INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely afreet reproduction.
In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book.
Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.
University Microfilms International A Bell & Howell Information Company 300 North Zeeb Road. Ann Arbor. Ml 48106-1346 USA 313/761-4700 800/521-0600
Order Number 9130584
Electron-stimulated desorption from aluminum surfaces
Whitten, James Edwaxd, Ph.D.
The Ohio State University, 1991
UMI 300 N. Zeeb Rd. Ann Arbor, MI 48106
ELECTRON-STIMULATED DESORPTION FROM ALUMINUM SURFACES
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in the Graduate School of The Ohio State University
By
Jam es E. Whitten, B.S.
The Ohio State University
1991
Dissertation Committee:
R.E. Gerkin A P.L. Jonas
C.W. McCurdy -7 » /vlftC-'iU Co-adviser C.E. Young Department of Chemistry ACKNOWLEDGMENTS
Due to the somewhat unusual situation of being both a student at Ohio State University and a laboratory graduate participant at Argonne National Laboratory, there are many people to whom I am deeply indebted. Many thanks go to Dieter Gruen, Mike Pellin, Wally Calaway, and Keith Lykke at Argonne for their continual encouragement and advice. Especially deserving of my gratitude is Charlie Young whose patience and guidance has helped me mature scientifically. I am also grateful to Patrick Jones, my adviser at Ohio State, who provided me with motivation and good advice during the last several years and who brought new insight to the experimental results. VITA
January 4,1963 ...... Born - Birmingham, Alabama
1985 ...... B.S. - Double Major in Chemistry and Mathematics - Magna Cum Laude Graduate from University of Alabama at Birmingham
1987-1991 ...... Department of Educational Programs Laboratory Graduate Participant, Argonne National Laboratory, Argonne, Illinois
PUBLICATIONS
J.E. Whitten, C.E. Young, M.J. Pellin, D.M. Gruen, and P.L. Jones, "Electron-stimulated desorption of neutrals from methanol-dosed Al(111)- velocity distributions and adsorbate decomposition determined by nonresonant laser ionization," Surf. Sci., vol. 241, pp. 73-90, 1991.
M.J. Pellin, C.E. Young, W.F. Calaway, J.E. Whitten, D.M. Gruen, J.D. Blum, I.D. Hutcheon, and G.J. Wasserburg, "Secondary neutral mass spectrometry using three-color resonance ionization-Os detection at the p.p.b. level and Fe detection in Si at the < 200 p.p.t. level," Phil. Trans. R. Soc. Lond. A, vol. 333, pp. 133-146, 1990.
C.E. Young, J.E. Whitten, M.J. Pellin, D.M. Gruen, and P.L. Jones, "Velocity distribution of laser photoionized neutrals ejected from methanol-dosed aluminum( 111) by electron-stimulated desorption," in: Desorption Induced by Electronic Transitions (DIET IV), G. Betz and P. Varga, Eds. Berlin: Springer-Veriag, 1990, pp. 187-192. iii C.E. Young, J.E. Whitten, M.J. Pellin, D.M. Gruen, R.L. Benbow, and P.L. Jones, "Electron-stimulated desorption of neutrals from 6063 aluminum- velocity distributions detected by 193 nm non-resonant laser ionization," Surf. Interface Anal., vol. 14, pp. 647-655, 1989.
V. Vijay Sen Reddy, J.E. Whitten, K.A. Redmill, A. Varshney, and G.M. Gray, "Synthesis and characterization of transition metal complexes of (2- ' (2-methoxyethoxy)ethyl)diphenylphosphine and ( 2 -(2- methoxyethoxy)ethyi)-dimethylarsine," J. Organomet. Chem., vol. 372, pp. 207-216, 1989.
G.M. Gray, J.E. Whitten, and J.W. Box, "Coordinated 2-halo-1,3,2- dioxaphosphorinane ligands. II. syntheses and 13C, 170 , 31P and 95Mo NMR and IR spectroscopic characterization of some Cr and Mo pentacarbonyl complexes of 2-substituted-4-methyl-1,3,2- dioxaphosphorinanes," Inorg. Chim. Acta, vol. 120, pp. 25-32, 1986.
G.M. Gray, J.E. Whitten, and J.W. Box, "Coordinated 2-halo-1,3,2- dioxaphosphorinane ligands. I. syntheses and 13C, 170 , 31P a n d 95Mo NMR and IR spectroscopic characterization of some molybdenum pentacarbonyl complexes of 2-substituted-5,5-dimethyl-1,3,2- dioxaphosphorinanes,” Inorg. Chim. Acta, vol. 116, pp. 21-30, 1986.
FIELDS OF STUDY Major Field: Chemistry Studies in Surface Chemistry with Professor Patrick L. Jones
iv TABLE OF CONTENTS
ACKNOWLEDGMENTS...... ii
VITA ...... iii
LIST OF T A B L E S ...... ix
LIST OF F IG U R E S ...... x
CHAPTER ...... PAGE
I. INTRODUCTION...... 1
Introduction to the desorption p ro b lem ...... 1
Major experimental observations to d a t e ...... 3
The scope of this th e sis ...... 6
R e fe ren c e s ...... 9
II. EXPERIMENTAL, PART I ...... 12
Overview of equipment and techniques used for the initial experiments ...... 12
v The UHV chamber and the ESD experimental setup . . . 14
Sample preparation ...... 22
Characteristics of the CHA as a TOF d e te c to r ...... 26
R e fere n c e s ...... 27
III. AI-6063 ESD EXPERIMENTAL RESULTS ...... 28
The surface conditions of the methanol-rinsed aluminum a llo y ...... 28
The ionic ESD channel from methanol-degreased 6063 Al allo y ...... 29
Neutral desorption from the degreased technical s a m p l e ...... 32
Neutral and ionic yields from the aluminum alloy s a m p l e ...... 38
R efe re n c e s ...... 40
IV. Al(111) CHA EXPERIMENTAL R E S U L T S ...... 43
An overview of the experiments carried out on the Al(111) s u r f a c e ...... 43
The interaction of methanol and oxygen with Al(111) . . . 43
The ionic ESD channel from methanol-dosed Al(111) . . . 54
Neutral desorption from methanol-dosed Al(111) .... 56
The ion and neutral desorbate yields ...... 65
vi Surface temperature and electron beam effects ...... 65
Conclusions from the CHA studies of methanol- dosed Al(11 1 ) ...... 71
R efe re n c e s ...... 72
V. EXPERIMENTAL, PART I I ...... 74
Overview of the SARISA e x p e r im e n ts ...... 74
Description of the operation of the SARISA apparatus . . 74
A demonstration of the sensitivity of the SARISA a p p a r a t u s ...... 83
The design of the SARISA electron g u n ...... 84
R eferen ces ...... ’ . . 93
VI. SARISA ESD EXPERIMENTAL R E S U L T S ...... 94
Introduction to the SARISA experimental results .... 94
Neutral ESD in SARISA using 1100 eV electron beam e n e r g y ...... 95
Detection of desorbed Al by resonant laser ionization . .102
Velocity/kinetic energy distributions of the desorbed n e u tr a ls ...... 105
Thermal effects on the neutral ESD sp e c tru m ...... 108
Comparison of the results from 1 keV and 3 keV electron bom bardm ent ...... 111
R eferen ces ...... 112 VII. MECHANISTIC IMPLICATIONS 114
Overview of the results that any proposed mechanism must explain ...... 114
The Menzel-Gomer-Redhead model of stimulated d e s o rp tio n ...... 114
The Knotek-Feibelman mechanism ...... 117
The Antoniewicz model for desorption of physisorbed s p e c i e s ...... 118
Desorption induced by a change in molecular orientation ...... 120
The nature of nonradiative decay processes of the excited s ta te ...... 121
Comparison of ionic and neutral desorption within the context of M G R ...... 124
The shape of the observed velocity distributions . . . .130
Intra-adsorbate bond cleavage ...... 131
Mechanistic implications of the observation of aluminum d e s o rp tio n ...... 131
R efe re n c e s ...... 132
VIII. CONCLUSIONS AND EXTENSIONS...... 136
The major findings of this thesis re se a rc h ...... 136
Proposed extensions of this research ...... 137
BIBLIOGRAPHY...... 140
viii LIST OF TABLES
TABLE PAGE
1.1 Existing neutral ESD kinetic energy studies ...... 8
4.-1 Studies of methoxy on aluminum ...... 44
ix LIST OF FIGURES
FIGURES PAGE
2.1 Experimental chamber used for the initial experiments. . . 15
2.2 The setup for the ESD TOF experiment ...... 16
2.3 Spatial and corresponding energy details of the ion (a) and neutral (b) experiments ...... 18
2.4 The timing sequence employed in the ESD TOF experiment ...... 19
2.5 a) Typical mass spectrum of laser ionized/fragmented neutrals, b) Velocity distributions of the carbon fragment. . 21
2.6 Auger of clean Al(111) ...... 24
2.7 a) Three rays entering at different angles differ in flight times; b) the total flight path of the zoom lens and hemisphere is only partially compensated for dispersions in flight times caused by variations of the energies of the ions. For clarity, the details of the behavior within the hemisphere are omitted in b), and the initial and final values of the dispersion in this region are simply connected by a dashed line ...... 25
3.1 Kinetic energy distributions of desorbed H+ from the 6063 aluminum alloy at different beam energies. . . . . 30
3.2 Kinetic energy distributions of desorbed F+ from the 6063 aluminum alloy at different beam energies. . . . . 31
3.3 Mass spectrum of laser ionized/fragmented neutrals desorbed from the AI-6063 sample ...... 34
x 3.4 Velocity distributions, at three electron beam energies, of the C+ fragment arising from laser ionization/ fragmentation of the neutrals desorbed from the AI-6063 sample. The continuous curves are least- squares fits to the data for the functional forms described in the text. Dashed curves: Boltzmann distribution; solid curves: planar-barrier/power law model ...... 35
4.1 The formation and thermal decomposition of the surface methoxy species on Al(111). As explained in the text, some assumptions are made about which surface sites are occupied ...... 46
4.2 0(1 s) and C(1s) study of the methanol dosing of Al(111)...... 48
4.3 0(1 s) core hole for O2 and CH 3O on Al(111) ...... 49
4.4 C(1 s) core hole of CH3O/AI0 11) at various temperatures ...... 50
4.5 The early stages of the oxidation of Al(111). Note that there are two kinds of threefold sites on the surface. The surface oxygen occupies the threefold hollow site. Subsurface oxidation is thought to proceed simultaneously with surface oxidation ...... 52
4.6 0(1 s) XPS study of the dosing of methoxy-saturated Al(111) with oxygen...... 53
4.7 Mass spectrum of the ions observed from 3 keV electron bombardment of methoxy-saturated Al(111) ...... 55
4.8 Kinetic energy distributions of H+ desorbed from 55 L CH3OH-dosed Al(11 1) at room te m p e ra tu re ...... 57
4.9 Mass spectra of laser-ionized/fragmented neutrals desorbed from methanol-dosed Al(111) ...... 58
4.10 Laser focal intensity dependences of the two mass regions observed in neutral ESD from methanol- dosed Al(111) ...... 60
xi 4.11 Velocity/kinetic energy distributions of photoionized fragments from neutral ESD from clean Al(111) dosed with 55 L CH3OH. Dashed curves (— ): Boltzmann fit to the distribution; dotted curves ( . . planar-barrier/ power-law model; solid curves: result of convolution of the dotted curves with the electron pulse width. The kinetic energy scale is for the parent methoxy species (mass 31)...... 62
4.12 Velocity/kinetic energy distributions of photoionized fragments from neutral ESD from pre-oxidized Al(111) dosed with 55 L CH 3OH. The legend for the caption of figure 4.11 applies to this figure also ...... 63
4.13 Temperature dependence of the neutral ESD signal as monitored by the two mass regions observed in laser ionization/fragmentation. For this experiment, the clean Al(111) sample has been dosed at ambient temperature with 25 L CH3OD...... 66
4.14 The effect of the electron beam on the neutral ESD signal from pre-oxidized Al(111) dosed with methanol to saturation coverage ...... 69
5.1 The general layout of SARISA IV ...... 76
5.2 SARISA operated in the traditional sputtering mode with the ion gun ...... 77
5.3 a) Schematic of the resistive disk analyzer device of the EARTOF, and b) examples of trajectories of ions with three very different energies. In b), E1 is allowed to pass, while E2 and E3 are not ...... 80
5.4 Representation of two photoion trajectories, denoted as "a" and "b", which have the same energy, but different angles of entry into the entrance window of the first analyzer. Note that the total flight paths of the two trajectories through the analyzers are identical due to the telescope lens between the disks...... 82
5.5 Resonance ionization schemes which have been used for ionizing sputtered Fe. The 3-color scheme on the left eliminates nonresonant ionization of the isobarically interfering silicon dimers ...... 85
xii 5.6 Fe depth profile of a Si wafer. The sample was prepared at Siemens, Inc. by Dr. H. Zieninger...... 86
5.7 a) The target region of SARISA IV. The circled portion has been simulated by the particles optics program EGUN. The equipotential lines calculated in this program are shown in b) ...... 88
5.8 Raytraces in the target region, for the same conditions shown in figure 5.7, of: a) electrons of various energies, and b) photoions leaving the laser ionization volume. . . 89
5.9 The SARISA multistage electron gun ...... 91
5.10 SARISA IV with the electron gun installed...... 92
6.1 ArF-ionized/fragmented ESD neutrals from methanol- dosed Al(111). The laser intensity is 230 MW/cm2. . . . 96
6.2 Laser ionization/fragmentation of ESD neutrals from methanol-dosed Al(111) at low and high laser intensities. . 97
6.3 Power study of the photofragments from ArF photo ionization of the neutrals desorbed by 1000 eV electron impact from methanol-dosed Al(111) ...... 98
6.4 Laser intensity dependence of desorbed and sputtered Al signals...... 100
6.5 Atomic energy level diagram of aluminum showing the transitions relevant to the desorption experiments described in the text ...... 101
6.6 The laser setup used for resonant ionization of desorbed aluminum ...... 103
6.7 Resonant scan of aluminum desorbed from methanol- dosed Al(111). The resonant excitation step is performed with frequency-doubled Rhodamine 610 dye light having a bandwidth of ~0.003 nm. The intensity of this doubled dye light is ~1.0 MW/cm2. KrF excimer radiation at ~248 nm is used to excite the aluminum from the resonant level to above the ionization potential ...... 104
xiii 6.8 Velocity distributions of laser-ionized/fragmented neutrals desorbed from methoxy-saturated Al(111). The ArF laser intensity is ~ 150 MW/cm2, and the electron beam energy used is 1100 eV ...... 106
6.9 a) Velocity distribution data of desorbed aluminum determined at 2.16 and 4.43 mm flight distance to the laser ionization volume, b) Boltzmann and planar-barrier fits to the combined data sets. . . 109
6.10 Neutral ESD from the methoxy/AI(111) system at room temperature and at 600 K ...... 110
7.1 Electronic excitation from the ground state metal- adsorbate potential surface, M + A, to the repulsive surface, (M + A)*, can lead to desorption if sufficient kinetic energy, E, is acquired on the upper state to overcome the binding energy of the adsorbate to the surface ...... 116
7.2 For physisorbed species, initial electronic excitation to an ionic state, M- + A+, which is bound closer to the surface than the ground state, leads to nuclear motion towards the surface. Deexcitation back to the ground state results in desorption only if sufficient kinetic energy has been acquired on the ionic potential curve ...... 119
7.3 a) charge transfer proceeding via resonant tunneling; b) charge transfer via Auger deexcitation. The circled numbers represent the steps explained in the text ...... 123
7.4 As described in the text, reneutralization of desorbing ions is expected to yield neutrals which desorb with substantially less kinetic energy than ions ...... 125
7.5 Efficient deexcitation channels on a surface make it improbable that desorption will occur via evolution along the entire initial excited state curve. Most neutrals will desorb following deexcitation to the ground state potential curve near the critical distance, Rc ...... 128
xiv CHAPTER I
INTRODUCTION
Introduction to the Desorption Problem When a surface is bombarded by particles, removal of the surface atoms and any adsorbed atoms or molecules may occur. Depending on the types and energies of the particles incident on the surface and on what is removed from the surface, this surface erosion may be classified as either sputtering or desorption. For example, sputtering results when energetic ions and atoms impinge on a surface, transfer their momentum to it, and remove the near-surface layers of the solid. Photons and low-to-medium energy electrons (e.g. ~10 to 10,000 eV) incident on a surface impart very little momentum and, in general, do not result in the removal of the atoms of the solid. However, they may readily induce adsorbate desorption and decomposition via electronic excitations which result in the breaking of the surface-adsorbate bond or of bonds within the adsorbate. Electron- and photon-stimulated desorption (ESD and PSD) have associated with them both useful and unwanted effects. The last fifteen years have seen the development of ESDIAD (Electron-Stimulated Desorption Ion Angular Distribution) as an analytical tool in surface chemistry for determining the orientation of adsorbates [1]. In this type of experiment, electron impact causes the breaking of either the surface- adsorbate bond or an intra-adsorbate bond and subsequent desorption along the direction of the broken bond. The orientations of the adsorbates are inferred from the angular distributions of the emitted ions which are readily detected visually by means of a phosphor screen or digitally by an array of microchannel plate multipliers in front of a resistive anode [2]. ESD and PSD have also been used to modify surface chemistry, with applications in the growth of electronic materials such as silicon nitride
1 2
(Si3N4) and silicon carbide (SiC) [3],[4]. Traditional means of producing these materials involve the reactions of Si with gases such as NH 3 and C2H4 at temperatures in the 800-1200°C range [5],[ 6]. In these types or reactions, the dissociation of the NH 3 or C2H4 leads to the formation of subsurface nitrogen or carbon atoms, respectively, and to surface hydrogen. The high temperatures are necessary to desorb surface hydrogen generated from the dissociation of the NH 3 or C2H4. Without these high temperatures, the reactions become self-limiting at less than a monolayer of product, with the hydrogen atoms tying up the Si surface dangling bonds and inactivating the surface. Such high temperatures, however, can lead to unwanted side-processes such as dopant diffusion. An alternative means of desorbing the surface hydrogen, and one which has the promise of providing the spatial control desirable in microelectronics, is the use of ESD or PSD. It has recently been demonstrated that thick layers of silicon nitride and silicon carbide may be produced by these methods even at temperatures as low as 90 K [3],[4]. Many commonly used techniques in surface science, such as LEED (Low Energy Electron Diffraction) and Auger spectroscopies, which will be discussed in some detail in later parts of this thesis, employ electron beams for surface analysis. An important example of an unwanted consequence of ESD is beam damage to the system being studied, in which the electron beams may facilely desorb or decompose the adsorbates they are probing. Another unfortunate effect of ESD and PSD is desorption from vacuum chamber walls. This is a particularly severe problem for synchrotron storage rings. Even after chemical cleaning and baking up to 300°C in vacuum, the surface layers of the vacuum wall materials may contain many equivalent monolayers of tightly bound adsorbed gases. The high binding energy of these adsorbates limits the thermal outgassing rate of the wall material at ambient temperature. When the synchrotron is in operation, however, the vacuum walls are exposed to energetic photons and photoelectrons which can desorb these tightly bound species. The desorption may be efficient enough to raise the base pressure of the vacuum chamber several orders of magnitude. An increase in background pressure is detrimental to the beam intensity because of scattering of the beam electrons by the molecules of 3
the residual gas. Because of the importance of minimizing the effects of beam-induced desorption, much research effort in the design and fabrication of synchrotrons is concerned with finding new vacuum wall materials and new methods of treating existing materials so as to produce surfaces with low desorption yields. In order to make the best use of ESD as a surface analytical tool, or as a technique for surface modification, or to successfully minimize deleterious electron-material interactions, it is necessary to obtain a fundamental understanding of the operative desorption mechanisms. The desorption of ions from adsorbate-covered metal surfaces has been studied in detail over the past two decades. Because of the difficulty of detecting neutral particles, very little work has been done on the neutral desorption channel. Since this channel usually dominates the ionic one by several orders of magnitude, understanding stimulated desorption hinges on understanding the neutral channel. The goals of this thesis are to compare the cross sections and dynamics of ionic and neutral electron-stimulated desorption channels from adsorbed molecules on technologically important materials, such as the aluminum alloy used in fabricating synchrotron vacuum wall chambers, and from adsorbates on well-characterized systems, such as single crystal aluminum, to assess whether or not there are differences in the mechanisms of ionic and neutral desorption, and to elucidate basic mechanisms in the desorption process.
Major Experimental Observations to Date A brief synopsis of the most important observations from electron- induced desorption experiments places this thesis work in context. More details about the history of ESD and the experimental techniques used may be found in several comprehensive reviews [7],[ 8],[9] and conference proceedings [ 10],[11 ],[12],[13]. 1) Electron impact on a clean metal surface does not yield any measurable desorption due to the strong coupling of the surface atoms to the bulk metal. 4
However, electron impact on an adsorbate-covered metal can lead to desorption of ions (both positive and negative), metastable neutrals, and ground state neutrals, with the ground state neutrals generally outnumbering the ions and metastables by several orders of magnitude. The adsorbates may be desorbed either intact or as fragments. 2) The flux density of desorbed particles is proportional to the flux density of the bombarding electrons and to the surface concentration of the species being desorbed. 3) On metal and semiconductor surfaces, the desorption cross sections are considerably smaller than analogous gas phase dissociation cross sections. Desorption cross sections range from 10 *18 to 10*23 cm2, while that of a typical gas phase dissociation is of the order of 10 ‘16 cm2. The lower cross section for desorption is representative of the ability of the surface to reduce excited-state lifetimes efficiently. 4) For systems in which desorption thresholds have been determined, the onset of desorption is found to be in the valence excitation/ionization region, i.e., for electrons with energies between 5 and 20 eV. There are often, however, strong enhancements of desorption yields at the energies of core hole formation. 5) Kinetic energy distributions of desorbing ions peak between 1 and 10 eV, while neutrals appear to be much less energetic, having peak kinetic energies <0.5 eV. Metastable neutrals, however, appear to be fast, similar to ions, and to have distributions which peak in the several eV region. Different bonding states of an adsorbate can lead to different kinetic energy distributions. In fact, multiple peaks in the kinetic energy distributions of desorbates have led to the identification of multiple adsorbate states [14],[15]. 6) Desorption cross sections appear to be very sensitive to the mode of binding of the adsorbate to the substrate, and, in general, cross sections for desorption via scission of internal bonds in adsorbed molecules are larger than for desorption via the breaking of the metal-adsorbate bond. In addition, there can be strong site dependences to desorption. For example, ESD cross sections for 0+ desorption from metals are larger for O adsorbed on step and defect sites than on atomically smooth surfaces [ 8]. 5
Although this thesis is mainly concerned with ESD processes, it is worth noting that many of the above observations apply also to PSD. While PSD is somewhat less studied than ESD, it is clear that the two proceed via similar mechanisms, with any different spectral dependences reflecting the differences in the physics of the initial excitation process [ 8]. The experimental observations stated above have led to the proposal of the following very general desorption mechanism [10],[16],[17]. Electron or photon impact causes an initial electronic excitation from the ground state of the adsorbate-substrate system to a repulsive neutral or ionic state (1 o-16 sec). This excited state decays by displacement of the atomic positions, but this recession from the surface is in competition with electronic decay mechanisms, which redistribute the electronic energy ( 10 * 15 to 10‘14 sec). Due to the high density of states in the solid, recapture of the particle as it is moving away from the surface may occur with transfer of the excitation energy to the solid. If it is not recaptured, then the particle is ejected, although it may be modified (e.g., energy, trajectory, charge state) as it escapes from the surface ( 10‘13 sec). Most of the information and theories available are based on the following experimental observables: 1) the identities of the desorbates; 2 ) energy and angular distributions: 3) thresholds for desorption; 4) charge states (for ions); and 5) electronic and vibrational energy distributions. All of these quantities are expected to be determined by the details of the initial and final electronic states of the bonding. It is important to note that although the literature describes hundreds of experiments which have investigated ionic desorption from surfaces, very little work has been carried out on the more predominant neutral ESD channel. This lack of neutral ESD experiments is a result of the difficulties involved in the detection of uncharged particles. Of those neutral desorption studies which have been performed, most have not detected the desorbates directly, but rather have looked at the loss of species from the surface. Such studies, though adequate for estimating total desorption yields, do not provide mechanistic information. The number of studies to date, including the results presented in this thesis, which have measured kinetic energy distributions of desorbing neutrals (excluding studies of metastable neutrals) is so small 6
that they may be summarized in Table 1.1.
The Scope of this Thesis As mentioned previously, this thesis is mainly concerned with investigating the mechanisms of ionic and neutral ESD. The initial system chosen for study is methanol-degreased 6063 aluminum alloy (0.4 at. % Si, 0.7 at. % Mg). This alloy has been chosen as the vacuum chamber material for the 7-GeV Advanced Photon Source (APS), which is to be built at Argonne National Laboratory and will produce the brightest beam of high energy x-rays anywhere in the world [18]. This alloy has the desirable properties of being economical to extrude and to machine, being strong, and of having good weldability. As extruded, it has a thick, relatively porous oxide layer (100-200 A) which may act as a trap for gases and adsorbates. In addition, the machining of this alloy may make use of hydrocarbon lubricants which must somehow be removed before final assembly. Although more sophisticated methods of cleaning this alloy have been explored [19], one possible degreasing agent is methanol. In this thesis research, actual disks of the proposed APS wall material have been rinsed with methanol and installed in our vacuum chamber where desorption experiments have been performed. The ionic ESD channel is readily detectable, while the neutrals have been detected via laser ionization coupled with time-of-flight (TOF) mass spectrometry. The observation of orders of magnitude differences in the energy distributions of the desorbing ions and neutrals from the 6063-AI system has raised the isssue of differences in desorption mechanisms and prompted the extension of the initial desorption experiments to more carefully controlled and well- characterized systems. Clean, carefully prepared single crystals provide the opportunity to study the chemistry and physics on well-defined surfaces dominated by one crystallographic plane. Because the chemistry of methanol on single crystal Al(111) has been the subject of recent investigations [20],[21] and is fairly 7 well understood, this substrate has been chosen for extension of the Al- 6063 experiments. As shown in Table 1.1, somewhat more energetic neutrals have been observed from the methanol-dosed Al(111) system than from the technological sample, although they are still more than an order of magnitude less energetic than the ions observed. Implications for desorption mechanisms are discussed in a later chapter. The experiments on the technologically important AI-6063 alloy have been carried out through a novel application of standard surface science equipment. Initial experiments on the Al(111) system have also been performed with this equipment. Because of the limitations of this setup, including poor mass resolution, part of this thesis research has included the building, testing, and extension of these neutral desorption experiments to a unique, state-of-the-art surface analytical machine which has proven to be extremely efficient for ESD studies. It is hoped that through such new instrumentation, future work on a variety of systems, such as those reported in this work, may be performed and a better understanding of the phenomenon of stimulated desorption may be obtained. 8
Table 1.1: Existing Neutral ESD Kinetic Energy Studies
Desorbate Svstem K.E. (eV) Year R eferenced
Physisorbed. Systems:
N20 N2O/Ru(001) 0.11 1984 [22]
Ar Ar/Ru(001) 0.06 1989 [23]
Kr Kr/Ru(001) < 0.10 1989 [23]
Chemisorbed Systems:
CO C0/Ru(001) 0.35 1984 [22]
n2o N2O/Ru(001) 0.22 1984 [22]
n 2 N2O/Ru(001) 0.10 1984 [22]
NO NO/Pt(111) 0.05 1987 [24],[25],[26]
CO CO/Pt(111) 0.50 1987 [24],[25],[26]
NO N 0 2/Pt(111) 0.10 1989 [27]
CH30 CH3O/AI-6063 0.01 1989 This Work
CH3O CH30/AI(111) 0.13 1990 This Work
Al or AIX CH30/AI(111) 0.11 (if Ai) 1991 This Work 9
References
[1] R. H. Stulen, "Recent developments in angle-resolved electron- stimulated desorption of Ions from surfaces," Prog, in Surf. Sci., vol. 32, pp. 1-38, 1989.
[2] J. T. Yates Jr., M. D. Alvey, K. W. Kolasinski, and M. J. Dresser, "Ion angular distribution of species desorbed from single crystal surfaces by electron impact," Nucl. Inst, and Meth. B, vol. 27, pp. 147-154, 1987.
[3] P. Avouris, F. Bozso, and R. E. Walkup, "Desorption via electronic transitions: fundamental mechanisms and applications," Nucl. Instr. and Meth. B, vol. 27, pp. 136-146, 1987.
[4] F. Bozso and P. Avouris, "Reaction of Si(100) with NH 3: rate-limiting steps and reactivity enhancement via electronic excitation.," Phys. Rev. Lett., vol. 57, pp. 1185-1188, 1986.
[5] J. A. Nemetz and R. E. Tressler, "Thermal nitridation of silicon and silicon dioxide for thin gate insulators," Solid State Techno!., vol. pp. 79-85, 1983.
[6] I. Kusunoki, M. Hiroi, T. Sato, Y. Igari, and S. Tomoda, "SiC film formation on Si(001) by reaction with C 2H2 beams," Appl. Surf. Sci., vol. 45, pp. 171-187, 1990.
[7] T. E. Madey and J. T. Yates Jr., "Electron-stimulated desorption as a tool for studies of chemisorption: a review," J. Vac. Sci. Technol., vol. 8, pp. 525-554, 1971.
[8] T. E. Madey, D. E. Ramaker, and R. Stockbauer, "Characterization of surfaces through electron and photon stimulated desorption," Ann. Rev. Phys. Chem., vol. 35, pp. 215-240, 1984.
[9] P. Avouris and R. E. Walkup, "Fundamental mechanisms of desorption and fragmentation induced by electronic transitions at surfaces," in Annual Review of Physical Chemistry, H. L. Strauss, G. T. Babcock and C. B. Moore, Eds. Palo Alto: Annual Reviews, Inc., 1989, pp. 173-206. 1 0
[10] Desorption Induced by Electronic Transitions (DIET I), vol. 24, N. H. Tolk, M. M. Traum, J. C. Tully, and T. E. Madey, Eds. Berlin: Springer-Verlag, 1983.
[11] Desorption Induced by Electronic Transitions (DIET II), vol. 4, W. Brenig and D. Menzel, Eds. Berlin: Springer-Verlag, 1985.
[12] Desorption Induced by Electronic Transitions (DIET III), vol. 13, R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988.
[13] Desorption Induced by Electronic Transitions (DIET IV), vol. 19, G. Betz and P. Varga, Eds. Berlin: Springer-Verlag, 1990.
[14] J. H. Craig Jr., "CO on Pt(111): TDS and ESD study," Appi. Surf. Sci., vol. 25, pp. 333-340, 1986.
[15] J. H. Craig Jr., "Characteristic energies of electronically desorbed ions from Rh(100)," Appl. Surf. Sci., vol. 35, pp. 520-526, 1989.
[16] D. Menzel and R. Gomer, "Desorption from metal surfaces by low- energy electrons," J. Chem. Phys., vol. 41, pp. 3311-3328, 1964.
[17] P. A. Redhead, "Interaction of slow electrons with chemisorbed oxygen," Can. J. Phys., vol. 42, pp. 886-905, 1964.
[18] Report No. ANL-87-15, p. II-6-2, Argonne National Laboratory, 1987.
[19] N. Kaufherr, A. Krauss, D. M. Gruen, and R. Nielsen, "Chemical cleaning of aluminum alloy surfaces for use as vacuum material in synchrotron light sources," J. Vac. Sci. Technol., vol. A8, pp. 2849- 2855, 1990.
[20] P. Basu, J. G. Chen, L. Ng, M. L. Colaianni, and J. T. Yates Jr., "Fragmentation of molecular adsorbates by electron and ion bombardment: methoxy Chemistry on Al(111),” J. Chem. Phys., vol. 89, pp. 2406-2411, 1988.
[21] J. G. Chen, P. Basu, L. Ng, and J. T. Yates Jr., "A comparative study ofthe reactivities of H2O, CH3OH, and CH 3OCH3 toward Al( 1 11)," Surf. Sci., vol. 194, pp. 397-418, 1988.
[22] P. Feulner, D. Menzel, H. J. Kreuzer, and Z. W. Gortel, "Kinetic-energy distributions of neutrals desorbed by electron impact from adsorbates on metal surfaces," Phys. Rev. Lett., vol. 53, pp. 671-674, 1984. 11
[23] E. Steinacker and P. Feulner, "Stimulated desorption of atoms from rare-gas monolayers on metal surfaces: dependence of yields and energy distributions on primary excitations," Phys. Rev. B, vol. 40, pp. 11348-11351, 1989.
[24] A. R. Burns, E. B. Stechel, and D. R. Jennison, "Desorption by electronically stimulated adsorbate rotation," Phys. Rev. Lett., vol. 58, pp. 250-253, 1987.
[25] E. B. Stechel, D. R. Jennison, and A. R. Burns, "Dynamics in neutral DIET from chemisorbed molecules," in Desorption Induced by Electronic Transitions (DIET ill), R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988, pp. 136-143.
[26] A. R. Burns, E. B. Stechel, and D. R. Jennison, "Rovibrational laser spectroscopy of ESD neutrals from chemisorbed species," in Desorption Induced by Electronic Transitions (DIET III), R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988, pp. 67-72.
[27] A. R. Burns, D. R. Jennison, and E. B. Stechel, "Electronically stimulated dissociation of NO 2 on Pt(111)," Phys. Rev.B, vol. 40, pp. 9485-9497, 1989. CHAPTER I!
EXPERIMENTAL, PART I
Overview of equipment and techniques used for the initial experiments Two experimental apparatuses have been used in the desorption studies performed in this thesis research. The two configurations differ mainly in the sophistication of the electron guns and of the ion detectors used for the time- of-flight experiments. The initial sets of experiments reported in this thesis have been performed mainly with surface science equipment that is commercially available. These experiments include ionic and neutral desorption from the methanol-degreased 6063 Al alloy, characterization of the methanol-dosed Al(111) system, and ionic and neutral ESD experiments from the single crystal Al system. This first experimental setup, while it has the advantage of being relatively simple to use, suffers from poor m ass resolution and low sensitivity (see chapter IV for further details). In order to improve the quality of the data achievable, a unique surface science machine (acronymed SARISA), especially designed for the detection of sputtered and desorbed neutrals, has been built and added to the vacuum chamber housing the initial setup. The details of the SARISA apparatus are discussed in chapter V. The present chapter is concerned with the equipment and techniques used for the initial sets of experiments. However, many of the procedures described in this chapter, such as the cleaning of the the Al(111) single crystal surface, apply to the later experiments as well. Before the details of the equipment and methods used are given, a brief overview of the major experimental considerations is worthwhile. First, the experiments must be carried out under ultrahigh vacuum (UHV) conditions to prevent sample contamination and to prevent gas phase molecules from interacting with the particles arriving at or leaving the surface (e.g., electron beam ionization of background gases). This is facilitated by performing the
12 13
experiments in a bakeable vacuum chamber in which care has been taken to minimize the use of any materials with high outgassing rates. Second, for the well-characterized single crystal work reported here, it is necessary to clean the surface and to verify its cleanliness and crystallinity. In addition, for an adsorbate-covered surface, surface reactions may take place, and it is desirable to have a method available to probe changes in surface chemistry. The most common surface science technique for verifying surface cleanliness, and one used in this work, is Auger electron spectroscopy (AES). This method makes use of the fact that atoms in the first few layers of a solid, when ionized by an electron beam, emit electrons with energies characteristic of the element and independent of the incident electron beam energy. By energy analyzing the emitted electrons, the surface composition may be determined. Surface crystallinity is determined by low energy electron diffraction (LEED). In such an experiment, the diffraction pattern of an incident electron beam is indicative of the arrangement of the surface atoms. Changes in surface chemistry are detectable via several methods. In this thesis, x-ray photoelectron spectroscopy (XPS) has been employed. Somewhat similar to the AES method, in XPS the energies of electrons emitted as a result of x-ray bombardment are analyzed. The energies of these electrons are relevant to the bonding states of the atoms in the adsorbed molecules. Detailed discussions of these standard electron spectroscopic techniques are available in several text books (see, for example, [1], [2]). The third major experimental aspect is time-of-flight (TOF) mass spectrometry, which constitutes the basic technique being used for the ESD experiments reported here. In the initial set of desorption experiments, the same energy analyzer used for electron spectroscopy is run in positive ion detection mode and used as the TOF detector. TOF detection requires that desorbing neutrals be ionized, and laser ionization is used for this purpose. 14
The UHV chamber and the ESD experimental setup The experiments reported here were carried out in a stainless steel UHV chamber equipped with tandem turbomolecular pumps (1500 I/s and 50 I/s Balzers models TPU 1500 and TPU 050, respectively) and a titanium sublimation pump. After bakeout at 150-200°C, the system achieves a base pressure of < 2x10 ‘10 Torr. The layout of the vacuum chamber, prior to the attachment of SARISA, is shown in figure 2.1. The electron gun (VSW Ltd., model EG-500) is capable of being operated in either continuous or pulsed mode with the use of a pulsing circuit of our own design. The continuous mode is used for Auger analysis and electron beam damage studies, while the pulsed mode is used for the time-of-flight desorption experiments. The electron beam energy used in all of the experiments reported in this work is in the range 900-3000 eV. At these energies, between 1 and 3 jiA can be delivered into a -0.5 mm diameter spot. Energy analyses of emitted electrons for surface characterization and of ESD ions (both from direct ionic desorption and from photoionized desorbing neutrals) are performed with a VSW model HA-100 concentric hemispherical analyzer (CHA). An x-ray source (VSW model TA10) is used for XPS studies, and a reverse view LEED (VG Scientific model 640-2 RVL) is used to assure that the surface is well-ordered. The XPS studies reported in this work are carried out with MgKa excitation (hv=1253.6 eV), with the x-ray source being run at 21 mA and 14.5 kV. The chamber is also equipped with a quadrupole mass spectrometer (Leybold-lnficon Quadrex 100) for residual gas analysis and for checking the purity of gases admitted into the chamber during dosing. The laser beam enters the chamber through a sapphire window and is brought to a focus 2.5 mm in front of the sample by an external fused silica lens of nominal focal length 320 mm (f « 240 mm at 193 nm). The zero of separation of the laser ionization volume from the sample surface is established prior to surface preparation, and later rechecked at the end of the experiment, by moving the sample with a Huntington model PM-275 precision manipulator into the high intensity focal region and observing a pressure rise as the sample edge approaches the focus. Figure 2.2 shows the setup used for the ESD TOF experiments. Both ions and neutrals are desorbed when the electron pulse strikes the surface. In studies of both neutral and positively charged desorbates, the 15
LEED Sample Manipulator Transfer Arm
To 1500 I/s Turbo Pump and later to SARISA
X-Ray Source e- Gun CHA
Figure 2.1: Experimental chamber used for the initial experiments. Sample Manipulator .Sample —== — Laser Beam Ions Pulsed Electron Beam
Zoom ___ Lens
Detector
1=100 m Concentric Hemispherical Energy Analyzer Resolution: AE/E = 0.02
Figure 2.2: The setup for the ESD TOF experiment. 17 sample is held at a nominal potential ~ 96 V above ground. Since the desorbing ions are formed at the surface, rather than 2.5 mm in front of it, their initial potential energy is higher than that of the neutrals from the laser ionization volume, and they enter the CHA with more kinetic energy. The energy separation between the surface ions and the photoionized neutrals is -20 eV and is more than adequate for the CHA, which has an energy window of -2 eV, to completely discriminate between them. By changing the kinetic energy window of the CHA, either ionized neutrals or desorbed ions may be selectively studied. Figure 2.3 demonstrates the spatial and corresponding energy details of the ionic and neutral desorption experiments. In both cases, the transmission energy through the hemisphere is held constant at 50 eV, with the desorbed ions and laser-ionized neutrals within the selected kinetic energy window (i.e., those having the pass energy) being preretarded to this energy before entering the hemisphere. For the ions, this amounts to about a factor of two in reduction from the nominal kinetic energy, 7i, with which they enter the zoom lens. The kinetic energy distributions of the desorbed ions are mapped out by scanning the pass energy, 72. of the CHA. For the case of laser-ionized neutrals, the ions are retarded by ~25 eV and velocity distributions, which may be converted to kinetic energy distributions, are determined by keeping Tz at a fixed value and varying the time between the laser and electron pulses. Because of the narrowness of the neutral energy distributions compared to the energy window of the CHA, the window at 72 is sufficiently wide to catch all of the laser-ionized neutrals. The sequence used in performing the time-of-flight laser-ionization mass spectrometry of the neutral ESD desorbates is: desorption of the neutrals from the methanol-dosed sample by an electron pulse, photoionization of the desorbates with 193 nm nonresonant laser radiation, and detection of the laser-formed ions by the detector of the CHA. As shown in figure 2.4, the delay, ti, between the electron and laser pulses is the quantity which establishes the velocity of the neutral desorbate, while the flight time, t 2, between the time of laser ionization and the time of detection of a single pulse within a pre-set gate, allows m ass identification. Therefore, by maintaining t 2 constant and varying ti, a velocity distribution for a particular mass can be obtained. For the velocity distribution experiments, a short electron pulse of Sample
Figure 2.3: Spatial and corresponding energy details of the ion (a) and neutral (b) experiments. (b) neutral and (a) ion of the details energy corresponding and Spatial 2.3: Figure Energy 0 Desorption Region Distribution Ion Energy Energy Ion Distance a (b) (a) Energy-Window (to Energy Analyzer) Energy (to Analyzer Zoom Lens Lens Zoom Sample Desorption Region Distribution Ion Energy Energy Ion Ionization Region Distance Laser * Energy Energy * Distribution Neutral Neutral (to Energy Analyzer) Energy (to Zoom Lens Lens Zoom Energy-Window Analyzer 19
Ion and Neutral TOF Mode x Neutral Mass Spectrometry E lectron puIse \ Mode
Laser Pulse
Mass Spectrometry Gate
D etector pulse
0 Time
Figure 2.4: The timing sequence employed in the ESD TOF experiment. 20
~1 j is provides acceptable resolution. For experiments in which such information is not sought, a sufficiently long electron pulse (~ 10 ps) is used in order to allow all velocities to be present in the photoionization volume. For the case of desorbed positive ions, an electron pulse width of ~ 1 ps is used. Since these species are already positively charged, no laser ionization is required, and the mass-resolved signal is identified by flight time t3. The kinetic energy distribution of the ions leaving the surface is mapped out by subtracting the nominal target bias from the kinetic energies detected by the CHA. For these ion experiments, the duty cycle of the laser is not a limiting factor, and repetition rates on the order of 10 4 Hz are typically employed. Laser ionization of the desorbed species is performed with an excimer laser (Questek model 2660) fitted with unstable resonator optics. The laser is operated with an ArF gas mixture and is capable of producing 20 ns pulses of light at 193 nm wavelength. Typical repetition rates for the experiments described here are 50-100 Hz. The laser ionization volume, determined by taking burn patterns at the laser focus, is estimated to be 2.4 x 10 -4 cm3. In some cases when it is desirable to reduce the laser power below the operating range of the laser, neutral density filters are used to attenuate the beam, with care being taken that there is no walk-off from the normal laser position. Two modes are used for data collection. Mass spectrometry runs, in which it is desirable to monitor all of the desorbing masses simultaneously, employs a computer-interfaced multichannel transient digitizer (LeCroy Corp. TR8828C). Because of gas phase background signal resulting from the efficient desorption of species from internal chamber surfaces by the 6.4 eV laser photons, the computer is operated in a subtraction mode, in which the electron beam is pulsed on alternate laser shots, and the difference spectrum is recorded. For velocity distribution measurements, where more efficient use of the data collection time is desirable, two gated counters are used for each m ass monitored, and the electron gun is toggled by a flip-flop circuit. The difference in counts between the counters is proportional to the ESD signal. Figure 2.5 provides examples of the types of data obtained in these experiments. Figure 2.5 a) shows a typical mass spectrum of laser a) 6 0 0 n '■ ■ T-p1 ■' i j 111' 11' 11 j i > i '|,,r i i ■ i j ■■ 'I-i | ■ 11 n 11
«♦—«tn — — "E 3 400 — — •Uo■ C0 L ] c 200 D) CO Q CO LLI 0 -H—----W ~ *....
■ " ■i■" 1 1■■■■i'■'■i ■' ■' i ■'' ■ ■ i' ■■' J ■1 ■11 ■■' ■ i' ■1' 0 10 20 30 40 50 Mass (m/z)
b) Electron Beam tn Energy (eV): c 3 3 0 0 0 "
CO
(0 c 03 03 1450 - + O 1200
0 0.4 0.8 1.2 1.6 Velocity (km/s)
Figure 2.5: a) Typical m ass spectrum of laser ionized/fragmented neutrals, b) Velocity distributions of the carbon fragment. 22
ionized/fragmented neutrals from ESD of methanol degreased aluminum-6063 alloy. Figure 2.5 b) displays velocity distributions of the C+ species shown in a) at a variety of electron beam energies. The details of these measurements are given in the next chapter.
Sample .preparation In order to preserve the condition of the AI-6063 alloy, no cleaning of this sample prior to or after insertion into the vacuum chamber has been performed. The alloy is used as received except for a rinse, before vacuum insertion, with research grade methanol. As opposed to the technical sample, extensive sample preparation is required for the single crystal studies. The Al(111) crystals used in this experiment are cut from a 99.999% pure boule purchased from Materials by Metron. The 8 mm diameter boule is oriented to within 2° of the (111) direction by Laue back reflection and cut by the spark erosion method. Each individual crystal is then oriented to within 0.5°, ground with 600 grit and 1200 grit paper, and polished successively with 30 pm, 15 pm, 6 pm, 3 pm, 1 pm, and 0.5 pm diamond abrasives. This polishing procedure leaves the crystal with a mirror-finish and a thickness of 1 -2 mm. The single crystal is mounted with a tantalum cup onto a commercial button heater (Spectra Mat., Inc.) which allows the sample to be resistively heated from room temperature to above 700 K. Extensive cycles of sputtering with 1.5 keV Ar+ at room temperature and annealing at 700 K are initially needed to remove carbon contamination presumably introduced in the polishing procedure. The sputtering is performed with a Varian (model 981 -2043) ion bombardment gun, during which the chamber is backfilled to a pressure of 8 x 10-5 mbar with 99.99% Argon (Matheson). Typical ion beam current at the target is ~ 5 pA. AES is used to monitor surface cleanliness and to follow the cleaning process. During the initial sputtering cycles of the aluminum, the crystal loses its mirror-finish and appears dull white (matte-finish). A similar observation was made by Crowell et al. [3] in their studies on Al( 111). 23
Once the initial carbon impurity is removed, the daily cleaning procedure is relatively short, consisting of one cycle of sputtering the crystal at 300 K for ~ 45 minutes followed by simultaneous annealing and sputtering at 700 K for an additional 45 minutes. To assure ordering of the surface atoms, the crystal is annealed at 700 K for 15 minutes. LEED spectroscopy of the sample reveals a sharp (1 x1) diffraction pattern characteristic of the AI(111) surface [4]. The carbon and oxygen impurity levels, based upon relative AES intensities, are <1 at.% after this cleaning procedure and before further experiments are conducted. A typical Auger of clean Al(111) is shown in figure 2.6. The peaks at 68 and 84 eV are due to Al, while C and O impurities, if present, would have major peaks at 272 and 510 eV, respectively. Since extensive sputtering destroys the mirror-finish of the crystal, several different Al(111) crystals have been used in the course of this investigation. In order to avoid breaking vacuum with the exchange of samples, a sample exchange system is employed in which a transfer arm can be admitted from outside the chamber to the vacuum system via an evacuable interlock. To facilitate this transfer procedure, thermocouples are not attached to the actual samples studied. Instead, a separate temperature calibration experiment has been performed under identical vacuum and sample mounting conditions. For this calibration experiment, a chromel- alumel thermocouple ( 0.01 inches diameter) is press-fitted into an aluminum sample, and the temperature is measured as a function of current applied to the sample button heater. Subsequently, sample temperatures at equilibrium are determined from the preset current, within an estimated error limit of ~ 20 K for this procedure. The methanol samples used in the course of the Al( 111) study are CH3OH, CH3OD, and CD 3OD. All are better than 99% pure and are handled under inert atmosphere conditions. They are purified with several freeze- pump-thaw cycles and introduced into the chamber via a leak valve connected to a stainless steel dosing tube inside the vacuum chamber. Dosing of the Al(111) sample is carried out by backfilling the chamber, and exposure of the surface to methanol is measured using a nude ion gauge and correcting for the relative sensitivity difference between methanol and air. All methanol doses are carried out at 300 K, and typical exposure pressures are dN(E)/dE 10 0 30 0 50 600 500 400 300 200 100 0 1 ■ « * 1 * 1 « 1 '■ 1 i 4 S t f V e 8 6 iue .: gro la Al(111). clean of uger A 2.6: Figure , , , I , 1 , 1 , 1 V e lcrn nry (eV) Energy Electron A ^ 'W 2 X j J . i . . . i 5 X ...... I . w . 24 a) 25
Concentric Hemispherical Energy Analyzer
Total net v energy dispersion -0.15
■a •a
-0.48--
■<— Zoom Lens Path Hemisphere > |
Flight Distance
Figure 2.7: a) Three rays entering at different angles differ in flight times; b) the total flight path of the zoom lens and hemisphere is only partially compensated for dispersions in flight times caused by variations of the energies of the ions. For clarity, the details of the behavior within the hemisphere are omitted in b), and the initial and final values of the dispersion in this region are simply connected by a dashed line. 26
5 x 10 "7 Torr. For some of the experiments, we wished to compare the results obtained from a methanol-dosed clean A!(111) surface to those from a surface which had been oxidized prior to methanol dosing. The procedure used in forming this oxide layer is the following. At room temperature, the chamber is backfilled to a pressure of - 1 x 10 ’6 Torr with oxygen of £ 99.6% purity (Matheson) until the aluminum has been exposed to several hundred Langmuir of oxygen. The chamber pressure is allowed to recover, and the crystal is then heated to 700 K. Typically,, two of these oxygen exposure- heating cycles are used in order to build up the oxide layer. Auger analysis of the surface after this procedure shows a strong oxygen signal at 510 eV and a distinct feature at 54 eV which is indicative of AI +3 from the formation of AI2O3. Similar methods of oxide buildup on Al(111) have been shown to be effective in producing an oxidized surface [3].
Characteristics of the CHA as a TOF Detector The concentric hemispherical analyzer is most often used in surface science for analyzing the energy of emitted electrons, such as in the Auger and XPS techniques. In addition to using it for these techniques, it has been employed in this thesis research as a time-of-flight detector for ESD ions and for laser-ionized neutrals. The mass resolution achievable with the setup is limited by uncompensated variations in the flight times through the hemispherical region arising mainly from the spread in the angles of entry. This effect is depicted in figure 2.7a). As can be seen in this figure, ions following the path labelled as "1" have a shorter time of flight to the detector than do ions entering normal to the plane of the CHA. Similarly, those entering at an angle such as that of "3", have a longer flight time. This dispersion in the time of flight for a given m/z, of course, leads to degradation of the mass resolution. Timing calculations (for which exact analytical formulas exist [5] for TOF on a pure R"1 potential, with R being the radial coordinate) give an angular 27
dispersion coefficient of 0.5 ps deg -1 for mass 12 at small angles of entry. Another effect which is detrimental to the mass resolution is the spread in the energies of the ions. Ion optic modelling shows that the energy dispersion is 0.33 ps eV '1 of variation of initial kinetic energy for the portion of the trajectory within the hemisphere, and -0.15 ps eV-1 for the entire path, which is partially energy compensated. The cumulative energy dispersion as a function of location along the flight path is depicted in figure 2.7 b). For reference, the protons detected in ionic ESD have a total flight time of "6.4 ps from the sample to the detector, while the time of flight from the laser ionization volume to the detector for the laser-ionized photofragment C+ species is ~23 ps. Since the energy distributions of the neutral desorbates are narrow, the dispersion in flight times, and, hence, the degradation of mass resolution is due mainly to the angular effect. Chapter V describes the construction of the SARISA apparatus, an advanced mass spectrometer which compensates for energy and angular spreads.
References
[1] L. C. Feldman and J. W. Ma yer,Fundamentals of surface and thin film analysis. New York: Elsevier, 1986.
[2] M. W. Roberts and C. S. McKee,Chemistry of the metal-gas interface. Oxford: Clarendon Press, 1978.
[3] J. E. Crowell, J. G. Chen, and J. T. Yates Jr., "Surface sensitive spectroscopic study of the interaction of oxygen with Al(111) - low temperature chemisorption and oxidation," Surface Sci., vol. 165, pp. 37-64, 1986.
[4] F. Jona, "Preparation and Properties of Clean Surfaces of Aluminum," J. Phys. Chem. Solids, vol. 28, pp. 2155-2160, 1967.
[5] L. D. Landau and E. M. Lifshitz, Mechanics. Reading: Addison- Wesley, 1960, p.38. CHAPTER III
AI-6063 ESD EXPERIMENTAL RESULTS
The surface conditions of the methanol-rinsed aluminum alloy As mentioned in the introduction, electron- or photon-stimulated desorption and ways to minimize their effects are problems to be considered in the design of synchrotron vacuum chambers. This chapter details a study of the use of methanol as a possible degreasing agent. The 6063 aluminum alloy, the vacuum wall material chosen for the Advanced Photon Source at Argonne National Laboratory, has been studied as extruded except as follows. First, a piece of the alloy has been machined down to a disk (~0.312 inches diameter, ~0.06 inches thickness) suitable to fit in a sample holder. Secondly, the sample has been rinsed, in air, with methanol in order to remove any surface contaminants, such as machine oil, which have low vapor pressures. This degreasing procedure, though it may wash heavier hydrocarbons from the surface, also may deposit methanol or methanol fragments onto the sample. If this type of agent is to be used as a degreaser, the hope is that when the vacuum chamber walls are baked the newly adsorbed species will either have a high enough vapor pressure to desorb thermally or will be converted to species which have small stimulated desorption cross sections. The interaction of methanol with this aluminum alloy has never previously been studied with standard surface science methods, and it has not been the goal of this research to elucidate the chemistry of this interaction. The goals of this research are more practical and are aimed at determining what the desorbates are and at estimating their yields both before and after the sample has been baked. Less practical, but of interest regarding mechanisms of desorption, is the goal of measuring velocity/energy distributions of the desorbates.
28 29
The ionic ESD channel from methanol-degreased 6063 Al allov The only ions detected from the degreased aluminum alloy are H+, H2+, and F+ approximately in the ratio 66:1:14. The desorbing hydrogen may arise from the methanol used to degrease the sample or from any hydrogen-containing species which may have contaminated the sample during the extrusion process, while it was exposed to air, or while it sat in the vacuum chamber. Fluorine is a known contaminant of aluminum surfaces and has been observed in ESD from aluminum films [1],[2], from water-dosed Al(100) single crystals [3], and from fluorine-dosed alumina [3]. Kinetic energy distributions of the H+ and F+ ions have been measured for several different electron beam energies between 1000 eV and 3000 eV and are shown in figures 3.1 and 3.2. These data have been acquired, as described in the previous chapter, by scanning the CHA energy window. For these ion energy distributions, an electron pulse width of 0.2 ps and a gate width of 0.5 ps are employed. The curves shown exhibit peak energy and FWHM values of several eV, typical of all ionic ESD species which have been studied to date. Surprisingly, very few studies exist in the literature in which kinetic energy distributions of desorbed ions have been measured over a large range of electron beam energies. The variation in the position of the peak in the energy distributions of figures 3.1 and 3.2 may be attributable to contributions to the total desorption signal by a group of excited state repulsive curves which have differing slopes in the region of Franck- Condon overlap. Variations in the probability of excitation to these upper states with excitation energy could explain the experimentally observed shifts. This sort of concerted shifting of the slopes of the repulsive curves might be visualized in terms of the desorption proceeding through a Coulombic ejection of the ion from the surface. The basic concept of this theory is that the impinging electrons or photons cause the formation of multiply-charged core hole excited states in the metal-adsorbate system. Inter- or intra-atomic Auger decay of theses states may then lead to Signal (Normalized) 6 - 160 0 2 1 80 40 0 © Figure 3.1: Kinetic energy distributions of d esorbed H+ from the the from H+ esorbed d of distributions energy Kinetic 3.1: Figure ' r ' i 6063 aluminum alloy a t different beam energies. beam different t a alloy aluminum 6063 4 6 4 2 o - S - iei Eeg (eV) Energy Kinetic : q ^ - b ■ r ■ i Electron Beam Beam Electron " O nry (eV): Energy ~ 0 ' 3000 1500 1300 1000 Signal (Normalized) 160 0 2 1 80 40 0
2 6 10 8 6 4 2 0 P Figure 3.2: Kinetic energy distributions of desorbed F+ from the from F+ desorbed of distributions energy Kinetic 3.2: Figure 1 | I [ I | ' ' < ( < j ' | ' | ' I ■ [ ■ I ' | ■ <6 nry (eV): Energy -. lcrn Beam Electron 0-0. 4 8 6 4 2 * 0 6063 aluminum alloy at different beam energies. beam different at alloy aluminum 6063 * iei Eeg (eV) Energy Kinetic iei Eeg (eV) Energy Kinetic „ o - - ° 3000 1300 1500 1000 ------r—
32 accumulation of positive charges on the originally core-ionized atom and neighboring atoms and to dissociation (i.e., desorption) via Coulombic repulsion. This type of mechanism, often referred to as the Knotek- Feibelman mechanism (see chapter VII), has been used to explain desorption from ionically bonded systems, such as 0 + desorption from T1O2 [4],[5], and has also been cited as a desorption mechanism operative in covalently-bonded systems [ 6], [7], [ 8], [9]. The exact details of the H+ and F+ desorption mechanism are not understood, but if the cross section for the formation of a multiply charged core hole increases more rapidy with decreasing beam energy than does that for a singly charged core hole, then the energy distribution of the repelled desorbate (for example, H+ desorbing from two methoxy carbon sites, C+ and C++) will contain a greater contribution from the desorption site which is doubly ionized as the beam energy decreases. Decreasing beam energies will, therefore, cause the desorbing positive ion to feel a greater contribution from the doubly ionized core hole and to desorb with higher kinetic energies. It must be admitted that this is a somewhat hand-waving argument and that very little is known about the energy dependence of the cross sections of interest. The kinds of experiments which are necessary to validate and quantify this theory are ones which search for the onset of ionic desorption and ones which measure the energy distributions as a function of beam energy down to the threshold energies of the core hole formation. Unfortunately, limitations in the beam energies accessible with the current setup did not allow for this kind of information to be obtained.
Neutral desorption from the degreased technical sample Nonresonant laser ionization of molecules, although it can be very efficient relative to other means of ionization, such as electron impact, may also cause extensive fragmentation of the molecule. This is especially true for laser intensities where photoionization becomes highly efficient, and the competition between photoionization and photofragmentation for 33
carbon-containing species is well documented in the literature [ 10], [11], [12], [13]. For the studies presented in this chapter, ArF excimer radiation (~193 nm) has been used to ionize ESD neutrals desorbing at room temperature from the aluminum alloy sample. Figure 3.3 shows a time-of- flight mass spectrum of the neutral species desorbed by 3 keV electrons. For this spectrum, long electron pulses (~10 |is) are followed by laser photoionization. The laser focal intensity is ~0.3 GW/cm2, with the dimensions of the effective laser volume in front of the sample (as determined by bum patterns taken at the focus) being approximately 0.15 x 0.4 x 4 mm. The spectrum has been acquired in a multichannel arrival time digitization mode (as described in chapter II), and the gas phase ionization background has been subtracted. The major species detected is C+, but m asses 13 (CH+), 24 (C 2+), 28 (CO+),and possibly 31 (CH 3 0 +) are also observed. As described later in this thesis, extensive studies of the dependence of fragmentation on laser focal intensity for the CH 30/AI(111) system show that C+ is efficiently formed as a photofragment of the methoxy (CH 3O) species. It is likely, then, that atomic carbon is not desorbing from the aluminum alloy surface, but that C+ arises from photofragmentation/ionization of a parent desorbate, such as methoxy. As pointed out earlier (see table 1.1), very few studies have been performed in which kinetic energy distributions of desorbed neutrals have been measured. Because of its relevance to desorption mechanisms, this type of information has been sought in the present study, and kinetic energy distributions have been determined by measuring velocity distributions via the method described in the previous chapter. In order to obtain efficient ionization of the neutral desorbates from the aluminum alloy sample, high laser focal intensities have been used. Under these conditions, it has been possible to obtain velocity distributions (from which energy distributions may be obtained by knowing the desorbates mass) with adequate signal-to-noise ratios only for the C+ photofragment. Figure 3.4 displays velocity distributions of the C+ species at three electron beam energies in the range 1200-3000 eV. The most remarkable feature of these distributions is the very low velocities observed, ~250 m/s at the peak, corresponding to a kinetic energy of only 0.01 eV if the parent is ESD Signal (arb. units) 200 400 600 Figure 3.3: Mass spectrum of laser ionized/fragmented ionized/fragmented laser of spectrum Mass 3.3: Figure 0 0 1 neutrals desorbed from the AI-6063 sample. fromthe desorbed neutrals as (m/z) Mass 20 040 30 50 34 Signal (arb. units) 1 1 iue34 Vlct itiuin, ttre lcrn em nris o h C+ the of energies, beam electron three at distributions, Velocity 3.4: Figure 0 fragment arising from laser ionization/fragmentation of the the of ionization/fragmentation laser from arising fragment neutrals desorbed from the Al-6063 sample. The continuous continuous The sample. Al-6063 the from desorbed neutrals curves are least-squares fits to the data for the functional forms forms functional the for data the to fits least-squares are curves described in the text. Dashed curves: Boltzmann distribution; distribution; Boltzmann curves: Dashed text. the in described oi cre: lnrbrirpwrlw model. law planar-barrier/power curves: solid 0.4 eoiy (km/s) Velocity 0.8 Electron Beam Beam Electron 1.2 nry (eV): Energy 3000 “ 3000 1450 - 1450 1200 1.6 - 36
assumed to be the methoxy species. Also notable is the apparent independence of the peaks and shapes of the distributions on the electron beam energy employed. This is in direct contrast to the ionic ESD channel studied in this work, in which the peak kinetic energy shifts as a function of electron energy. Velocity distributions of the higher mass region (m/z=28- 31) have been attempted and, though they confirm the low value of the mean velocity seen in the runs monitoring the C+ species, the noise level is too high to extract significant information about the shape of the distribution. In the velocity distribution studies, extensive searching of the time domain towards shorter flight times reveals no highly energetic components of the velocity distribution. As an independent verification that there are no energetic photoionized desorbates which are being overlooked, the CHA energy-window has been scanned in search of any energetic fragment ions. This energy-domain search is carried out with an electron pulse width of 10 jis, which is long relative to the range of possible flight times from the sample to the laser ionization volume. When the width of the obtained scan is compared to the intrinsic energy window of the CHA instrument, it is concluded that there is no significant signal arising from photoionized species having a kinetic energy greater than about 0.2 eV, the effective resolution of this test. The transmission window characteristic of the CHA is mapped by bombarding the aluminum target with electrons and scanning the elastic peak, which is essentially monochromatic at the resolution employed. As discussed in chapter II and depicted in figure 1.4, two modes of electron pulse length have been employed in the neutral desorption studies. In standard TOF mass spectrometry runs, in which information is sought regarding the identities of the desorbates, the electron beam is typically operated up to or past the firing time of the laser. This type of operation (long- pulse limit) provides a steady-state distribution of velocities in the laser volume and the maximum possible signal. At the other extreme, for velocity distribution measurements, short electron pulses are desirable (short-pulse limit) so that only a narrow band of velocities is present in the laser volume during the laser pulse. The delay time between the electron pulse and the laser fire can then be varied, and the velocity distribution can be determined. 37
The data in figure 3.4 have been fit to two types of functions: a standard Maxwell-Boltzmann distribution and a planar-barrier/power-law model. Limiting cases of standard velocity distribution functions have been described recently [14], and for the short-pulse limit, the Boltzmann form of the distribution is given by
g(u)=iA exp(-i£) (3.1)
where the laboratory velocity v is converted to the reduced velocity u=v/vm, and vm ='J(2kT/M), with k being the Boltzmann constant, 7" the temperature, and M the mass of the desorbed species. In the planar-barrier/power-law model, there are only two assumptions: ( 1) that there exists a one dimensional energy barrier to desorption of magnitude £t>. and ( 2) that there is an isotropic distribution of flux just below the surface with a power-law dependence on kinetic energy E, i.e. flux ~E'n. The short-pulse limit of the distribution corresponding to equation (3.1) is
g(u) = u*/(1 (3.2)
where u = v/v^ and vb = V( 2 Et>//W). Both the Boltzmann and the planar- barrier/power-law distributions peak at a value up = V2. As determined from the data of figure 3.4, the velocity distributions peak at a value vp = 250 m/s, giving values of vb or vm = 177 m/s. Assuming that the desorbate is the methoxy species (mass 31), then the value of Ep or kT obtained is 0.005 eV. The functions given by equations (3.1) and (3.2) have been fit (with n=2) to the data of figure 3.4, the two free parameters being a vertical scaling factor and either i/b or vm. In all cases, the Boltzmann function given by equation (3.1) is unable to fit the high-velocity tail of the experimental data, while the function specified by equation (3.2) represents the data much better. In addition, when the power-law index n is allowed to be a free parameter, fitted values close to 2 are obtained. This value is familiar from applications of the planar-barrier/power-law model to sputtering [15], where the value of 2 arises from two-body momentum transfer collisions. Direct momentum transfer from electrons to the desorbate cannot be the mechanism for removal of 38
chemisorbed species in ESD [16], so the value 2 for the exponent in equation (3.2) cannot be explained on the basis of the cascade of secondary electrons [17],[18],[19],[20]. The observation of very low energy neutral desorption from the degreased aluminum alloy surface, though in agreement with other existing neutral ESD investigations, is quite different than results from the many studies of kinetic energy distributions of desorbing ions which have been performed. A discussion of possible differences in ionic and neutral desorption mechanisms will be deferred until chapter VII. The ESD results from this technical sample have prompted the extension of these studies to more carefully controlled surfaces, and the next chapter describes the ESD results from methanol-dosed Al(111) surfaces.
Neutral and ionic yields from the aluminum alloy sample The desorption yield is of interest in assessing the magnitude of the ESD problem for synchrotrons and in evaluating possible desorption mechanisms. The detected signal S (in counts per laser shot) originating from neutral ESD is related to the density n of the neutral desorbate in the laser photoionization volume V by
S=nVfe|£T£D (3.3)
where the factors e\, cr. and cd represent the effieciencies of photoionization, transmission through the zoom lens and CHA, and detection by the electron multiplier, respectively. The gas phase density, n is given by
n =[(/Y)/(4vb)]
with vb being the characteristic velocity obtained in the fit of the data to the planar-barrier/power-law model, I the electron beam current (in 39
electrons/sec) onto the sample, 0 the angle of desorbate ejection relative to the surface normal, and R the distance from the surface to the laser ionization region. In this approximation, the average over the experimental geometry, indicated by the angled brackets in equation (3.4) can be ignored, and the cos0 factor can be set equal to one. Also, since so is close to unity for the electron multiplier conditions used, this factor may also be set to one. Choosing a value for £| is somewhat more approximate, but a value of one represents an upper limit. Though the actual value of e\ may be somewhat lower, this choice is not unreasonable for the high laser power conditions employed in obtaining the data used for this estimate. Ion optic modeling of the zoom lens system has been used to obtain a value for er of 5.0 x 10'4. Experimentally, a net count rate of at least 1 count per laser shot (S = 1) is observed for laser-ionized ESD neutrals. This, combined with typical values / =1.9 x 1013, vb = 1.77 x 10 4 cm/s, and R = 0.25 cm, gives a yield, Y, of ~1 x 10 '3 desorbed neutrals per incident electron. In the case of ionic desorption, an equation analogous to equation (3.3) may be written as
S=IAtYer£D (3.5) with At representing the width of the electron pulse used. Experimentally, At =0.85 ps. If £d is again assumed to be unity, and a value of 2.5 x 10 '4 is taken for the transmission efficiency (a value a factor of two lower than in the case of the neutrals has been chosen to account for rejection of some of the ions by the CHA energy window which is narrower than the width of the ionic energy distribution), then a yield estimate of ~1 x 10'6 desorbed ions per incident electron is obtained. These yields are not atypical of studies carried out on other chemisorbed systems, where the neutral yields are found to be orders of magnitude larger than the ionic ones. The experiments described in this chapter have been carried out both before and after baking the entire vacuum chamber (with the AI-6063 sample installed) at ~170° C. The observed velocity distributions and yields are, within the accuracy of the 40
measurements, identical. This independence of yields on whether or not the sample has been baked has important implications for the use of this material and this kind of degreasing procedure in the design of synchrotron vacuum chambers. In such designs, it is hoped that a bake of this kind would remove any species which may be desorbed by electrons or photons. The findings of this research indicate that baking alone is not sufficient to desorb the vacuum-wall adsorbates. Typical strategies used in bringing a synchrotron on line are to bake the system and then to condition the walls of the vacuum chamber by allowing the x-rays produced (and their accompanying phtoelectrons) to desorb the species gradually. A total yield in the range of 10 '5 - 10-6 desorbates per photon is generally considered tolerable in a synchrotron, although higher beam quality and lifetime could be achieved with lower desorption rates. A yield of ~1 x 10'6 is the best that can currently be achieved (see [ 21], for example), and this results only after extensive beam-conditioning. Clearly, the total yield of ~1 x 10 -3 (the ions are negligible compared to the neutrals) observed from the degreased aluminum alloy study of this thesis is far too large to be acceptable under steady state synchrotron running conditions.
References
[1] P. Williams and G. Gillen, "Direct evidence for coulombic ejection of electron-desorbed ions," Surf. Sci., vol. 180, pp. L109-L112, 1987.
[2] C. Park, M. Bujor, and H. Poppa, "Electron-stimulated desorption study of hydrogen-exposed aluminum films," Thin Solid Films, vol. 113, pp. 337-344, 1984.
[3] M. L. Knotek, R. H. Stulen, G. M. Loubriel, V. Rehn, R. A. Rosenberg, and C. C. Parks, "Photon-stimulated desorption of H+ and F+ from BeO, AI2O3 and Si 0 2 : comparison of near edge structure to photoelectron yield," Surf. Sci., vol. 133, pp. 291-304, 1983. 41
[4] M. L. Knotek and P. J. Feibelman, "Ion desorption by core-hole auger decay," Phys. Rev. Lett., vol. 40, pp. 964-967, 1978.
[5] M. L. Knotek, V. O. Jones, and V. Rehn, "Photon-stimulated desorption of ions," Phys. Rev. Lett., vol. 43, pp. 300-303, 1979.
[6] R. Franchy and D. Menzel, "Adsorbate core ionization as primary process in electron- and photon-stimulated desorption from metal surfaces," Phys. Rev. Lett., vol. 43, pp. 865-867, 1979.
[7] M. Q. Ding, E. M. Williams, J. P. Adrados, and J. L. de Segovia, "Energy distribution of H+ ions with ESD of water adsorbed at aluminum and tungsten surfaces," Surf. Sci., vol. 140, pp. L264-L268, 1984.
[8] D. E. Ramaker, C. T. White, and J. S. Murday, "On Auger induced decomposition/desorption of covalent and ionic systems," Phys. Lett. A, vol. 89, pp. 211-214, 1982.
[9] D. E. Ramaker, "Models for desorption in covalent systems," in Desorption Induced by Electronic Transitions (DIET I), N. H. Tolk, M. M. Traum, J. C. Tully and T. E. Madey, Eds. Berlin: Springer-Verlag, 1983, pp. 70-89. '
[10] H. J. Neusser, "Multi-photon mass spectrometry and unimolecular ion decay," Int. J. Mass Spectrom. Ion Processes, vol. 79, pp. 141-181, 1987.
[11] R. B. Bernstein, "Systematics of multiphoton ionization-fragmentation of polyatomic molecules," J. Phys. Chem., vol. 8 6, pp. 1178-1184, 1982.
[12] R. L. Whetten, K.-J. Fu, R. S. Tapper, and E. R. Grant, "Highly efficient production of neutral carbon atoms in the ultraviolet multiphoton fragmentation of aromatic molecules," J. Phys. Chem., vol. 87, pp. 1484-1487, 1983.
[13] R. C. Sausa, A. J. Alfano, and A. W. Miziolek, "Efficient ArF laser production and detection of carbon atoms from simple hydrocarbons," Appl. Optics, vol. 26, pp. 3588-3593, 1987.
[14] H. F. Arlinghaus, W. F. Calaway, C. E. Young, M. J. Pellin, D. M. Gruen, and L. L. Chase, "High-resolution multiphoton laser-induced fluorescence spectroscopy of zinc atoms ejected from laser-irradiated ZnS crystals," J. Appl. Phys., vol. 65, pp. 281-289, 1988. 42
[15] P. Sigmund, "Theory of sputtering. I. Sputtering yield of amorphous and polycrystalline targets," Phys. Rev., vol. 184, pp. 383-416, 1969.
[16] D. Menzel and R. Gomer, "Desorption from metal surfaces by low- energy electrons," J. Chem. Phys., vol. 41, pp. 3311-3328, 1964.
[17] P. A. Wolff, "Theory of secondary electron cascade in metals," Phys. Rev., vol. 95, pp. 56-66, 1954.
[18] E. N. Sickafus, "Linearized secondary-electron cascades from the surfaces of metals. I. Clean surfaces of homogeneous specimens," Phys. Rev. B, vol. 16, pp. 1436-1447, 1977.
[19] E. N. Sickafus, "Linearized secondary-electron cascades from the surfaces of metals. II. Surface and subsurface sources," Phys. Rev. B, vol. 16, pp. 1448-1458, 1977.
[20] E. N. Sickafus and C. Kukla, "Linearized secondary-electron cascades from the surfaces of metals. III. Line-shape synthesis," Phys. Rev. B, vol. 19, pp. 4056-4068, 1979.
[21] H. J. Halama and C. L. Foerster, "Comparison of photodesorption yields from aluminum, stainless and Cu-plated beam tubes," Vacuum, vol. 42, pp. 185-188, 1991. CHAPTER IV
Al(111) CHA EXPERIMENTAL RESULTS
An overview of the experiments carried out on the Al(111 > surface The surface studied in the previous chapter, though technologically important, is complex and ill-defined. This chapter discusses investigations carried out on atomically clean single crystal Al(111) and on heavily oxidized Al(111). The work described here has been motivated, in part, by the results from the methanol-rinsed 6063 aluminum alloy, and by the fact that the interactions of methanol and oxygen with Al(111) are fairly well understood. This chapter is organized into several parts. The first section summarizes current knowledge about the chemistry of methanol on aluminum and describes XPS experiments which have been performed in this thesis work to verify and add to this knowledge. Included here are investigations into the thermal stability of the methanol-dosed system, studies of the oxidation of Al(111) via oxygen dosing, and experiments exploring the site dependence of methanol dosing. The second and third sections describe experiments on the ionic and neutral desorption channels, respectively. The last section is concerned with demonstrating the use of neutral ESD as a probe of changes in surface chemistry.
The interaction of methanol and oxygen with Al(111) Considerable published evidence leads to the conclusion that exposure of Al(111) to a saturation dose of methanol at 300 K results in monolayer coverage by the methoxy species, CH3O, with the O atom adjacent to the metal. Table 4.1 summarizes the existing literature on the subject. Yates and
43 44
Table 4.1: Studies of methoxy on aluminum
Surface Significance Referenced
AI(111) Surface is dosed with methanol [1] at 90 K. Methoxy is the stable surface species over the range 143->400 K.
Al(110) Surface is dosed with methanol [2] at both 150 K and 300 K. The methoxy species is adsorbed at these temperatures and is stable until > 400 K.
polycrystalline Room temperature dosing with [3] methanol leads to surface methoxy formation. The methoxy is stable until ~500 K.
polycrystalline Room temperature dosing with [4], [5] methanol leads first to oxidation of the aluminum surface. Further methanol exposure leads to surface methoxy formation. 45
co-workers [1] have followed the chemistry of methanol-dosed Al(111) as a function of temperature using temperature programmed desorption (TPD) and high-resolution electron energy-loss spectroscopy (HREELS) and have shown that at 90 K multilayers of methanol adsorb on the surface. At 143 K, TPD shows the desorption of methanol and hydrogen as the multilayers of molecularly adsorbed methanol are removed and a monolayer of surface methoxide is formed. The hydrogen observed in the TPD apparently arises from the scission of the OH bond of the methanol as the methoxy binds to the surface. HREELS spectra in the temperature range 200 - 400 K are assigned to the methoxy species. By 700 K, the methoxy has decomposed leaving aluminum oxide and aluminum carbide on the surface. No CH 3O is observed to desorb in TPD studies, but the evolution of methane at ~ 445 K is related to the decomposition of the surface methoxy species. Waddill and Kesmodel [2] have also studied the interaction of methanol with single crystal aluminum using HREELS. These investigators found that for methanol-exposed Al(110) at 150 K (and also at 300 K), methoxy forms and is stable until the temperature is above 400 K. Studies of methanol adsorbed on polycrystalline aluminum [3],[4] also provide evidence for the formation of surface methoxide at room temperature. XPS and SIMS data, along with the release of methane, have been interpreted as evidence that some initial oxidation occurs during the first few Langmuirs of methanol dosing of clean aluminum at room temperature [4],[5]. The picture that emerges, then, as depicted in figure 4.1, is that the methanol dosing of Al(111) at room temperature proceeds through an initial phase in which the surface is oxidized prior to the formation of the methoxy species. Further methanol dosing leads to the formation of the surface methoxy species, which is stable until temperatures are between 400 and 500 K, where it decomposes to surface carbide and oxide. As shown in figure 4.1, the methoxy is believed to be tilted with respect to the surface normal, although the bend angle is unclear [1],[3]. Note that some assumptions have been made in this diagram. In reality, the particular surface sites occupied by methoxy, its surface coverage, and the sites for, and degree of, oxygen underlayer formation are not known. 300 K
Figure 4.1: The formation and thermal decomposition of the surface methoxy species on Al(111). As explained in the text,. some assumptions are made about which surface sites are occupied. 47
In this thesis work, XPS has been used to monitor the carbon and oxygen surface concentrations as a function of methanol exposure. XPS is preferable to AES analysis for adsorbate-covered systems due to the lessened effects of beam damage. As shown in fig. 4.2, both the C(1s) and 0 (1 s) core hole signals rise rapidly with initial dose, then level off, indicating that the AI(111) surface is approaching saturation coverage of methoxy between 4 and 12 L exposure. XPS has also been used to follow changes in adsorbate composition as the dosed sample is heated and to gain some information about the pyrolysis products. For these studies, excitation is at the MgKa energy (1253.6 eV), and the observed binding energies are calibrated by reference to known values of the Al(2p) core level of clean Al (B.E. = 72.6 eV) and of the 0(1 s) core level for oxygen on A!(111) (B.E. = 531.3 eV) [6], The XPS data displayed in figures 4.3 and 4.4 are consistent with the observations that between 400 and 500 K, methoxy is pyrolyzed to Al-C and AI-0 species. In these studies, a saturation coverage of methoxy has been used. As demonstrated in figure 4.3, the 0.9 eV shift in the 0(1 s) core level observed from the pyrolysis of adsorbed methoxy is the sam e as the difference in 0(1 s) binding energies between adsorbed methoxy and 1 L of oxygen dosed onto clean Al(111) in a separate experiment. As shown in figure 4.4, the 3.6 eV shift in the C (1s) core level is indicative of the thermal decomposition of adsorbed methoxy to an aluminum carbide species. Note in both figures 4.3 and 4.4 that the methoxy species is stable until at least 354 K. The large shift in the C(1s) core level between methoxy and the carbidic species is representative of the change in environment of the carbon atom in going from a situation in which it is bonded to oxygen and hydrogen atoms to one in which it is attached to aluminum. The oxygen shift is smaller because of its less drastic change in chemical environments. Though comparable XPS observations have been used to help confirm the surface chemistry on polycrystalline aluminum [3], the XPS results described in this thesis are the first for methoxy on the Al(111) substrate. Although the location and coverage of methoxy on Al(111) are not determinable from the literature or from this thesis research, the oxygen/AI(111) system is well-studied. There appear to be three stages of oxidation of Al(111) [ 6]. The first is ordered overlayer formation, with the Peak Area (arb. units) 2 -- - 3 - - 5
j - - iue42 0(1Figure C(1s)and studyofs) methanolthe4.2: dosingof Al(111). 0 4 ehnl xoue (L) Exposure Methanol 8 f 12 16 (1s) C ) s 1 ( O 20 48 Intensity (arb. units) 4 56 3 58 2 520 524 528 532 536 540 Figure 0(1 s) core4.3:holefor 0 =. eV 8=0.9 idn Eeg (eV) Energy Binding 2 and K 0 0 3 t a O g H C K 4 5 3 t a O o H C CHgO K 1 1 7 t a O n H C K 0 0 3 t a 0 0 r —A mi, onAl( 1 11). 49 Intensity (arb. units) 9 20 8 20 7 270 275 280 285 290 295 -1- 4- iue44 C(1s)Figurecore hole of 4.4: CHgO/AI(111) various at I 1 I 1 i -- | M | - I 1 I-1 I 1 | 1 M | l-l-i 1 1 i 1 II | 1 I idn Eeg (eV) Energy Binding temperatures. 36 eV =3.6 I K 0 0 3 K 4 5 3 50 51
oxygen molecule dissociatively adsorbing and occupying the threefold hollow sites on the surface. The second stage is underlayer formation, although the site occupied by the subsurface oxygen is unclear. The last stage of Al( 111) oxidation is bulk formation of alumina, AI 2O3. The kinetics of these stages of oxidation are dependent on the oxygen pressure and surface temperature. Figure 4.5 depicts the early stages of oxidation of Al(111). Interestingly, it appears that the second stage of oxidation, underiayer formation, begins prior to formation of a full monolayer of oxygen on the surface [7]. In order to gain some insight into how methoxy resides on the Al(111) surface, the following XPS experiment has been performed. The 0(1 s) core level has been monitored for changes in intensity as room temperature Al(111), saturated with methoxy, is exposed to oxygen. At issue is whether or not the methoxy blocks the uptake of oxygen. The results of this investigation are displayed in figure 4.6. Clearly, as evidenced by the growth of the 0 (1 s) intensity, the methoxy does not completely block the incorporation of oxygen into the surface. Note also the 0.9 eV shift in the position of the peak away from the position of methoxy and towards that of oxygen. Separate experiments, for intensity comparison, have been carried out on oxygen dosed onto clean Al(111) at about the same exposure (114 L). However, because of differences in the optimized signal from one day to another, it is difficult to ascertain the degree of site blocking. What seem s clear, though, is that methoxy does not occupy every threefold hollow site on the Al(111) surface, as does oxygen [ 6], for if it did, there could be no further incorporation of oxygen onto the surface and into the bulk. The possiblity that the methoxy is somehow displaced by oxygen dosing may be dismissed by comparison of the FWHM of the 0(1 s) peak from the coadsorbed system of figure 4.6 to the 0(1 s) peak from oxygen dosed onto a clean Al(111) surface. The former displays a FWHM of 3.5 eV, while the latter has a FWHM of only 3.0 eV. The wider FWHM of the coadsorbed system is indicative of the presence of more than one "type" of oxygen. Two types of methanol-dosed surfaces have been used in the ESD studies presented in this thesis work. The first is Al(1 11) which is shown to be clean, prior to dosing, by AES and XPS. The second type of surface is Al(111) which has been sputter cleaned and then dosed with several hundred Clean Al(111) Partially O xidized A l ( 1 1 1 )
Figure 4.5: The early stages of the oxidation of Al(111). Note that there are two kinds of threefold sites on the surface. The surface oxygen occupies the threefold hollow site. Subsurface oxidation is thought to proceed simultaneously with surface oxidation. Intensity (arb. units) 540 iue46 0(1Figure XPSs) study4.6: of dosing the of methoxy-saturated 3 3 2 2 515 520 525 530 535 A!(111) with The oxygen.surface is room at temperature. idn Eeg (eV) Energy Binding en - lean C 53 54
Langmuir of oxygen, as described in chapter II of this dissertation, prior to applying a saturation dose of methanol. Subsequently, we will refer to the two types of dosed surfaces as "clean", and "pre-oxidized", respectively. However, as described previously, XPS and SIMS data, along with the release of methane during the methanol dosing, have been interpreted as evidence that some initial oxidation occurs during the first few Langmuirs of methanol dosing of a clean aluminum surface at room temperature. The Al(2p) core level is a good monitor of the degree of oxidation of the aluminum surface [6], and we have monitored this core level for both methanol-dosed Al(111) and oxygen-dosed Al(111) in an attempt to ascertain the degree of oxidation brought about when methanol is dosed onto clean Al(111). The oxygen- dosed sample gives an XPS spectrum indicative of the formation of AI 2O3, through a peak in the spectrum at a binding energy of 2.74 eV relative to Al(2p) on the clean surface. The methanol-dosed sample shows much less oxidation, with very little or no indication of the formation of AI 2O3. This difference in the Al(2p) core level spectrum indicates that the electronic properties of the clean and pre-oxidized surfaces are different. It should be pointed out, however, that the escape depth of the Al(2p) electrons, when excited by the MgKa radiation, is - 20 A [8],[9],[10] and that the XPS technique is sensitive only to the first few surface layers. Thus, after completion of the methanol dosing, both the clean and pre-oxidized surfaces may have a more similar degree of oxidation than indicated by the XPS spectra.
The ionic ESD channel from methanol-dosed AIM 11) Figure 4.7 is a mass spectrum, taken at 300 K, of the ions desorbed by 3 keV electron bombardment of clean Al(111) dosed with 55 L of CH 3OH to give a saturation coverage of methoxy. Notable is the lack of any detectable methoxy ion, CH 3O+, or any of the fragments seen in laser photofragmentation/ionization of the neutral desorbate channel (as described in the next section). As will be discussed in chapter VII, this has important Ion Signal (arb. units) iue47 Massspectrum Figureof 4.7:ionsthefrom observed keV3 electron 0 bombardment of methoxy-saturated Al(111). 10 as (m/z) Mass 20 Mass 30 40 55 56
implications regarding whether or not reneutralization of desorbing ions is a dominant channel in neutral ESD. A similar spectrum is obtained from CH 3 OD dosing of clean Al(111), with the major ionic desorbate still being H+ This is evidence that the protons observed in ESD arise from the methyl group of the methoxy and that the hydrogen atoms dissociated from the hydroxyl groups either are not present on the surface or are ESD inactive. If the mechanism proposed by Yates for methoxy formation at 143 K, as discussed in the last section, is valid at room temperature, then the hydrogen atoms from the hydroxyl group are not on the surface, but have recombined and thermally desorbed as hydrogen molecules. Figure 4.8 displays kinetic energy distributions of the desorbing protons for electron beam energies between 2 and 3 keV. The shift in the peak of the energy distribution as a function of beam energy is most likely due to the same mechanism described in the last chapter and used to explain similar observations from the AI-6063 sample.
Neutral desorption from methanol-dosed Al(111 > Figure 4.9 demonstrates mass spectra of the laser-ionized neutrals desorbed from a 50 L CH 3 0 D-dosed aluminum surface at various laser intensities. The electron beam energy employed is 3 keV, the surface is at 300 K, and ArF excimer radiation (193 nm) is used for ionization of the neutrals. Although the CHA provides valuable discrimination as an energy analyzer, mass resolution is low (m/Am = 12), largely because of uncompensated variation in flight time with angle of entry into the 180° hemispherical sector, as discussed in chapter II. The ground state of gas- phase methoxy appears to be only marginally stable [ 11] relative to rearrangement (CH 2OH) and/or fragmentation (H 2 + HCO). At high laser intensity (the 430 MW/cm 2 spectrum of figure 4.9), ionization and fragmentation of the parent ESD species give rise to an asymmetric peak in the mass 12 to 13 range, attributable mainly to C+, along with some unresolved CH+, and a broad peak centered at mass 29. Since the centroid of the mass peak shifts to mass 30 when CD 3OH is used instead of CH 3OH, this + x Signal (Norm.) 16 - iue48 KineticFigureenergydistributions 4.8: H+desorbedof from L55 CHgOH-dosedAl(111) room temperature.at 6 10 8 6 4 iei Eeg (eV) Energy Kinetic : y g r e n E n u G n o r t c e l E V e 0 0 0 3 V e 0 0 5 2 V e 0 0 0 2 >>4 U\ ESD Signal (rel. units) 5 0 5 0 5 0 5 40 35 30 25 20 15 10 5 0 Figure 4.9: Mass spectra oflaser-ionized/fragmented Massspectra Figureneutrals 4.9: "X 80 desorbed frommethanol-dosed Al(111). as (m/z) Mass 3 MW/cm _ 430 0 MW/cm 30 58 59
peak is nominally assigned to the HCO+ (DCO+) fragment. It is recognized, however, that this higher mass peak also likely contains mass 28 (CO+). The other two spectra of figure 4.9 are recorded at much, lower laser intensities. In these two spectra, the C+ signal has dropped significantly, and the major observable species in the lower mass region is the CH+ photoion. The centroid of the broad feature in the higher mass region seems to have shifted from m ass 29 to mass 31, indicative of the CH 3 0 + species. At these lower focal intensities, the parent ESD species, assumed to be methoxy, is apparently not efficiently photolyzed all the way to the C+ fragment. Similar mass spectra and power dependencies were obtained from methanol dosed onto a pre-oxidized Al(111) surface. Figure 4.10 shows the laser power dependencies of the peak-integrated signals from the C+ and HCO+ regions. Notice that at the lower laser intensities, the signal levels drop drastically. In the region of high intensity, however, C+ signal has reached a level exceeding that from HCO+ by more than an order of magnitude, and shows evidence of saturation, presumably at a photoion yield of one C+ ion for each methoxy parent. This regime, as discussed later, is useful in estimating total yields. These results showing extensive photodecomposition are consistent with multiphoton ionization studies of gas phase organic molecules such as methanol and benzene, which show efficient fragmentation to C+ at high laser intensities [12],[13]. There is also a gas phase study indicating high sensitivity for methanol detection using 193 nm excimer radiation and the resonance ionization emission spectroscopy (RIES) detection method [14]. Those experiments utilize the population of C atoms arriving in the metastable 1D2 and 1S 0 states after photolysis, which is evidently quite efficient at intensities comparable to those used in the present work. In the case of methanol, the 193 nm excitation is known to overlap a 1-photon absorption [15]; in the case of the methoxy radical, the absorption spectrum in this spectral region is not known, but a similar resonant enhancement is a plausible explanation for the ease of photodissociation observed in the present study. For velocity distribution measurements of the desorbing neutrals, the magnitude of the ArF laser intensity has been selected as a compromise Signal (arb. units) 10000 00 r-, □ ^ r 1000 0 - » ■ t i ■ i » ■ - I 10 1 e 0 0 1000 100 10 Figure 4.10: Laser focal Laser Figure ofintensity 4.10: the two dependences Slop© = = Slop© Rgo • • • • Region + C ----- methanol-dosed Al(111). mass regions observed in regions observed neutralESD mass from " ? ' o Focal Intensity (MW/cm2) Intensity Focal 1 —1 — ■1,1 ■ * r \ r .2 2 Slope = 0.98 = Slope HCO+ Region HCO+ 03' n □ ' 3 .0 _____ i j . i
i - 60 61
between photoionization efficiency and photofragmentation to the nominal species C+ and HCO+, and laser intensities of ~300 MW/cm 2 have been used. Because of low signal intensity in the velocity-resolved mode, data have been taken with a detection gate width of several mass units, sufficient to integrate the available signal. Figures 4.11 and 4.12 show velocity distributions from methanol dosed onto a clean and a pre-oxidized Al(111) surface, respectively, with an electron beam energy of 3 keV. Both surfaces give distributions which peak at -900 m/s, corresponding to a kinetic energy of 0.13 eV if the neutral desorbate is taken to be the methoxy species. If the HCO+ signals from the two surfaces are compared, the velocity distributions are indistinguishable, within experimental statistics. A better signal-to-noise ratio is achieved by monitoring the C+ region and, in that case, there is evidence for slightly more flux at high velocities from the clean surface, relative to that from the pre-oxidized case. The possibility that C+ could acquire additional kinetic energy in the photofragmentation process must be considered. Differences in photophysical behavior could arise if the parent desorbate left the surfaces in different states of excitation, for example. However, the similarity of the distributions obtained by monitoring either the C+ or HCO+ fragments from either surface is noteworthy. Apparently, little additional kinetic energy, relative to CO+/HCO+, is acquired by the C+ in the process of fragmentation. Such behavior might be expected on kinematic grounds, since, if repulsive forces are acting in the center-of-mass system, fragments acquire velocities inversely proportional to their masses. Thus, the major portion of any kinetic energy release would be carried off by the departing H atoms during the fragmentation to CO+and HCO+. These facts, as well as the low values of kinetic energy observed, are strong evidence that kinetic energy distributions characteristic of the parent CH 3O desorbate are obtained in these measurements. The velocity distributions of figures 4.1 1 and 4.12 have been fit to the same two types of functions which are used (and described) in chapter III: (1) the standard Boltzmann function with temperature as free parameter, and (2) a simple "planar-barrier/power-law model", in which a power-law distribution of the desorbing flux and a planar energy barrier to desorption are assumed. The distribution functions are defined in the limit of an electron pulse which is Figure 4.11: Velocity/kinetic energy distributions of photoionized fragments fragments photoionized of distributions energy Velocity/kinetic 4.11: Figure
ESD Signal (arb. units) 0 31). The kinetic energy scale is for the parent methoxy species (mass (mass species methoxy parent the for is scale energy kinetic The of convolution of the dotted curves with the electron pulse width. pulse electron the with curves dotted the of convolution of curves (•• ••): planar-barrier/power-law model; solid curves: result result curves: solid model; (••••): planar-barrier/power-law curves Dashed curves (— ): Boltzmann fit to the distribution; dotted dotted distribution; the fit to Boltzmann ): (— curves Dashed LCH 55 with dosed Al(111) clean from ESD neutral from 1 0 . 1 5 2 . 0 2 iei Eeg (eV) Energy Kinetic eoiy (km/s) Velocity 3 0 . 2 7 4 0 . 8 0 . 4 C Region HCO 5 Rgo - Region + C 6 3 OH. 62 8 63
Kinetic Energy (eV)
0.25 1.0 2.0 4.0 8.0 tiiv i
03 “izf "c
■ C Region _Q X 03
CO c CD 03 Q HCO Region 03 LLI
Velocity (km/s)
Figure 4.12: Velocity/kinetic energy distributions of photoionized fragments from neutral ESD from pre-oxidized A!(111) dosed with 55 L CH3OH. The legend for the caption of figure 4.11 applies to this figure also. 64
arbitrarily short (relative to the flight time from sample surface to photoionization volume); a convolution with the actual finite-duration electron pulse is then performed. A variable-width gate function is used for the electron beam: 1.6 |is duration for long flight-time, contracting to zero for the shortest TOF intervals. Flight times are taken from the center of the nominal electron-beam pulse, for proper display of model and convolved distributions on the same time base in figures 4.1 1 and 4.12. The 20 ns scale of the laser pulse is negligible relative to the transit times in this work. In all cases, the Boltzmann function is unable to fit the high-velocity tail of the experimental data, while the planar-barrier/power-law model (with n=2, as in chapter III) represents the data much better. The observation of very similar velocity distributions from methanol adsorbed onto clean and pre-oxidized Ai(111) surfaces supports the concept that room temperature exposure to methanol leads to oxidation of the aluminum and subsequent surface methoxy formation. As already mentioned, Tindall and Vickerman [4] have studied the adsorption of methanol on an aluminum foil at 273 K using XPS, SIMS (secondary-ion mass spectrometry), and thermal desorption spectroscopy and have concluded that methanol initially decomposes to oxidize the surface. Further methanol exposure leads to methoxy adsorption on the partially oxygenated surface. The similarities of the velocity distributions and electron beam-induced damage cross sections from the clean and pre-oxidized surfaces (see the discussion later in this chapter) suggest that the observations of reference [4] apply to the methanol- dosed Al(111) system. The peak of the neutral ESD velocity distributions, corresponding to a kinetic energy of about 0.13 eV, is low compared to the values obtained in ionic ESD where peak kinetic energies of 1-10 eV are typically observed [16],[17],[18]. The most probable energy for desorbed protons from chemisorbed CH 3O in the present study is in the 4-6 eV range, for example. A review of available studies on neutral desorption from chemisorbed systems reveals that kinetic energies quite generally lie in a much lower region, typically a few tenths of an electron volt. It also appears generally true that the observed neutral desorption yields are several orders of magnitude larger than the values for ionic channels, and are large enough to require 65
substantial dwell times in the excited electronic states which induce desorption. Chapter VII discusses mechanisms which might be operative to give the differences in the kinetic energy distributions between the ions and neutrals.
The ion and neutral desorbate yields For the neutral ESD channel, typical net C+ signal observed from the methanol-dosed clean and preoxidized AI(111) surfaces is about one count per laser shot, quite similar to that from the AI-6063 surface. In obtaining this number, the high laser intensity regime is employed in order to photolyze all of the parent desorbates to the carbon ion. As discussed in chapter III, one C+ count per laser shot corresponds to a desorbate yield of about 10'3 neutral/electron. For the ionic channel, the yield is, within an order of magnitude, also the same as for desorption from the technical sample. Thus, the H+ yield from the methanol-dosed Al(111) surfaces is about 10 ’6 ions per incident electron.
Surface temperature and electron beam effects The ESD dependence on temperature over the range 300-710 K of the nominal C+and HCO+photofragments is presented in figure 4.13 for methanol-dosed Al(111). Again, 193 nm ArF excimer radiation is employed, and at room temperature both the C+ and HCO+ mass regions are representative of the methoxy species. The HREELS studies [1] and the XPS work presented earlier in this chapter show that as the temperature increases, methoxy is thermally degraded to other carbon-containing species; thus the C+/CH+ signal may no longer be indicative of the methoxy desorbate alone at temperatures exceeding ~470 K. However, the CO+/HCO+ mass region 66
w C+ Mass Region *c 3 \ ^ - *...... m 0 3 L - 1-
CO C o CO - 5 - - ^ a co f Lii HCO+ Mass Region (x 4) 4------1------1------200 300 400 500 600 700 800 Temperature (K)
Figure 4.13: Temperature dependence of the neutral ESD signal as monitored by the two mass regions observed in laser ionization/fragmentation. For this experiment, the clean A!(111) sample has been dosed at ambient temperature with 25 L CH3OD. 67
signal should be representative of the surface concentration of this species over the temperature range studied if it is assumed that the methoxy does not thermally decompose to a CO- or HCO-containing molecule. The relatively small number of data points in figure 4.13 makes it difficult to ascertain the exact behavior of the parent mass region as a function of temperature. Generally, a surface species such as methoxy is thought to be stable on a surface until the decomposition temperature is reached. This sort of behavior implies that the slope of a line connecting the first two data points in figure 4.13 should be zero, and then, once the decomposition temperature (somewhere between 400 and 500 K for the methoxy/AI(111) system) is reached, the signal should drop off rapidly. It is not obvious that this is the behavior exhibited by the HCO+ mass region, with the ESD signal in this region appearing to decrease somewhat linearly with increasing temperature. The details of the thermal decomposition of methoxy on aluminum are not completely understood. One possible explanation for the experimentally observed behavior is that the CO+/HCO+ signal does not originate solely from desorbing methoxy. If over part of the temperature range investigated, species with higher or lower cross sections for desorption and/or ionization than methoxy give rise to signal in this mass region, then deviations from the expected behavior could occur. An alternative explanation is that the desorption of methoxy is site-dependent. The site-dependence of methoxy adsorption on Al(111) is not presently well understood, but if the occupation of certain ESD-active sites is temperature dependent, then the occupation by methoxy, at higher temperatures, of adsorption sites having lower ESD cross sections could lead to the experimentally observed behavior. In figure 4.13, the increase in C+/CH+ signal at about 470 K is consistent with our XPS results and the observations of Yates and co-workers [1] that in this temperature range the surface adsorbate composition is undergoing rapid changes as the surface methoxide is thermally decomposed. The increase in carbon signal indicates that a change in surface chemistry is occurring in which a surface species is produced which either has a higher cross section for desorption under electron impact than the surface methoxy species and/or has a higher cross section for C+/CH+ formation from interaction of the desorbate with the laser beam. The decrease in this signal between 500 and 68
700 K is likely due to the formation of a strongly bound surface carbide species which has a small ESD cross section. The XPS and HREELS studies cited above also reveal that in this temperature range, methoxy is completely degraded to aluminum carbide and aluminum oxide species. It should be emphasized that the data of figure 4.13 are difference spectra, i.e., any quasi steady-state gas phase signal is subtracted by the alternation procedure described in chapter II. Variations with temperature exhibited in this figure must be due to changes in the stimulated-desorption cross section, or the photofragmentation/photoionization properties of the desorbing species. In addition, the individual data channels (corresponding to electrons + laser, or laser only) show no evidence of thermal desorption peaks over the entire range covered. Under the experimental conditions employed, the density of molecules in the photoionization region due to thermal desorption from even a minority site should vastly exceed the corresponding quantity produced by ESD. Consequently, individual-channel data are equivalent to standard temperature-programmed desorption with a laser-ionization detector. Our results confirm other studies cited above that the CH3O is strongly chemisorbed on the surface and decomposes prior to appreciable thermal desorption. In order to examine the effect of electron beam damage on adsorbed methoxy, the electron gun is operated in the continuous mode, and the ESD signal is monitored as a function of electron beam exposure time. After a given exposure, the electron gun is switched to pulsed mode, and mass spectrometry of the ESD neutrals performed. Figure 4.14 shows a semilogarithmic plot of signal against electron beam dose (3 keV energy) for methanol dosed onto pre-oxidized Al(111) at room temperature. The decrease in the HCO+ mass region signal, S, from the surface may be modeled with first order kinetics:
d InS/dF = -a (4.1) where F is the electron fluence onto the surface (electrons/cm2), a being the cross section (cm2) for the process. With the assumption that the signal in this mass region is due solely to the surface methoxy species, the cross section Figure 4.14: The effect of the electron beam on the neutral ESD signal ESD neutral the on beam electron of the effect The 4.14: Figure In Intensity 10 0 2 4 8 6 C Ms Region Mass HCO coverage. from pre-oxidized AI(111) dosed with methanol to saturation saturation to methanol with AI(111)dosed pre-oxidized from Electron Fluence [1017cm2] Fluence Electron + as Region Mass C+ 69 70
derived from equation (4.1) is that for all disappearance channels, desorption plus destruction, of methoxy as a result of electron beam bombardment. The HCO+ mass region data from figure 4.14 give a cross section for surface methoxy loss of 2 ± 1 x 10 ‘17 cm2 from the pre-oxidized surface. Similar experiments and analysis for methanol dosed onto clean Al(111) give a cross section of 3 ± 1 x 10 ‘17 cm2. These two cross sections are equivalent within the uncertainty of our experiment. The calculated value of a is almost independent of the electron beam spatial profile, since we use the initial slope, corresponding to a low value of the damage parameter, i.e., F a« 1 . At much higher damage parameter values, the apparent slope decreases, as the most intense regions of electron bombardment become "burned out". Yates and co-workers [19] have performed electron beam damage studies of this system, using HREELS to monitor the C-H stretch and have concluded that the methoxy is decomposed to surface CHX species and eventually to surface carbide and surface oxide. This conclusion is fully supported by our observations. The C+ mass region of figure 4.14 is much more resilient to electron beam damage than the HCO+ region, indicating that the species which are formed from the electron beam decomposition of methoxy can also give C+/CH+ signal, albeit with a substantially reduced cross section for electron-induced removal. It is of interest to compare the cross section for electron beam-induced destruction/desorption of surface methoxy obtained in our work, by monitoring the ESD signal in the HCO+ region, with the corresponding cross section obtained by Yates and colleagues [19], where the loss of adsorbed methoxy was followed by monitoring an HREELS C-H stretch intensity. Using an electron beam energy of 300 eV, these authors obtained a cross section of 5 x 10-15 cm2. Here, we report a corresponding cross section of 3 x 10" 17cm2 using an electron beam energy of 3 keV. Because of limitations in the electron beam energy accessible with the electron gun used in our experiments, no direct comparison between the two cross sections is possible. However, the difference in the values is related to the more general issue of the relative contributions of primary and secondary electrons to the decomposition or desorption processes. How much these secondaries contribute to the loss of surface methoxy would depend on the threshold for 7 1
the desorption/decomposition processes, and at this time this value is unknown. The two orders of magnitude difference between the cross sections obtained in this work and in the work of Yates et al. is most likely due to differing primary eiectron-adsorbate interaction cross sections. In related gas-phase electron-molecule studies [ 20],[21], dissociation and electronic excitation cross sections peak for electron beam energies in the range of 100 eV and fall off quickly - one or more orders of magnitude - by the kilovolt region. It thus appears that the majority of the electron beam damage is due to the primary electrons, with an unknown degree of contribution from the secondaries. Electron beam damage experiments at a beam energy of ~1000 eV, described in chapter VI, indicate a cross section for methoxy dissappearance intermediate to the values obtained for beam energies of 300 and 3000 eV.
Conclusions from the CHA studies of methanol-dosed Alf1111 The work in this chapter represents the first attempt to use neutral electron-stimulated desorption to monitor changes in surface chemistry and demonstrates that it can successfully complement standard surface science techniques. It also points out the generalities of many of the findings from the AI-6063 work of the previous chapter. Namely, 1) the neutrals desorb with much less kinetic energy than the ions, 2) neutral yields are large compared to ionic ones, and 3) surface temperatures greater than 470 K are needed to remove methoxy from aluminum (recall that the AI-6063 sample was baked to only ~440 K). It is clear, however, that the data sets presented in this chapter suffer from the poor mass resolution inherent in the use of a CHA as an ion detector in time-of-flight experiments. The next two chapters discuss the extension of the neutral ESD experiments to a more advanced mass spectrometer and demonstrate results which add to the understanding of those of this chapter. 72
References
[1] J. G. Chen, P. Basu, L. Ng, and J. T. Yates Jr., "A comparative study of the reactivities of H2O, CH3OH, and CH 3OCH3 toward Al(111)," Surf. Sci., vol. 194, pp. 397-418, 1988.
[2] G. D. Waddill and L. L. Kesmodel, "Identification of methoxide species on Al(110) by high-resolution EELS," Surf. Sci., vol. 182, pp. L248- L252, 1987.
[3] J. W. Rogers Jr., R. L. Hance, and J. M. White, "An electron spectroscopic investigation of the interaction of methanol with polycrystalline aluminum," Surf. Sci., vol. 100, pp. 388-406, 1980.
[4] I. F. Tindall and J. C. Vickerman, "The adsorption and reactions of methanol on clean and oxidised aluminium surfaces," Surf. Sci., vol. 149, pp. 577-591, 1985.
[5] D. Al-Mawlawi and J. M. Saleh, "Interaction of alcohols with evaporated metal films," J. Chem. Soc. Faraday Trans, i,vol. 77, pp. 2965-2976, 1981.
[6] I. P. Batra and L. Kleinman, "Chemisorption of oxygen on aluminum surfaces," J. Electron. Spectrosc. Relat. Phenom., vol. 33, pp. 175- 241, 1984.
[7] U. Yxklinten, M. Limback, and B. I. Lundqvist, "Initial stages of oxidation of A!(111)," in Many-Atom Interactions in Solids, R. M. Nieminen, M. J. Puska and M. J. Manninen, Eds. Berlin: Springer- Verlag, 1990, pp. 315-318.
[8] C. J. Powell, "Attenuation lengths of low-energy electrons in solids," Surf. Sci., vol. 44, pp. 29-46, 1974.
[9] C. J. Tung, J. C. Ashley, R. D. Birkhoff, and R. H. Ritchie, "Electron slowing-down flux spectrum in AI 2O3," Phys. Rev. B, vol. 16, pp. 3049-3055, 1977.
[10] C. J. Tung, J. C. Ashley, and R. H. Ritchie, "Electron inelastic mean free paths and energy losses in solids II," Surf. Sci., vol. 81, pp. 427- 439, 1979. 73
[11] H. R. Wendt and H. E. Hunziker, "Electronic absorption spectrum of CHgO," J. Chem. Phys., vol. 71, pp. 5202-5205, 1979.
[12] R. L. Whetten, K.-J. Fu, R. S. Tapper, and E. R. Grant, "Highly efficient production of neutral carbon atoms in the ultraviolet multiphoton fragmentation of aromatic molecules," J. Phys. Chem., vol. 87, pp. 1484-1487, 1983.
[13] H. J. Neusser, "Multi-photon mass spectrometry and unimolecular ion decay," Int. J. Mass Spectrom. Ion Processes, vol. 79, pp. 141 - 181, 1987.
[14] R. C. Sausa, A. J. Alfano, and A. W. Miziolek, "Efficient ArF laser production and detection of carbon atoms from simple hydrocarbons," Appl. Optics, vol. 26, pp. 3588-3593, 1987.
[15] M. B. Robin and N. A. Kuebler, "Excited electronic states of the simple alcohols," J. Electron Spectrosc. Relat. Phenom., vol. 1, pp. 13-28, 1972.
[16] T. E. Madey and J. T. Yates Jr., "Electron-stimulated desorption as a tool for studies of chemisorption: a review," J. Vac. Sci. Technol., vol. 8, pp. 525-554, 1971.
[17] W. L. Clinton and R. E. Jutila, "Ion energy distributions from photon- and electron-stimulated desorption: reflection approximation," Phys. Rev. B, vol. 31, pp. 6441-6446, 1985.
[18] W. L. Clinton, S. Pal, and R. E. Jutila, "Ion energy distributions from photon- and electron-stimulated desorption. II. The quasiclassical final state and reneutralization," Phys. Rev. B, vol. 36, pp. 4123- 4134, 1987.
[19] P. Basu, J. G. Chen, L. Ng, M. L. Colaianni, and J. T. Yates Jr., "Fragmentation of molecular adsorbates by electron and ion bombardment: methoxy Chemistry on Al(111)," J. Chem. Phys., vol. 89, pp. 2406-2411, 1988.
[20] L. J. Kieffer, Report No. 6, JII.A Information Center, 1969.
[21] D. E. Donohue, J. A. Schiavone, and R. S. Freund, "Molecular dissociation by electron impact: optical emission from fragments of methane, ethylene, and methanol," J. Chem. Phys., vol. 67, pp. 769- 780, 1977. CHAPTER V
EXPERIMENTAL, PART II
Overview of the SARISA experiments The remainder of the ESD experiments described in this dissertation make use of both chambers of the SARISA (Surface Analysis by Resonance Ionization of Sputtered Atoms) machine. The entire apparatus is depicted in figure 5.1. The CHA is no longer operated in its ion detection mode, but is still used in the electron mode for surface characterization. A typical experimental run might consist of the following steps: 1) sputter/anneal cleaning of the Al(111) single crystal in the surface characterization chamber of the machine; 2) Auger or XPS analysis of the sample to verify surface cleanliness; 3) LEED analysis to verify that the surface is well-ordered; 4) dosing of the sample with the adsorbate of interest; 5) transfer of the Al(111) from the characterization side to the SARISA side via a transfer arm which can lock and unlock the sample holder from the sample manipulators and move it from one chamber to the other; and 6) performance of the actual SARISA experiment of interest. The first four steps listed above have been described in chapter II, while details of the SARISA apparatus and its operation follow.
Description of the operation of the SARISA apparatus As pointed out in the previous chapters, the concentric hemispherical energy analyzer, although it provides useful discrimination between ions and neutrals, suffers from relatively poor sensitivity and mass resolution as an ion detector in TOF experiments. A major part of this thesis research has been devoted to the construction and testing of an advanced surface analytical machine, acronymed SARISA IV. SARISA
74 75
was originally developed for performing trace analysis of surfaces by, as its title implies, the resonant ionization of neutral atoms sputtered from the surface of the material being studied. As in ESD, neutrals removed from the surface by sputtering dominate the ions by several orders of magnitude, and the sputtered neutral flux is representative of the surface composition. Resonant ionization has the advantage over nonresonant ionization in that it is immune to isobaric interferences and can differentiate between species of the same mass. The technique, however, is not limited to resonant laser ionization, and nonresonant ionization may be performed as well, and is often desirable for some experiments. SARISA IV is an improved version of earlier models of the instrument. My research efforts at Argonne have involved the building and testing of this instrument and the extension of it to ESD. While assembling and testing SARISA, I have performed the early ESD experiments in the surface analytical chamber of the machine, as described in previous chapters. Once it was clear that the new apparatus was working, a multistage electron gun was constructed and installed, and the desorption experiments were begun in the new machine. SARISA is an advanced mass spectrometer specifically designed to facilitate the high efficiency detection of laser-ionized neutrals departing from a surface. Although the instrument, as applied to the detection of sputtered neutrals, has been described in the literature [ 1],[2], a somewhat detailed discussion follows, with attention given to the differences relevant to the desorption experiments. The SARISA apparatus may be conveniently grouped into five subunits: the surface characterization chamber (detailed in chapter II), the SARISA electron gun, the primary ion source, the energy and angle refocusing time-of-flight (EARTOF) mass spectrometer, and the ionizing lasers. The general layout of the instrument is depicted in figure 5.1, while details of ion gun and the mass spectrometer are best demonstrated by figure 5.2. The SARISA electron gun is of our own design and is the source of the electrons for the ESD experiments. For sputtering studies, the primary ion beam is generated by a commercially available ion source (Colutron, Inc.). This ion gun is differentially pumped by two cryopumps. Lasers Out EARTOF Mass High Voltage Spectrometer Sample Einzel Lens Manipulators Sample Transfer Arm \
Primary ion Bending X-Ray Plates Source Electron Gun Lasers In SARISA Colutron Electron Ion Gun Gun
t i t 1 Surface Characterization Side SARISA Side
o ON Figure 5.1: The general layout of SARISA IV. — Target Laser High Voltage Einzel Lens Photoion Flight Einzel Lens Path Horizontal and Vertical Deflection Deflection Plates Plates Pulsed Deflection Primary Ion Beam Plates Bending Plates f1! Resistive Disk Energy ^ Analyzer
Photoion Ion Beam Flight MicroChannel Aperture Path Plates Colutron Ion Source Resistive Collector Disk Energy Analyzer
Figure 5.2: SARISA operated in the traditional sputtering mode with the ion gun. 78
Its output beam Is mass analyzed via a Wien filter and may be pulsed by sweeping across an aperture using pulsed deflection plates. Deflection plates downstream of the gun are used to guide the ions, and since the ion gun is m 2l in line of sight of the sample, bending plates are used to steer the primary ions towards the target. On their way to the target, after being turned by the bending plates, the primary ions pass through another set of deflection plates for steering and through an einzel lens for focusing. This steering and focusing allows the beam to reach the center of the target with a tight focus. For the sputtering studies performed for this thesis research, a 3.5 keV Ar+ beam has been employed, with 1-2 pA of ion current typically being delivered to the target in an approximately 150 pm spot. The electron gun, unlike the ion source, is in line of sight of the target, and typical electron current at 1000 eV beam energy is 1 pA in a spot size of about 1 mm. The details of the focusing and delivery of electrons to the target is discussed later in this chapter. The sample is held at a potential of 1100-1300V , with the value chosen depending on the distance of the laser from the surface. As the surface is bombarded by electrons or ions, both ions and neutrals may leave. Although the neutrals are generally many orders of magnitude more abundant than the ions, the ions are detectable without modification, while the neutrals require laser ionization, which may occur with less than unit probability. The desorbed or sputtered positive ions are created at the surface of the target and have the full 1100-1300 eV worth of kinetic energy, while the laser-ionized neutrals, formed some distance (0.5 - 4.0 mm) in front of the target, acquire much less kinetic energy. This difference in kinetic energies is adequate to discriminate the ions from the neutrals, since the potentials of the resistive disk energy analyzers of the EARTOF are chosen such as to allow ions of nominal energy 1000 eV to pass through the analyzer to the detector. The band-pass of the EARTOF is 200 eV, and, in actuality, ions of energies 900-1100 eV are allowed to pass. As mentioned previously, the value chosen for the target potential depends on the distance from the sample to the laser beam, with an appropriate choice allowing the laser-ionized neutrals to be formed with the nominal 1000 eV needed to pass through the energy analyzer of the 79
EARTOF. The ions leaving the surface have too much kinetic energy and do not reach the detector. The ionic desorption or sputter channel may be studied simply by not using the lasers and by lowering the target potential to around 1000 eV in order to give the desorbed or sputtered ions the correct pass energy. After laser ionization of the neutral atoms or molecules, the photoions are accelerated away from the target and into the EARTOF mass spectrometer by the large potential field of the target. On their way to the EARTOF, the laser-formed ions travel through the same einzel lens and deflection plates that the primary ions (in the case of sputtering) go through on their way to the target. This high voltage einzel lens now serves the purpose of focusing the photoions through the photoion flight path. On their way to the aperture of the first resistive disk analyzer, these ions pass through the primary ion bending plates and several lenses. The photoions then continue through the EARTOF mass spectrometer and are detected. For sputtering studies, the high voltage einzel lens mentioned above has the dual function of focusing the primary ion beam onto the sample with minimal aberration and extracting the photoions with high efficiency. The middle element of this einzel lens is held at a large negative potential (~-20,000 V) for sputtering in order to focus the primary ion beam effectively. For desorption studies in which no primary ions are used, it is unnecessary to run this middle element at such a large negative potential, with -5,000 V being adequate to extract the photoions from the laser volume. In fact, as discussed later, the electric field resulting from this lens being run at -20,000 V does not permit electrons from the SARISA electron gun to reach the target effectively. In addition, when this lens is run at voltages exceeding ~-15,000 V, electrons released by field emission may cause degradation of an adsorbate-covered surface. At a potential of -5,000 V, this emission is negligible. The resistive disk analyzers of the SARISA EARTOF are hemispherical energy analyzers designed to have large energy and angular acceptance windows. Details of one of the analyzers are given in figure 5.3. The main components are the inner ball, the outer hemisphere, 80
a) Screen Equipotential Lines Resistive Film on Ceramic Disk
Ceramic Backing Disk Ion Exit Ion Entrance Window Window
Figure 5.3: a) Schematic of the resistive disk analyzer device of the EARTOF, and b) examples of trajectories of ions with three very different energies. In b), E1 is allowed to pass, while E2 and E3 are not. 81
and the disk which supports the assembly. A major feature of this assembly is the use of a fine mesh screen as the outer hemisphere. This screen allows sources of noise, such as ions with the incorrect pass energy and stray electrons, to pass through the screen and out of the analyzer, never reaching the detector. Energy discrimination is due to the 1/R equipotential lines within the hemisphere's concentric ball and screen, with R being the distance from the common center of the analyzer electrodes. The supporting disk is fabricated out of ceramic onto which has been deposited a resistive film. The purpose of this resistive film is to preserve the 1/R potential lines at the entrance and exit windows of the analyzer, thereby eliminating edge effects [3]. The combination of two resistive disk analyzers, with a telescopic lens in between, is employed in order to provide refocusing of the photoions in time and space, and, hence, enhanced mass resolution. The angle refocusing provided by this combination is in direct contrast to the CHA detector employed in the early ESD experiments (see chapter II) in which poor mass resolution resulted from uncompensated variations in the angle of entry to the analyzer. Figure 5.4 demonstrates this refocusing of the photoions. The SARISA instrument achieves a mass resolution in excess of 200 and has an efficiency (neutrals detected/neutrals desorbed) of around twenty percent. Ions which pass through the combination of energy analyzers strike a chevron microchannel plate assembly. The resultant electron pulse is then transmitted outside the vacuum chamber through a shielded lead and detected by a gated pulse-counting system or by a transient digitizer system (Lecroy model 8828C). Another major difference between the CHA setup used in the experiments of the previous chapters and the SARISA apparatus is the range of desorption angles monitored. Only desorbates which leave the surface within one or two degrees of the surface normal are ionized and enter the zoom lens of the CHA. In contrast, the SARISA machine is designed to detect a wider spread of desorption angles, with species desorbing at angles as large as ~30 degrees from the surface normal being detectable. 82
Second Energy - Analyzer
Exit Window
Telescope Lens
Entrance Window
First Energy Analyzer
Figure 5.4: Representation of two photoion trajectories, denoted as "a" and "b", which have the sam e energy, but different angles of entry into the entrance window of the first analyzer. Note that the total flight paths of the two trajectories through the analyzers are identical due to the telescope lens between the disks. 83
Depending on the particular system under investigation, various lasers are used for the ionization of the neutral desorbed or sputtered flux. A majority of the results reported in this dissertation have been obtained through the use of nonresonant ionization. Nonresonant ionization has been chosen because it allows for the detection of all of the desorbed or sputtered species that can be ionized by the laser radiation. ArF excimer radiation is particularly convenient for this purpose because of its short UV wavelength (~193 nm, 6.4 eV) and, hence, its ability to ionize, with one or two photons, most neutral species. Other excimer gases which have been employed are KrF (~248 nm, 5.0 eV) and XeCI (~308 nm, 4.0 eV). For the cases in which resonant ionization has been performed, excimer-pumped dye lasers have been used, with the visible output radiation of the dye being used either straightup or being frequency doubled with a doubling crystal. For most of the experiments described in this thesis, lenses have been employed to bring the laser beams to a focus 1-4 mm in front of the SARISA target. If multiple lasers are used for a particular analysis, the laser beams are overlapped in front of the target both temporally and spatially.
A demonstration of the sensitivity of the SARISA apparatus As mentioned earlier, SARISA IV was originally designed for the detection of sputtered neutrals. Before performing the desorption experiments, in which the yield, defined as the number of atoms or molecules released from the surface per incident particle, is several orders of magnitude lower than from sputtering, the new apparatus has been tested through several sputtering experiments. A stringent test of the new machine's performance is provided by an iron-implanted silicon sample in which the Fe in the bulk Si is at an atomic concentration of 3 ± 3 p.p.b., as verified by deep layer transient spectroscopy (DLTS). Fe in Si is a technologically important case with implications for the semiconductor industry due to the fact that Fe impurity atoms form deep level traps in Si, 84
changing bulk electrical properties, even at concentrations approaching 1 p.p.t. Though techniques such as DLTS are able to detect trace contaminants at the p.p.b level, depth profiling of a sample with monolayer resolution cannot be performed by such techniques [4]. The determination of the concentration of 56Fe in a silicon matrix, by sputtering, is complicated by the isobaric interference from sputtered 28Si2 molecules. Resonance-enhanced multiphoton ionization provides a means of eliminating this isobaric interference. By employing the three step ionization scheme detailed in figure 5.5, the specificity to ionize Fe and not Si2 is achieved. Figure 5.6 shows the depth profile of Fe-implanted Si obtained using this method. The procedure involves sputtering with a continuous ion beam (e.g., the Colutron ion source operated without pulsing) to remove layers of the silicon wafer and taking Fe concentration analysis runs using a pulsed ion beam. The point labeled "surface" is an Fe determination made with less than one monolayer sample removal. The "bulk" value of 3 p.p.b. is obtained ~1 pm below the surface. The point designated "separate run" represents an Fe concentration measurement made on a different spot of the Si wafer. For this determination, more than a micron of the Si surface has been sputter removed. It is a check on the bulk Fe concentration obtained during the depth profile. Even for the measurement of the bulk concentration (the most difficult one of the profile because of the lower numbers of Fe atoms present in the matrix), only 2.5 nm of sample surface has been removed in obtaining this datum point. This kind of analysis demonstrates the ability of the SARISA apparatus to detect efficiently atoms and molecules sputtered or desorbed from a surface.
The design of the SARISA electron gun Some special considerations have gone into the design of the electron gun which has been installed in SARISA IV. The problem is complicated by the existence of strong electric fields in the target region 85
77777777777777?/, / Autoionization - Ionization ' resonance Potential ' 64279 cm'1 63480 ±500 cm 7.870 ± 0.06 eV 510.0 nm 308 nm 308 nm 44677 cm'
42912 cm
640.00 nm
29056 cm k 33096 cm
344.06 nm 302.06 nm 232.96 nm
3-Color Resonance Single Resonance Ionization Ionization j = o Scheme Schem es
0 cm"
Figure 5.5: Resonance ionization schemes which have been used for ionizing sputtered Fe. The 3-color scheme on the left eliminates nonresonant ionization of the isobarically interfering silicon dimers. Figure 5.6: 5.6: Figure
[Fe]/[Si] concentration (ppb at.) 1000 100 Siemens, inc. by Dr. H. Zieninger. H.Dr. by inc. Siemens, h t p e d e F Surface point profile of a Si wafer. The sample was prepared at at prepared was sample The wafer. Si ofa profile Interface i s y c i s et (nm) Depth atrd o depth to Rastered 100 Separate Separate fno fno profile) 1000 run
86 87
and by physical constraints imposed by the placement of the gun in an existing apparatus. As described previously, the target is held at a potential in excess of 1000 V, while the middle element of the high voltage einzel lens is normally at -20,000 V, and its other two elements are at potentials of 0-2000 V. These kinds of electric fields can effect the trajectories of electrons, especially low energy ones. Prior to building the electron gun, I performed electron trajectory computer simulations for the target region through the use of the particle optics program EGUN [5j. This particle optics program is designed to calculate charged particle trajectories in electrostatic and magnetostatic fields. The data input involves formulating the boundary conditions, which are determined by the locations, dimensions, and potentials of the electrodes in the region of interest, in terms of cylindrical coordinates, and designating these boundaries as either Dirichlet boundaries (the metal surfaces of the electrodes) or Neumann boundaries (the gaps between the surfaces). Data input also allows the user to start particles at various locations, directions, and energies within the boundary region. The output of the program consists of drawings of the equipotential lines within the region and of ray traces of the particles. Figure 5.7 a) shows the target region of SARISA IV, while b) depicts EGUN's output of the equipotential lines for conditions approximately the same as for the ESD experiments. Note in this figure that the density of the equipotential lines is chosen as an output parameter for clarity of presentation and has no physical significance. Also note that the middle element of the high voltage einzel lens (used only for extraction of the photoions in the ESD experiment) is run at about -5000 V, instead of at the -20,000 V potential required in sputtering experiments. Figure 5.8 a) demonstrates, for the same conditions as in the previous figure, the difficulty of getting low energy electrons to the target. In this figure, rays, representing electrons of energies 100 eV-1000 eV, have been started at the output of the cup in which the end of the gun sits. The rays have been started at an angle of 0° relative to the line which would connect the center of the cup and the center of the target. The 100 eV electron never arrives at the target; it instead runs into the plate that surrounds the High Vo/Iage Lens Electron Gun Cup
Deflection. Plates
r - Target
1000 V
1000 V Ground IfI f " > ’i
Ground
Ground -5000 V
S j S i AI / * f/ •11* W*' ,&v" F '
1000 V
Figure 5.7: a) The target region of SARISA IV. The circled portion has been simulated by the particle optics program EGUN. The equipotential lines calculated in the program are shown in b). 89
1 /1
y-\ji _ / IVIUHttimn. lY r 'ifffTlHi / I / * ,fe*ntnt / I /'wwriiJ / , "i nwn / (/ ./'"III'I M.' nil 1 I lint * / l & J I / • / /M'A r r c *, *.Vhi »•* > ->.v I.V’/< »> 1.1 "V' ' n i i V .I i n ’ 1
Figure 5.8: Raytraces in the target region, for the same conditions shown in figure 5.7, of: a) electrons of various energies, and b) photoions leaving the laser ionization volume. 90
target. The 200-1000 eV electrons are successful in reaching the target, but only the 700-1000 eV rays strike the center. In reality, angular spreads of the electrons leaving the gun would allow some fraction of all the electrons, even of energies 100 eV and less, to reach the target. However, as is clear from this diagram, a price is paid in terms of electron current to the target for using lower energy electrons. Figure 5.8 b) is the output of a simulation of photoions that are created in the laser ionization volume and are traveling away from the target and towards the EARTOF. All of these rays have been given the sam e initial kinetic energy of 0.1 eV, but have been started at various angles. Note the focusing of these rays by the high voltage lens. Because of the complications of the electric fields of SARISA, the relatively large distance that the end of the gun has to sit from the target (~2.90 inches), and the desire to be able to obtain high electron currents in small spot sizes, a multistage design for the electron gun has been chosen. Multistaging provides the versatility to focus with some of the stages and, especially important for very low energy electrons, not to be limited by space charge effects in front of the cathode. Electrons emitted from a hot filament have kinetic energies on the order of 0.1 eV. Mutual repulsion of these electrons (space charge) causes divergence of the beam and a loss of current. To avoid this effect, it is beneficial to accelerate the electrons to higher energies, where space charge is less of a factor, allow them to travel much of their distance through the gun at this higher energy, and then to decelerate them to the desired energy. A diagram of the SARISA electron gun is shown in figure 5.9. Depending on the energies of electrons desired at the target, various focusing and accelerating/decelerating conditions may be applied to the gun. Briefly, however, the electrons are produced by a tungsten filament and accelerated by the first few elements of the gun. The electrons then pass through a set of up/down deflection plates which are used for steering the electrons through the aperture and through the remaining stack of elements. This set of deflection plates is also used to pulse the beam. This pulsing is accomplished by applying a voltage to one pair of the plates suitable to cause the electrons no longer to pass through the X-Y Deflection P l a t e s f o r
Tungsten Filament Deflection Plates Steering Electrons Used for Pulsing to the Target t h e B e a m
■rtj i«i
\ _ n l
A p e r t u r e
Figure 5.9: The SARISA multistage electron gun.
VO 92
Target
Laser High Voltage Einzel Lens Pulsed Electron Gun Einzel Lens Horizontal and Vertical Deflection Plates
Resistive y Disk y Analyzer Photo-ion Flight Path Micro Channel Plates Einzel Lens —
Resistive Disk V Collector \ ' I p i r Energy Analyzer V III < L \ L W
Figure 5.10: SARISA IV with the electron gun installed. 93
aperture. Downstream, the electrons are decelerated and pass through the final set of deflection plates. This set of plates serves the purpose of centering the electron beam on the target. Figure 5.10 depicts SARISA IV with the new electron gun installed. The gun is installed approximately 70° from the center line of the photoion flight path. For the experiments reported in this thesis, full use has not been made of the potential of this electron gun. The SARISA ESD experiments reported here have been performed with ~1000 eV electrons in order to bridge the experimental gap between the 3000 eV electrons used in the early CHA experiments and future lower energy (<300 eV) studies. The versatility of the gun’s design should allow for further ESD experiments in SARISA to be performed with lower energy electrons.
References
[1] M. J. Pellin, C. E. Young, and D. M. Gruen, "Multiphoton ionization followed by time-of-flight m ass spectroscopy of sputtered neutrals," Scanning Microsc., vol. 2, pp. 1353-1364, 1988.
[2] C. E. Young, M. J. Pellin, W. F. Calaway, B. Jorgensen, E. L. Schweitzer, and D. M. Gruen, "Trace surface analysis via RIS/TOF mass spectrometry," Inst Phys. Conf. Ser., vol. 84, pp. 163-168, 1987.
[3] M. W. Siegel and M. J. Vasile, "New wide angle, high transmission energy analyzer for secondary ion mass spectrometry," Rev. Sci. instrum., vol. 52, pp. 1603-1615, 1981.
[4] N. M. Johnson, "Measurement of semiconductor-insulator interface state by constant-capacitance, deep-level transient spectroscopy," J. Vac. Sci. Technol., vol. 2 1 , pp. 303-314, 1982.
[5] W. B. Herrmannsfeldt, Report No. SLAC-226, UC-28 (A), Stanford Linear Accelerator Center, Stanford University, Stanford, CA, 1981. CHAPTER VI
SARISA ESD EXPERIMENTAL RESULTS
Introduction to the SARISA experimental results This chapter presents the results of ESD experiments performed in the SARISA instrument described in the previous chapter. The higher sensitivity of this new apparatus, as compared to the CHA ion detector, allows for the use of the lower electron beam currents associated with lower electron beam energies and for the use of lower laser focal intensities for laser ionization of the desorbed neutrals. In addition, its high mass resolution enables clear identification of the fragmentation patterns associated with the laser ionization process. In this chapter, the results of experiments using electrons of energies in the 1000-1100 eV range are described. Of interest are comparisons of the ESD results at these lower beam energies with those obtained at 3 keV (chapter IV). Of particular concern are whether there are differences in: 1) kinetic energy distributions of the desorbing neutrals, 2) electron beam decomposition cross sections, and 3) neutral desorption yields. In addition, the improvements in the ESD experimental apparatus have led to the observation of a new, and heretofore unexpected, desorption channel. Data are presented which show unequivocally that aluminum metal is being desorbed from the methoxy/AI(111) system. Detection of a metal desorbate has not previously been described in the literature.
94 95
Neutral ESD in SARISA using 1100 eV electron beam energy Figure 6.1 displays a mass spectrum of laser-ionized/fragmented neutrals desorbed from Al(111) dosed with CH 3OH to give a saturation coverage of methoxy. The techniques used to clean and dose this sample are identical to those described in chapter II, and the ArF excimer radiation fluence employed is 230 MW/cm2. The most obvious difference between previous spectra, acquired with the CHA, and this one is the much higher mass resolution, with the mass peaks in figure 6.1 being completely resolved. The major laser-formed ions/fragments present in this TOF spectrum are: C+ (m/z=12), CH+ (m/z=13), CH 2+ (m/z=14), CO+ (m/z=28), HCO+ (m/z=29), CH3 0 + (m/z=31), and AI+ (m/z=27). The identities of these masses have been verified via 13CH3 0 H-dosing of Al(111). The photofragmentation/ionization pattern obtained from ArF laser irradiation of the neutral ESD desorbates is highly dependent on the laser intensity employed. As demonstrated in figure 6.2, at a high laser fluence of ~420 MW/cm2, the major species observed are C+, CH+, and CO+. However, at ~10 MW/cm2, the AI+signal dominates the spectrum. The 10 MW/cm2 spectra shown in figure 6.2 is normalized to the signal level of the higher laser fluence spectrum for clarity of presentation. In actuality, the signal acquired at 10 MW/cm 2 is smaller by a factor of 100-1000 than that taken at the higher laser fluence, where a lower detector gain has been used. Figure 6.3 presents laser power studies for C+, CH+, CO+ and HCO+. The CH3 0 + signal is not adequate to allow a similar study of this species. This figure illustrates the following points. First, above ~100 MW/cm2, C+ is the most prevalent photofragment in the mass spectrum. At the highest intensity studied, ~420 MW/cm2, additional C+ signal is obtained at the expense of the CH+ and HCO+ species, with these latter species apparently being completely photolyzed to the C+ fragment. Secondly, at low laser powers, CO+ and CH+ dominate the laser-formed methoxy fragments. Recent studies [1], [ 2] have placed the ionization potential of methoxy at around 10.77 eV, and, hence, two ArF photons are required for ionization. Assuming the parent desorbate is methoxy, it is apparent from figure 6.3 that photofragmentation dominates the iue .: r-oie/rgetd S nurl fo methanol- from neutrals ESD ArF-ionized/fragmented 6.1: Figure ESD Signal (arb. units) 0 1 dosed Al(111). The laser intensity is 230 MW/cm2. 230 intensityis laser The Al(111). dosed 15 as (m/z) Mass 20 25 30
35 96 ESD Intensity (Norm.) 1 Figure 6.2: Laser ionization/fragmentation of ESD neutrals from neutrals ESD ionization/fragmentationof Laser 6.2: Figure methanol-dosed Al(111) at low and high laser intensities. laser highlowand at Al(111) methanol-dosed as (m/z) Mass 20 25 30 1 MW/cm 417 0 MW/cm 10 35 97 iue .: oe td fte htfamns rsn fo ArF from arising photofragments the of study Power 6.3: Figure ESD Signal (rel. units) 4 0 0 0 0 0 0 1 10000 A: the HCO+ fragment. HCO+ the A: fragment; □ : the CH+ fragment; X: the CO+ fragment; CO+ the X: CH+fragment; the : □ fragment; r J - 0 0 1 matfo mehnldsdA(1) Lgn: h C+ the : • Legend: Al(111). ethanol-dosed m from impact photoionization photoionization 10 4 ------J ------f o 1 ------te etas eobd y 10 V electron eV 1100 by desorbed neutrals the ae Itniy M/m ) (MW/cm Intensity Laser 1— —I 1— h H-l l - H l- - -H ..... I I 11 I I I 100 9 • ------1 ------—}_++ + _ } j— —I I ji j I I H I I— 1— A □ X
1000 98 i 99
photoionization of the parent species when ArF radiation is used. A unique feature of SARISA IV is the ability to perform both sputtering and electron impact desorption experiments in the same apparatus. Figure 6.4 displays the desorbed Al signal as a function of ArF laser fluence, along with data, obtained in a separate experiment, in which clean Al(111) is sputtered with 3.5 keV Ar+. Of interest here is whether or not the desorbed species is atomic aluminum or an aluminum-containing molecule. The major component of the sputtered flux from metal surfaces consists of monomers of the metal [3], If an aluminum-containing molecule is being desorbed in the ESD experiment, then its ionization might be expected to exhibit a different laser power dependence than the atomic aluminum ejected in sputtering. Although fewer data points have been obtained in the desorption study than in the sputtering investigation, it appears from figure 6.4 that the ESD and sputtered Al signals have different ArF laser intensity dependences, with the desorbed signal approaching saturation somewhat slower than the sputtered signal. This is indirect evidence that the species being desorbed in the ESD experiment is not atomic aluminum, but rather some molecule that contains aluminum and is efficiently photofragmented by the 193 nm ArF laser radiation. Possible candidates for the desorbed molecule include AIOCH 3, AIH, AIO, and AI 2O3. As of this writing, our attempts to observe these species, even at very low laser intensity, have been unsuccessful. Noteworthy from figure 6.4 is the fact that the sputtered and desorbed Al signals begin to saturate by 10 MW/cm2, a much lower laser power than that for which the hydrocarbon fragments (figure 6.3) are even detectable. This is consistent with aluminum having an ionization potential of 5.985 eV [4] and requiring only a single ArF photon for ionization. As discussed earlier, the desorption of metals from metallic surfaces is expected to be immeasurably small [5]. The apparent ease with which aluminum is detected in our experiment is facilitated by the somewhat fortuitous choice of ArF as the ionizing laser radiation. As shown in figure 6.5, aluminum has two transitions from the 2P ground state to an autoionizing level at 193.162 and 193.582 nm. These transitions are well Al Signal (rel. units) 1 1O'3 4 1
0 10 0 1 O'2 4 4 - *-1 ° ° 0 4
iue .: ae nest eedne fdsre and desorbed of dependence intensity Laser 6.4: Figure i 1i| i 1nj i irrj " "i| i i 111 i ~i ii"iin| r i" i iririj i"i 111nij i i i11 i in| i I I I I r Lsr nest (MW/cm2) Intensity Laser ArF 0.01 m ill ill m ptee Alsignals. sputtered
l-.l..l-LI.nil 0.1 1..
I I □ m ill ill m 1
□ 1 I I I sre Al esorbed D □ ptee Ai Sputtered • Legend: □ 10 mil t. t. mil
I 11 I □ • 100
100
in il I Hill 101
‘55725
2s ^ Ionization Limit ' 3s3p Z^V 48278 2 p o
co
32436 32435
25348
(£>
Energy [cm'1] 112.061 Figure 6.5: Atomic energy level diagram of aluminum showing the transitions relevant to the desorption experiments described in the text. 102 within the ~1 nm gain curve [ 6] of the ArF excimer radiation, and excitation to this autoionizing level via a single ArF photon is expected to be a very efficient mode of ionization. In order to determine the absolute desorption yield of the ESD aluminum species, the desorbed Al signal has been compared to the Al signal obtained when clean Al(111) is sputtered with 3.5 keV Ar+. Identical photoion extraction and detection conditions have been used for the two types of experiments, with the sputtering measurement being performed prior to dosing the sample with methanol. The sputtering yield of aluminum at 3.5 keV primary beam energy is 2.8 Al atoms per incident Ar+ [7]. Comparison of signal levels between the sputtering and desorption measurements and analysis in light of equations 3.3-3.5 give a desorption yield of 3.9 x 10 ‘6 aluminum atoms or aluminum-containing molecules per incident electron. Also, comparison of the desorbed Al signal to signal levels for high laser intensity photofragmentation of all carbon-containing species to C+ verifies the yield estimate of 1.0 x 10 ‘3 desorbed carbon-containing molecules per incident electron obtained independently in chapter IV. Detection of desorbed Al bv resonant laser ionization Also notable from the Grotrian diagram of figure 6.5 are the three transitions from the ground state to the 2D state. In order to prove unequivocally that the species observed at m/z=27 is aluminum, a counting gate has been placed around the mass 27 signal, 1000 eV electrons have been used to initiate desorption, and these three transitions have been scanned with the frequency-doubled output of a tunable visible dye laser. KrF excimer radiation (~248 nm) has been used to excite the aluminum from the 2D state to above its ionization potential. The laser configuration employed is demonstrated in figure 6.6, and the results of the dye laser scan are displayed in figure 6.7. The result is clear: the three aluminum transitions cause an increase in m/z=27 signal which is Questek Lumonics Hyperdye KDP Crystal 308 nm. Excimer Laser Laser running with Frequency XeCI (308 nm) Output Rhodamine 610 dye s Doubler Tunable Tunable light at light at ~616 nm ~308 nm Questek To Excimer Laser 248 nm SARISA KrF (248 nm) Output \- Tunable light at ~308 nm and 248 nm light Figure 6.6: The laser setup used for resonant ionization of desorbed aluminum. Figure 6.7: Resonant scan of aluminum desorbed from methanol-dosed methanol-dosed from desorbed aluminum of scan Resonant 6.7: Figure Al signal [counts] 100 150 200 250 616.40 oeta. h KFitniyepoe s 1 MW/cm2. ~14 is employed intensity KrF The potential. aluminum from the resonant level to above the ionization ionization the above to level resonant the from aluminum MW/cm2. KrF excimer radiation at ~248 nm is used to excite the the excite to used is nm ~248 at radiation KrFexcimer MW/cm2. of ~0.003 nm. The intensity of this doubled dye light is ~1.0 ~1.0 is light dye doubled this of intensity The nm. ~0.003 of frequency-doubled Rhodamine 610 dye light having a bandwidth bandwidth a having light dye 610 Rhodamine frequency-doubled Al(111). The resonant excitation step is performed with performed is step excitation resonant The Al(111). 308.2151 616.45 (y lsr [nml laser) (dye X 616.50 616.50 618.50 |309.271 618.55 [309.284 Nonresonant 618.60 Signal Resonant Signal 618.65 104 105 factors of three to four above the background signal from non resonant ionization by the KrF radiation. The fits through the data of figure 6.7 are to guide the eye and are obtained by fitting the experimental points to Gaussian distributions. The differing widths of the peaks are likely an artifact of the noisiness of the data and are not necessarily physically meaningful. This experiment provides indisputable evidence that aluminum is indeed being desorbed by electron impact from the methoxy/AI(111) system. It is worthwhile noting that thermal-induced desorption of aluminum due to electron beam heating can be discounted. The duty cycle of the electron beam is low; typical repetition rates are <50 Hz and the pulse width is only a few microseconds. Furthermore, the beam is not tightly focused onto the sample, having an estimated spot diameter of ~1 mm. Calculation shows that under these conditions the heating of the surface with the electron beam is negligible, with the maximum power deposited onto the surface being ~1.9 x 10 '5 W/cm2. In addition, no aluminum desorption is observed from clean Al(111). If heating of the surface were responsible, then thermal desorption of aluminum atoms might be expected from this surface as well as from the methoxy-covered one. Velocity/kinetic energy distributions of the desorbed neutrals In the same manner as performed with the CHA detector, velocity distribution measurements can also be carried out in SARISA. In SARISA, however, the individual m ass peaks are completely resolved, and velocity distributions can be measured for each one. Figure 6.8 shows velocity distributions for the C+, CH+, and CO+ photofragments believed to originate from the photofragmentation of desorbing neutral methoxy; the HCO+ signal-to-noise ratio is too poor to obtain a corresponding measurement for this species. ArF excimer radiation is used here for ionization, and the distance from the surface to the laser ionization volume is 2.25 mm. The data for these three photofragments have been collected 106 Kinetic Energy (eV) 0.1 0.2 0.5 1.0 2.0 3.0 T3 CD N ^ =^= 75 CO Signal E o c 32: CO CH+ Signal c 0 5 * Q CO C Signal LU ■Jr, l ■ I I I I I H I I I I »-I ■ » 1 f » I » * » «- I „ I,. I., 0 12 3 4 Velocity (km/s) Figure 6.8: Velocity distributions of laser-ionized/fragmented neutrals desorbed from methoxy-saturated Al(111). The ArF laser intensity is ~150 MW/cm2, and the electron beam energy used is 1100 eV. 107 simultaneously, and it is noteworthy that the actual noise in the data may be somewhat higher than statistical noise (represented by the error bars) due to fluctuations in the laser intensity. The planar-barrier fit described in chapter III is included, and as for the previous data sets for desorption from the AI-6063 and Al(111) samples, this model appears to fit the data quite well. The velocity distributions of figure 6.8 peak at ~ 1100 m/s, a value which corresponds to a kinetic energy of 0.19 eV if methoxy is assumed to be the parent species. Note that this is somewhat more energetic than the 0.13 eV (velocity of 900 m/s) result obtained in the CHA apparatus (figures 4.11 and 4.12). While this difference may be an experimental artifact (note that a small change in the velocity causes a larger change in the kinetic energy), there are some differences between the two experiments which are noteworthy. First, for the data presented figure 6.8, the desorbing electrons are of an energy of 1100 eV, instead of the 3000 eV energy used for the experiments in chapter IV. Secondly, as discussed in chapter V, the SARISA apparatus has been designed so as to maximize the ratio of detected to total number of desorbed neutrals, and it ionizes neutrals ejected over a much wider angular spread than does the CHA apparatus. The consequence of this is that the data obtained in SARISA are much more affected by any dependence of velocity distributions on ejection angles. Very little work has been performed to date on angular distributions of neutrals desorbed by photon or electron impact, and angular-resolved velocity distributions, because of the low signal levels involved, have not been investigated at all. Hence, how important the differences in collection angle are to the data presented in this thesis is unknown. There appears to be no laser power regime where the parent methoxy ESD signal is very large, and the apparently small signal of this species (see, for example, figure 6.1) is a point that should be addressed. It appears that the competition between photoionization and fragmentation is dominated by the latter effect. However, this thesis work cannot rule out the possibility that the parent desorbate is some species other than methoxy. From the literature and the XPS work of this thesis (see chapter IV) it appears that methoxy is overwhelmingly the dominant 108 surface species; however methoxy may not desorb, and the desorption of a methoxy fragment should be considered as a possibility. One possible scenario is that the impinging electrons cause simultaneous desorption and fragmentation. Hence, a fragment such as HCO or CO could be reaching the laser ionization volume. It is important to note, though, that these species have smaller masses than methoxy, and, since the energy is directly proportional to mass, this would mean only that the kinetic energies are smaller than those calculated assuming that methoxy is the parent desorbate. Figure 6.9 is a velocity distribution study of the desorbed Al species. As a test of the effect of different flight distances from the surface to the laser ionization volume on the velocity distribution, the experiment has been performed at flight distances of 2.16 and 4.43 mm. As can be seen from Figure 6.9 a), this effect is negligible within the experimental noise. Figure 6.9 b) displays the data points fitted to both the Boltzmann and the planar-barrier models. Interestingly, both the planar-barrier and Boltzmann models appear unable to fit the data of figure 6.9. Recall that the planar-barrier model fits the methoxy fragment data quite well. The differences in velocity distributions between the desorbed aluminum and the methoxy fragments indicate that there are two separate neutral ESD desorption channels. The first, and the more dominant one, consists of desorbed neutral methoxy. The second, which is down two to three orders of magnitude from the methoxy channel, is comprised of the aluminum- containing species. Thermal effects on the neutral ESD spectrum It seems appropriate here to emphasize that no aluminum or methoxy desorption is detected from clean Al(111); dosing of the clean surface with methanol is a prerequisite to observing these species. Figure 6.10 demonstrates the effect of heating the CH 3 0 /AI(111) system from room temperature to ~600 K, as monitored via neutral ESD. For this 109 E • 2.16 mm i — o a 4.43 mm 2 cd c 03 CO O CO Lit < 0 1 2 3 4 5 6 Velocity (km/s) b) O) 0 1 2 3 4 5 6 Velocity (km/s) Figure 6.9: a) Velocity distribution data of desorbed aluminum determined at 2.16 and 4.43 mm flight distance to the laser ionization volume, b) Boltzmann and planar-barrier fits to the combined data sets. ESD Signal (rel. Units) 1 1 2 2 3 35 30 25 20 15 10 5 iue61: eta EDfo te ehx/I11 ytm at system methoxy/AI(111) from the ESD Neutral Figure6.10: room temperature and at 600 K.room temperature at 600 and as (m/z) Mass 0 K : K 300 0 K 600 110 111 study, 1000 eV electrons are used in the desorption process, and ArF excimer radiation (~50 MW/cm2) is used to ionize the desorbed neutrals. From the experiments presented in chapter IV, the disappearance of photofragments associated exclusively with methoxy is expected at temperatures in excess of ~500 K. This is observed in figure 6.10, as the CO+ fragment signal becomes reduced in intensity at the elevated temperature. Somewhat surprisingly, the AI+ signal also disappears at 600 K. Thus, the adsorbed methoxy is somehow responsible for the aluminum desorption signal. This result may be an indication that the aluminum- containing desorbate is a species such as AI-OCH 3. The next chapter discusses other possible mechanisms of how methoxy on the surface could be responsible for the aluminum desorption. Also notable from figure 6.10 is the increase in the CH+/C+ ratio upon heating the sample from room temperature to 600 K. This further supports the contention, put forth in chapter IV, that an intermediate in the thermal decompostion of CH 3 0 /AI(111) to aluminum carbide and aluminum oxide is a CHX species which is present on the surface at elevated temperatures. Comparison of the results from 1 keV and 3 keV electron bombardment Chapter IV presents neutral ESD results from methanol-dosed Al(111) when this system is bombarded by 3 keV electrons and reports a cross section of 3 x 10 ’17 cm2 for loss, through either desorption or decomposition to other surface species, of methoxy from the surface. The experiments performed in SARISA, and reported in this chapter, are carried out with electrons of 1000-1100 eV of energy. Analysis, similar to that presented in chapter IV, gives a cross section for loss of methoxy from the surface at 1100 eV of 5.5 x 10 -17 cm2. This cross section is obtained by monitoring the CO+ fragment ESD signal as a function of electron beam exposure time. Note that this value is intermediate to that obtained at 3000 eV (this thesis work) and 300 eV (reference [ 8]). 112 In general, this chapter has verified the neutal ESD results obtained using the CHA apparatus and reported in chapter IV. Some differences, however, are apparent. First, the low mass resolution results of the CHA do not clearly show the presence of aluminum, although reference to figure 4.9 shows that there is, indeed, signal at m/z = 27. Certainly, the aluminum desorption channel would have been detected in the old setup had higher mass resolution and sensitivity been available. Secondly, the higher m ass resolution of SARISA does not reveal as much CH 3 0 + or HCO+ ESD signal as might have been expected from the low mass resolution results of the earlier chapter. Whether or not this is a result of the differences in desorption angles sampled by the two setups or of differences in electron beam energies, or just a consequence of the poor mass resolution of the CHA remains unclear. Both experiments, however, reveal the low velocity distributions which seem to be typical of all ground state neutrals studied to date. The next chapter investigates possible mechanisms which could account for the low kinetic energies observed in this dissertation and discusses the observation of aluminum desorption. References [1] B. Ruscic and J. Berkowitz, "Photoionization mass spectrometric studies of the isomeric transient species CD 2OH and CD 3O," to be published, 1991. [2] L. A. Curtiss, D. Kock, and J. A. Pople, "Energies of CH 2OH, CH3O, and related compounds," to be published, 1991. [3] H. Gnaser and W. O. Hofer, "The emission of neutral clusters in sputtering," Appl. Phys. A, vol. 48, pp. 261-271, 1989. [4] G. Herzberg,Atomic spectra and atomic structure. New York: Dover, 1944. 113 [5] D. Menzel, "Electronically stimulated desorption: mechanisms, applications, and implications," in Surface and Interface Characterization by Electron Optical Methods, A. Howie and U. Valdre, Eds. New York: Plenum Press, 1987, pp. 285-299. [6] J. C. White, J. Bokor, R. R. Freeman, and D. Henderson, "Tunable ArF excimer-laser source," Opt. Lett., vol. 6, pp. 293-294, 1981. [7] N. Matsunami, Y. Yamamura, Y. Itikawa, N. Itoh, Y. Kazumata, S. Miyagawa, K. Morita, R. Shimizu, and H. Tawara, Report No. IPPJ- AM-32, Institute of Plasma Phyics, Nagoya University, 1983. [8] P. Basu, J. G. Chen, L. Ng, M. L. Colaianni, and J. T. Yates Jr., "Fragmentation of molecular adsorbates by electron and ion bombardment: methoxy Chemistry on Al(111)," J. Chem. Phys., vol. 89, pp. 2406-2411, 1988. CHAPTER VII MECHANISTIC IMPLICATIONS Overview of the results that any proposed mechanism must explain The work presented in this thesis verifies the observations of other workers (see table 1.1) that ground state neutrals desorbed from surfaces by electron impact have kinetic energy distributions which are narrower and peaked at lower energies than those of desorbed ions. It also demonstrates that the neutrals are several orders of magnitude more abundant than the ions, a result which is also in agreement with existing studies. Any proposed mechanism must somehow be able to explain these experimental observations. In addition, this dissertation presents data which show that metals can be desorbed from an adsorbate-covered metai surface, and any operative desorption theory must be able also to account for this finding. The goals of this chapter are to discuss the theories which have been put forth in the literature to explain the results from neutral and ionic ESD experiments and to evaluate the relevance of these theories to the results of this thesis. The Menzel-Gomer-Redhead model of stimulated desorption In the mid 1960's, some twenty years before the direct detection of ESD neutrals was performed, Menzel and Gomer [1] and Redhead [2] independently proposed the first theoretical model (the MGR model) which was successful in explaining qualitatively the major observations of ESD and PSD experiments. Their theory is based on elementary kinematic 114 115 considerations which show that direct momentum transfer from electron or photon impact is insufficient to lead to desorption and that the desorption must be due to electronic excitation of the surface-adsorbate bond. Within the context of their model, electron or photon impact causes the metal- adsorbate (M - A) system to undergo a vertical electronic transition from the bonding state to an excited state which is repulsive in the region of Franck-Condon overlap. As the system evolves along the excited state curve, potential energy is converted into nuclear motion, and the adsorbate moves away from the surface. However, the large density of states of the metal can lead to deexcitation, and only if the system remains on the excited curve long enough to acquire adequate kinetic energy to overcome the metal-adsorbate bond energy can desorption occur. The nature of the desorbed species is determined by the asymptotic character of the excited curve (for example, excitation from the ground state to a repulsive curve which behaves asymptotically as M- + A+ can lead to the desorption of positive ions). Figure 7.1 depicts an energy level diagram typical of those used in discussions of the MGR mechanism. In this schematic, the initial excitation is to an antibonding curve, denoted as (M + A)*. If the kinetic energy, E, acquired by the adsorbate as it evolves on the excited state potential energy surface before being deexcited, is sufficient to overcome the barrier for desorption, then the adsorbate will desorb with a kinetic energy of E - E'. Hence, the observed kinetic energy distribution is representative of the details of both the ground and excited electronic states and of the nature of the deexcitation of the excited state. Note that if E is less than E', then the would-be desorbate is recaptured, and the acquired kinetic energy is absorbed by the substrate. In other words, there exists some critical distance along the excited state curve beyond which the system must evolve in order for desorption to take place. The MGR mechanism represents a good starting point for examining the dynamics of stimulated desorption. It must be mentioned, however, that it is a generalized model which is not specific about the details of the initial excitation, the nature of the ground and excited state curves, or the nature of the deexcitation processes. We will return to a discussion of Figure 7.1: Electronic excitation from the ground state metal-adsorbate metal-adsorbate state ground the from excitation Electronic 7.1: Figure Potential Energy the adsorbate to the surface. the to adsorbate the acquired on the upper state to overcome the binding energy of energy binding the overcome to state upper the on acquired a la odsrto i ufcetkntceeg, , is E, energy, kinetic ifsufficient desorption to lead can potential surface, M + A, to the repulsive surface, (M + A)*,+ (M surface, repulsive the A,to M+ surface, potential R M A) + (M M A' + A + M 116 117 these very relevant points after briefly examining some of the other desorption mechanisms which have been proposed. The.Knotek-Feibelman mechanism As discussed above, the MGR mechanism is not specific about the nature of the initial excitation. As originally formulated, however, it was applied in the context of valence excitations and was used successfully to explain results of ionic desorption from covalently bonded systems. Knotek and Feibelman [3] were the first to point out the relevance of core hole excitations as applied to desorption from surfaces, and they originally used this concept to explain the ejection of positive ions from maximal valency ionic compounds. The impetus for the construction of the Knotek- Feibelman (KF) mechanism was the observation of the desorption of positively charged species which are ionically bonded on the surface as negative ions. In the case of T 1O2, for example, the oxygen is bonded in a nominally O2- state. The oxygen, however, is observed to desorb as 0+ and to have a desorption threshold at the Ti(3p) core level ionization potential (34 eV). The three electron transfer process which is required in the transformation from 02- — 0+ is explained in the KF mechanism as being initiated by the ionization of the Ti(3p) core level. The formation of this core hole causes an interatomic Auger transition to occur in which a valence electron from the O2- is transferred to the Ti(3p) core hole. The excess energy is then released via the ejection of two electrons from the oxygen. This set of events just described has the result of transforming the O2- to 0+ The titanium, however, is still in the Ti +4 state, and the change of charge on the oxygen causes the oxygen to now be "bound" to the surface in a totally repulsive potential (there has effectively been a reversal of the Madelung potential) and to desorb as 0+ This type of ejection of the species from the surface has often been referred to as desorption via Coulombic repulsion. 118 It is clear that this mechanism is relevant to desorption from maximal valency ionic compounds, and since its original conception it has found applications to a variety of systems [4], including covalently bonded ones [5], [6], [7],[ 8]. In chapter III, this mechanism was suggested as possibly playing a role in the desorption of the H+ and F+ desorption observed in that chapter. In general, whether valence or core transitions initiate desorption appears to depend on the excitation energy and on the details of the particular system under investigation. The Antoniewicz model for desorption of phvsiorbed species Antoniewicz [9] has described a variation of the MGR mechanism which appears to be applicable to the desorption of physisorbed species. His so-calied "bounce" model is based on recognition of the fact that a physisorbed species is less closely bound to the surface than is the ionized physisorbate, which is strongly attracted to the surface via its image potential. In the Antoniewicz model, electron or photon impact excites the metal-adsorbate system to an ionic state (e.g., excitation from the (M + A) ground state to the (M- + A+) ionic state). Since the ionic state has its equilibrium distance closer to the metal substrate, the adsorbate accelerates towards the surface and gains kinetic energy. As it gets closer to the surface, the ion experiences an increasing probability of reneutralization due to the high density of states, and, at some distance above the surface, is reneutralized, without losing the kinetic energy it acquired as an ion. On the ground state potential curve, the neutral particle continues running towards the surface until it reaches its classical turning point where it has lost all of its kinetic energy. There it begins receding ("bouncing") from the surface, and, if sufficient kinetic energy has been acquired in the accelaration towards the surface to overcome the van der Waals attraction of the ground state curve, then the neutral species desorbs. The sequence of events which might take place is depicted in figure 7.2. This model has been invoked, with differing degrees 119 t >» 03 a> c LU aJ 4- R M A Figure 7.2: For physisorbed species, initial electronic excitation to an ionic state, M‘ + A+, which is bound closer to the surface than the ground state, leads to nuclear motion towards the surface. Deexcitation back to the ground state results in desorption only if sufficient kinetic energy has been acquired by the adsorbate on the ionic potential curve. 120 of success, to explain the results of the stimulated desorption studies of the physisorbed systems presented in table 1.1. It is worth noting that the Antoniewicz model is likely not appropriate for the methoxy/AI(111) system. The TPD and HREELS work of Yates and colleagues [10], as well as the XPS studies of the thermal decomposition of the surface methoxy species presented in this thesis, provide strong evidence that the methoxy is stable on the surface at temperatures in excess of 400 K. In addition, a recent theoretical study by Curtiss et al. [11] has. placed the energy of the AI-0 bond in the AI-OCH 3 isolated molecule at 5.44 eV. Though this energy might be somewhat lower for methoxy adsorbed on the aluminum surface, it is quite clear that the systems being studied in this dissertation are chemisorbed ones. Desorption induced bv a change in molecular orientation Burns, Stechel, and Jennison [12],[13],[14] have used time-of-flight laser resonance ionization to study state-selectively the translational energies of neutral NO and neutral and metastable CO desorbed from Pt(111) by electron impact, and have formulated a mechanism for NO desorption based on "desorption by electronically stimulated adsorbate rotation". According to their mechanism, if a molecule such as NO undergoes excitation from a hindered rotor ground state to a free rotor excited state, then upon deexcitation it is allowed to access regions of the ground state potential which are only weakly bound. An obvious requirement of this mechanism is that there be a change in the rotational geometry in going from the ground to the excited state. NO is chemisorbed and terminally bound through the nitrogen atom to Pt(111). Upon excitation of an electron from the 5a to the 2k orbital, the NO molecule electronically resembles O 2, which is bound to Pt(111) side-on by only 0.3 eV. The NO, therefore, undergoes a change in geometry from being rotationally hindered to being a relatively free rotor. CO does not desorb by this mechanism because it does not undergo such a change in 121 rotational geometry; electronic excitation of an electron from the 5a orbital in this case causes CO to resemble NO, which is also rotationally hindered. The applicability of this mechanism to the methoxy/AI(11 1) system is uncertain, as it depends upon the details of the electronic excitation of the adsorbed methoxy and on whether or not methoxy undergoes a change in rotational geometry upon electronic excitation. This mechanism, however, appears to be specific to certain molecules and can not be considered a general mechanism. The nature of nonradiative decay processes of the excited state As pointed out in the introductory chapter of this thesis, the cross sections for desorption from metals and semiconductors are several orders of magnitude lower than the analogous gas phase dissociation processes. This is a result of the ability of the substrate to reduce substantially the excited state lifetime. The possible routes by which an excited molecule near a surface may decay include: 1) radiative decay (fluorescence or phosphorescence), 2 ) bond breaking via transformation of potential energy into nuclear motion (this includes chemical transformation as well as desorption), and 3) nonradiative deexcitation through energy or charge transfer to the substrate. Because of the relevance of the lifetime of the excited state to the desorption problem, it is useful to review what is known about the decay of excited molecules near surfaces. In a now classic series of experiments, Drexhage, Kuhn, and coworkers [15],[16] studied the fluorescence decay time of an europium dibenzoylmethane complex as a function of its distance from an evaporated metal surface and found that at very large distances from the surface, the decay rate of the excitation exhibits its free space value. As the molecular distance from the surface is decreased, becoming on the order of the molecular emission wavelength, the excitation lifetime oscillates about the free space value as a function of distance. The oscillatory behavior of the excited state lifetime is 122 understood in terms of an interference (field coupling) model [17] in which the radiating molecule interacts with its own radiation field which is partially reflected by the metal surface. At smaller distances (~ < 1 00 A), however, the lifetime of the excited state drops precipitously as nonradiative decay channels open and energy is transferred to the metal. In this "near field" region of the metal, two regimes of excited state quenching may be delineated. The first is that where the field coupling described above dominates the deexcitation. In this regime, the dipole field of the excited molecule interacts through space with the metal, and the energy of the molecule is transferred to the solid via excitations of plasmons, phonons, and electrons. Depending on whether volume or surface contributions of the metal dominate, the quenching of the excited state is found to have a cubic or quartic dependence, respectively, on the distance from the surface [17],[18]. The second regime is dominated by charge transfer between the metal and adsorbate. In this case, the overlap between the adsorbate and substrate wave functions allows for a more efficient route of deexcitation as charge, rather than energy, as in the case of field coupling, is transferred. Since the metal wave functions fall off approximately exponentially with increasing distance from the surface [19], the quenching of the excited state via charge transfer also has this distance dependence. To understand this process of charge transfer [18], consider the simple case of a molecule with two nondegenerate electronic energy levels, denoted as |1> and |2). Assume that level |1> is doubly occupied in the ground state and that level |2> is empty and is located above the Fermi level. Next consider excitation of one of the electrons from level |1) to level |2>. Due to the resulting positive hole in level |1), level |2) will be pulled down in energy and may end up either above or below the Fermi level. Figure 7.3 a) depicts the events that follow if level | 2> ends up above the Fermi level. The coupling of the adsorbate electronic state |2> to the metal orbital, |k), in the unoccupied part of the metal conduction band allows for the excited state to decay via resonant tunneling into this level. The resulting adsorbate ion may then relax via Auger neutralization, with a metal electron filling level |1) and an Auger electron a) b ) Ik) Fermi Level Fermi Level |2> |1> Metal Adsorbate Metal Adsorbate Figure 7.3: a) charge tranfer proceeding via resonant tunneling; b) charge transfer via Auger deexcitation. The circled numbers represent the steps explained in the text. 124 from the metal being ejected. What happens if the excited state |2> ends up below the Fermi level? In this case, resonant transfer from |2) to |k) is forbidden by the Pauli principle. The excited state may then decay either by the field coupling described earlier or by nonresonant charge transfer, such as Auger deexcitation, in which an electron from the conduction band occupies the hole in |1> and the electron in |2> is ejected. This process of Auger deexcitation is depicted in figure 7.3 b). It is important to emphasize that field coupling does not require orbital overlap between the adsorbate and substrate and that field coupling is the only mode of deexcitation for adsorbates in physisorbed multilayers or for adsorbates separated from the metal surface by spacer layers that do not have delocalized wave functions. The longest-lived excitations are those which do not decay by resonant electron transfer [20],[21]. Typical lifetimes for states which deexcite via resonant electron transfer are on the order of 10 ' 15 s. Excited states which do not have this decay path open to them may have lifetimes which are longer by factors of 10-100. Since nuclear motion occurs on the timescale of ~ 10’13 s, it is clear that, depending on the deexcitation channels available, large variations in desorption yields are expected. Comparison of ionic and neutral desorption within the context of MGR Menzel, Feulner, and colleagues [22], [23] have explained the low mean value of their kinetic energy distribution measurements of desorbed neutrals as being due to reneutralization of desorbing ions. The basic idea behind this mechanism is shown in figure 7.4. In order to illustrate better the resulting difference in kinetic energies between the ions and neutrals, quenching of the excited state is represented not by transition arrows to the ground state, as in figures 7.1 and 7.2, but by curve crossing to an excited copy of the ground state potential energy curve. Direct excitation to the excited state curve labeled M’ + A+ and continued evolution on this Figure 7.4: As described in the text, reneutralization of desorbing ions is ions desorbing of reneutralization text, the in described As 7.4: Figure Potential Energy t iei eeg hn ions. than energy kinetic expected to yield neutrals which desorb with substantially less less substantially with desorb which neutrals yield to expected M A) + (M R A -> MA + A + M " A' + M" E I E o 125 126 curve will lead to the desorption of positive ions, denoted as A+. The observed ion kinetic energy distribution will have a peak kinetic energy of E+. However, reneutralization, which corresponds roughly to resonant tunneling from the metal bands into the positive hole of the ion, will result either in desorption of the neutral adsorbate or in recapture of this species, depending upon where the reneutralization occurs along the M- + A+ excited state ionic curve. Once reneutralization occurs, the adsorbate will find itself in an excited copy of the M + A curve and will have to climb up the wall of this ground state curve in order for desorption to occur. If reneutralization along the M- + A+ curve occurs too soon (before the critical distance has been reached), the adsorbate will lose all of its kinetic energy in the ground state potential energy well and there will be no desorption. As discussed in the previous section, since the metal wave functions decline approximately exponentially with increasing distance from the surface, the probability of reneutralization increases exponentially with decreasing distance to the surface (i.e., with decreasing M-A distance). As a result of this dependence, most escaping neutrals will arise from reneutralization close to the critical distance. Since reneutralization at the critical distance yields neutrals with zero kinetic energy, it follows that the desorbed neutrals will have low kinetic energies and that their energy distributions will be narrow due to the exponential falloff in the probability of ionic reneutralization. As illustrated in figure 7.4, curve crossing from the initial ionic state to an excited copy of the ground state curve, labeled M** + A, will give a kinetic energy E0. Within the context of this reneutralization mechanism, it is easy to understand qualitatively the low yields of ionic, relative to neutral desorption. In order for an ion to desorb, the system must stay on the ionic curve for the whole duration of nuclear motion. Given the small lifetimes of electronic excitations, it is clear that ionic desorption is a low probability event. Feulner, Menzel, and coworkers have strong evidence that this type of reneutralization mechanism is operative in the chemisorbed systems they have studied (see table 1.1). Their strongest evidence 127 comes from studies in which desorption thresholds have been monitored as a function of electron beam energy [24]. Large enhancements in neutral desorbate yield at the threshold energy of ion desorption are taken as evidence that this mechanism is prevalent. For example, neutral CO desorption from Ru(001) is seen to increase significantly at the same electron beam energy as the onset of CO+ desorption is detected, indicating that reneutralization of CO+ is the major ESD mechanism for neutral CO formation. As pointed out in chapter IV, no ions of methoxy (or any ions at any of the photofragment masses observed in the neutral ESD studies) are detected in this thesis work. If this type of reneutralization mechanism is operative in the systems studied in this dissertation, then there must be complete reneutralization of the methoxy ions, at least at the level of our ability to detect them. The lack of parent ions in this methoxy/aluminum work, unlike in the studies of Menzel's group, where ions as well as neutrals of the same desorbate are detectable, indicates that reneutralization may not be the major desorption channel for the methoxy/aluminum systems studied in this dissertation. It is not necessary to invoke reneutralization of desorbing ions in order to explain the differences in kinetic energies and yields between ESD ions and neutrals. The same exponential dependence on distance from the surface that leads to desorption via reneutralization near the critical distance along an ionic excited state will also cause deexcitation to the ground state near the critical distance along a repulsive neutral excited state. Consider figure 7.5. The further the system evolves along the repulsive excited state curve, the more kinetic energy the desorbate acquires. Due to the ability of the surface to quench efficiently the electronic excitation, however, it is improbable that desorption of a neutral will result from complete evolution along the excited state repulsive curve. This event, approximately speaking, might be as probable as is the desorption of ions in ESD (several orders of magnitude down from that of neutrals). It is much more likely that neutrals will result from deexcitation off of this potential curve. If this deexcitation occurs before the critical distance, labeled Rc, then the would-be desorbate is recaptured. 128 * ^ ca p tu re >* 0 3 i _ 0) c LL! CO c 0 o Q_ R M A Figure 7.5: Efficient deexcitation channels on a surface make it improbable that desorption will occur via evolution along the entire initial excited state curve. Most neutrals will desorb following deexcitation to the ground state potential curve near the critical distance, Rc. 129 Deexcitation near, but after, the critical distance, leads to desorption with low kinetic energies, as most of the kinetic energy acquired on the repulsive curve is reconverted to potential energy as the system evolves in the ground state well after being deexcited. In figure 7.5, for example, the desorbate has acquired E' worth of kinetic energy by the time it crosses over (is deexcited) to the curve labeled (M* + A). However after having evolved along the ground state curve, it desorbs with kinetic energy equal only to E. Clearly, deexcitation events closer to Rc , which are more probable, will lead to even less energetic desorption. In order to know whether this concept of deexcitation near the critical distance will explain the observed experimental result(i.e., can it explain both the energy distribution and yield of the ESD neutrals?), detailed modeling of the excitation/deexcitation events is required. This kind of theoretical work is quite complicated for a molecule, especially one as large as methoxy, chemisorbed on a surface. However, the relative simplicity of physisorbed systems (as compared to chemisorbed ones) allows for detailed comparisons between theory and experiment. These kinds of theoretical studies are important in assessing the accuracy and appropriateness, in general, of MGR-type models. Recently, Gortel et al. [25],[26] have modeled the neutral ESD N 2O/Ru( 001) experimental results (see table 1.1) with the Antoniewicz model. Using realistic potential energy curves, these workers have shown that while it is possible theoretically to reproduce the kinetic energy distribution data (peak energy at 0.112 eV), it is impossible to reconcile this low peak kinetic energy with the high experimental desorption yield ( 0.1 molecules/incident electron) without invoking seemingly unphysical neutralization parameters. Similarly, Hiibner and Bennemann [27], modeling the desorption from the Ar and Kr/Ru(001) study (again, see table 1.1), have been unable to fit simultaneously the energy distributions and yields with expected reneutralization rates. The point of all this is that while reneutralization near the critical distance appears to be the operative mechanism, there are still many unanswered questions concerning the nature of electronic deexcitations. 130 The shape of the observed velocity distributions In chapters III, IV, and VII, it was pointed out that attempts to fit the parent neutral velocity/energy distribution data to a Boltzmann distribution failed, while fits to the so-called "planar-barrier/power law" model (with n=2, see equation 3.2) appeared to reproduce the data much better. These two functional forms represent extreme cases of possible models to the data, and caution should be exercised in drawing definite conclusions about their physical significance when applied to the case of stimulated desorption. The planar-barrier/power law model is a functional form that arises from sputtering theory [28], [29], in which it is assumed that there is a one-dimensional barrier, the binding energy, to ejection of the atoms from the surface. In this theory, momentum transfer from the impinging ion causes a collision cascade within the solid that results in the release of the sputtered surface atoms. As mentioned previously, momentum transfer from slow electrons, such as those used in this thesis work, is negligible. Within the framework of current ESD theory, then, there is no reason to expect to obtain a distribution similar to that obtained from sputtering. It is most likely that the observed distribution arises fortuitously as a result of the details of the electronic excitation/deexcitation process. In stimulated desorption, the shape of the velocity/kinetic energy distributions of the desorbates are representative of the details of the ground state and excited state curves. This can be best visualized within a simple reflection approximation [30],[31] in which the velocity/energy distribution is a reflection of the square of the ground state vibrational wave function. The repulsive portion of the excited state serves as the "mirror" which reflects this probability distribution. Clearly, any curvature in the excited repulsive state potential will cause the observed distribution to appear skewed. The observation of a distribution which is totally symmetric, such as the Boltzmann distribution, would mean that: 1) there is negligible anharmonicity in the ground state, and 2 ) the portion of the 131 excited state curve within the region of Franck-Condon overlap is linear. Neither of these conditions is generally fulfilled, and the experimentally measured velocity/energy distributions are not expected to be symmetric. Detailed knowledge of the identities and shapes of the initial and final state potential energy curves for the methoxy/aluminum system would be necessary in order to explain the observed distributions. Intra-adsorbate bond cleavage This chapter has discussed ESD in terms of breaking of the surface- adsorbate bond. Desorption can also occur via rupture of intra-adsorbate bonds, and in some cases, due to the lessened influence of the surface, this desorption is more efficient than that occurring directly from the metal. Consider the H+ desorption channel detected from the methoxy/AI(111) system. As isotopic substitution studies have verifed, these protons originate from the methyl group of the methoxy species. This implies that rupture of the C-H bond is occurring on the surface due to electron impact. The fact that protons are observed instead of methoxy ions is due, in part, to the extra distance from the surface provided by the oxygen and carbon atoms of methoxy. Since the process of charge transfer deexcitation drops off quickly with distance from the metal, the initial excitations of the C-H bond of methoxy are likely to stay localized longer than excitations that would occur in the Al-C bond of the surface methoxy species. Mechanistic implications of the observation of aluminum desorption The detection of metal desorption from an adsorbate-covered surface, reported in chapter VI, represents the first observation of its kind. Because of the expected efficiency of excited state quenching within the valence bands of the metal, this discovery of aluminum desorption from the 132 methoxy/AI(111) system was quite unexpected. From the discussions of mechanisms in this chapter, the low kinetic energy observed can be rationalized in terms of deexcitation near the critical distance. The real mystery, however, is how the excitation can stay localized long enough for the system to evolve even as far as the critical distance of the excited state curve. From the yield measurements, the desorption of the neutral aluminum atom or aluminum-containing species is several orders of magnitude less probable than is the desorption of the methoxy species. It is, however, as probable as typical ionic desorption channels. Several possibilities exist within the context of current desorption theories. First, as described in chapter IV, it is believed that some degree of oxygen underlayer formation occurs during the dissociative adsorption of methoxy onto the aluminum surface. This underlayer of oxygen may be effective in allowing electronic excitations of the surface layer to be somewhat more long-lived than if they occurred in the bulk. Detailed band-structure calculations are needed for the case of lightly oxidized Al(111) in order to be able to evaluate this possibility. Secondly, it is possible that the aluminum-containing species (it would be very useful if its identity can be obtained) is desorbing from a defect site on the single crystal surface. As Menzel has pointed out [32], surface imperfections, causing changes in the electronic structure of the surface, can result in localized electronic states. Defects on the surface may make desorption of a species such as AI-OCH 3 possible. The last chapter of this thesis includes a brief discussion of future experiments which might aid in obtaining a better understanding of the operative aluminum desorption mechanism. References [1] D. Menzel and R. Gomer, "Desorption from metal surfaces by low- energy electrons," J. Chem. Phys., vol. 41, pp. 3311-3328, 1964. 133 [2] P. A. Redhead, "Interaction of slow electrons with chemisorbed oxygen," Can. J. Phys., vol. 42, pp. 886-905, 1964. [3] M. L. Knotek and P. J. Feibelman, "Ion desorption by core-hole auger decay," Phys. Rev. Lett., vol. 40, pp. 964-967, 1978. [4] M. L. Knotek, "Stimulated desorption," Rep. Prog. Phys., vol. 47, pp. 1499-1561, 1984. [5] R. Franchy and D. Menzel, "Adsorbate core ionization as primary process in electron- and photon-stimulated desorption from metal surfaces," Phys. Rev. Lett., vol. 43, pp. 865-867, 1979. [6] M. Q. Ding, E. M. Williams, J. P. Adrados, and J. L. de Segovia, "Energy distribution of H+ ions with ESD of water adsorbed at aluminum and tungsten surfaces," Surf. Sci., vol. 140, pp. L264-L268, 1984. [7] D. E. Ramaker, C. T. White, and J. S. Murday, "On Auger induced decomposition/desorption of covalent and ionic systems," Phys. Lett., vol. 89A, pp. 211-214, 1982. [8] D. E. Ramaker, "Models for desorption in covalent systems," in Desorption Induced by Electronic Transitions (DIET I), N. H. Tolk, M. M. Traum, J. C. Tully and T. E. Madey, Eds. Berlin: Springer-Verlag, 1983, pp. 70-89. [9] P. A. Antoniewicz, "Model for electron- and photon-stimulated desorption," Phys. Rev. B, vol. 21, pp. 3811-3815, 1980. [10] J. G. Chen, P. Basu, L. Ng, and J. T. Yates Jr., "A comparative study of the reactivities of H2O, CH3OH, and CH 3OCH3 toward Al(111)," Surf. Sci., vol. 194, pp. 397-418, 1988. [11] L. A. Curtiss and D. Kock, private communication. [12] A. R. Burns, E. B. Stechel, and D. R. Jennison, "Desorption by electronically stimulated adsorbate rotation," Phys. Rev. Lett., vol. 58, pp. 250-253, 1987. [13] A. R. Burns, E. B. Stechel, and D. R. Jennison, "Rovibrational laser spectroscopy of ESD neutrals from chemisorbed species," in Desorption Induced by Electronic Transitions (DIET III), R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988, pp. 67-72. 134 [14] E. B. Stechel, D. R. Jennison, and A. R. Burns, "Dynamics in neutral DIET from chemisorbed molecules," in Desorption Induced by Electronic Transitions (DIET III), R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988, pp. 136-143. [15] K. H. Drexhage, M. Fleck, H. Kuhn, F. P. Schaefer, and W. Sperling, "The influence of a mirror on the fluorescence of a europium chelate," Ber. Bunsenges. Phys. Chem., vol. 70, p. 1179,1966. [16] K. H. Drexhage, H. Kuhn, and F. P. Schaefer, "Variation of the fluorescence decay time of a molecule in front of a mirror," Ber. Bunsenges. Phys. Chem., vol. 72, p. 329,1968. [17] R. R. Chance, A. Prock, and R. Silbey, "Comments of the classical theory of energy transfer," J. Chem. Phys., vol. 62, pp. 2245-2253, 1975. [18] P. Avouris and B. N. J. Persson, "Excited states at metal surfaces and their nonradiative relaxation," J. Phys. Chem., vol. 88, pp. 837-848, 1984. [19] H. D. Hagstrum, "Theory of auger ejection of electrons from metals by ions," Phys. Rev., vol. 96, pp. 336-365, 1954. [20] P. Avouris and R. E. Walkup, "Fundamental mechanisms of desorption and fragmentation induced by electronic transitions at surfaces," in Annual Review of Physical Chemistry, H. L. Strauss, G. T. Babcock and C. B. Moore, Eds. Palo Alto: Annual Reviews, Inc., 1989, pp. 173-206. [21] D. R. Jennison, E. B. Stechel, and A. R. Burns, "DIET and the electronic structure of chemisorbed molecules and physisorbed rare gases," in Desorption induced by Electronic Transitions (DlETill), R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988, pp. 167-172. [22] P. Feulner, "ESD neutrals from chemisorbed and physisorbed species: angular and energy distributions," in Desorption Induced by Electronic Transitions (DIET II), W. Brenig and D. Menzel, Eds. Berlin: Springer-Verlag, 1985, pp. 142-151. [23] P. Feulner, W. Riedl, and D. Menzel, "Angular distributions of neutrals desorbed by electron impact for chemisorbed and physisorbed layers on metal surfaces," Phys. Rev. Lett., vol. 50, pp. 986-989, 1983. 135 [24] P. Feulner, R. Treichler, and D. Menzel, "Thresholds and mechanism in electron-stimulated desorption of ions and neutrals from covalent adsorbates on metals," Phys. Rev. B, vol. 24, pp. 7427-7430,1981. [25] Z. W. Gortel, H. J. Kreuzer, P. Feulner, and D. Menzel, "Electronically stimulated desorption from physisorbed layers on metal surfaces: kinetic-energy distributions of desorbed neutral atoms," Phys. Rev. B, vol. 35, pp. 8951-8968, 1987. [26] Z. W. Gortel, H. J. Kreuzer, P. Feulner, and D. Menzel, "Theory of desorption of neutrals by temporary ionization of physisorbed species at metal surfaces," in Desorption Induced by Electronic Transitions (DIET III), R. H. Stuhlen and M. L. Knotek, Eds. Berlin: Springer- Verlag, 1988, pp. 173-181. [27] W. Hubner and K.-H. Bennemann, "Stimulated desorption of rare gas neutrals - calculation of yield and kinetic energy distributions," Z. Phys. B, vol. 78, pp. 131-136, 1990. [28] P. Sigmund, "Theory of sputtering. I. Sputtering yield of amorphous and polycrystalline targets.," Phys. Rev., vol. 184, pp. 383-416, 1969. [29] M. W. Thompson, "The energy spectrum of ejected atoms during the high energy sputtering of gold," Philos. Mag., vol. 18, pp. 377-414, 1968. [30] W. L. Clinton and R. E. Jutila, "Ion energy distributions from photon- and electron-stimulated desorption: reflection approximation," Phys. Rev. B, vol. 31, pp. 6441-6446, 1985. [31] W. L. Clinton, S. Pal, and R. E. Jutila, "Ion energy distributions from photon- and electron-stimulated desorption. II. The quasiclassical final state and reneutralization," Phys. Rev. B, vol. 36, pp. 4123- 4134, 1987. [32] D. Menzel, "Desorption induced by electronic transitions," Nucl. Instrum. Methods Phys. Res. vol. B, 13, pp. 507-517, 1986. j CHAPTER VIII CONCLUSIONS AND EXTENSIONS The major findings of this thesis research The main conclusions of this thesis research may be summarized as follows: 1) The total ESD yield from methanol-degreased AI-6063 is on the order of 10"3 molecules per incident electron, and this yield is not changed upon baking the sample at 170° C for more than 50 hours. These findings indicate that substantial beam conditioning would be required if this type of degreasing agent were used in the fabrication of synchrotron vacuum chambers. 2) The desorption yield of ions from the methanol-degreased AI-6063 alloy and from methanol-dosed Al(111) is several orders of magnitude lower than the neutral yield. 3) Existing studies in the literature and photoelectron spectroscopy experiments performed in this dissertation indicate that the major species present at room temperature on methanol-dosed Al(111) is the methoxy species. Heating of the surface to temperatures near 500 K causes the pyrolysis of this surface species to aluminum carbide and aluminum oxide. 4) The major neutral desorbate from the two aluminum systems studied is a hydrocarbon which, upon laser ionization, is photolyzed to fragments in the mass 12-13 and 28-31 atomic mass unit ranges. This desorbing hydrocarbon is thought to be the adsorbed methoxy species. 5) Neutral ESD can be used to study changes in surface chemistry. In this work it has been used to monitor adsorbed methoxy decomposition as a function of temperature and electron beam fluence. 6) Electron beams can facilely destroy and alter adsorbed species, with the cross sections for loss of methoxy from the Al(111) surface ranging from 136 137 10--I5 to 10-17 cm 2f depending on the electron beam energy used. The pulsed nature of the time-of-flight experiments performed in this thesis has allowed these sensitive systems to be investigated without their being appreciably altered during the course of the investigation. 7) Desorbing neutrals are between 10 and 100 times less energetic than desorbing ions. Unlike other chemisorbed systems studied with ESD, no ions of the parent neutral species are observed to desorb. These findings indicate that the ions result from sustained evolution in their initial excited state potential, while neutrals, on the other hand, arise predominantly from deexcitations (curve crossings) from the initial excited state potential surface to the bound ground state potential. It is not necessary to invoke reneutralization of an ion, such as CH 3 0 +, as deexcitation from a repulsive neutral curve can also lead to low-energy neutrals. 8) Under some conditions, metal atoms or metal-containing molecules may be desorbed by electronic excitations. No aluminum desorption is detectable from Al(111) which is shown to be clean by standard surface science techniques. A neutral aluminum desorbate, however, is detectable from the methoxy/AI(111) system, as proven through the use of laser resonant ionization. Though the yield of this aluminum species is 2-3 orders of magnitude lower than the methoxy yield, it is as high as that of ions typically observed in ESD studies. The details of the mechanism which can lead to this desorption channel are yet unclear, but the answer, no doubt, involves ways in which the excited state can stay localized long enough for nuclear motion to ensue. Possible explanations are that oxygen underlayer formation or defect surface sites provide the needed extension in excited state lifetime. Proposed extensions of this research Though the experiments presented in this thesis have provided new insight into the mechanisms operative in stimulated desorption processes, it is clear that further work needs to be done in the field. There are three 138 main avenues of further investigation which future work from the Al(111) system could address. The first is internal state distributions of the desorbing neutrals. Resonant ionization provides the possibility to investigate the rotational-vibrational distribution of the desorbates. For example, these kinds of experiments couid yield information as to whether a change in rotational geometry is taking place in the desorption process (as discussed in the previous chapter). Unfortunately, this is a rather difficult experiment due to the size of the methoxy species (the population of the desorbates may be spread over a large number of states), to the high ionization potential (10.77 eV) of the molecule, and to the small gas phase density resulting from desorption. The second avenue of experiments that would prove useful in elucidating operative desorption mechanisms is threshold work. To understand whether the initial electronic excitation is valence or core level in nature requires measuring ESD signal as a function of electron beam energy down to the desorption threshold. This requires beam energies as low as 10 to 20 eV. It is still unclear how low in energy the SARISA electron gun can operate with good current to the target. But if these kinds of studies could be performed, the initial step in the desorption process would certainly be better understood. Perhaps the most pressing need is to determine the origin of the desorbing aluminum metal. As mentioned in chapter VI, attempts, to date, to identify the aluminum-containing molecule have been unsuccessful. Assuming that the desorbate is not aluminum atoms, the problem is most likely that the ArF laser radiation is efficiently cracking the parent desorbate. One possible solution is to search for the parent species with resonant laser ionization. Some candidates for this search are Al-O, Al-H, and AI 2O3. Though AI-OCH3 is also a likely desorbate, the spectroscopy of this species is apparently not known. As an alternative to this resonant ionization search, frequency-tripling to produce vacuum-ultraviolet (VUV) radiation (~ 100 nm) might allow for the ionization of the parent without the unwanted competing process of cracking dominating. Interestingly, preliminary experiments show that aluminum desorption is observed (using ArF radiation) from Al(111) which is dosed 139 heavily with oxygen. This is a significant finding since aluminum oxide is a very common material. As discussed in chapter IV, oxygen forms underlayers on Al(111), and the detection of metal desorption from this surface provides further support for the theory that underlayer formation is effective in slowing down the excited state quenching process of desorption. It is obvious that there is much more work to be done, not only in the field of stimulated desorption, but also on the Al(111) system investigated in this dissertation. It is hoped, however, that the work of this thesis has helped to provide a better understanding of some of the processes important in electron and photon beam interactions with surfaces and will serve as an impetus for further research. BIBLIOGRAPHY P. A. Antoniewicz, "Model for electron- and photon-stimulated desorption," Phys. Rev. B, vol. 21, pp. 3811-3815, 1980. Report No. ANL-87-15, p. II-6-2, Argonne National Laboratory, 1987. H. F. Arlinghaus, W. F. Calaway, C. E. Young, M. J. Pellin, D. M. Gruen, and L. L. Chase, "High-resolution multiphoton laser-induced fluorescence spectroscopy of zinc atoms ejected from laser-irradiated ZnS crystals," J. Appl. Phys., vol. 65, pp. 281-289, 1988. P. Avouris and B. N. J. Persson, "Excited states at metal surfaces and their nonradiative relaxation," J. Phys. Chem., vol. 8 8, pp. 837-848, 1984. P. Avouris and R. E. Walkup, "Fundamental mechanisms of desorption and fragmentation induced by electronic transitions at surfaces," in Annual Review of Physical Chemistry, H. L. Strauss, G. T. Babcock and C. B. Moore, Eds. Palo Alto: Annual Reviews, Inc., 1989, pp. 173-206. P. Avouris, F. Bozso, and R. E. Walkup, "Desorption via electronic transitions: fundamental mechanisms and applications," Nucl. Instr. and Meth. B, vol. 27, pp. 136-146, 1987. P. Basu, J. G. Chen, L. Ng, M. L. Colaianni, and J. T. Yates Jr., "Fragmentation of molecular adsorbates by electron and ion bombardment: methoxy Chemistry on Al(111)," J. Chem. Phys., vol. 89, pp. 2406-2411, 1988. I. P. Batra and L. Kleinman, "Chemisorption of oxygen on aluminum surfaces," J. Electron. Spectrosc. Relat. Phenom., vol. 33, pp. 175- 241, 1984. R. B. Bernstein, "Systematics of multiphoton ionization-fragmentation of polyatomic molecules," J. Phys. Chem., vol. 8 6, pp. 1178-1184, 1982. 140 141 F. Bozso and P. Avouris, "Reaction of Si(100) with NH 3: rate-limiting steps and reactivity enhancement via electronic excitation.," Phys. Rev. Lett., vol. 57, pp. 1185-1188, 1986. Desorption Induced by Electronic Transitions (DIET II), vol. 4, W. Brenig and D. Menzel, Eds. Berlin: Springer-Verlag, 1985. A. R. Burns, D. R. Jennison, and E. B. Stechel, "Electronically stimulated dissociation of N0 2 on Pt(111)," Phys. Rev. B, vol. 40, pp. 9485-9497, 1989. A. R. Burns, E. B. Stechel, and D. R. Jennison, "Desorption by electronically stimulated adsorbate rotation," Phys. Rev. Lett., vol. 58, pp. 250-253, 1987. A. R. Burns, E. B. Stechel, and D. R. Jennison, "Rovibrational laser spectroscopy of ESD neutrals from chemisorbed species," in Desorption Induced by Electronic Transitions (DIET III), R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988, pp. 67-72. R. R. Chance, A. Prock, and R. Silbey, "Comments of the classical theory of energy transfer," J. Chem. Phys., vol. 62, pp. 2245-2253, 1975. J. G. Chen, P. Basu, L. Ng, and J. T. Yates Jr., "A comparative study of the reactivities of H2O, CH3OH, and CH 3OCH3 toward Al(111)," Surf. Sci., vol. 194, pp. 397-418, 1988. W. L. Clinton and R. E. Jutila, "Ion energy distributions from photon- and electron-stimulated desorption: reflection approximation," Phys. Rev. B, vol. 31, pp. 6441-6446, 1985. W. L. Clinton, S. Pal, and R. E. Jutila, "Ion energy distributions from photon- and electron-stimulated desorption. II. The quasiclassical final state and reneutralization," Phys. Rev. B, vol. 36, pp. 4123- 4134, 1987. J. H. Craig Jr., "Characteristic energies of electronically desorbed ions from Rh(100)," Appl. Surf. Sci., vol. 35, pp. 520-526, 1989. J. H. Craig Jr., "CO on Pt(111): TDS and ESD study," Appl. Surf. Sci., vol. 25, pp. 333-340, 1986. J. E. Crowell, J. G. Chen, and J. T. Yates Jr., "Surface sensitive spectroscopic study of the interaction of oxygen with Al(111)- low temperature chemisorption and oxidation," Surface Sci., vol. 165, pp. 37-64, 1986. 142 L. A. Curtiss and D. Kock, private communication. L. A. Curtiss, D. Kock, and J. A. Pople, "Energies of CH 2OH, CH3O, and related compounds," to be published, 1991. M. Q. Ding, E. M. Williams, J. P. Adrados, and J. L. de Segovia, "Energy distribution of H+ ions with ESD of water adsorbed at aluminum and tungsten surfaces," Surf. Sci., vol. 140, pp. L264-L268, 1984. D. E. Donohue, J. A. Schiavone, and R. S. Freund, "Molecular dissociation by electron impact: optical emission from fragments of methane, ethylene, and methanol," J. Chem. Phys., vol. 67, pp. 769- 780, 1977. K. H. Drexhage, H. Kuhn, and F. P. Schaefer, "Variation of the fluorescence decay time of a molecule in front of a mirror," Ber. Bunsenges. Phys. Chem., vol. 72, p. 329,1968. K. H. Drexhage, M. Fleck, H. Kuhn, F. P. Schaefer, and W. Sperling, "The influence of a mirror on the fluorescence of a europium chelate," Ber. Bunsenges. Phys. Chem., vol. 70, p. 1179, 1966. L. C. Feldman and J. W. Mayer, Fundamentals of surface and thin film analysis. New York: Elsevier, 1986. P. Feulner, D. Menzel, H. J. Kreuzer, and Z. W. Gortel, "Kinetic- energy distributions of neutrals desorbed by electron impact from adsorbates on metal surfaces," Phys. Rev. Lett., vol. 53, pp. 671-674, 1984. P. Feulner, "ESD neutrals from chemisorbed and physisorbed species: angular and energy distributions," in Desorption induced by Electronic Transitions (DIET il), W. Brenig and D. Menzel, Eds. Berlin: Springer-Verlag, 1985, pp. 142-151. P. Feulner, R. Treichier, and D. Menzel, "Thresholds and mechanism in electron-stimulated desorption of ions and neutrals from covalent adsorbates on metals," Phys. Rev. B, vol. 24, pp. 7427-7430, 1981. P. Feulner, W. Riedl, and D. Menzel, "Angular distributions of neutrals desorbed by electron impact for chemisorbed and physisorbed layers on metal surfaces," Phys. Rev. Lett., vol. 50, pp. 986-989, 1983. 143 R. Franchy and D. Menzel, "Adsorbate core ionization as primary process in electron- and photon-stimulated desorption from metal surfaces," Phys. Rev. Lett., vol. 43, pp. 865-867, 1979. H. Gnaser and W. O. Hofer, "The emission of neutral clusters in sputtering," Appl. Phys. A, vol. 48, pp. 261-271, 1989. Z. W. Gortel, H. J. Kreuzer, P. Feulner, and D. Menzel, "Electronically stimulated desorption from physisorbed layers on metal surfaces: kinetic-energy distributions of desorbed neutral atoms," Phys. Rev. B, vol. 35, pp. 8951-8968, 1987. Z. W. Gortel, H. J. Kreuzer, P. Feulner, and D. Menzel, "Theory of desorption of neutrals by temporary ionization of physisorbed species at metal surfaces," in Desorption Induced by Electronic Transitions (DIET III), R. H. Stuhlen and M. L. Knotek, Eds. Berlin: Springer- Verlag, 1988, pp. 173-181. H. D. Hagstrum, "Theory of auger ejection of electrons from metals by ions," Phys. Rev., vol. 96, pp. 336-365, 1954. H. J. Halama and C. L. Foerster, "Comparison of photodesorption yields from aluminum, stainless and Cu-plated beam tubes," Vacuum, vol. 42, pp. 185-188, 1991. W. B. Herrmannsfeldt, Report No. SLAC-226, UC-28 (A), Stanford Linear Accelerator Center, Stanford University, Stanford, CA, 1981. G. Herzberg,Atom/c spectra and atomic structure. New York: Dover, 1944. W. Hubner and K.-H. Bennemann, "Stimulated desorption of rare gas neutrals - calculation of yield and kinetic energy distributions," Z. Phys. B, vol. 78, pp. 131-136, 1990. D. R. Jennison, E. B. Stechel, and A. R. Burns, "DIET and the electronic structure of chemisorbed molecules and physisorbed rare gases," in Desorption Induced by Electronic Transitions (DIETIII), R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988, pp. 167-172. N. M. Johnson, "Measurement of semiconductor-insulator interface state by constant-capacitance, deep-level transient spectroscopy," J. Vac. Sci. Technol., vol. 21, pp. 303-314, 1982. F. Jona, "Preparation and Properties of Clean Surfaces of Aluminum," J. Phys. Chem. Solids, vol. 28, pp. 2155-2160, 1967. 144 N. Kaufherr, A. Krauss, D. M. Gruen, and R. Nielsen, "Chemical cleaning of aluminum alloy surfaces for use as vacuum material in synchrotron light sources," J. Vac. Sci. Techno!., vol. A8, pp. 2849- 2855, 1990. L. J. Kieffer, Report No. 6, JILA Information Center, 1969. M. L. Knotek and P. J. Feibelman, "Ion desorption by core-hole auger decay," Phys. Rev. Lett., vol. 40, pp. 964-967, 1978. M. L. Knotek, R. H. Stulen, G. M. Loubriel, V. Rehn, R. A. Rosenberg, and C. C. Parks, "Photon-stimulated desorption of H+ and F+ from BeO, AI2O3 and Si 0 2 : comparison of near edge structure to photoelectron yield," Surf. Sci., vol. 133, pp. 291-304, 1983. M. L. Knotek, "Stimulated desorption," Rep. Prog. Phys., vol. 47, pp. 1499-1561, 1984. M. L. Knotek, V. O. Jones, and V. Rehn, "Photon-stimulated desorption of ions," Phys. Rev. Lett., vol. 43, pp. 300-303, 1979. I. Kusunoki, M. Hiroi, T. Sato, Y. Igari, and S. Tomoda, "SiC film formation on Si(001) by reaction with C 2H2 beams," Appl. Surf. Sci., vol. 45, pp. 171-187, 1990. L. D. Landau and E. M. Lifshitz, Mechanics. Reading: Addison- Wesley, 1960, p.38. T. E. Madey and J. T. Yates Jr., "Electron-stimulated desorption as a tool for studies of chemisorption: a review," J. Vac. Sci. Techno!., vol. 8, pp. 525-554, 1971. T. E. Madey, D. E. Ramaker, and R. Stockbauer, "Characterization of surfaces through electron and photon stimulated desorption," Ann. Rev. Phys. Chem., vol. 35, pp. 215-240, 1984. N. Matsunami, Y. Yamamura, Y. Itikawa, N. Itoh, Y. Kazumata, S. Miyagawa, K. Morita, R. Shimizu, and H. Tawara, Report No. IPPJ- AM-32, Institute of Plasma Phyics, Nagoya University, 1983. D. Al-Mawlawi and J. M. Saleh, "Interaction of alcohols with evaporated metal films," J. Chem. Soc. Faraday Trans. I,vol. 77, pp. 2965-2976, 1981. D. Menzel and R. Gomer, "Desorption from metal surfaces by low- energy electrons," J. Chem. Phys., vol. 41, pp. 3311-3328, 1964. 145 D. Menzel, "Desorption induced by electronic transitions," Nucl. Instrum. Methods Phys. Res. vol. B, 13, pp. 507-517, 1986. D. Menzel, "Electronically stimulated desorption: mechanisms, applications, and implications," in Surface and Interface Characterization by Electron Optical Methods, A. Howie and U. Valdre, Eds. New York: Plenum Press, 1987, pp. 285-299. J. A. Nemetz and R. E. Tressler, "Thermal nitridation of silicon and silicon dioxide for thin gate insulators," Solid State Technol., vol. pp. 79-85, 1983. H. J. Neusser, "Multi-photon mass spectrometry and unimolecular ion decay," Int. J. Mass Spectrom. Ion Processes, vol. 79, pp. 141 - 181, 1987. C. Park, M. Bujor, and H. Poppa, "Electron-stimulated desorption study of hydrogen-exposed aluminum films," Thin Solid Films, vol. 113, pp. 337-344, 1984. M. J. Pellin, C. E. Young, and D. M. Gruen, "Multiphoton ionization followed by time-of-flight m ass spectroscopy of sputtered neutrals," Scanning Microsc., vol. 2, pp. 1353-1364, 1988. C. J. Powell, "Attenuation lengths of low-energy electrons in solids," Surf. Sci., vol. 44, pp. 29-46, 1974. D. E. Ramaker, C. T. White, and J. S. Murday, "On Auger induced decomposition/desorption of covalent and ionic systems," Phys. Lett.A, vol. 89, pp. 211-214, 1982. D. E. Ramaker, "Models for desorption in covalent systems," in Desorption Induced by Electronic Transitions (DIET I), N. H. Tolk, M. M. Traum, J. C. Tully and T. E. Madey, Eds. Berlin: Springer-Verlag, 1983, pp. 70-89. P. A. Redhead, "Interaction of slow electrons with chemisorbed oxygen," Can. J. Phys., vol. 42, pp. 886-905, 1964. M. W. Roberts and C. S. McKee, Chemistry of the metal-gas interface. Oxford: Clarendon Press, 1978. M. B. Robin and N. A. Kuebler, "Excited electronic states of the simple alcohols," J. Electron Spectrosc. Relat. Phenom., vol. 1, pp. 13-28, 1972. 146 J. W. Rogers Jr., R. L. Hance, and J. M. White, "An electron spectroscopic investigation of the interaction of methanol with polycrystalline aluminum," Surf. Sci., vol. 100, pp. 388-406, 1980. B. Ruscic and J. Berkowitz, "Photoionization mass spectrometric studies of the isomeric transient species CD 2OH and CD 3O," to be published, 1991. R. C. Sausa, A. J. Alfano, and A. W. Miziolek, "Efficient ArF laser production and detection of carbon atoms from simple hydrocarbons," Appl. Optics, vol. 26, pp. 3588-3593, 1987. E. N. Sickafus and C. Kukia, "Linearized secondary-electron cascades from the surfaces of metals. III. Line-shape synthesis," Phys. Rev. B, vol. 19, pp. 4056-4068, 1979. E. N. Sickafus, "Linearized secondary-electron cascades from the surfaces of metals. I. Clean surfaces of homogeneous specimens," Phys. Rev. B, vol. 16, pp. 1436-1447, 1977. E. N. Sickafus, "Linearized secondary-electron cascades from the surfaces of metals. II. Surface and subsurface sources," Phys. Rev. B, vol. 16, pp. 1448-1458, 1977. M. W. Siegel and M. J. Vasile, "New wide angle, high transmission energy analyzer for secondary ion mass spectrometry," Rev. Sci. instrum., vol. 52, pp. 1603-1615, 1981. P. Sigmund, "Theory of sputtering. I. Sputtering yield of amorphous and polycrystalline targets," Phys. Rev., vol. 184, pp. 383-416, 1969. E. B. Stechel, D. R. Jennison, and A. R. Burns, "Dynamics in neutral DIET from chemisorbed molecules," in Desorption induced by Electronic Transitions (DIET Hi), R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988, pp. 136-143. E. Steinacker and P. Feulner, "Stimulated desorption of atoms from rare-gas monolayers on metal surfaces: dependence of yields and energy distributions on primary excitations," Phys. Rev. B, vol. 40, pp. 11348-11351, 1989. Desorption Induced by Electronic Transitions (DIET III), vol. 13, R. H. Stulen and M. L. Knotek, Eds. Berlin: Springer-Verlag, 1988. R. H. Stulen, "Recent developments in angle-resolved electron- stimulated desorption of Ions from surfaces," Prog, in Surf. Sci., vol. 32, pp. 1-38, 1989. 147 M. W. Thompson, "The energy spectrum of ejected atoms during the high energy sputtering of gold," Philos. Mag., vol. 18, pp. 377-414, 1968. I. F. Tindall and J. C. Vickerman, "The adsorption and reactions of methanol on clean and oxidised aluminium surfaces," Surf. Sci., vol. 149, pp. 577-591, 1985. Desorption Induced by Electronic Transitions (DIET I), vol. 24, N. H. Tolk, M. M. Traum, J. C. Tully, and T. E. Madey, Eds. Berlin: Springer-Verlag, 1983. C. J. Tung, J. C. Ashley, and R. H. Ritchie, "Electron inelastic mean free paths and energy losses in solids II," Surf. Sci., vol. 81, pp. 427- 439, 1979. C. J. Tung, J. C. Ashley, R. D. Birkhoff, and R. H. Ritchie, "Electron slowing-down flux spectrum in AI 2O3," Phys. Rev. B, vol. 16, pp. 3049-3055, 1977. Desorption Induced by Electronic Transitions (DIET IV), vol. 19, G. Betz and P. Varga, Eds. Berlin: Springer-Verlag, 1990. G. D. Waddill and L. L. Kesmodel, "Identification of methoxide species on Al(11 0 ) by high-resolution EELS," Surf. Sci., vol. 182, pp. L248- L252, 1987. H. R. Wendt and H. E. Hunziker, "Electronic absorption spectrum of CHgO," J. Chem. Phys., vol. 71, pp. 5202-5205, 1979. R. L. Whetten, K.-J. Fu, R. S. Tapper, and E. R. Grant, "Highly efficient production of neutral carbon atoms in the ultraviolet multiphoton fragmentation of aromatic molecules," J. Phys. Chem., vol. 87, pp. 1484-1487, 1983. J. C. White, J. Bokor, R. R. Freeman, and D. Henderson, "Tunable ArF excimer-laser source," Opt. Lett., vol. 6, pp. 293-294, 1981. P. Williams and G. Gillen, "Direct evidence for coulombic ejection of electron-desorbed ions," Surf. Sci., vol. 180, pp. L109-L112, 1987. P. A. Wolff, "Theory of secondary electron cascade in metals," Phys. Rev., vol. 95, pp. 56-66, 1954. 148 J. T. Yates Jr., M. D. Alvey, K. W. Kolasinski, and M. J. Dresser, "Ion angular distribution of species desorbed from single crystal surfaces by electron impact," Nucl. Inst, and Meth. B, vol. 27, pp. 147-154, 1987. C. E. Young, M. J. Pellin, W. F. Calaway, B. Jorgensen, E. L. Schweitzer, and D. M. Gruen, "Trace surface analysis via RIS/TOF mass spectrometry," Inst. Phys. Conf. Ser., vol. 84, pp. 163-168, 1987. U. Yxklinten, M. Limback, and B. I. Lundqvist, "Initial stages of oxidation of Al(111)," in Many-Atom Interactions in Solids, R. M. Nieminen, M. J. Puska and M. J. Manninen, Eds. Berlin: Springer- Verlag, 1990, pp. 315-318.