
cdSn
^  20 
0
0 1 2 3 4 5 6  7
Charge s t a t e
6000 
5000 
m 4000 
5  3000 
u  2000 
1000 
0
- 2  0 2 4  6  8  10 12 14
Re ta rd ing  voltage (V)
Figure 4.7 Comparison between the simulation of isolated Ar and the experimental results 
in (a) charge states distribution, and (b) the RFE spectra.
i I ' | ' I ' I ' I I I I I r
1,1,1"< ltlIf-Ip;'- •
' " ' , . 1I
Simulation
Vi'ii,
'i1 mil
Tl
'H. (b)
—  - I  , -t u  - V i 1 VirtT ttri K  i-
I , I , I I I , I , I , I , I , I
I  I I I I I
H Experim ent 
I l S im ulation
(a)
JO_____L
the ions, Le., the highly charged ions would have higher energy and these ions would not 
show up in the high voltage end of the RFE spectra because of the multiple retarding 
nqV. This assumption was tested by comparing the width of the ToF distribution with the 
retarding potential set to 0.5 V (which would allow all Cl ions to enter the detector) with
90
the width of the distribution obtained by setting the retarding potential to 6.0 V (which 
would allow only singly charged Cl ions to enter the detector). Both distributions were 
obtained in coincidence with KLL Auger electrons. There was evidence of about 1 0 - 15  
% narrowing but the statistics were poor because the source had decayed considerably by 
the time this measurement was conducted.
Ion E nergy  f rom  ToF
12000
8000
"O  4 0 0 0
K ine tic  E nergy  (eV)
RFE S p e c trum  ■
6000
5000
, ,  4 0 0 0  
S  3000
U  2000
R e ta rd in g  vo ltag e  (V)
Figure 4.8 The maximum kinetic energy, Ek obtained by (a) ToF spectrum and (b) RFE 
spectrum. The shadowed region could be the contribution of higher charged ions.
91
Results on Multiple and Mixed Lavers
The discrepancy between the expected charge-state distribution and the RFE 
experimental results and the higher maximum kinetic energy indicate that the 37Cl ions 
have interacted and exchanged charges with their neighbors during the desorption 
process. In order to understand these interactions, several experiments were carefully 
designed. These experiments include studies of 37Cl ion desorption processes from 
multiple layers of 36737Ar gases and from mixed layers of 36737Ar and 40Ar gases.
Figure 4.9 shows the RFE spectra Of37Cl ions from different coverage (1L, 2L, 4L 
and 7.4L) o f 36737Ar.gases on the cold Au/Si(l 11) substrate. The sticking coefficient of Ar 
is about I at 16 K for up to 10 ML thickness.13 This is confirmed by observing the linear 
increase of the KLL Auger electron counts (at ~ 2.4 keV energy) as a function of 
exposure, and also by comparing the count rate with that expected from the initial total 
decay rate of 2 x IO4Zs per monolayer of the 36737Ar mixture (Chapter 3). Here a negligible 
loss of the KLL Auger electrons is assumed as they penetrate through a few Ar layers. 
Reproducibility of Ar coverage is the biggest unknown in the measurements. As the 
thickness of the Ar film increases, both the charge state and the energy distribution of the
37Cl ions change. This is shown in Fig. 4.9(b) - Fig. 4.5(d), where the highest kinetic 
energy of the. 37Cl ions, Ek(max) increased to above 15 eV and the profile of the RFE 
spectra changed considerably over a range of ~7 L of 36737Ar exposure.
92
R e ta rd in g  field  ion  en e rg y  sp e c trum
..... ...%12000
9.54 eV
7.4 L
4 0 0 0
12000
2000
q 7000
2000
10000
5000
R e t a r d i n g  v o l t a g e  (V)
Figure 4.9 Retarding field energy (RFE) spectra of the desorbing 37Cl ions at different 
coverage (I L to 7.4 L). Note that as thickness increases there is an increase in Ek(max), 
and a change in the profile of the RFE spectra.
93
Figure 4.10(a) shows two superimposed RFE spectra from a monolayer 36737Ar 
film. In one case (full line), the active film was physisorbed on the top of a 3L 40Ar film 
and in the other case (dashed line) the active 36737Ar film was deposited first on the Au 
substrate and followed by 3L of 40Ar film. The curves are spline fits to the data points
3 L Ar  + I L'
- I L
R e ta rd in g  Voltage (V)
m 500
^ 400
Total
Ion C oun ts 
Ek (m ax)
Ar Exposu re  (L)
Figure 4.10 (a)Two superimposed RFE spectra from IL 36737Ar layer. Dashed line: IL 
36737Ar at the bottom of 3L passive 40Ar film. Full line: IL 36737Ar on the top of 3L 40Ar 
film, (b) Dashed line: total ion intensity from a I L 36737Ar as a function of 40Ar coverage. 
Full line: Ek(max) from a I L 36737Ar file as a function of the 40Ar coverage.
94
(circles). Notice that the maximum kinetic energy, Ek(max), of the 37Cl ions is about 13.5 
eV in both cases but that the profiles and the intensities are very different. The profile of 
the top curve is similar to that of Fig. 4.9(a), while the profile of the bottom curve is 
similar to that of Fig. 4.9(d). The 40Ar between the active 37Ar and the substrate (top 
curve) served as a buffer layer to prevent charge exchange between the 37Cl ions and the 
substrate. The fact that the charge state distribution did not change appreciably when the 
active layer was separated from the Au substrate by an additional three layers of 40Ar 
means that the Au substrate is unlikely to be the significant source of charge exchange for 
the desorbing 37Cl ions.
The variation in total ion yield and Ek(max) of a single layer of active 36z37Ar is 
shown in Fig. 4.10(b) as functions of passive 40Ar physisorption exposure over the active 
layer. The total 37Cl ion intensity decreases linearly at a rate of ~ 12 % per 40Ar layer, and 
Ek(max) increases linearly at a rate of ~ 10 % per 40Ar layer within 6 L of 40Ar deposit. 
There is about a 5% uncertainty in determining E(max) from the RFE spectra, because of 
low counts at the thresholds.
The broadening of the energy distribution spectra (EDS) determined by ToF 
spectroscopy is shown in Fig. 4.11. The two normalized EDS correspond to IL and 5L 
36z37Ar mixture physisorbed on the AuZSi(Ill) substrate. Note that the 5L spectrum 
ranges from 5 to 16 eV while the IL  spectrum ranges from 5 to 13 eV. The EDS are 
rather symmetric in shape and the majority of the broadening takes place, on the high-
95
energy side of the EDS. There are some low kinetic energy ions below 5 eV but their 
contribution is rather small. The broadening in energy is explained in terms of charge 
exchange and Coulomb repulsion in the discussion section.
1200
900
>
CD
Sm
CD
^  600 
W
"ti3ou
300
0
0 5 10 15 20
Kinetic Energy (eV)
Figure 4.11 Energy distributions from a IL 36737Ar film (full circles) and 5L 36737Ar film 
(empty circles). The counts are normalized. Notice the increase in Ek(max) and counts.
Desorption Probability of 37Cl ions
Ion Energy
O 5 L 36/37.'Ar
.  I L
4
4 t
* A"
/
X j f r
\
The probability (3 that a 37Cl ion is desorbed as a result of the neutrino emission 
accompanying the decay of its parent 37Ar will be examined in this section. This 
probability can be determined reasonably well, as (3 can be measured independently by
96
two methods: (I) the direct measurement using RFE spectroscopy, and (2) the indirect 
measurement using coincidence techniques. They produced the same result as expected.
Direct measurement of the probability P was performed by measuring first the 
number of KLL Auger electrons emitted from a I L 36z37Ar mixture. A retarding'voltage 
of -400 V (selecting only KLL electron) was applied to MCPl detector which was 
operated in electron detection mode for this measurement. Background counts (typically 
1-2 cps) were carefully measured to find the net KLL Auger counts. This was followed by 
direct 37Cl ion counts measurement in the same MCP detector biased now for positive ion 
detection. With a screen bias of +0.5 V, the MCP detector would detect all positive ions 
plus x-ray photons, if any. The background counts varied around 3 cps for the ion 
detecting mode and no x-rays were observed. The MCPs have a detection efficiency of 
2+1 % for x rays, 85+10% for 37Cl ions at 2.7 keV, and 85+2 % for electrons at 2.5 
keV. Since only 9 % of the total decays were accompanied by K x rays, only
about 0.2 % of the total electron counts were due to x rays, and these could be ignored. 
Taking into account that there are 0.81 KLL Augers for every 37Cl EC decay (90 % of 
capture occurs from the K shell and 90 % of the K holes decay by Auger emission), the 
KLL Auger electron and the 37Cl ions counts for a total N0 EC decay events should be
Nkll = N0 Pk (I -w) fK>4oo (AQ/4k) Tk Sr Vk fiive (4.2)
N io n  -  No PioNfioN>o.5 (AQ/4tz) Tion £/0n Tj10M fiive (4 .3)
97
where Pk = 0.9 is the K-capture probability, w = 0.1 is the radiative transition probability 
for a K hole,, f K>4oo =fioN>o.5 ~  I are the portion of a KLL Auger electron whose energy is 
greater than 400 eV and an 37Cl ion whose energy is greater than 0.5 eV when leaving the 
surface (scattering effect), (AQ/4k) is the solid angle of the MCP detector, Tk = T10n = 
0.72 are the transmissions of the three screens in front of the MCPs for KLL Auger 
electrons and ions, £K ~ S1on are the efficiencies of the MCP and the electronics system 
for 2.5 keV KLL Auger electrons and 2.7 keV ions (accelerated), tjk = TI10n are the 
chances of escape for KLL Auger electrons and ions from anything which might cover 
them during the measurement and f Uve is the computer efficiency (live time/dead time). 
Some of the values given above may not be exact and there might be some other factors 
which will affect the yields, but most of these factors will cancel each other when 
calculating the desorption probability P10n by dividing Eq, (4.2) by Eq. (4.3)
P ioN = 0.81 N10n/  Nki.j. (4.4)
From Eq. (4.4) the desorption probability of 37Cl ions is calculated as P10n = 9.4 ±1 .2  %. 
This result suggests that, at most, 90.6 % of the Cl atoms are recaptured or desorbed in 
the neutral charge state. The largest uncertainty (12 %) in the measurement is due to 
uncertainty in the detection efficiency of the 37Cl ions.
The second method of measurement makes use of the coincidence techniques. The 
basic idea is to compare: (I) the number of coincidences between the 37Cl ions and the 
KLL Auger electrons, (i-e), and (2) the coincidence between the LMM and KLL Auger
98
electrons (e-e). Typical (i-e ) and (e-e) spectra are shown in Fig. 4.2 (a) and Fig. 4.12, 
respectively. The FWHM of the peak in Fig. 4.12 is ~ 2 ns. Most of the broadening of the 
peak comes from the energy spread of the LMM Auger electrons ranging from 100 eV to
' I ' '
e—e Coincidence S p ec trum
- - * •  -  l  I I  H i I f <  I i i w i w x i i i i M  I I i i i i i i  I r f I ^ m r T A i iH i I i i n i i
50 60 70 80 90
Time (ns)
Figure 4.12 KLL and LMM e-e coincidence and ToF spectrum. Most of the broadening of 
the peak is due to a spread in the energy of LMM electrons (100-200 eV).
200 eV (~ 4 ns spread in 8 cm). The absolute time resolution of the system is better than I 
ns. The detailed analysis of Fig. 4.12 will be in next section. Here, only the coincidence 
counts of the (i-e), Nk,on and (e-e), Nkl are needed.
Nkl = NoPk (1-w) [ 2 fKLL + f KLM ] f K>4oofL>ioo (AQ/4n)2 Tk Tl £k £l r\K Tji fuve (4.5)
NfiiON = No Pk (1-w) PionfiON>o.5 (AT2/4k)2 Tk Tion £k £ion t)k t]ionfive (4.6)
where 2 fKLL + f KLM =1.84 is the number of LMM electrons created per KLL Auger decay 
and fKLL = 0.85 and fKLM= 0.15 are chances of a K hole decay through KLL and KLM
99
Auger, respectively. Other parameters are self-explanatory with the subscript K  and L 
standing for KLL and LMM Auger. The desorption probability can be expressed as
P io n= 1-84 N kion /  N kl (4.7)
The resulting desorption probability of 37Cl ions is found to be, Pion = 9.6 ±1.2 %, in 
excellent agreement with the result of the first method.
Discussion of the Desorption Results
The discussion of the desorption results will be centered around two related 
topics: the local environment surrounding an 37Ar atom, and the interaction of the 37Cl 
ion with its surroundings and possible mechanisms for energy broadening.
In order to form a solid Ar film on the substrate, the sample had to be cooled 
down to 16 K in 3 hours. The entire sample surface facing the MCPs was uniformly flat 
with no physical defects. It is quite likely that there was at least one layer of residual gas 
physisorbed on the Au surface during the time the sample was cooled down, although it 
could not be detected by Auger spectroscopy. The 36/37Ar film was most likely 
physisorbed on top of this layer. For one (or less) monolayer o f36/37Ar film, a typical 37Ar 
will be surrounded by 36Ar atoms on the surface and residual gas molecules underneath 
(Fig 4.13(a)). Typical residual gases in this UHV system were H2, CHn, H2O, CO, CO2, 
and 36/37Ar mixture. An experiment for I L exposure was typically run for 8 hours, with 
ToF spectra saved every two hours automatically before desorbing the old layer and
100
adsorbing a fresh layer of 36737Ar. Time-dependent analysis of ToF and RFE spectra 
showed no noticeable differences in the count rates or the shapes of the spectra during the 
8-hour period. The residual gases (if any) on top of the 36737Ar film introduced no 
observable effect. As mentioned earlier, layers of 40Ar physisorbed over and/or under an 
active layer of 36737Ar film caused the same amount of increase in Ek(max) (Fig. 4.10) 
from 10 to 13.5 eV. On the other hand the RFE spectra vs. V resembled Fig. 4.2(a) if the 
active layer was on top Of40Ar layers, and looked like Fig. 4.9(d) if the active layer was at 
the bottom of the 40Ar layers. Fig. 4.10(b) shows a linear increase in Ek(max) as 40Ar 
layers pile up on top of the 36737Ar active layer. This behavior is puzzling, since it suggests 
that not only the near-neighbor atoms but also atoms far from the active center interact 
with the 37Cl ions during desorption. A second puzzling behavior is the large penetration
Figure 4.13 Local environment of 37Cl ions at (a) IL 36737Ar coverage and (b) more 
realistic case for 40Ar on the top of IL active 36737Ar mixtures— the morphology of the 
films is a porous structure.
101
depth of the 37Cl ions, as indicated in Fig. 4.10(b). It is possible that the morphology of 
the films is not close-packed but rather a porous structure, which would explain the 
behavior shown in Fig. 4.10(b). The Au substrate, which forms the base for the Ar films, 
is not an atomically flat surface, and could thus cause a porous film structure (Fig. 
4.13(b)). This would increase the effective surface area and would explain the large 
penetration depth and the long-range interaction of 37Cl ions during desorption.
As mentioned earlier, the EC decay of 37Ar and the subsequent Auger cascade 
leave behind a positively charged 37Cl ion. The majority of the 37Cl ions are initially in an 
n= 3 charge state (Table 4.1), which makes them highly reactive and ready to accept 
electrons from the sample environment during their motion, since the electronic 
transitions are very fast (typical lifetime of IO'15 s) and can easily occur before the ion 
leaves its initial environment (typical speed of I cm / jxs). Figure 4.6 suggests that 53 % 
of the total ions are in the n = I state, 21 % in n = 2, 15% in n = 3 and 11% in n > 4 as 
opposed to about <1 % in n = 0, 6 % in n = I, 16% in n = 2, 39% in n = 3, and 38% in n 
> 4, in the gas phase.22,23 Essentially all the recoiling 37Cl should be positively charged, 
but only 10% of them were being detected. The question one then asks is what happened 
to the other 90 %? The fact that 53 % of the observed ions have a charge of +1, whereas 
only 6 % are predicted to have charge +1, could have two explanations: (I) Most of the n 
= 2 and higher charge states do not desorb, thus increasing the apparent fraction of n =1 
desorbed ions, and/or (2) most of the ions are neutralized by charge transfer before
102
desorbing. The second argument is clearly more plausible as the cause of the high fraction 
of +1 ions. One piece of evidence for this is that coincidences were observed between 
charge +1 ions and KLL Augers. In the gas phase the n = I ions result only from K holes 
which decay by X- ray emission22 and should not be in coincidence with the KLL Auger 
electrons, while the overwhelming majority of Auger decay of K-holes yield ions of 
charge 2 or higher. Therefore, the observation of charge I ions in coincidence with KT T 
Augers suggests that the ions became singly charged by picking up electrons after the 
Auger decay. This shows that the final charge state is a strong function of the local 
environment of the 37Cl ions.
Whether a 37Cl ion can transfer charge between itself and the neighboring atoms 
(or substrate) is determined by the ionization potentials of the interacting species. The 
ionization potentials of 37C r ne ions are about 13, 23, 40, 54, and 68 eV for n = 0, 1 , 2 , 3  
and 4, respectively. The ionization potentials of the surrounding species are 16 eV for Ar, 
4.5 eV for the Au substrate, between 10 and 13 eV for CHn (n < 4), and between 13 and 
15 eV for CO, CO2, H2O, and H2. According to the discussion in chapter 2, the available 
energy has to satisfy Eq. (2.18), E > I  - A + T, for singly charged ions to transfer charge. 
These numbers indicate that energetically a single charged 37CI+6 ion can easily receive an 
electron from the Au substrate and be neutralized, but that it cannot receive an electron 
from a nearby Ar atom. The neutralization of a singly charged 37CI+1 ion (either comes 
directly from the 10% x-ray relaxation or from a higher charge ion losing its charge) will 
be discussed in next paragraph.
103
In general, both resonant and direct Auger mechanisms may contribute to the 
neutralization. Since the work function of the surface is about 5 eV, a 13 eV deep 37CI+16 
vacancy is enough for the direct Auger transition to occur. The direct Auger transition is 
the major channel for the neutralization where another electron is ejected from the 
substrate when the 37CI+16 is neutralized. Hagstrum26 has deduced that essentially all 
neutralization occurs by this mechanism for ions with energies less than 10 eV. The 
survival rate of singly charged ions approaching a metal surface has been studied by 
theoretical and experimental groups.26,27 In some cases, the metal is modeled as “jellium”, 
Le., as an electron gas embedded in a constant positive background. A good approxima­
tion to the probability density for electron transfer from the metal surface to an ion on an 
incident trajectory is given phenomenologically by Hagstrum and Snowdon et. uZ.26,27 as
/  „ \ Z X
A A
exp — e — ubI h J I a HIJ e ( s - f o ) (4.8)
where S is the distance of the ion from the surface, v is the velocity in direction 
perpendicular to the surface, 0(x) is the step function and S0,, a and A  are constants 
determined by fitting experimental data. Since only the absolute value of v enters the 
equation, this relation should be applicable to the case of outgoing ions, as in this 
experiment. For this application, it is appropriate to set S0 =S, the mean distance beyond 
which the ion will not be neutralized.
S/v) = v t +  rAr (4.9)
104
where Tis the mean life time of the 37Cl excited state and rAr is radius of 37Ar atom prior 
to decay. The velocity of a 9.54 eV 37Cl ion is 0.071 Ms, and the mean life time is about 
9 fs, which is the sum of the times filling holes in K-shell (~lfs) and the L-shells (~8fs) 
sequentially. Taking the radius of an Ar atom on the surface to be 1.86 A,24 one obtains Sz 
= 2.5 A. By using the values suggested by Hagstrum for the parameter a = 2.29 A"1 and A 
=210 fs"1, the 37Cl ion charge transfer probability density Pt vs. S is calculated as shown in 
Fig. 4.14(a). The probability of survival of Gl+ ion at distance S is then
M s H  = 1^ n(S-M) = 1 -JJPM ',M M S ' (4.10)
which is shown in Fig. 4.14(b). From Fig. 4.14 (b) one can see the survival probability of 
Cl+ ions beyond 5 A is predicted to be very small. This explains the experimental result 
that up to 90% desorbed species were neutral. After neutralization, an Antoniewicz 
mechanism as illustrated in Fig. 2.3 might add additional energy to these desorbing 
neutrals on the top of the 9.54 eV recoil energy.
i
^0.75
o
'°-5
0 .2 5
0
D is tan ce  (A) D is tan ce  (A)
Figure 4.14 (a) The Cl+ ion charge exchange probability density vs. distance, (b) The Cl+ 
ion survival probability vs. distance.
105
One way to prevent the neutralization, is to put several layers of stable Ar (e.g., 
40Ar) between the substrate and the 37Ar. As discussed above, a 37CI+1 ion cannot transfer 
charge from an Ar atom nearby, so therefore neutralization will not occur this way. The 
effect of such a “buffer, zone” has been observed by Dujardin, et. a l68,69 in their PSD 
experiment of Ar on Pt. They report the yield of Ar+ ions increases rapidly with the 
addition of each monolayer of Ar between the second and fifth monolayer, but reaches 
saturation for thicker layers. In fact the effect of the extra layers in between can be 
expressed as
Sz(v) = vz+  rAr+ M  rAr (4,11)
where M  is number of stable monolayers and rAr- is the effective thickness of one 
monolayer of Ar (2.63 A, assuming a fee  structure of Ar)16. Using the numerical value and 
simple relationships given in Eq. (4.11), the survival probability from the substrate charge 
transfer neutralization, Ps is near I for M > 3.
The charge transfer of a multiple charged 37Cl+ne (n > 2) ion can occur through 
different channels. Beside being neutralized by the conductive substrate, it can also 
receive an electron from any species surrounding it and reduce its charge state by one 
charge. This charge transfer is mainly due to the resonant tunneling mechanism. Similar 
to the “Coulomb explosion” model for covalent systems discussed in Chapter 2, the 
multiple vacancies left in the final state of the Auger cascade in the 37Cl ions can be filled 
by the electrons from neighboring atoms. The energy required for a 37CI2+ ion to transfer a 
charge with a neighboring Ar atom is E > (Vn - Vrfci -(Vi)a,- -T, where V1 and Vn are the
106
first and second ionization potential, and T is the kinetic energy gained in the process (Eq. 
(2.19)). The data shown before indicate that this process can happen for n > 2 ions. For 
example, a 37CI+2 ion can release 23 eV energy when it receives an electron to become 
37CI+1 , while a neighboring Ar atom only needs 16 eV energy to supply the electron. For 
this 37CI+2 ion, after the initial charge transfer, further charge exchange is strongly 
inhibited because the 37CI+1 ion cannot exchange charge with neighboring Ar atoms. This 
gives an extra long life time for the two adjacent positive ions. Part of the 7 eV excess 
energy released in this charge exchange will become kinetic energy of the interacting 
species and the rest will most likely remain as internal energy (e.g., excited states) (Fig. 
4.15). For an 37CTh3e ion, charge exchange would release even more kinetic energy. This 
effect would lead different charge states to have different energy distributions.
Vacuum
level
Coulumb
repulsior
Figure 4.15 (a) The charge exchange of a 37CI2+ with the neighboring Ar atom, (b) The 
Coulomb repulsion between the two holes resulting from the charge exchange
107
As explained earlier in this chapter, the expected recoil energy of the 37Cl ions is 
9.54 eV with a broadening of 20 % (2 eV), due primarily to the recoil induced by Auger 
electrons. The measured ion energy distribution for the I ML active layer on Au is a 
Gaussian with a maximum at 9 eV and a FWHM ^ 3 eV. The maximum possible 
uncertainty in the energy distribution measurement is less than 4 %. Hence while the 
energy centroid is close to the expected position, the energy broadening is at least 50 % 
larger than expected. This means that charge exchange and the subsequent Coulomb 
repulsion shape the energy distribution. In the case of the I ML active layer it is expected 
that the kinetic energy of the 37CFne (n > 2) ions is greater than that of the 37CFe ions as 
shown in Fig, 4.8. The reason for this is that most of the 37Cl ions have already exchanged 
charges with the residual gas layer or with the Au substrate. Since the 37Cl ions are on top 
of the residual gas layer, the charge exchange and the subsequent Coulomb repulsion give 
more energy to multiple charges than to single ones. In the case of multiple layers the 
situation is different: most of the active atoms have passive layers above them. In this 
case it is possible that a 37CFne ions could transfer most of its kinetic energy and part of 
its charge to an Ar atom. The resulting Coulomb explosion would give1 additional kinetic 
energy to the Ar+e ion, which would then have a chance to be detected in coincidence 
with an Auger electron from the 37Cl atom. A single-charged Ar detected in this way 
would have more or less the same energy as the multiply-charged 37Cl ions. The detector 
cannot differentiate between a 37Cl and an Ar ion entering the detector. This idea will be 
tested in the future using Xe overlayers on an active 36z37Ar layer. The mass difference is 
large enough to differentiate whether a 37Cl or a Xe ion is entering the detector.
108
Auger Relaxation of a 37Cl Atom
The electron capture decay of 37Ar leaves the 37Cl atom with a K-shell hole 90% 
of the time and an Li-shell hole 10% of the time.8 These highly excited initial states of the 
37Cl atom will decay through radiative (x-ray emission) or non-radiative (Auger) 
transitions. The K-hole created as the result of EC process will decay 90.3% of the time 
via Auger and 9.7% of the time via x-ray emission.8 As discussed in Chapter I, the EC 
decay creates an excited 37Cl atom whose initial state is unusual in three ways. First, there 
is a single hole created in the K-shell while the rest of the shells are essentially intact, 
leaving the 37Cl atom with the electronic configuration of Ar. Second, the initial state is 
neutral, unlike the ionized state obtained by electron or photon bombardment. Third, with 
90% of the electron capture occurring in K-shell, this initial state is almost a pure K-hole 
state, while the initial states obtained from the conventional ways are predominantly L- 
hole states. For example, when Ar is bombarded with 5 keV electrons, the fraction of K- 
holes created is over two orders of magnitude less than that of L-holes. It might be 
possible to improve this ratio, with a photon beam having variable energy, but multiple 
ionization associated with photo-absorption might obscure the multiple ionization 
associated with decaying of the K-hole. Therefore, this unique initial state brought about 
by the EC process allows the direct observations of some interesting relaxation processes 
which are amenable to many-body theoretical calculations, but essentially impossible to 
probe experimentally using conventional techniques. This section is devoted to a 
discussion of the two phenomena which have been predicted theoretically,4’30’31 but not
109
observed directly: the shift in energy of an normal LMM Auger line due to a multiply 
ionized L shell and the decay of a K-hole by double Auger emission in the 37Cl atom.
37Cl LLM and LMM Auger Peaks
As discussed in Chapter 2, right after the EC decay, the core levels of an 37Ar 
atom change immediately to the core levels of a 37Cl atom with the K-shell at 2822 eV, 
the L1 shell at 270.2 eV, the L2 shell at 201.6 eY, the L3 shell at 200.2 eV, the M1 shell at 
17.5 eV, and the M2 and M3 shells at 6.8. eV (Table 2.2). From Eq. (2.26), the KLL Auger 
energies of a 37Cl atom could be calculated as from 2232 eV for KL1L1 (6.9%) to 2377 
eV for KL2j3L2j3 (55.4%) Apger. Unfortunately, these energies are out of the measuring 
range of the CMA used in this work which can only detect a kinetic energy up to 2000 
eV. Therefore, the emphasis of study was put on the LLM and LMM Auger cascades 
following a KLL Auger. The LMM Auger energies were calculated as from 150 eV 
(L3M1M1) to about 250 eV (L1M2j3M2j3) with major peaks at around 170 eV. Two 
independent experiments were conducted to study the LMM and LLM Auger energies: 
the retarding field energy distribution measured by the channeltron or MCP detectors and 
the differential energy distribution (dN/dE) measured by the CMA.
Figure 4.16 shows the RFE spectrum of a 4 ML 36737Ar mixtures on the 
AuZSi(Ill) substrate taken in the early stage of the experiment by using a channeltron 
detector with retarding screen (Fig. 3.5). The main feature here is the sudden change of
no
slope from 150 V to 180 V retarding voltage (The absolute voltage values are shown in 
this section. For electron measurements, the retarding voltages are negative.), hence 150
Sam p le  b locked
R e ta rd in g  Voltage (V)
Figure 4.16. The RFE spectrum of 37Cl LMM Auger. The lower curve was the result 
when the sample was blocked.
eV to 180 eV kinetic energy that corresponds to Cl atom LMM Auger peaks centered 
around 170 eV. This result was the first indication that confirmed the adsorption of 37Ar 
atoms on the cold substrate. The spectrum is flat after 200 V retarding voltage and would 
stay this way until it reaches the KLL Auger energies. A maximum of 1000 V retarding 
voltage has been applied in similar experiments and no surprising result was found above 
200 V retarding voltage. The result of the measurement during which the sample was 
mechanically blocked is also shown at the lower part in Fig. 4 .1. This curve shows the
I l l
very low background count rate in this experiment. The poor resolution of this spectrum 
is due to the intrinsic limitation of the RFE detectors.
M 1000
. LLM
Energy (eV)
Figure 4.17 Comparison of varies Auger spectra. Full line: 37Cl LMM following EC 
decay. Dotted lines: regular Auger from 40Ar and CsCl excited by electron bombardment, 
with their background subtracted. LMMh labels the hypersatellite peak.
Figure 4.17 shows three superimposed Auger spectra in the LLM and LMM 
regions of Cl and Ar. The full line is the 37Cl Auger spectrum associate with 37Ar EC 
decay. No background subtraction or signal smoothing has been carried out to present the
112
data. The spectrum was taken with a double pass CMA. The peak labeled as TTM and 
LMM are those associated with known Cl Auger lines in the Auger spectrum taken from 
CsCl.salt, also shown in Fig. 4.17. The CsCl spectrum was taken by Phi-595 scanning 
Auger system and was normalized and background-subtracted to fit to the scale of the
37Cl Auger. For comparison, it also shows the LMM Auger peak of natural 40Ar gas 
physisorbed on the same substrate and bombarded by a 1000-eV electron beam. It is 
worth noting that the 37Cl spectrum was taken with only 5x10"5 effective monolayer, 
while the natural 40Ar spectrum was taken with multiple layer coverage. In Fig. 4.17, one 
can compare the Cl Auger peaks from the 37Cl and from CsCl. The LMM and LLM 
Auger peak energies (-170 eY and -40  eV) and shapes from these two different sources 
are similar except the different backgrounds and an extra peak labeled as LMMh, a 
hypersatellite peak. This extra peak became one of the subjects of this work. A higher 
resolved spectrum with the LMMh peak is shown in Fig. 4,18. This spectrum was taken 
with the new radioactive 37Ar source made from the 40Ca(n,a)37Ar nuclear reaction and 
the new substrate of pyrolytic graphite. The fact that LMMh peak appears in both 
experiments with different radioactive sources and substrates indicates that it is not a 
solid state or surface contamination effect. The LMMh peak with such a large shift (about 
22 eV higher energy than the normal LMM line) has not been reported before. Notice that 
LMMh peak does not appear in the normal Cl Auger spectra. A high resolution gas phase 
study on Ar LMM Auger29 did not report such peak either. The new peak was only seen 
in this experiment with the 37Ar sources.
113
5 0 0 0
_  4 0 0 0
m 3 0 0 0
o 2000
Energy  (eV)
Figure 4.18 17Cl LMM Auger peaks measured from the new radioactive source and 
substrate.
Discussion Of 37Cl LMM Auger Results
The discussion of the LMM Auger results will be centered around two related 
topics: the peak assignments of the 37Cl LMM Auger spectrum and the hypothesis 
explaining the extra LMMh peak. As discussed in Chapter 2, the LMM Auger peaks of 
the 37Cl following the EC decay would likely have energies of normal Cl Augers and 
structures of normal Ar Augers. Table 2.4 shows the energies of twelve major Auger lines 
from a highly resolved Ar LMM Auger spectrum.29 Unfortunately, there is no Auger 
spectrum of Cl with such high resolution found in the reference. To get the corresponding 
Cl Auger peak energies, each group of the Ar Auger lines in Table 2.4 was shifted
114
5000
_  4000
d 3000
o 2000
E n e rg y  (eV)
Figure 4.19 The normal Cl LMM Auger energies marked on the top of the 37Cl LMM 
Auger spectrum. Note that no normal Auger peak is corresponding to the LMMh peak.
according to the calculated energy differences between Ar and Cl Augers. For example, 
the calculated L2M2^ M2J Auger energies were 212 eV and 182 eV for Cl and Ar, 
respectively. Therefore, the Cl L2M2 2M2J  Auger energy was obtained by shifting 30 eV 
from the corresponding Ar Auger energy. The resulting Cl LMM Auger energies are 
marked in Fig. 4.19 on top the 37Cl LMM Auger spectrum. The Auger transitions with 
their final state spectral terms corresponding to these twelve Auger lines are, from higher 
energy to lower energy, L2M2i3M2j3(3P0iu), L2M2j3M2j3(1D2), L3M2i3M2i3(3P0iIi2), 
L3M2i3M2i3( lD2), L2M2i3M2i3 (1S0), L3M2i3M2i3(1S0), L2M,M2i3(3Poj), L2MiM2i3(1Pi),
115
LgM iM^(1Pi), Lz Mi M i(1Sq) and Lg Mi M i (1Sq). Figure 4,19 shows the good agreement 
between the energies of these normal Cl LMM Auger peaks and the 37Cl LMM Auger 
spectrum except the extra LMMh peak.
The explanation of the LMMh peak requires a detailed understanding of this 
many-body relaxation process. A hypothesis is given in the following paragraphs 
regarding this mystery peak. The area under the LMMh peak is about 12+3% of the total 
LMM Auger of 37CL Another interesting characteristic of the LMMh peak is the 
sharpness of the peak (~5 eV FWHM). As discussed earlier, normal Auger peaks are 
broadened by the interactions of the two holes in the final states through different 
coupling of the spin and orbital momentum. For normal Cl LMM Augers, the S-L 
splitting can be as wide as 5 eV. Transitions from different initial states in L2 and L3 can 
also have different Auger energies and further broaden the LMM Auger peak. The total 
FWHM of the LMM Auger peak in Fig. 4.19 is in the order of 20 eV. Therefore, the 
sharpness of LMMh Auger peak might indicate that there is no hole-hole interaction in 
its final state.
One way to achieve a non-interactive final state is to have only one electron left in 
the M23 subshell after the LMM Auger transition. This final state must lose five of the 
valence electrons in the M23 subshell in the Auger cascades. Figure 4.20 illustrates such 
an Auger cascade channel. The K-hole first decays by a KLiL23 Auger transition, which 
happens 22% of the time. The Li hole then decays by a Coster-Konig Auger transition
1 1 6
LlL2i3M2i3, which happens about 46% of the time. The two L2i3 holes then decay through 
L2i3M2i3M2i3 and four more electrons are emitted. Five electrons are lost this way and the 
last L2i3M2i3M2i3 Auger transition has a non-interactive final state. The Auger decay 
probability of this combination is about 8% of all the LMM Auger transitions. This is 
somewhat lower than the 12+3 % area under the LMMh peak. The energy shift is also 
believed to be the result of the core level shifts due to the multiple ionization by this 
particular Auger cascade.
(c) LggMggMgg (d) L23M23M23
Figure 4.20 (a)-(d) Auger cascade for the hypothesis explaining the LMMh peak.
117
The reason the LMMh peak would not have been observed in the LMM Auger 
spectra produced by electron bombardment is that the initial states created by electron 
excitation are mostly L-holes. As mentioned above, the ratio of L to K shell holes created 
by a 5 keV electron beam bombardment is SjJSk =120. Therefore the LMMh peak in such 
experiment would be 120/12%. =1000 times smaller than the normal LMM peak and 
would be effectively buried in the background.
Double Auger Decay of a 37Cl Atom
The deep K-shell hole created by the EC process in the 37Cl atom can sometimes 
decay through another Auger process where the energy released by filling the K-hole 
electron is shared by two ejecting electrons simultaneously. This is called a double Auger 
decay process.4,40"42 As discussed in Chapter 2, the double Auger decay process is caused 
by three electron-electron correlation effects:4 the virtual inelastic scattering effect where 
an outgoing normal Auger electron in an intermediate state is inelastically scattered by 
another electron; the cascade mechanism where the initial K-hole decays through a 
normal Auger transition, producing a vacancy in one of the intermediate shell and then 
decaying via another Auger transition; the shake-off model where the initial K-hole 
decays through normal Auger decay, then the atomic field caused by this Auger transition 
shakes off another electron. In each of these models, the probability of the double Auger 
decay process is the product of the probabilities of each step in the process. The energy 
distribution between the two outgoing electrons is expected to be highly asymmetric due
118
to the first and third mechanism. This section is devoted to the study of the double Auger 
decay probability and the energy distribution between the two double Auger electrons.
37Cl Double Auger Probability
In most double Auger processes, the two emitted electrons will share the energy 
provided by the filling of a K hole by an L electron. These two electrons would have 
continuous energy spectra instead of discrete Auger lines as in normal Auger 
spectroscopy. The probability of the double Auger process cannot be very large because it 
is a second-order process. Therefore, in conventional Auger spectroscopy, these two 
double Auger electrons would be buried in the huge background of secondary electrons, 
and hence essentially impossible to measure directly. A breakthrough in the study of the 
double Auger process has been achieved in this work by using the coincidence techniques 
to measure the time correlation between the two electrons emitted by the process. The 
two double Auger electrons are always emitted simultaneously and always share a fixed 
total kinetic energy, E m a x  = 2200 eV (Ek -3El). Their correlation in time and energy 
makes them ideal candidates for the coincidence measurements.
After an EC process, a K-shell hole will relax mostly by the Auger cascades of the 
KLL and LMM Auger transitions. In a normal Auger decay, a KLL Auger electron whose 
energy is about 2400 eV is always in coincidence with the cascading LMM Auger 
electrons whose energies are about 170 eV. Therefore, two MCP detectors will always
119
register coincidence counts if one of the retarding screen voltages is set below 170 V (Fig. 
4.12 and Fig. 4.21). However, if the retarding voltages at both detectors are set above the 
LMM Auger energy (e.g., at 250 V), no electron-electron coincidence should be 
registered from a normal Auger process. Therefore, the recording of the coincidence 
events between two electrons having energy above the LMM Auger energy in this work 
provides the first direct evidence of the double Auger decay process (as discussed early, 
the coincidence contribution between x-ray and KLL Auger electron is essentially zero).
Vp^=IOOeV
Vlpft=SSOeV
cu 50
T im e (c h an n e ls )
Figure 4.21 Coincidence spectrum of KLL Auger and LLM Auger electrons taken with 
MCP screen voltages at 250V and 100V.
Figure 4.22 shows two electrons having kinetic energy more than 250 eV in 
coincidence with each other. This is the first direct evidence of the double Auger decay 
process. The measurement was taken with 10 L radioactive 37Ar source made from
120
40Ca(n,a)37Ar reaction and deposited on the graphite substrate for about 5 hours. The two 
MCP detectors (L and PSD) were biased for electron detection and their single counts 
were also recorded simultaneously. The distances from both detectors to the sample were 
6 cm. The shortest time in the coincidence peak (left edge) was due to the coincidence 
starting by a 250 eV slow electron and stopping by a 1950 eV fast electron (2200 eV-250 
eV), whereas the longest time in the coincidence peak (right edge) was due to starting by 
the fast electron and stopping by the slow one. The calculated time difference between the 
left and right edge should be 7.35 ns. This time difference corresponding to about 70 
channels is in good agreement with the coincidence peak as shown in Fig. 4.22.
T im e  ( c h a n n e l s )
Figure 4.22 Coincidence spectrum of two electrons emitted in the double Auger process.
121
The probability of the double Auger decay was measured by two independent 
experiments both utilizing coincidence techniques. The first experiment measured the 
coincidence counts, Nda, i-e, the area under the coincidence peak in Fig. 4.22, the 
coincidence counts, Nkl, Le., the area under the coincidence peak in Fig. 4.21 and their 
single counts of the KLL Auger electrons, NK(da) and NK(Kl) which were simultaneously 
recorded. The Nk and Nkl have been expressed in Eq. (4.2) and Eq. (4.6) and are listed 
here as Eq. (4.7) and Eq. (4.8)
Nk = N0 Pk (I -w) f K>2so (AQ/4k) Tk eK Thcfiive (4.7)
Nkl = NoPk (1-w) [ 2 fKLL + fKLM ] fK>25ofL>wo (AQ/4tz)2 Tk Tl £k £l TJk rjLfiive (4.8) 
where all the parameters has been explained following Eq. (4.5). The Nda have the form 
Nda = NoPk (1-w) PDAfEi>25ofE2>250 (AQ/4k)2 Tei Tei £ei £e2 t]ei T]E2 fiive • (4.9)
where Ej and E2 stand for the energies of the two double Auger electrons (E1 < E2). The 
factors fEi>2sofE2>250 < 7 are the portion of double Auger decays with both double Auger 
electrons having more than 250 eV kinetic energy. This factor takes into account the 
effects of intrinsic energy distribution and the slowing down of electrons on the way out 
by inelastic scattering. The energy-dependent electronics system efficiency e and escape 
probability rj are taken here with their average values because E1 and E2 are not fixed 
energies any more, unlike the case in KLL and LMM Auger. From Eq. (4.7) and Eq, (4.8) 
one can get
Nkll /  NK(KL) = (2fKLL +fKLM)fL>ioo(AQ/4jz) Tl £l rjL (4.10)
From Eq. (4.8) and Eq. (4.9) One can get
122
n DA / NK(DA) = Pda (fEl>25ofE2>25o/fK>25o) (AQ/47V) TE1 &E1 VEI (4.11)
where the parameters associated with higher energy electrons are assumed to be equal, 
i.e.,TE2 ~Tk, £e2 ~ e^and T]e2 ~ Dividing Eq.(4.11) by Eq. (4.10), one can get
where I /kel +fiMM =1.84 is the average number of LMM Auger electrons following a K- 
hdle that has decayed by KLL or KLM Auger transitions. In general, the electronics 
response factor e and the escape probability p  are larger if  electron energy is higher. For 
this measurement, Ej > 250 eV >EL, and therefore £Ei > £t, and T}ei > t)l- However, each 
pair of these parameters are not far away from each other. It is also known that f L>1o (fK>25o 
< I and/^>250 ~ I. Now defining the probability, P ’DA as the probability of the measured 
portion of double Auger decay (Ej > 250 eV and E2 > 250):
P tDA = (fEl>25ofE2>25o) P m  . (4.13)
Then the upper limit of P ’DA can be calculated as
The data used to calculate P ’DA were obtained from Nda ,NK(da), Nk l, and NK(KL). They 
were measured as close together in time as possible to ensure that similar radioactive 
source and electronics conditions were maintained invariant during the experiment. 
Substituting the data from Fig. 4.21 and Fig. 4.22, one can calculate the upper limit of the 
double Auger decay probability to be 16.1+0.6%. Two other measurements with similar
KLL f  LMM ) (4.12)
P ’DA <M .8 4 (4.14)
123
conditions yield upper limits results as 15.4+0.7% and 11.9+0.6%. The weighted average 
result for the upper limit of double Auger decay probability is 14.9+0.4%
too V
250 V
R e ta rd in g  Voltage (V)
Fig. 4.23 RFE spectrum of 37Cl LMM Auger. The counts C2 were from KLL Augers 
while counts C, were the contribution from both KLL and LMM Augers.
The lower limit of the double Auger decay probability can be obtained by 
combining the discussion above with the retarding field energy (RFE) spectrum of the 
37Cl LMM Auger. Figure 4.23 shows a RFE spectrum taken with the same MCP detector 
under the same experimental conditions as the coincidence measurement. The counts C2 
at 250 V retarding voltage were from KLL Auger electrons while the counts C, at 100 V 
were the contribution from both KLL and LMM Auger electrons. They have the forms
Ci = No Pk (1-w) (AQ/4n) Tftive [ / k > 100 £k %  + ( 2 /kll  + /klm ) fi.>wo£L (4.15)
124
C2 — No Pk (1-w) (AQ/4jt) Tfuve f K>25o £k Vk (4.16)
Their ratio, R ’= C1VC2 will be
R ’ =  fK>10(/fK>250 + (2 / klL +fKLM) (fL>100£L £k Vk) (4.17)
where/^>/00 ~ f K>250 and 2fKLL + fuLM  = 1.84. From Fig. 4.23, the counts Ci = 965 and C2 
= 393, therefore
R ’ -  I +1.84 (fL>100£L Vl / fK>250 £r  Vk) =2.46 (4.18)
This gives
(fL>ioo£i V l / / k >250 £k  V k )  = 0.8 
From Eq.(4.19) one can get
1-25 = (ftoiso £k  V k / fL>ioo£L V l )  > (1/fioioo) ( £ e i  V e i  /  fL > io o £ L  T|l) 
Comparing Eq. (4.20) and Eq. (4.12),. one can then get
P ’da  -  ( f s i  >250 fE2>25o) P DA
r u m
I V
J l\ J  Z,>100 /  V
e E J l  E1
> — 0.8
1.84
(4.19)
(4.20)
(4.21)
This gives
0.8 #1.84 3 3 #1.84 (4.22)
Substitute the results as when calculating the upper limit, one can calculate the lower 
limit of the double Auger decay probability to be 11.9±0.3%.
The second experiment designed to measure the double Auger probability used a 
indirect approach. As shown in Fig. 4.21, if  one of the MCP screen voltages is higher 
than 250 V and the other is lower than 170 V, a coincidence peak from KLL and LMM
125
Auger electrons can be registered. If the high and low retarding screen voltages are 
switched, the coincidence peak would be centered at a different channel (time). Now 
imagine reducing both screen voltages below 170 V then one would expect to see three 
coincidence peaks appearing on the spectrum: the KLL starting and LMM stopping 
coincidence peak on the right (late time); the LMM starting and KLL stopping 
coincidence peak on the left (early time) and the LMM starting and: LMM stopping 
coincidence in the middle. If the double Auger process happens, the triple vacancies left 
in the L-shell by this process would largely enhance the LMM Auger transitions, and thus 
increase the middle peak. Therefore, the size of the middle peak compared to the left or 
right peak size could yield some information about the double Auger decay probability.
Figure 4.24 shows coincidence spectra taken with different MCP detector screen 
voltages. The upper one (Fig. 4.24(a)) was taken with the starting (MCP detector) screen 
voltage set at 250 V and the stopping screen voltage set at 100 V, with the coincidence 
peak centered to the right (late time). The lower one (Fig. 4.24(c)) was taken with the 
starting screen voltage set at 100 V and the stopping screen voltage set at 250 V, with the 
coincidence peak centered to the left (early time). The middle one (Fig. 4.24(b)) was 
taken with the starting and stopping screen voltages both set at 100 V, and the 
coincidence peak is centered at the middle. Part of the tails of Fig. 4.24(a) and Fig. 
4.24(c) were extended to the middle range below the center peak. The background of the 
center peak in Fig 4.24(b) was calculated by scaling these tails. The contribution of the 
right peak to the center background was calculated by taking the area ratio between the
1 2 6
right peak and its tail in the center region in Fig 4.24(a) and multiply the inverse of this 
ratio to the right peak area in Fig 4.24(b). The contribution of the left peak to the central 
background was obtained in the same way. The background was then subtracted from the 
center peak in Fig. 4.24(b) to get the net area of the peak. The resulting net areas of the 
left peak, Al_k and the center peak, Al-l in Fig 4.24(b) are 1653 and 993, respectively.
£  4 0
(D BO
o  4 0
Time (ch ann e l s )
Figure 4.24 Coincidence spectra with different MCP detector screen voltages. 
(a)KLL(start) /LMM(stop), (b)LMM/LMM and (c) LMM/KLL.
127
This gives the ratio Rexp=  {Al.k/AL-l) =1.66 which contains the contribution from double 
Auger process. On the other hand, each Auger electron emitted by filling a K-hole (KLL 
or KLM) will be followed by 2fKLL + f KLM = 1 .8 4  LMM electrons and each LMM 
electron can only find 1.84 -I = 0.84 LMM electrons to be in coincidence with. If a 
double Auger decay happens, the filling of the three extra L vacancies left by the double 
Auger process will create three more LMM Auger electrons. The slower of the double 
Auger electrons will also contribute the same as an LMM Auger electron. These four 
electrons would have six different combinations of coincidence to contribute to the center 
peak but only four ways to contribute to the left peak. Assuming the double Auger decay 
probability to be Pda,, the contribution of the double Auger process to the center peak 
would be 6xPDa , whereas the contribution to the left peak would be 4xPDA. Therefore the 
double Auger decay probability can be calculated as
Jggp= ^84 + 4A  - 1.66
0.84+ 6A (4.23)
This gives a result Pda = 8+3% which is somewhat lower but still comparable to the 
results from the other independent experiments which set the double Auger decay 
probability to be between 12% to 15%.
37Cl Double Auger Energy Distribution
The energy distribution between the two emitted electrons is expected to be highly 
asymmetric because of the virtual inelastic scattering and shake-off mechanisms in the
128
double Auger process.4 However, the energy has to be conserved during the process. This 
means the total kinetic energy, Etot of the two emitted electrons must be equal to the 
energy released by the initial transition from L-shell to the K-hole
E tot = (Ek -El ) - 2EL (4.24)
The last term in Eq. (4.24) is the energy required for two L shell electrons to escape above 
the vacuum level. Substituting the energy level values from Table 2.1, one can get the
CU
C
a
Cti
-aO
G
GO
U
CUOC
CU
O
G
O
U
300 /1300
J-1L 1 [ 4 0 0 /1 300
500 /1300
600 /1300
700 /1300
800 /1300
900 /1300
320 340  360 380 400 420
T im e  (c h a n n e ls )
Figure 4.25 Coincidence spectra with different MCP screen voltages as indicated by the 
numbers to the left of the peaks.
129
total kinetic energy of the two emitted electrons, Exot ~ 2200 eV if most transitions 
involve the L2i3 level. Due to this energy correlation, the kinetic energies of the two 
emitted electrons are no longer independent, e.g., limiting one of them to be greater than 
250 eV would also restrict the other one to be less than 1950 eV. Therefore, the 
electronics would register coincidence contributions from double Auger pairs with 
different combination sharing the total 2200 eV energy if different retarding screen 
voltages are applied in the coincidence measurements. This provides a possible way to 
map the energy dependence of the double Auger probability by scanning the retarding 
screen voltage.
In the actual experiment, the kinetic energy of one of the electrons was kept above 
1300 eV by the retarding voltage of the MCP detector and the other one was changed 
from 200 eY to 1300 eV in 100 eV step. This way the faster one of the two double Auger 
electrons would always enter the high biased detector, thus simplifying the measured 
coincidence spectra. The experiments were conducted by using 6L of 36737Ar source on 
the AuZSi(Ill) and 10L 37Ar source from 40Ca(n,a)37Ar reaction on the graphite 
substrate. Figure 4.25 shows the coincidence spectra with different MCP screen voltages 
taken for 20 hours each with 6L 36737Ar source on the Au substrate. The single counts on 
both MCP detectors were recorded simultaneously to provide the normalization reference.
Figure 4.26 shows the measured energy dependence of the probability distribution 
for one of the two electrons emitted in the double Auger process. This is the first time
130
such measurement has been performed. The probability was measured by simply 
following the areas under the coincidence peaks like the ones in Fig. 4.25 as a function of 
the kinetic energy of one of the two electrons, as determined by the MCP screen voltages. 
Each area was normalized by its single counts The kinetic energy of the other electron 
was kept above 1300 eV by placing a retarding screen voltage of 1300 V. Note that the 
spectrum in Fig 4.26 is an integrated spectrum. For example, all the double Auger 
electron pairs with one of the pair having kinetic energy from 400 eV to 1000 eV and the
CD 400
CO 300
C 200
500 BOO 1100
R e ta r d in g  Voltage V1 (V)
1400
Figure 4.26 Double Auger probability as the function of one of the electron energies.
other one from 1200 eV to 1800 eV will contribute to the double Auger coincidence yield 
at the 400 V point in the x-axis The energy distribution, dN/dE is obtained by taking a
131
derivative of the curve in Fig. 4.26 with respect to the retarding voltage, V. Figure 4.27 
shows the resulting 37Cl double Auger decay energy distribution superimposed with the 
theoretical results of the energy distribution of the atomic Ne double Auger decay.4 
Notice that there is qualitative agreement between the experimental result in this work 
and the theoretical predictions (the 2 Ry to 25 Ry part of the Ne spectrum is shown for 
comparison). The preferred energy distribution of the double Auger emission is for one of 
the electrons to take most of the energy with the second one receiving the small 
remaining balance. This is due to the first and third double Auger mechanisms discussed 
above.
Ne theore tical 
37Cl experim ental
(0 1.5
Energy E1 (eV)
Figure 4.27 double Auger energy distributions by this experiment (solid line) and 
theoretical calculation for a Ne atom with 5 Ry to 25 Ry kinetic energies.
132
In the experiment with the new 40Ca(n,(X)37Ar source on the graphite substrate, 
the absolute double Auger decay probabilities vs. retarding voltages were measured. One 
of the retarding voltage was set at 1100 V and the other was scanned from 250 V to 1100 
V. Figure 4.28 shows the measured absolute double Auger probabilities. Each data point, 
{X,Y) in Fig. 4.28 stands for the contribution to the integrated double Auger probability 
(Y) from the double Auger electron pairs with one of them having kinetic energy from X 
eV to 1100 eV and the other 1100 eV to (2200 - X) eV. The coincidence counts at 
different retarding voltages were converted to the double Auger probabilities (%) by 
using the same calculations as in last section. It shows the double Auger probability for 
250 WllOO V screen voltages is 6±0.3 % and thus the 250 V/250 V screen voltages 
should be 12±0.6%, which is in good agreement with the value achieved before.
Vp=IlOO V
R e ta rd in g  Voltage V1 (V)
Figure 4.28 double Auger probability as the function of electron energy of one of the two 
emitted electrons. One of the screen voltage was always set at 1100V.
133
Discussion of 37Cl Double Auger Decay Results
To back up the claim that this work observed directly for the first time the double 
Auger decay process, it is appropriate to have several different experiments to confirm the 
observation independently. In this section, another positive evidence of the double Auger 
decay process will be discussed. Other topics discussed here include the factors which 
will affect the measurement of the probability and energy distribution of the double Auger 
decay, especially the scattering effects of the multiple layers of gases to the outgoing 
double Auger electrons.
The first direct double Auger evidence provided in this work is the recording of 
true coincidence counts between two electrons above LMM Auger energy of Cl atom. 
These coincidence counts were attributed to the emission of two double Auger electron 
sharing the total energy released by filling the initial K-hole by an L-electron. The excess 
coincidence counts between LMM and LMM Auger electrons in Fig. 4.24(b) provided 
more evidence of the double Auger decay process. If the double Auger decay process 
exists, the filling of the K-shell by an L-electron and the emission of two electrons from 
L-shell in the double Auger process would have left three L-shell vacancies 
simultaneously. This intermediate configuration qould not be achieved by any other 
transition except the double Auger decay process. The decay of this particular 
intermediate state would result in the emission of three LMM electrons simultaneously.
134
Therefore, the third direct evidence of the double Auger decay would be the recording of 
the triple coincidence of these LMM Auger electrons in a new experiment.
v :
IuC
Cto
I)
E
P
QJOC
QJ
rEO
g
O
U
K
KLL/LMM
LMM/LMM
LMM/KLL
h - j
3 -C o in c id e n c e  
L PSD R 
A: LMM/KLL/LMM 
B: LMM/LMM/KLL 
C: KLL/LMM/LMM
%
S
S
- J
S
- J
- J
- J
L /PSD  C o in c id en ce  T im e (c h an n e ls ) )
Figure 4.29 Schematic diagram of a triple coincidence spectrum. The solid region is 
where a LMM/LMM/LMM coincidence appears.
The experimental logic diagram has been shown in Fig. 3.16. A multiple layer of 
active 40Ca(n,(X)37Ar source was deposited on the cold graphite substrate. Three MCP 
detectors were biased with retarding screen voltages all set at 100 V. The coincidence 
between left (L) and middle (PSD) detectors and the coincidence between L and right (R) 
detectors were recorded simultaneously on the magnetic tape. The triple coincidence 
results would be then extracted by replaying the magnetic tape and plotting the data on a 
two dimensional diagram with the coincidence L and PSD as x-axis and L and R as y-
135
axis. The experimental results are shown schematically in Fig. 4.29. Most of the 
coincidence counts were the double coincidences (L/PSD or UR) which stay on the x or y 
axis. It means that when a double coincidence was registered, there was no coincidence 
between one of the two registered MCP signals from this double Auger with the third 
MCP signal. A point off the axis, however, represents three MCP signals arriving within 
a very short time interval, Le., a triple coincidence. Most of the triple coincidence counts 
were expected to be KLL/LMM/LMM coincidences come from a regular KLL and LMM 
Auger cascade. These coincidence counts are shown in the shadowed regions in Fig. 4.28. 
The fact that there are six counts in the middle (solid) region which are above the 
accidental background (near zero) provides an evidence of a triple LMMyLMMZLMM 
coincidence, and hence the third direct evidence of the double Auger decay process.
The techniques employed in this work to measure the double Auger probability 
were not as universal as one would like. Because the double Auger decay process depends 
strongly on the energy partitioning between the two double Auger electrons and on their 
relative angle, it is difficult to obtain the probability of this process relative to the normal 
Auger process from a measurement conducted over a limited energy range and at one 
relative angle. The preferred energy configuration of a double Auger emission is for one 
of the electrons to take most of the energy with the second one receiving the small 
remaining balance. The energy selected by the retarding voltage in this measurement 
could not be lower than 200 eV, otherwise the KLL and LMM Augers would contribute 
the coincidence counts. Therefore, those double Auger electron pairs with one of them
136
having energy less than 200 eV and the other greater than 2000 eV would not be 
measured by this measurement. This would affect the upper limit of the double Auger 
decay measured in this experiment. The angular correlation between the two electrons is 
predicted to be peaked in the forward direction for the preferred configuration and in the 
backward direction when energies are close to each other. In the present experimental 
setup, the MCP detectors were 45° from each other which is at the middle range of the 
curve of probability vs. angular separation predicted by theory.4 The position sensitive 
detector (PSD) in this experimental setup did not have the necessary range of angle yet to 
measure the angular distribution of the double Auger process. Such a measurement is 
definitely needed in order to calibrate the double Auger decay probability result. Further 
experiments will apply a moveable detector or a PSD detector whose angular separation 
is large enough to sense the angular effect.
The last point to discuss in this section is the effects of the scattering of the double 
Auger electrons escaping the substrates. This attenuation effect was clearly visible in Fig. 
4.24(b) where the left coincidence peak was smaller than the right coincidence peak. The 
left peak was caused by a fast (KLL) electron to stop at the middle (0°) PSD detector and 
a slow electron (LMM) to start at the 45° left detector. Since the attenuation effect was 
much more severe for a slow electron than for a fast electron, the actual effect was 
determined by how many layers of adsorbed gase's the slow electrons have to penetrate 
before leaving the surface. In the left coincidence peak, the slow electrons had to reach 
the 45° left detector, and hence had to penetrate ~1.5 times as many layers of gases,
137
whereas in the right peak the slow electron only needed to reach the O0 middle detector. 
The ratio of these two peaks can serve as a good reference of the attenuation effects 
denoted as 77 in the probability calculation equations (e.g., Eq. 4.14).
Besides the attenuation, another influence of the scattering effect is the slowing 
down of an outgoing double Auger electron. This would alter the measured spectra of the 
double Auger energy distribution. To study this effect, it is helpful to revisit the 
coincidence peak in Fig. 4.22 which was taken with both MCP screen voltages set at 
250V. As discussed above, this coincidence peak contained the contributions from double 
Auger electron pairs having kinetic energies from 250 eV to 1950 eV for one electron and 
1950 eV to 250 eV for the other electron. Each-energy combination will be registered at a 
different channel (time) in the coincidence spectrum. Therefore, the whole energy 
distribution information could be extracted from the Fig. 4.22 alone if the time resolution 
was adequate. Unfortunately, this is not the case because the time separation 
corresponding to a fixed energy difference changes when energy changes. It is much 
easier to tell two electrons with 100 eV and 200 eV kinetic energy than two with 2100 eV 
and 2200 eV. The time resolution was not good enough for fast electrons.
A simple Monte Carlo simulation was conducted to study the effects of the 
scattering (Fig. 4.30). The double Auger electron pairs in coincidence were generated 
according the energy distribution predicted by the theoretical work on Ne atom.4 The 
kinetic energy was scaled from 5 - 25 Ry to 200 - 1100 eV. The probability of generating
138
e E x p e r im e n t  
___ S im u la t io n
T im e  (c h a n n e ls )
Figure 4.30 Comparison between the experiment result and simulations on double Auger 
coincidence spectrum.
a double Auger electron with energy E was proportional to the corresponding double 
Auger yield, y(E) as shown in Fig. 2.10. Each electron generated was allowed to keep its 
original energy 80% of the time and the remaining 20% of the time to have a flat energy 
distribution (due to scattering) from this energy to the minimum detectable energy 
determined by the screen voltage. The other member of the double Auger electron pair is 
generated in the same way with a kinetic energy 2200 eV -E (Etot = 2200 eV). 
Furthermore, generating a random point on the source region of the substrate for the pair 
of double Auger electrons and allowing them to reach at random points on the two MCP 
detectors, the time-of-flight would then be calculated from the known energies and
139
calculated distances. A total number of one thousand events were generated to compare 
with the thousand or so coincidence counts in Fig. 4.22. Good agreement was achieved by 
this simple, simulation as shown in Fig. 4.30. Comparing Fig. 4.30 and Fig. 2.10, the 
scattering of the low energy electrons has totally distorted the energy distribution. To 
reduce this effect, higher concentrated radioactive 37Ar sources are required in the future 
experiment to reach the same amount of activity with less coverage.
140
CHAPTER 5
CONCLUSIONS
Conclusions
The emphasis of this thesis report is on the results obtained in the fields of surface 
science and atomic physics rather than the study in the possible mass limit of neutrinos. 
The nuclear reaction of 37Ar electron capture decay and the following relaxation of the 
core holes created by the electron capture process have provided exciting opportunities to 
study some novel physics phenomena which previously have been amenable only to 
theoretical calculations but impossible to probe experimentally. The study on the 
desorption of 37Cl ions due to recoil induced by neutrino emission and the study on 37Cl 
atom Auger relaxation yielded several surprising results in this research work.
Using coincidence techniques, the desorption of 37Cl ions from cold substrates 
due to recoil induced by neutrino emission in the 37Ar electron capture decay was 
observed for the first time in this experiment. The kinetic energy distribution of the 
desorbing 37Cl ions was accurately measured. The resulting energy distribution of 37Cl 
ions ranges from 5 eV to 13 eV with a maximum at around 9 eV. The spread of this
141
energy distribution (FWHM) is about 3 eV. The center of this energy distribution is close 
to the 9.54 eV calculated value while the spread of energy is much larger than the value 
for isolated 37Ar atoms. The charge state distribution of the desorbing 37Cl ions was also 
measured by using retarding field energy analyzers in addition to the coincidence 
measurements. The resulting charge state distribution is: 53% of the total ions have 
charge +e, 21% have charge +2e and 26% have charge +ne, where n > 3. This charge 
distribution is, again, substantially different from that expected for isolated decay atoms. 
The Cl ions desorption probability was also measured by the using of retarding fields 
energy analyzers and coincidence techniques. The resulting 37Cl ion desorption 
probability is 9.6±1.2%. This gives an upper limit of neutral desorption probability of ~ 
90 %. The recoil energies, the charge state distribution and the desorbing probability of 
the 37Cl ions have been found to be greatly affected by the charge exchange and the 
following Coulomb repulsion between the 37Cl ions and their surrounding atoms. This 
interaction dominates the broadening on the recoil energy spectrum, significantly alters 
the charge states of the desorbing 37Cl ions, and is responsible for the large neutral 
desorption probability. Further experiments to reduce these effect will be discussed in the 
next section.
The Auger relaxation of the K-hole initial state created by the 37Ar EC decay, 
especially the LMM Auger transitions, was studied by using coincidence techniques and 
conventional surface analysis techniques. The measured 37Cl Auger peaks following the 
EC decay of 37Ar were compared with the normal Cl Auger peaks obtained by
142
conventional means. The 37Cl LMM Auger transitions had most of the peaks with 
energies of the normal Cl LMM Augers and structures resembling normal Ar LMM 
. Augers. An extra peak, the hypersatellite LMMh peak with 20 eV ,higher energy than the 
normal Cl was observed for the first time in this work. It. is hypothesized that it is the 
result of the unusual initial state created by the EC decay and a special multiple-hole final 
state in the Auger transitions. Further work to understand this LMMh peak includes 
energy differentiated coincidence measurements between different Augers to identify the 
correlation of this strange peak.
The first direct evidence of 37Cl double Auger decay process was reported in this 
research work. In a double Auger decay process, the filling of the initial K-hole created 
from the EC decay by an L-shell electron results in the emission of two Auger electrons 
simultaneously. Using the coincidence techniques, the probability of the double Auger 
decay and the energy distribution between the two double Auger electrons were, 
measured. The experimental results in this work achieved good agreement with the 
theoretical predications.4 The probability of double Auger decay measured in this work 
are from 12+0.3% to 15+0.4% of the normal LMM Auger process. The preferred energy 
distribution between the two double Auger electrons is one of them taking most of the 
available energy and the second one taking the small remaining balance. Further work in 
this respect includes the measuring of the angular correlation between the two double 
Auger electrons and using this correlation to adjust the total probability of the double 
Auger decay measured in this work.
143
Perspective on Neutrino Mass Limit Study
It is demonstrated in this work that the surface interactions and atomic relaxation 
processes greatly complicate the simple picture of the two-body problem on the recoil 
processes described at the beginning of Chapter 2. The broadening of the recoil energy 
spectrum due to these surface and atomic processes circumscribed the resulting mass 
limit measured by this experiment. However, this pioneering effort of combining nuclear 
physics and surface analysis techniques to attack one of the most fundamental problem in 
physics is still of great merit and, with certain optimism, promising from the results 
obtained in the past three years. The information reported in this thesis work may provide 
valuable assistant in further experiments to search for neutrino mass limits. In this 
section, the resulting limits on the neutrino mixture ratio on certain mass range will be 
reported. Finally, the perspective on how to improve these experiments will be addressed..
The information on a massive neutrino in the EC decay has been extracted from 
the time-of-flight spectrum of the desorbing 37Cl ions.' The detailed data processing 
procedures are presented in the Appendix A. Here only a brief description will be given. 
The physically relevant quantity one tries to measure here is the mixing ratio (%) between 
the massless neutrino and the massive neutrino in this EC reaction. Tb obtain this ratio, 
the coincidence peak in the time-of-flight spectrum in Fig. 4.3(a) is first fitted with a
144
Gaussian and a constant background. A flat region to the right of the peak representing 
slower recoil ions hence massive neutrino emission is then fitted with a Gaussian and a 
flat background with the centroid and width fixed. The area ratio between the fitted 
"second peak" and the main peak provides the information on the neutrino mixing ratio. It 
is found that this experiment does not have the sensitivity for neutrinos with mass below 
Hiv = 600 keV. For neutrino with mass mv = 600 keV, the measured upper limit of the 
mixing ratio is 2+0.4 %. This limit is twice as high as the results obtained by other means 
in this mass region (~ 1%).
To get better results on neutrino mass limit, Le., to obtain a lower mixing ratio 
limit and at a lower mass region, a narrower recoil peak with better statistics will be 
needed. This would require the reducing of the broadening effects in the energy spectrum 
of the desorbing 37Cl ions in future experiments. In an EC decay process, the coincidence 
between the 37Cl ions and the x-ray photons would select only the singly charged 37CI+1 
ions with a negligible recoil due to photon and LMM Auger electron emission (3.5%). 
These 37CI+1 ions created by x'-ray relaxation instead of the Auger relaxation of the initial 
K-hole would be less likely to exchange charges with their neighbors due to the 
restrictions of energy conservation. This would significantly reduce the charge exchanges 
and the following Coulomb repulsion interactions, and thereby reduce the dominant 
broadening effect. Several buffer layers of the stable 40Ar will also be adsorbed before the 
active Ar is dosed on the cold substrate to deduce the charge exchanges between the
37Cl ions and the substrate or the residual gases per-adsorbed as the substrate being
145
cooling down. Simulations have shown that with the broadening effects reduced, the 
recoil time-of-flight coincidence peak would be appropriate to extract information on the 
mixing ratio between the massless neutrino and the 200 keV massive neutrino to achieve 
a limit much better than the results obtained by other means.
146
REFERENCES
147
1. V. Barger, F. Halzen, M. Marshak and K. Olive, Neutrino Masses and Neutrino 
Astrophysics (Telemark IV, World Scientific, Ashland, Wisconsin, 1987).
2. H.V. Klapdor and B. Povh, Neutrino Physics, Proceedings o f an International 
Workshop Held in Heidelberg, October 20-22,1987 (Springer-Verlag, Berlin, 1988).
3. F. Boehm and P. Vogel, Physics o f Massive Neutrinos (Cambridge University Press, 
London, 1992).
4. M.Ya. Amusia, IS . Lee and V.A. Kilin, Phys. Rev. A 45, 4576 (1992).
5. J.A. Miskel and M.L. Perlman, Phys. Rev. 94, 1683 (1954).
6. I. Kelson, H. Dumont, B. Perraillon, D. Nir and A. Zidon, Phys. Lett. A 139, 406 
(1989); I. Kelson, I. Phys. D 20, 1049 (1987).
7. E. Glikman, I. Kelson and N.V. Doan, J. Vac. Sci. Technol. A 9, 2776 (1991).
8. C.M. Lederer and V.S. Shirley, Table o f Isotopes, 7th ed., (Wiley, New York, 1978).
9. T.E. Madey and IT . Yates, J. Vac. Sci. Technol. 8, 525 (1971).
10. T E. Madey and J.T. Yates, Surface Sci. 63, 203 (1977)
11. M.L. Knotek, Rep. Prog. Phys., 77, 1499 (1984)
12. T.E. Madey, D.E. Ramaker and R.Stockbauer, Ann. Rev. Phys. Chem. 35, 215 (1984)
148
13. D.M. Yuang and A.D. Crowell, Physical Adsorption o f Gases (Butter Worths, 
Washington, 1962)
14. S. Ross and J.P. Olivier, On Physical Adsorption (Interscience Publishers, New York, 
1964)
15. J.G. Dash, Films on Solid Surfaces (Acadamic Press, New York, 1975).
16. O.G. Petersen, D.N. Batchelder and R.O. Simmons, Phys. Rev. 150, 703 (1966).
17. P.J. Kuntz and J.J. Hogreve, J. Chem. Phys. 95, 156 (1991).
18. GA . Kimmel, Z.H. Levine and B.H. Cooper, Phys. Rev. B 43, 9403 (1992).
19. E.E. Nikitin and D.V. Shalashilin,, Soviet J. of Chem. Phys. 6 1051 (1990).
20. M.J. Ashwin arid D.P. Woodruff, Surface Sci. 244 247 (1991)
21. L. Galdyszewski and G. Galdyszewski, Surface Sci. 247 274 ( 1991).
22. A H. Snell and F. Pleasonton, Phys. Rev. 55, 1396 (1955).
23. O. Kofoed-Hansen, Phys. Rev. 96, 1045 (1954).
24. R.C.Weast, HandBook o f Chemistry and Physics (CRC Press, Cleveland, 1976).
25. R. Franchy and D. Menzel, Phys. Rev. Lett. 43, 865 (1979).
26. H.D. Hagstrum, Phys. Rev. 96, 325 (1954). .
149
27. R. Hentschke, P.Hertel and W. Heiland, Surface Sci. 173, 565 (1986).
28. K.J. Snowdon, R. Hentschke, and W. Heiland, Surface Sci. 173, 581 (1986).
29. L.O. Werme, T. Bergmark and K. Siegbahn, Physica Scripta 8, 149 (1973)
30. G. Omar and Y. Hahn, Phys. Rev. A 43, 4695 (1990)
31. G. Omar and Y. Hahn, Phys. Rev. A 44, 483 (1991).
32. G. Omar and Y. Hahn, Z. Phys. D 25, 31 (1992)
33. G. Omar and Y. Hahn, Z. Phys. D 25,41 (1992)
34. J.C. Levin, N.Keller and D.W. Lindle, Nuc. Ins. Met. in Phys. Res. B 56, 124 (1991).
35. W.M. Ariyasinghe, H.T. Awuku and D. Powers, Phys. Rev. A 42, 3819 (1990).
36. F. Xu and A. Bonanno, Surface Sci. 273,414 (1992).
3 7 .1. A. de Gouw, J. van Eck, and H.G.M. Heideman, Jour, of Phys. 25, 2007 (1992).
38. P. van der Meulen, M.O. Krouse and CA . de Lange, Phys. Rev. A 46, 2468 (1992).
39. J.G. McCaffrey, H. Kunz and N. Schwentner, Jour, of Chem. Phys. 96, 155 (1992).
40. TA . Carlson and M.O. Kraus, Phys. Rev. Lett. 14, 390 (1965)
41. Li. Becker, B.Langer and R.Wehlitz, Jour. Phys. B 22, 749 (1989)
150
42. V.A. Kilin, I.S. Li and O.V. Li, Soviet Physics 36, 146 (1991).
43. P. Moretto-Capelle and H J. Andra, Z. Phys. D 21, 347 (1991).
44. G.W. Rodeback and J.S. Allen, Phys. Rev. 86 446 (1952).
45. RA . Rubenstein and J.N. Snyder, Phys. Rev. 99, 189 (1955).
46. W. Neumann, Phys. Rev. Lett. 42 1514 (1979).
47. J.P.Renier, H. Genz, K.W.D. Ledingham and R.W.Fink, Phys. Rev. 166, 935 (1968).
48. D. Menzel and R. Gomer,, Jour. Chem. Phys. 41, 3311 (1964).
49. PA . Readhead, Can, lour. Phys. 42, 886 (1964).
50. M.L. Knotek and P.J. Feibelman, Phys. Rev. Lett. 40, 946 (1978); Surf. Sci. 90, 78 
(1979).
51. P.Feulner, T.Mueller and D. Menzel, Phys. Rev. Lett. 59, 791 (1987).
52. E.Steinacker and P.Feulner, Phys. Rev. B 40, 11348 (1989).
53. G.Rocker, P.Feulner, L.Zhu and D.Menzel, Physica Scripta 41, 1014 (1990)
54. R.Scheuerer, P.Feulner, L.Zhu and D.Menzel, Desorption Induced by Electronic 
Transitions, DIET TV, p 235, Edited by G.Betz and P.Varga (Springer Ser. Surface Sci. 
19 Berlin, 1990),
151
55. D.Menzel, Appl. Phys. A 51, 163 (1990).
56. Y.Baba, G. Dujardin and D. Menzel, Phys. Rev. Lett. 66, 3269 (1991).
57. T. Schwabenthan, R.Scheuerer and P. Feulner, Solid State Commun. 80,773 (1991).
58. E.Hudel, E. Steinacker and P. Feulner, Phys. Rev. B 44, 8972 (1991).
59. H.Schlichting and D.Menzel, Surface Sci. 272, 27 (1992).
60. W.Wurth, G. Rocker, L.Zhu and D. Menzel, Phys. Rev. B 47, 6697 (1993).
61. G.Ehrlich and F.G. Hudda, Jour, of Chem. Phys. 30, 493 (1959).
62. T.Engel and R. Gomer, Jour, of Chem. Phys. 52, 5572 (1970).
63. D.E. Sullivan, Phys. Rev. B 20, 3991 (1979).
64. G.P. McTauge, J. Bohr and M. Nielsen, Phys. Rev. B 25, 7765 (1982)
65. E.E. Johnson and M.Inokuti, Nucl. Instr. Meth. 206, 289 (1982)
66. B.J. Garrison and R.E. Johnson, Surface Sci. 148, 388 (1984).
67. R.E. walkup, P.Avouris and N.D. Lang, Phys. Rev. Lett. 63, 1972 (1989).
68. D.Dujardin, L.Hellner and RAzria, Phys. Rev. Lett. 64, 1289 (1990).
69. D.Dujardin, L.Hellner and P. Cirkel, Phys. Rev. B 48, 14529 (1993).
152
70. W. Bambyneki Rev. Mod. Phys. 44, 716 (1972),
71. C.C. Chang, Analytical Auger Electron Spectroscopy (Plenum Press, New York,
1974) .
72. TA . Carlson, Photoelectron and Auger Spectroscopy (Plenum Press, New York,
1975) ,
73. L.E. Davis, G.E. Riach and R E. Weber, Handbook of Auger Electron Spectroscopy 
(Physical Electronics Industries, Inc, Eden Prairie, MN, 1976)
74. G.E. McGurie, Auger Electron Spectroscopy Reference Manual (Plenum Press, New 
York, 1979).
75. K.D. Sevier, Low Energy Electron Spectroscopy (Wiley-Interscience, New York 
1972).
76. Chen, Craseman and Mark, K- and L-Shell Radiationless Transition Probabilities, 
Atomic Data and Nuclear Data Tables, 24, 17 (1979).
77. R.R. Rye and J.E. Houston, Jour. Chem. Phys. 78, 4321 (1983).
78. R.Platania and F. Maracci, Chem. phys. 163, 381 (1992).
79. D. Burch, L. Wilet and W.E. Meyerhof, Phys. Rev. A 9, 1007 (1974).
80. T. Darko, H.Siegbhan and P. Kelfve, Chem. Phys. Lett. 81, 475 (1981).
81. J.C. Levin, LA. Sellin and D.W. Lidle, Phys. Rev. Lett. 65, 988 (1990).
153
82. N. Vaeck and J.E. Hansen, Jour. ofPhys. B 25, 3613 (1992).
83. N. Saito and I.H. Suzuki, Physica Scripta 45, 253 (1992).
84. L.M. Brown, History o f Physics (American Institute of Physics, 1985)
85. E. Fermi, Ricercha Scient. 2, 12 (1933)
86. E. Fermi, Z. Physik. 88, 161 (1934)
87. M. Cardona and L. Ley, Photoemission in Solid !  (Springer-Verlag, Berlin, 1978)
88. P.R. Antoniewicz, Phys. Rev. B 21, 3881 (1980)
89. Q.J. Zhang and R.Corner, Surface Sci. 109, 567 (1981).
90. Q.J. Zhang, R.Comer and D.R. Bowman, Surface Sci. 129, 535 (1983).
91. R B .Kay and M.J. van der Weil, Jour. Phys. B 10, 2521 (1977)
92. M.Cini, Solid St. Comm. 20, 605 (1976).
93. GA . Sawatsky, Phys. Rev. Lett. 39, 504 (1977).
94. D.E. Ramaker, Phys. Rev. B 21,4608 (1980).
95. D.R. Jennison and R.R. Ray, Phys. Rev. B 25, 1384 (1982)
96. D.E. Ramaker, C.T. White and J.S. Murday, J. Vac. Sci. Technol. 18, 748 (1981)
I97. P J . Feibelman, Surface Sci. 102, L51 (1981)
98. M.L. Knotek, Nature 291, 452 (1981).
99. D.R. Jennison, J. Vac. Sci. Technol. 20, 548 (1982).
100. C.F. Melius, R.H. Stulen and J.O. Poell, Phys. Rev. Lett. 48, 1429 (1982).
101. M.F. Chung and L.H. Jenkins, Surface Sci. 22, 479 (1970)
102. G Wendin, Breakdown o f the One-Electron Pictures in Photoelectron Spectra 
(Springer-Verlag, Berlin, 1981)
103. TA . Carlson and M.O. Kraus, Phys. Rev. 151, 41 (1966)
104. D.A. Church, Phys. Rev. A 36, 2487 (1987)
105. H.W. Haak, GA . Swatasky and T.D. Thomas, Phys. Rev. Lett. 41, 1825 (1978);
106. H.W. Haak, GA . Swatasky and T.D. Thomas, Rev. Sci. Instrum. 55, 696 (198,4)
107. E. Jensen, R. Bartynski, E. Johnson and R. Garret, Phys. Rev. Lett. 41, 1825 (1989)
108. S.Thurgate, B. Todd and A. Stelbovics, Rev. Sci. Instrum. 61, 3733 (1990)
109. E. lost, Jour. Phys. E 12, 1001 (1979)
HO. Galileo Electro-Optics Co., Electron Multiplier Handbook for Mass Spectrometry 
Applications (1991).
154
155
111. Kurz, Channel Electron Multipliers (American Laboratory, March 1979)
112. M.Lampton and C.W. Carlson, Rev. Sci. Instrum. 50, 1093 (1979)
113. R.S. Gao and R.F. Stebbings, Rev. Sci. Instrum. 55, 1756 (1984)
114. W.G. McMullans and M.L.W. Thewalt, Rev. Sci. lustrum. 58, 1626 (1987)
115. S.H. Courtney and W.L. Wilson, Rev. Sci. Instrum. 62, 2100 (1991)
116. S. Charbonneau and B.J. Kyle, Rev. Sci. Instrum. 63, 5315 (1992)
117. CA . Murrar and S.B. Dierker, Jour. Opt. Soc. Am. A 3, 2151 (1986)
118. Y. Talmi and C.W. Simpson, AppL Opt. 13, 1401 (1980)
119. D. Rees and S J . Kellock, Jour. Phys. E 13, 763 (1980)
120.1. McWhirter and D. Rees, Jour. Phys. E 15, 745 (1982)
121. W.M. Augustyniak and H.P. Pie, IEEE Trans. NucL Sci. NS 19, 196 (1972)
122. K.S.Krane, Introduction o f Neuclear Physics (John Wiley & Sons, New York, 1988)
123. C.N. Burrous, A.J. Liebber and V.T. Zaviantseff, Rev. Sci. Instrum. 38, 1477 (1967).
124. C.W. Kim and A. A. Pevsner, Neutrino Physics, Contemporary Concepts in Physics 
Series (John Wiley & Sons, New York, 1994)
APPENDICES
157
APPENDIX A
INTRODUCTION TO NEUTRINO PHYSICS 
AND NEUTRINO MASS LIMIT STUDY
158
Introduction to Neutrino Physics124
The idea of a neutrino was derived from experiments demonstrating a continuous 
beta decay spectrum. Experiments by Chadwick in 1914 produced the surprising result 
that beta ray spectra, in contrast to alpha and gamma ray spectra, were continuous rather 
than discrete. The missing energy was explained by Pauli in 1930 by proposing a new 
particle—a neutral fermion called “neutron”. In 1933 Fermi dubbed Pauli’s particle 
“neutrino” to distinguish it from the heavy neutron found by Chadwick a year previously. 
In 1934, Fermi formally developed his theory of beta decay in the framework of the 
quantum electrodynamics (QED) of Dirac, Heisenberg and Pauli. In the next section, 
there will be a brief discussion on the Dirac representation of neutrino derived from the 
QED theory.
Description of Neutrinos
hi the QED theory, a Dirac state can be visualized in terms of a top spinning as it 
moves along. A particle’s state is right-handed or left-handed according to whether its 
spin is along or opposite the direction of motion, respectively, th e se  right- and left­
handness are defined as positive and negative helicity. The spinor solutions of the Dirac 
equation, yXx), are interpreted as quantum field operators and the electromagnetic (EM)
159
field is represented by the vector potential field Afi (x). Particles such as electrons, 
neutrinos and quarks are also presented by fields yAx) that carry the particle’s quantum 
numbers and are given by the linear superposition of creation and annihilation operators. 
The operator AJx) creates or annihilates a quantum of the electromagnetic field at a space 
time point x. The electron field y/(x) annihilates an electron or creates a position at x,
while xj/(x) creates an electron and annihilates a positron The interaction term between 
an electron and a EM field can be described by a Lagrangian density as
£ ( x )  =  - j j e> W Y x )  =  - ( - i e )  y/e(x) Jfi y/e(x) A 11(X ) (A l.I)
—^
where the gamma (y) are the Dirac matrixes, y
Z
0 CJ «;0 GO - n 5 (I O^
->
Iv-CJ o ,
> r  = Oj , 7  = < o - I z
and = (-ie) y/e(x) Jfl y/e(x) is the electron current. By simply replacing by jJN)(x) = 
WP(x) YitW x), the neutron-proton nucleon current and A tl(X ) by jJ l)(x)= y/e(x) Yfl K  W, 
the electron-neutrino current, the Lagrangian density of neutron beta decay (n ->p + e~ + 
v ) can be expressed as
^  (A l.2)
-G 1, —  —___
where Gf = 1.19939 XlO'5 /GeV is the couple constant. Similarly, the Lagrangian density 
of the electron capture decay in this work ( p  + e  —> n  + v )  will be
-G c
V2 y, (^ )y (^ )y , wL(x) — (A l.3)
160
Once the Lagrangian density is determined, one can uses it to set up the Euler- 
Langrangian equation of motion. The solutions of this equation can be then used to 
describe a Dirac state.
The Lagrangian describing a free neutrino, v (x)can be expressed as 
Lfxj= V(x) (iy-d - m) v(x) (A l.4)
where d = d Zdx^ l and m is the mass operator. The Euler-Lagrangian equation of motion 
from Eq. (A l.4) becomes
( iy d - m) v(x) = 0 (A l.5)
The solution of Eq. (A l.5) is a superposition of plane waves 
—>
v(x) = J  d p y  2 ,  [ u(p,s) a(p,s) e ipjc + vfp,s) b+(p ,s) eipx] (A l.6)
(2 ^ )^  T
Eq. (A l.6) is called the Dirac representation of neutrino, where s = ±1/2 denotes the spins 
and afp,s)  and b (p,s)  are, respectively, the neutrino annihilation and anti-neutrino
—^ —>
creation operators. The u(p,s)  and v(p,s)  are four-component Dirac spinors. They are the 
solutions of Eq. (A1.7) as
(iyd - m) u(p,s)  = 0, (iyd - m) v(p,s) = 0 (A l.7)
—> —>
The u(p,s)  and v(p,s)  are given by Eq. (A l.8)
u(p , s )  = E + m 
2m
—^ —> 
cr- p l (E  + m): Xs
—^
v(p , s ) E+m
2m
G- p  I (E + m) 
I
(ALS)
161
where %s is the spin matrix given by
I—
i V
Il \Pj for  s = + 1 /2  
for s = - l / 2 ■Xs='
—W
r r
( i / A
for s = +1/2  
for s = - H 2 (A1.9)
For a free neutrino, the Dirac spinors u ( p , s )  and v ( p , s )  are the eignstates of the helicity 
operator. Therefore, The left-handed and right-handed neutrino are defined by
Vl -
' 1 - 7 5 v, zIlT sx
I  2
(ALIO)
They are also called Weyl neutrino fields. In fact, there are four Weyl neutrino fields, Vl, 
Vr , (vl )C and (Vr)c . Because Vr and (Vr)c have not been observed in the nature, they are 
absent from the Standard Model of the electroweak theory. Neutrinos and electron are 
called leptons-. Each one of them is characterized by its own lepton number which is 
conserved during weak interaction. There are three generations of leptons and quarks. 
Table A l. I shows the three generations of leptons and quarks.
Table A l.I  The Three Generations of Quarks and Leptons
Generations 1st 2nd 3rd
U C ' t quarks with charge +2/3
d S b quarks with charge -1/3
Ve K neutrinos with charge 0
e P T leptons with charge -I
162
The quarks and leptons listed in Table A l.I  are called “conventional” quarks and 
leptons. There might also exist a forth generation with very heavy neutrino. But they have 
not yet been observed. In 1937, Majorana proposed Vm, another representation of neutrino 
which is self-conjugate, Vm = (Vmf- Both Dirac and Majorana representations do not 
require neutrino to be massless. Neutrino mass will be discussed in the next section.
Neutrino Mass and Neutrino Oscillation
There are many models that describe the masses of neutrinos. For simplicity, the 
following discussion is restricted to the case of only two generations of neutrinos, e.g. ve 
and V11. The neutrino mass term in the Lagrangian is
2
L-mass-- —(V^Mvi  +  VzMv£) (Al. 11)
where
V ,' '  mee me^
and M =
L J nlIe m/l/t J (Al. 12)
Notes that the v L is not the mass eigenstate and the mass matrix M  is real and symmetric. 
By diagonalizing the mass matrix M, one can find the two eigenvalues of mass Tn1 and m2. 
The correspondent mass eigenfunctions are %i and %2- For the left-handed system, the 
electron and muon neutrino are
Ve,L = COSd X l lL  + s i n d  X2,L (Al. 13)
163
V  = -sin9 Zil  + cos 6 X2, (Al. 14)
Note that the electron and muon neutrinos are not the eigenstate of mass but the linear 
superposition of mass eigenstates Xi and According to quantum mechanics, if  a 
system starts at f = 0 in a state that is not an energy eigenstate, then at later times it can 
oscillate into another orthogonal state. This brings about oscillation between electron and 
muon neutrinos. The time dependent of the electron neutrino state is
VcCO = Cosflx1 e~iE'‘ + s in f lX2 (AT 15)
where E1 and E2 are the neutrino energies associate to two different masses. Using the 
relativistic limit and assuming p »  Mi , Ei = p  + mz2 /2p. The probability of an electron 
neutrino becoming an muon neutrino after traveling a distant L is
P(ve -+Vtl.) = ICV10 | ve (t)> I 2 = sin226sin2 ((Am)2 L/4p) (A l.16)
where Am is the mass different between m; and m2..
Most of the neutrino mass research relies on the study of the nuclear beta decay ' 
(n—> p + e + Vg ). Since there are three particles resulting from this decay, the energy 
sharing determines that the maximum energy of the beta electron, E  to be E0 - m^, where 
E0 is the reaction energy and Tov. is the neutrino mass. Therefore, the focus point of the 
study is to analysis the details of the high energy end of the beta ray. Laboratories around 
the world have performed countless experiments using weak decay process with neutrinos 
as final products. The current experimental limits of neutrino masses are: mVe < 9 eV, mv
164
< 250 keV and mv < 35 MeV. The least stringent limit for the mixing of Ve with Vr is for 
100 < /My < 1000 keV, where the current limit is 0.7%.
The study of beta decay requires the understanding of a very complex three body 
problem. Therefore, another approach is sometimes used: the electron capture (EC) decay 
to probe of neutrino mass. As discussed in Chapter 2, neutrinos with different masses will 
induce different recoil energies in the daughter nuclei in the EC decay. If this energy 
difference could be measure experimentally, the neutrino masses could then be deduced 
from the recoil energy spectra.
Perspective on Neutrino Mass Limit Study
The information on the mixing rate of massless neutrino and massive neutrino 
was extracted from the time-of-flight spectrum of the recoiled 37Cl ions. The coincidence 
peak in the time-of-flight spectrum in Fig. 4.3(a) was first fitted with a Gaussian and a 
constant background. The centroid obtained from the fitting was at channel 122.96 and 
the sigma (the sigma in the Gaussian expression; FWHM = 2.348 sigma) was 9.8 
channels. The fitted area was 14390 ± 139 counts. A flat region to the right of the peak 
which represents slower recoil ions hence massive neutrino recoil is then fitted with a 
Gaussian and a flat background with the centroid and width fixed. The width was 
assumed to scale with the centroid, i.e., it is assumed that AV/V remains constant as the
165
velocity decreases. There is no physical basis for this assumption. It could very well be 
that AV/V (or AE/E) remains constant for lower velocity, in which case the errors below 
would be larger (i.e., the limits would be worse). The distance between the sample and 
the center detector was 7.62 cm. The time scale was I channel = 0.0885 ms. From this 
information one gets the absolute velocity of the recoil 37Cl ions for each channel number. 
The results of the fit is shown in Table A l.2.
Table A 1.2. Fitting, Result on the Neutrino Mass Mixing.
Channel
(cha.)
Sigma
(cha.)
Recoil E 
(eV)
V-Mass
(keV)
Area
(count)
%Area %Mix
150 12.0 6.5 464 1069+64 7.2+0.4 10.6+0.6
160 12.7 5.7 519 910+56 6.1+0.4 10.3+0.7
170 13.5 5.0 560 625±54 4.2+0.4 8.1+0.8
180 14.3 4.5 592 270+53 1.8+0.4 3.8+0.9
190 15.1 4.0 618 33+54 0.2+0.4 0.5+1.0
200 15.9 3.6 639 -113+54 -0.7+0.4 -1.9+1.1
210 16.7 3.3 657 -204+55 -1.2+0.4 -4.1+1.2
Where the percent area is the ratio of the fitted "second peak" to the main peak. As one 
can see from the table, this experiment does not have the sensitivity below In v = 600 keV 
because the tail of the main peak there gives a positive answer outside the statistical error 
for the area of the "second peak". Therefore, only the last three rows of data can be used.
166
The physically relevant quantity one tries to measure is the mixing ratio: if the 
neutrino emitted in the EC capture is a mixture of two mass eigenstates
\mEc > = C0 \m0> + Ci ITTijr> (Al. 17)
then the mixing ratio is Rm = C12 /Co . Since C 2 + Co =1 and C 2 «  C02, Rm = C12. The 
probability P7 of decay to a given mass state is proportional to C2 q 2
f i  =  C72 g /  (711.18)
where qi is the momentum of the neutrino. The factor q7 is just the phase space available 
for the neutrino. (Fermi's Golden Rule contains the density of final states). Therefore
f /P o  = CQ2 g72) / (Q 2 (A l.19)
hence the mixing ratio
Rm = C 2 /Co =  P/Po qo /  qi (A l .20)
In Table A l.2, the next to the last column is the ratio of the massive neutrino "peak" to 
the main peak and the last column is that ratio multiplied by q02 /  q 2. Note that the ratio 
of the areas is not exactly the same as the ratio of probabilities because one has to account 
for the fact that the desorbing probability for 3-4 eV Cl37 ions might be different from 
those for 9.6 eV Cl37 ions. The dependence of the desorbing fraction on energy in this 
experiment is still unknown. If one assumes that they are the same, then the last column 
gives the mixing ratio. The last column is really C 2 /Co x (desorbing probability for 
recoils at energy E) / (desorbing probability for recoils at energy 9.6 eV). Usually one 
quotes a 90 % confidence level (about 2 sigma). Therefore, with the above assumptions, 
the mixing ratio is about < 2% at the 2 sigma level for neutrino with Hiv ~ 600 keV. This 
result is a little higher than the results obtained by different ways for this mass region. To
167
achieve better result, Le., to obtain lower mixture rates and for lower mass region, a 
narrower recoil time-of-flight coincidence peak with better statistics will be needed. This 
would require to reduce the broadening effects in the neutrino recoil induced 37Cl ion 
desorption process.
To get better results on neutrino mass limit study, future experiments will measure 
coincidence between 37Cl ions and X-ray photons which would select charge +1 ions only 
with negligible recoil due to photon and LMM Auger electron emission (3.5%). More 
important, those charge +1 37Cl ions created by X-ray relaxation instead of Auger 
relaxation from the initial K-hole would be less likely to transfer charges with their 
neighbors due to the restrictions of energy conservation. This would significantly reduce 
the charge exchanges and the following Coulomb repulsion interactions, and therefore 
reduce the dominant broadening effect. Several buffer layers of stable 40Ar will also be 
used before the active 37Ar is dosed on the cold sample to deduce the charge exchange 
with the substrate and the residual gases pre-adsorbed as the substrate is being m oling 
down. Simulations have shown that with the reduced broadening effects, the recoil time- 
of-flight coincidence peak would be more appropriate for extracting information on the 
mixture rates of the 200 keV massive neutrino than the results obtained by other means.
168
APPENDIX B
DESIGNING DIAGRAMS OF THE DETECTORS
169
U n it: m m
Ceramic rings
B ias v o l ta g e
5 MO
Channeltroi
S ig n a l
Ceramic rods
W -mash 90%
tran sparency
Ta sheet
C h an n e l L ron B ased  D e te c to r  #2
170
RAE-PSD 01
RAE h o ld e r  
Teflon I pc ± 45  ±135 \  4 x 00.085
4 x M 0 -8 0 4 x 00.40
5 x M 1 /7 2
171
RAE-PSD 02 
Screen  and  
MCP Holder 
Teflon Ipc
t h r o u gh
e v e ry  45 /4 x  00.40 dep th  0.08
/d e p th  0.20
/  e v e ry  11.25
16x M 0 -8 0 , th ro u g h
~ T "  " \ ^ \8—s lo t f r o m  o th e r  sit^e 
s t a r t i n g  11.25° e v e ry  45 
d e p th  0.30
0.70
172
RAE-PSD 03 
D e te c to r  h o ld e r  
Al I pc
Bx M l / 7 2  45  e a c h  0.28 d e e p  on  t h e  sh e ll;
0 3 ( 02.90 02.70
0.10 th ic k
4 x 00 .085 90 e a c h  
on  th e  s h e ll
173
RAE-PSD 04 
Bottom  p la te  
Al 2pc
8x00.085
( g .  01.00 ^
24x00.10
174
r=0.75i
00.08
-‘0.19
0.005 a n d
|  0.007 Ta / 0 .0 4  T aflon
RAE-PSD 05 
Ta ring 10 pc 
Teflon ring  10 pc 
Ta holder Ipc
, 0 .04 Ta
01.295
175
RAE-PSD 06 
C apac ito r h o ld e r  
Teflon 5 pc 
MCP ho ld e r  
Teflon 5 pc Ta I pc
c ap a c ito r
0.250
Gold
Zr=IOOO
0.675
0 .475
0.04 Taflon
I
176
APPENDIX C
PLASMON DECAY INTO 
MULTI-ELECTRON-HOLE PAIRS IN SI(ITl)
1
177
P Ia sm on  D ecay  in to  M u lt i-E le c tron -H o le  P a irs  in  S i ( I l l )
Lin Zhu,(1) Recep Avci,(1) M. M. H indi,(2) and Gerald J. Lapeyre(1)
^  Department of Physics, Moiiiana State University, Bozeman, MT 59717 
W Department of Physics, Tennessee Technological University, Cookeville, TN 38505
(December 20, 1992)
Abstract
The decay of plasmon excitations in S i(Ill) into multi-electron-hole pairs has 
been investigated. We observed coincidences between a ~1000 eV-electron 
beam, inelastically scattered from a S i(Ill) surface, and ~  8.3 dh 0.3 eV- 
electrons resulting from the plasmon decay. The system geometry allowed 
only those coincidences in which a minimum of three plasmons with momenta 
<? < <?c — I A-1 were excited, one of which subsequently decayed into two 
electron-hole pairs.
178
When a high-energy electron interacts with a solid, it gives rise to elementary excitations 
in that solid. The collective excitations, such as plasmons, and their subsequent decay have 
been a topic of interest since the early 1960’s [1-5]. The decay of plasmon excitations can 
occur through a variety of channels; the one of interest here is the creation of electron-hole 
pairs with the emission of the electron into the vacuum. An impediment to the study of 
this phenomenon is that the kinetic energies of the emitted electrons are expected to fall 
between 0 and ~15 eV, which puts them in the region of intense secondary electrons. As 
a result the electrons emitted following plasmon decay are overwhelmed by the secondaries 
and their detection using conventional techniques is problematical.
In this paper we describe a coincidence method. [6-9] of detecting just those electrons 
emitted as a result of plasmon decay. The coincidence is between (I) a primary electron 
that has created one or more plasmons and emerged from the sample surface with an energy 
reduced by a certain number of plasmon energies and (2 ) an electron which is created by 
the decay of one of these plasmons and is then emitted from the surface. The method 
discriminates against ordinary secondary electrons, which exhibit no correlation with the 
inelastically scattered primaries. The high selectivity of the method then permits detection 
of the electrons created in plasmon decay and a study of this process. Below we present 
for the first time the application of the coincidence technique to the study of the decay of 
plasmon excitations having a nonzero momentum in a Si crystal [10].
In general an electron beam transfers part of its energy and momentum to plasmon exci­
tations in the solid. A typical dispersion relation between the energy hu>(q) and momentum 
q transferred to the bulk plasmons is given to a first-degree approximation by [11]
7kn(fy) ~  Tkjp +  a { T r /m ) q 2 +  0 ( q 4). (I)
The parabola determined by Equ. (I) intercepts the single-pair excitation continuum of a 
simple metal at g =  qc, where qc is called the critical momentum [2], The significance of qc 
in the plasmon decay process is that if q <  qc the decay process must involve a minimum 
of two electron-hole pair excitations in order to conserve energy and momentum during the
179
When a high energy electron interacts with a solid, it gives rise to elementary excitations 
in that solid. The collective excitations, such as plasmons, and their subsequent decay have 
been a topic of interest since the early 1960's [1-5]. The decay of plasmon excitations can 
occur through a variety of channels; the one of interest here is the creation of electron-hole 
pairs with the emission of the electron into the vacuum. An impediment to the study of 
this phenomenon is that the kinetic energies of the emitted electrons are expected to fall 
between 0 and ~15 eV, which puts them in the region of intense secondary electrons. As 
a result the electrons emitted following plasmon decay are overwhelmed by the secondaries 
and theii detection using conventional techniques is problematical.
In this paper we describe a coincidence method [6-9] of detecting just those electrons 
emitted as a result of plasmon decay. The coincidence is between (I) a primary electron 
that has created one or more plasmons and emerged from the sample surface with an energy 
reduced by a certain number of plasmon energies and (2 ) an electron which is created by 
the decay of one of these plasmons and is then emitted from the surface. The method 
discriminates against ordinary secondary electrons, which exhibit no correlation with the 
inelastically scattered primaries. The high selectivity of the method then permits detection 
of the electrons created in plasmon decay and a study of this process. Below we present 
for the first time the application of the coincidence technique to the study of the decay of 
plasmon excitations having a nonzero momentum in a Si crystal [10].
In general an electron beam transfers part of its energy and momentum to plasmon exci­
tations in the solid. A typical dispersion relation between the energy %w(g) and momentum 
q transferred to the bulk plasmons is given to a first-degree approximation by [11]
Tiu{q) ~  Tiup + a(Ti2/m ) q 2 +  0 ( q 4). (I)
The parabola determined by Equ. (I) intercepts the single-pair excitation continuum of a 
simple metal at g =  qc, where qc is called the critical momentum [2], The significance of qc 
in the plasmon decay process is that if q <  qc the decay process must involve a minimum 
of two electron-hole pair excitations in order to conserve energy and momentum during the
180
decay, while for q >  qc energy and momentum can be conserved in a single electron-hole 
pair excitation. Theoretical calculations [12] suggest that two-electron-hole-pair excitation 
dominates the decay process for q .<  qc.
In our setup we were restricted in the placement of detectors. As a result, we were able 
to observe only those processes in which multiple plasmons with momentum q <  qc were 
created. Within the restrictions imposed by the detection geometry it was found that the 
best signal-to-noise ratio is obtained when the kinetic energy of the emitted electrons is 
around 8 eV, which appears to be a density-of-states effect.
A schematic diagram of the experimental setup is shown in Figure I. The sample and 
the detectors were placed in an ultrahigh vacuum system at a pressure of 10 10 torr. The 
sample, a p-type S i(Ill) crystal with ~20 fi-cm resistivity, was cleaned using conventional 
techniques to obtain a (7x7) reconstruction. Such a sample shows strong multiple plasmon 
energy loss peaks at ~17 eV spacings (see Fig. 2). A double-pass cylindrical mirror analyzer 
(CMA) detected slow electrons (< 15 eV) with a 0.6 eV energy window. A channeltron C2 
with a high-pass energy filter and a small acceptance angle (±1.4°) detected electrons above 
a predetermined kinetic energy. Primary electrons were generated by a co-axial electron gun 
in the CMA. The primary electrons were incident at 42° with respect to the sample normal. 
The shaped and amplified channeltron pulse from C2 started a time-to-amplitude converter 
(TAC), while the shaped and amplified signal from channeltron Cl in the CMA stopped the 
TAG. The TAC output was fed into a multichannel analyzer (MCA). An MCA spectrum 
then represents the coincident counts vs. time within a preset time window. A typical MCA 
spectrum is shown in the inset in Fig. I. One spectrum typically takes about six hours to 
collect. As evident in Fig. I, the statistical error is dominated by the accidentals. Because 
the accidental count rate is proportional to the square of the primary electron beam intensity, 
whereas the true coincidence rate is proportional to the primary electron beam intensity, 
increasing the primary beam intensity does not substantially improve the statistical error for 
a fixed counting time. (See, for example, Ref. [8].) Further improvement in the statistical 
error can only result from increasing the counting time and/or improving the detection
181
efficiency.
Figure 2 shows a coincidence spectrum (CS) superimposed on ail electron energy loss 
spectrum (ELS). The coincidence is between E k ~  8.3 ±  0.3 eV-electrons (detected by Cl) 
and primary electrons whose energy loss is smaller than A E  (detected by C2). Also shown 
in Fig. 2 is a spline fit to the CS spectrum. Each point and its error bar in the CS spectrum 
are determined by the background-subtracted coincidence peak area of the MCA spectrum, 
as shown in Fig. I. For the ELS the A E  in the figure represents the energy loss by the 
primary electrons; for the CS the A E  represents the difference between the primary energy 
(E0=IOOO eV) and the retarding voltage applied to the screen of C2. For the last point in 
Fig. 2, having A E  =  200 eV and ~1600 counts per six hours (cp6h), the background counts 
were about 10,000 cp6h, Cl counted ~  10s cp6h, and C2 counted 4 x IO7 cp6h. About 5% 
of the C2 counts (~  2 x IO6 cp6h) were due to plasmon losses. Of these only a handful of 
~  1600 contributed to the coincidence in our setup.
The arrows in Fig. 2 indicate the locations of A E  = Tihup, corresponding to the energy 
loss due to multiple plasmon creation. Also shown are the locations of the Si 2p and Si 2s 
core levels. The CS shows no detectable coincidence counts below the 50-eV energy loss 
which corresponds to n  ~  3. Above the 50-eV loss the coincidence counts increase in a 
step-like fashion at energy loss locations corresponding roughly to the creation of multiple 
plasmons with n  ~  4 and 5, and to excitation of the Si 2p and Si 2s core levels.
The coincidence threshold for n >  3 can be explained as follows: Detector C2 is located 
12° above the scattering plane; therefore, specularly reflected electrons cannot reach it. 
Excitations of one or more plasmons with <7 ~  I A-1 are needed to deflect the beam by 
~  12°. A single plasmon loss requires a momentum transfer larger than qc in order to bend 
the beam into C2. The cross section for exciting plasmons with q >  qc is very small [11] and 
hence results in a negligible coincidence count rate. On the other hand, if the momentum 
transfer to the plasmon is </ < qc, the plasmon can decay via two electron-hole pair creation 
with one of the electrons staying in the solid and the second ejected from the solid in the 
direction of Cl. However, for momentum transfers of q <  qc three inelastic scatterings are
182
required to bend the beam 12° out of the plane. The details of these arguments are given 
below.
The relationship among the momentum k of the incident beam, the momentum q of the 
excited plasmon, and the energy Tio j ( q )  lost to the plasmon is fundamental to understanding 
our observations. Reference [11] reports the measured q c and the dispersion coefficient a 
(see Eq. I) for Si as ~  1.1 A-1 and a  ~  0.41 for q  in the (1 00 ) direction and a  ~  0.32 
in the (1 1 1 ) direction. We assume a  ~  0.4 of the nearly free electron gas value in the (lTO) 
direction. Taking into account Tilop  =  17 eV, this means an electron will transfer an energy 
of Tko( q )  ~  20 eV at momentum q  ~  q c in the (110) direction. It can easily be shown that 
for q — qc most of the momentum transfer by a ~  1000 -eV incident electron is perpendicular 
to the travel direction of the electron. With this momentum transfer the incident electron 
beam will deviate from its original direction by ~  3.6°.
Figures 3a and 3b depict the scattering processes and the coincidence geometry. The 
CMA accepts electrons falling into a polar angle of 42 ±  6° from the CMA axis (the direction 
of the incident electrons). Detector C2 is located about 12° off the scattering plane defined 
by the incident k vector and the sample normal. A coincidence can take place in one 
of two ways: (I) An electron scatters elastically (but non-specularly) within the 3° cone 
about detector C2, followed by a single plasmon loss with momentum transfer q to reach 
the detector. Since no coincidence signal is observed for plasmon losses corresponding to 
Ti =  I or ?i =  2 we assume that this is a rather weak contribution, owing to the fact that 
non-specular elastic scattering has a rather weak intensity. (2) An electron excites multiple 
plasmons, each with q ~  </c, on the way into and/or out of the solid, following or preceding a 
specular reflection. We think the latter channel is responsible for our observations. As shown 
in Fig. 3a, the multiple plasmon loss provides a multiple momentum transfer, bending the 
beam by about 12° into detector C2. For an electron beam of 1000-eV energy a minimum 
of three consecutive plasmon losses, each bending the beam by ~  3.6°, is required for a 
coincidence. The observed coincidence rates are consistent with rough estimates of the 
probability of these multi-scattering processes. For ?? > 3 some of the losses may have ~  0
183
momentum transfers causing the step-like increase in coincidence counts every time a new 
loss channel is introduced (see Fig. 2). After the core level transitions have taken place, the 
relaxation process generates a cascade of secondary electrons, resulting in a further step-like 
increase in the coincidence counts.
For single plasmon excitation (n  = I) a momentum transfer q 3 A-1 is required for 
bending the beam by 12°. The coincidence rate for this case is very small because (I) the 
cross section of plasmon excitation for ~  3 A"1 is almost nil [11], (2 ) the decay of plasmons 
with q >  qc is mostly via single electron-hole creation; energy and momentum conservation 
limit the possible initial states from which the electron can be ejected. Similar arguments 
can be made for two plasmon excitation (n =  2 ).
Figure 3b shows the vector diagram for the conservation of momentum in the case of 
plasmon decay to two electron-hole pairs: Ic1 +  Ic2 =  q where vectors Ie1 and k2 are changes 
in the momenta of the two electrons excited as a result of plasmon decay. Here.we assumed 
that the electrons originated at the F point. The electron with Ic1 is detected by Cl while 
the electron with k2 stays in the solid because of the direction of its momentum. Vector q is 
assumed to be nearly parallel to the surface along the (110) direction. The conservation of 
energy requires that huj(q) — E \  +  S12 ~  20 eV, where E \  and S12 are the changes in energy 
of the electrons with momenta ki and k2, respectively. The kinetic energy Ek  (=8.3 eV) 
of the emerging electron is selected such that E 2 matches the difference between the bulk 
initial and final density-of-state maxima, which are separated by ~  6.5 eV [14]. This means 
— 13.5 eV, and an electron initially at the top of the valence band maximum (F point) 
will come out of the solid with Ek — 8.5 eV. However, E i  and E 2 can vary due to variations 
in initial states, such that E \  +  E 2 — 20 eV and Ek  = E \  — $ ~  8.3 eV, where <3? ~  5 eV 
is the work function. The electron ejected from the crystal will be detected by Cl provided 
that it falls into the acceptance angle of Cl. For example, an 8.3 eV electron making a 
45° angle with q will need a perpendicular component of about I A-1. The component of 
ki perpendicular to the surface cannot easily be related to the component inside the solid 
because the translational invariance is broken in that direction. Knowledge of the band
184
structure in the various k-directions, the surface reconstruction, the initial states of the 
ejected electrons in relation to the plasmon decay, the orientation of the crystal, and the 
angular and energy acceptances of both detectors is necessary to pinpoint exactly the initial 
and final states of the ejected electrons accompanying a plasmon decay.
In summary, we report the first direct measurements of multiple electron-hole pair cre­
ation resulting from the decay of plasmons having momentum near qc. The system geometry 
only allowed observation of coincidences between incident electrons which have suffered a 
minimum of three multiple plasmon losses and electrons ejected as a result of the decay of 
these plasmons. The technique has a promising future for studying the unexplored elemen­
tary excitations of solids provided that proper angular and energy resolving detectors and 
a precision sample manipulator are used.
We thank J. Anderson for critical editing of the manuscript with many helpful comments 
and corrections. We thank J. Hermanson and A. Equiluz for helpful discussions. The work 
is supported by NSF Grant No. DMR 9102854. M.M.H. acknowledges support by the U.S. 
Department of Energy, Nuclear Physics Division, via Grant No. DE-FG05-87ER40314.
185
REFERENCES. .
[1] D. F. DuBois1 Annals of Physics 8 , 24 (1959), and D. F. DuBois and M. G. Kivelson, 
Phys. Rev. 186, 409 (1969).
[2] Elementary Excitations in Solids, by David Pines (W. A. Benjamin, Inc., 1964).
[3] M. Hasegawa and M. Watabe, J. of the Phys. Soc. of Japan 27, 1393 (1969).
[4] P. K. Aravind, A. Holas, and K. S. Singwi, Phys. Rev. B 25, 561 (1982), and A. Holas 
and K. S. Singwi, Phys. Rev. B 40, 158 (1989).
[5] M. E. Bachlechner, W. Macke, H. M. Miesenbdck, and A. Schinner, Physica B, 168 104 
(1991).
[6] H. W. Haak1 G. A. Sawatzky1 L. Ungier, J. K. Gimzewskiv and T.D. Thomas, Rev. Sci. 
lustrum. 55, 696 (1984).
[7] S. Thurgate1 B. Todd, B. Lohmann, and A. Stelbovics, Rev. Sci. Instrum. 61, 3733 
(1990),
[8] E. Jensen, R. A. Bartynski, S. L. Hulbert, and E. D. Johnson, Rev. Sci. Instrum. 63, 
3013 (1992).
[9] F. J. Pijper and P. Kruit1 Phys. Rev. B 44, 9192 (1991).
[10] An excellent work by Pijper and Kruit in Ref. [9] reports coincidence measurements 
between the secondary electrons and the energy-loss events in C foils using 80 keV 
electrons in transmission. Some of the low-energy electron production is attributed to 
the plasmon decay process in C-foil.
[11] J. Stiebling and H. Raether, Phys. Rev. Lett. 40, 1293 (1978).
[12] See, for example, Ref. [5] and references therein.
[13] The broadening in the time resolution is mostly due to different flight paths in the
186
cylindrical minor analyzer. For details see reference [6].
[14] Handbook o f  the B an d  S tru c tu re  o f  E lem en ta l  So lids , by D. A. Papaconstantopoulos 
(Plenum Press, New York, 1986).
187
FIGURES
FIG. I. Schematic of the coincidence spectrometer. A typical channeltron pulse is shaped and 
amplified by TFA and then CFD before it arrives at the TAG input. A typical TAG spectrum is 
shown in the inset. The area under the background-subtracted peak determines the coincidence 
counts.
FIG. 2. The coincidence spectrum (open circles with error bars) superimposed on the electron 
energy loss spectrum (dotted line). Arrows indicate energy positions corresponding to multiple 
plasmon loss energies for n =  I through ?i =  5 and the Si 2p and Si 2s core level positions. The 
(A S ) for the coincidence spectrum is AE =  E0 -  eV,  where V is the voltage applied to the 
retarding grid in front of C2. The line through the points is a spline fit to the coincidence data.
FIG. 3. The model explaining the data shown in Fig. 2: (a) Inelastic process scattering the 
incident electron beam with Iq into detector C2. The model assumes the incident beam transfers a 
momentum q perpendicular to k;. Each time a momentum q — qc pointing to the left is transferred 
to a plasmon the beam bends ~  3.6° to the right. Three successive momentum transfers (n =  1,2 
and 3) to a bulk plasmon will bend the beam into the detector. The plasmon losses can take 
place on the way into the solid and/or on the way out of the solid following a specular reflection 
represented by ks in the figure, (b) Conservation of momentum during the plasmon decay process. 
Vector q represents the decaying plasmon while vectors ki and kg represent the excited electrons, 
one of which (/q) is detected by Cl in coincidence with the inelastically scattered electron (k/)  
detected by C2 in part (a).
TFA: T im in g  F i l t e r  A m p l i f i e r
CFD: C o n s t a n t  F r a c t i o n  D i s c r im i n a t o r
TAG: T im e - t o - A m p l i t u d e  C o n v e r t o r
MCA: M u l t i c h a n n e l  A n a ly z e r  
TD: T im e  Delay
CMA: C y l in d r ic a l  M ir ro r  A na ly z e r
C
oi
nc
id
en
ce
 C
ou
nt
s 
/ 
6 
ho
ur
s
2000 4 5 0 0 0
1500
5 0 0
0
Si 2s
P lasm on  loss=nhcj
E =1009 eV
.' Ek=8.3±0.3 eV Si 2p
Coin. Counts 
CMA Counts
100
4 0 0 0 0
3 5 0 0 0
3 0 0 0 0
2 5 0 0 0
20000
15000
AE (eV)
0 5 0 150 200
CM
A 
C
ou
nt
s 
(c
ps
)
C2
a. in e la s tic  sca tte rin g . b. p la sm on  decay.