CHOPRA -Characterization of Semiconductor Surfaces by Appearence Potential Spectroscopy

6

Characterization of Semiconductor Surfaces

by Appearance Potential Spectroscopy

Dev R. Chopra and Anil R. Chourasia

INTRODUCTION

The unoccupied electronic states in the range 0 to 10 eV above the Fermi level (EF) are considered important for understanding many physical and chemical properties of materials. Techniques such as appearance potential spectroscopy (APS) [l-3], bremsstrahlung isochromat spectroscopy (BIS) [4-61, inverse photoemission (IPE) [7-91 and x-ray absorption spectroscopy (XAS) [lo-121 have appeared to be the most promising approaches for measuring the electronic density of states (DOS) above EF. Among these techniques, APS has the advantages that of being the simplest one, requires relatively inexpensive instrumentation for its undertaking, and provides rich information about the conduction band DOS of the surfaces of materials. In this threshold spectroscopy the surface of a solid is bombarded by electrons in the O-2000 eV range. When the energy of the incident electron equals to that of a particular core level, the incident electron imparts its energy to the core level electron as a result of the inelastic collision. The system in question is then left in a final excited state consisting of a core hole and two electrons above EF in the conduction band. The yield (X-ray or Auger electrons) of the relaxation process is measured as a function of energy of the incident electron. The relaxation of the core hole can be measured in several ways:

1. when the intensity of the emitted x-rays is measured the method is called soft x-ray APS or SXAPS [13],

2. when the total current of the secondary electrons is measured the method is called Auger electron APS

or AEAPS [14,15],

290 Characterization of Semiconductor Materials

3. when the current of the elastically reflected electrons is measured, the number of electrons that are scattered inelastically at the threshold energy of core level excitation increases. They disappear from the measured current, and the method is, therefore, called Disappearance Potential Spectroscopy or DAPS [16].

Other types of APS are x-ray photoelectron APS &PAPS) and resonant photoelectron APS (RPAPS). In XPAPS, the sample is bombarded by photons [17,18]. The spectra obtained with this technique are much weaker than those of SXAPS and depend upon the material used for the anode of the x-ray tube. RPAPS, developed by Hua et al. [19], is a combination of SXAPS and XPAPS. In it the photocathode is made of the same material as the target. Using this, they have found RPAPS to be more sensitive to the sub-peaks for principal elements and peaks for the impurities present. However, these spectroscopies are not commonly used for surface analysis.

Appearance potential spectroscopy measures the probability for electronic excitation of a core level as a function of incident electron energy. The energy of the incident electrons is gradually increased, and the dependence of the total signal strength on this energy is measured. At certain energies a sudden change (increase or decrease) in the signal is observed, which corresponds to the excitation of a given energy level of the sample. The signal is extracted from the background with the help of electronic differentiation technique which enhances the signal-to-noise ratio. The intensity of the features in the APS spectrum at the threshold energy and above depends on the core hole excitation rate at that energy. The final state in the excitation process will consist of a core hole, an excited core electron and the scattered incident electron. The fact that the shape of an APS feature is independent of the relaxation mechanism suggests the utilization of a non-dispersive scheme for this technique. Since neither an electron energy analyzer nor an X-ray monochromator is needed to obtain the experimental data, APS is undoubtedly the simplest method for studying the unoccupied DOS of solid surfaces. The surface sensitivity of APS is due to the short inelastic mean free path for the primary electrons in the energy range (O-2 kV). A primary electron with an energy close to the threshold that experiences a characteristic energy loss upon penetrating the sample is no longer able to excite atoms. The depth of information is, therefore, of the order of 10 A.

Appearance potential spectroscopy is a classical technique. It was

Appearance Potential Spectroscopy 291

introduced in 1921 to determine the binding energy (BE) of core levels of atoms in solids [20]. The onset structure was extracted from a large background emission by graphical differentiation. In subsequent years the BEs of the core states of several materials were determined. Difficulties resulting from inadequate sample preparation techniques and poor vacuum conditions were encountered at that time. Due to impurities present in the sample the number of threshold structures exceeded the number of available core levels of the material. However, the most serious problem was that the differentiation method was not very accurate. With the development, in the 1930’s, of dispersive x-ray analyzers which could filter out continuum radiation the attention of the researchers was diverted to the field of X-ray spectroscopy and the appearance potential measurements were abandoned. In the 1950’s, it was briefly revived by Shinoda et al. [21]. They ramped the accelerating potential with a saw- tooth wave and, using electronic differentiation, were able to observe appearance potential spectra with an oscilloscope display. In 1967 Liefeld [22] used more sophisticated electronic differentiation. Later in the 1970’s it was developed as a practical tool for the study of the electronic structure and composition of solid surfaces by Park et al. [2,13,23]. The availability of ultrahigh vacuum (UHV) and the techniques for cleaning the surface in situ together with electronic differentiation using potential modulation has made APS a more versatile spectroscopy for the study of surfaces of materials.

Because of its experimental and conceptual simplicity and because of its ability to obtain detailed information from structures superimposed on a large background [1,2,24] APS has attracted considerable attention. As a tool for chemical analysis it has its merits when compared with other techniques, especially when applied to multicomponent alloys [25]. APS is sensitive to adsorption phenomena [26-281. Chemisorption and oxidation phases can be distinguished [29]. An important aspect of APS is that it reveals the localized DOS because the matrix element governing the core hole production involves the very short range wave function of the initial core electron state. Since electronic excitation of atomic core levels does not follow dipole selection rules, APS reveals information regarding the total DOS above EP. As stated above, APS does not require a dispersive analyzer in contrast to other techniques. It thus measures energy rather than momentum. This accounts for the extreme simplicity of the APS spectrometer.

Information which can be derived from an APS spectrum consists of the following:

1. The identification of elements is easily accomplished by means of the known BEs of the core electrons.


Only a few, narrow lines per element are observed in the AI’S spectrum. The complications therefore arising due to line overlap are not present.

2. Chemical information about the elements under consideration is available in the form of chemical shifts. The threshold energy changes when the electron density surrounding a particular atom is changed by the formation of a chemical bond.

3. The spectral line shape also provides information about the empty DOS which is also influenced by the chemical bonds and surface effects [30]. The intrinsic empty DOS may be recovered by deconvolution.

In addition to these, the analysis of the 3d transition metals indicates the following characteristics:

4. In the derivative spectrum the width of the positive peak approximates the width of the unfilled portion of the 3d band, and

5. In the absence of the 4sp band, the height of the negative peak should equal the positive peak. The decrease in the negative peak thus measures the contribution of the 4sp states to the 3d band.

PRINCIPLE

Low energy electrons interact with atoms by elastic collisions, by the emission of electromagnetic radiation, and by inelastic collisions. At the threshold energy for core hole excitation a certain fraction of the incident electrons are involved in ionization. These electrons give up their energy and can, therefore, no longer produce Bremsstrahlung radiation in the sample. Thus, along with the sharp increase in characteristic radiation, the Bremsstrahlung radiation decreases. The AI’S spectrum consists of step-like features superimposed on Bremsstrahlung background. When the incident electron is captured by a state cl above EP the energy may be conserved by the excitation of a core electron into a state c2 such that cl + 62 = E - Eb. Here E is the energy of the incident electron and Eb is the BE of a particular core level relative to EB. The final state consists of an excited atom and two quasi-free electrons in energy states E 1 and e2. The excess energy E - Eb will now be shared between the two electrons, both landing in empty energy states above EB, as shown in fig. 1. Simplifying for


SAMPLE

Figure 1. Electron excitation in APS. The core electron may scatter into a state c2 = eV + ed + kT - E l - Eb. The core hole may subsequently decay by the emission of characteristic X-rays.

the value of E, the incident electron energy, we get:

E2 =eV+e$C+kT-E, -E,, (1)

where V is the potential applied between a thermionic emitter and the sample, edc is the emitter work function, and kT is the average thermal energy of the emitted electrons.

The features in the APS spectrum are proportional to the core

294 Characterization-of Semiconductor Materials

hole excitation rate P(E). According to Dev and Brinkman [31] and Park and Houston [2,32] the excitation rate depends on the excitation cross- section, the spatial distribution of atoms in the solid, the penetration depth of the incident electrons, and the width of the core state involved. It is then given by:

s t P(E) - F(E - BJ f(B,B,J a, (2) 0

where F(E - Eb) represents the core level Lorentzian function. The function f(E,Eb) must take into account both the core electron which is promoted to the Fermi level and the scattered incident electron, and is given by:

f(E,EJ = IE-Eb P,&E) P,(+E) N@J Wz) dc,- (3) 0

Here N(E )

i

and N(E ) d2

are the empty DOS at energies E

factors pl cl,E) an p$ c2,E) .1

and ~2 The representing the respective transition

probabilities, may depen on selection rules which are unknown in solids. If these factors are assumed to be independent of E and E over a small range of E values above the threshold value Eb, then they can be taken as simply proportional to N(E), the empty DOS. Equation 3 then reduces to:

s

E-Eb

f(E,EJ = N(B - Et,- Q) W2) (&I (4) 0

which is simply the self-convolution of the density of conduction band states. In APS, the derivative APS(E) of the yield is determined experimentally. Therefore, differentiating eqn. (4) and taking the core level DOS as the Dirac delta function, we get:

AI’S(E) = f’ (E,E,,) = N(O)N(E - EJ +

s

E-Eb dN(E - E,- I$

dE d+

0

(5)

At EF the first term vanishes. The structure in the APS spectrum is then given by the second term. For a simple step-like density of empty states the derivative dN(E - Eb - e2)/dE can be approximated [33] by NEF6E, where N

Ef is the DOS at the threshold and SE is the Dirac delta

function. T erefore:

AI’s (E) = NEF . N (E - Eb). (6)

The intensity at the threshold is then given by:


APS (E) = NEF2, (7)

i.e., the height of the structure above the background is proportional to the square of the DOS at EF

To get a qualitative picture of an AI’S spectrum we have shown in fig. 2a a schematic of the model generally considered valid for the 3d transition metals. In this model, the Fermi level lies in a fairly sharp d band which is superimposed on a broad free-electron like 4sp band. The

n(E)

APPROXIMATE DENSITY OF

CONDUCTION BAND STATES

CONVOLUTION OF DENSITY

OF EMPTY STATES

E-E b

n(E ) n(E - Eb)dE

ENEMY DERIVATIVE

d E -Eb

Z J

n(E) n(E -Eb)dE

0

E

E (cl

Figure 2. Schematic representation of the DOS of 3d transition metals. (a) The dotted line shows the approximation of the conduction band DOS. (b) The self-convolution of the empty DOS is shown. (c) The derivative of the convoluted function is also shown.


self convolution of the unfiied portion of the band can be approximated by a step function, as shown in fig. 2b. To obtain an APS spectrum we differentiate this function. In fig. 2c is shown the derivative of the convoluted function. Thus in an APS spectrum we expect a positive peak followed by a negative peak. The width of the positive peak corresponds to the unf’iied portion of the d-band. The reduction of the negative peak with respect to the positive peak height measures the relative contribution of the s-p states to the total DOS. In actual practice, the APS spectrum is broadened by the core-level lifetime width, the instrument response function, the energy spread of the incident electrons, and the modulation voltage.

The BE in APS is obtained directly from the recorder plots by applying the correction for the work function of the thermionic electron source. To avoid the uncertainty introduced due to this correction in BE measurements, Fukuda et al. [34] have used a field-emission source. In earlier measurements in APS, the BE was determined in a simple way by the intersection of the extrapolated projection of the background and positive going low energy slope of the peak. Since the APS yield is proportional to the self-convolution of the density of the final electron states broadened by the finite lifetime of the core hole and other effects stated earlier, precise knowledge about BE can be obtained by using deconvolution techniques. Successful deconvolution techniques have been developed by Fukuda et al. [34], Dose et al. [35,36], and Schulz et al. [373.

EXPERIMENTAL

Soft X-ray Appearance Potential Spectroscopy

In SXAPS the total soft x-ray intensity emitted by a sample under electron bombardment is measured as a function of incident electron energy. The schematic of the SXAPS spectrometer fabricated in our laboratory [38] is shown in fig. 3. Inside the spectrometer chamber which is constructed of stainless steel the target (S) and the filament (F) are mounted on a high-vacuum feed through flange. The filament is a fine tungsten wire mounted close to the target for stable operation of the spectrometer at low accelerating potentials. The detector assembly is mounted on another flange on the opposite end of the chamber and is screened from the filament-target assembly by a nickel wire mesh grid. This grid is biased negative relative to the filament. The function of the grid is to prevent the thermal electrons from reaching the detector assembly. The chamber walls act as a photocathode.


ref. PHASE

sig. LOCK -

Figure 3. Simplified schematic of the soft X-ray appearance potential spectrometer. S is the sample and F is the filament. Signal is extracted by the potential modulation technique.

Electrons from the filament are accelerated towards the sample which is biased positively. Upon impinging on the target, the electrons produce soft X-ray radiation. The filament is operated by an emission control unit which maintains a constant emission current irrespective of the target potential. The potential on the target is varied from 0 to 2000 V by a variable slope ramp generator. The same potential drives the abscissa of an x-y recorder. In order to record the spectrum, the target potential is linearly varied through the desired voltage range. The resulting x-rays pass through the grid and strike the chamber wall, generating photoelectrons. These electrons are collected by the detector. As the target potential approaches a threshold for core-level excitation of surface atoms, the x-ray production increases abruptly with increase in incident electron energy. Extraction of the signal from the background is accomplished by differentiation of the signal. This is done by superimposing a low voltage (- 0.3 VP-J, high frequency sinusoidal signal on the accelerating voltage (V). This signal causes the x-ray intensity, I, to vary at that frequency rate, the amplitude and phase of the variation being proportional to AI/AV. The signal is detected by filtering the photoelectron current to extract the above frequency component which is then amplified with the help of a phase-lock amplifier and synchronously rectified to provide a dc level corresponding to the slope of the x-ray intensity. This signal drives the ordinate of the x-y recorder. The APS spectrum is plotted in terms of the detector output versus the accelerating potential. The emission current, modulation voltage, and time constant of the phase-lock amplifier are adjusted to record the precise spectra as determined by the signal-to-noise ratio and peak width at half-maximum. Other details of the technique are given elsewhere [2,13,38].


Auger Electron Appearance Potential Spectroscopy

The AEAPS spectrometer, also fabricated in our laboratory, consists of a triode arrangement [39] and is shown in fig. 4. The anode is cylindrical in shape and completely surrounds the filament except for a one mm diameter exit aperture on the top. The target sample is mounted directly above the aperture as shown. Electrons emitted from the filament are accelerated to the anode by a constant potential. The emission current between the filament and anode is typically less than 1 mA. The emission current leaving the exit aperture is therefore much smaller, and on the order of 5 PA. The target is also at a positive potential which is linearly varied by a programmable ramp generator. The ano.de potential is held at a voltage that is higher than the maximum voltage of the ramp for a certain core state, therefore the anode serves as a collector of backscattered electrons. The current in the anode-sample circuit is then:

where I is the primary electron current, and I, is the secondary electron current.‘Because the anode potential is kept constant, the primary current remains constant. An increase in I, due to the appearance of Auger

Figure 4. Schematic diagram of Auger electron appearance potential spectrometer. Thermionic electrons passing through an aperture in the anode impinge on the sample. Signal is extracted by the potential modulation technique.


electrons following the onset of core level excitation results in a sudden decrease in I. Variations in I then reflect changes in the secondary current only. Extraction of this information is accomplished by differentiating the secondary current with respect to the target voltage. This is done, as in the case of SXAPS, by superimposing a low voltage, high-frequency sinusoidal signal on the filament. The secondary current in the anode-sample circuit that varies at this frequency is then synchronously detected with a conventional phase-lock amplifier. This detector output, which is proportional to AI/AV, is plotted as a function of the sample potential on the x-y recorder. Because the secondary electron emission does not exhibit a linear dependence on incident electron energy, it is generally advantageous to make measurements in the second derivative mode. This method of measurement enhances the signal-to-noise ratio considerably. Other details of the experimental technique can be found elsewhere [39].

Papagno and Scarmozzino [40] have modified a Varian Auger electron spectrometer to include AEAPS. Such a system enables one to use very low beam current and has the advantage of performing in situ on the same sample both AES and AEAPS measurements. Euler [41] has constructed a simple APS spectrometer out of the commercially available ionization gauges. The tubes containing the samples were evacuated and sealed off, so that no vacuum equipment was needed to maintain UHV conditions necessary during the course of measurements. The spectrometer could be used for both SXAPS and AEAPS measurements.

Disappearance Potential Spectroscopy

At low energies the AEAPS spectra of single-crystal materials are complicated by a low energy electron diffraction (LEED) structure. This structure is formed only by elastically scattered electrons. To separate the elastic and inelastic contributions to the APS spectra, Eckertova and Pavluch [42,43] have used a 3-grid hemispherical system in a special arrangement. The schematic of this system is shown in fig. 5. The grid G1 has the same potential as that of the sample. The grid G2 operates on a positive potential and collects the backscattered electrons. Some of these electrons strike G2 while others pass through the space between G2 and G3. The grid G3 has a negative potential and controls the operation of the spectrometer. By changing the potential of G3 the contributions of electrons with different energies can be obtained. At a particular potential applied to G the secondary electron current at the threshold decreases. The elastica ly reflected electrons then constitute the DAPS signal. This ? spectrometer could also be used for AEAPS measurements. The advantages of such a modified spectrometer are that it reduces the effect of the primary current on the spectra and also reduces the non-linearity of


Lock in

Figure 5. Schematic of the DAPS method. Grid G3 controls the operation of the spectrometer. The same set-up could be used for AEAPS by applying proper voltage to GS. The figure is reproduced by permission from ref. [42].

the electron gun. With these modifications the signal-to-background ratio is considerably enhanced, thus facilitating detailed interpretation of the spectra, especially in the case of single crystals. Due to the smooth background, the fine structure appearing on the high energy side of a particular core level spectrum can be easily detected and analyzed to obtain important information. Kirschner and Losch [44] have shown the feasibility of combining the DAPS and AEAPS techniques with AES by modifying a commercially available cylindrical-mirror-analyzer system.

The major problem encountered in APS is the signal-to-noise ratio. In the SXAPS, broad-band noise is present due to Bremsstrahlung photons, and this increases steadily with the primary electron energy [45]. In AEAPS, the yield of low energy secondary electrons is not a simple


function of the primary electron energy and depends sensitively on the surface conditions. In DAPS, the backscattering cross section of the electrons in the solid is low. This makes the total reflection coefficient rather small (of the order of 10m3 to 10W2) which in turn reduces the signal- to-noise ratio. Andersson et al. [46] have designed a low noise SXAPS spectrometer to improve the signal-to-noise ratio. It utilizes a silicon surface-barrier diode detector cooled with liquid nitrogen. They have also used an Al window to filter out low energy Bremsstrahhmg photons. Lee [47j has discussed the signal-to-noise performance of an SXAPS spectrometer in the cases of quantum and energy detectors. He found that in the useful electron energy range, the sensitivity of energy detectors is superior to that of quantum detectors using an x-ray filter. He has suggested the use of an energy detector with unity quantum efficiency and the largest possible collection efficiency in order to achieve maximum sensitivity in an SXAPS spectrometer. The sensitivity can further be enhanced by geometrical arrangement of the detector, the electron beam and the sample.

APPLICATIONS

In this section, the application of APS to the study of surface phenomena will be discussed. The section is divided into three parts. In the first part, the elucidation of electronic structure of the surfaces of semiconductors and metals by APS is described with suitable examples. The second part deals with the phenomenon of adsorption of gases on metallic surfaces leading to the formation of compounds. The third and final part examines the determination of local structure of semiconductor surfaces from the tine structure observed on the high energy side of an appearance potential edge.

Electronic Structure of Semiconductors, Metals, and Semiconductor-Metal Interfaces

In this section we discuss the electronic structure of semiconductors, metals, and semiconductor-metal interfaces as determined by APS. First, we describe the APS studies on Si. Then we take the example of elemental Ti studied by DAPS, AEAPS, and SXAPS. After this, the SXAPS results on intermetallics TixNil_, (x=0,0.3,0.5, 0.7, 1.0) are discussed. These are included to give an idea of the types of information available from AI’S spectra. The reaction of silicon upon titanium deposition which is used as a metallization material in microelectronics is then discussed. The study also includes the effect of temperature on Ti-Si interface. Finally, the applications of APS for band


structure determination in some other semiconducting compounds are described.

The strength of the signal in APS depends upon the density of unoccupied states at EF of a material. Semiconductors and insulators are devoid of DOS at EF due to the location of Fermi level in the band gap. This led Tracy [45] to believe that either a low DOS above EF or a low fluorescence yield makes the APS spectra of semiconductors and insulators undetectable. However, commenting on this work, Park and Houston [4S] showed that the AI’S spectrum of Si was observable. Also, they found [49,50] the spectra of other semiconductors, like NiO, and insulators, like Sc203, to be much stronger than the corresponding spectra from the pure metals. Among the class of compounds yet to be fully investigated by APS are the semiconductors and insulators. Some work has been done on the fine structure following the appearance potential edge. This is discussed in the section on “Extended Appearance Potential Fine Structure.”

Nishimori et al. [51] have studied the Ti L3 threshold by DAPS, AEAPS, and SXAPS. It must be noted that the probing depth of DAPS is about half of that in AEAPS and SXAPS. The spectra taken in the first derivative mode are shown in fig. 6. The DAPS spectrum shows a double peak for the

7 level while the AEAPS and SXAPS spectra show a

shoulder on the ow energy side of the main %

peak. This shoulder is due to the surface state just above EF of Ti. T ey subtracted the SXAPS spectrum from the DAPS spectrum of the Ti L3 level in order to investigate the surface effect due to the different probing depths of the techniques. The resulting curve showed a strong peak followed by a weak peak. Comparison of this curve with the calculated linear DOS of the first layer of Ti (0001) film by Feibelman et al. [52] shows good agreement. Nishimori et al. have also observed the behavior of the Ti

s DAPS

spectra as a function of oxygen exposure at room temperature wrth a view to study the surface effects. The low energy peak disappeared at an oxygen exposure of 10 L while the second peak remained unaffected. This disappearance of low energy shoulder upon oxygen exposure confirms the existence of surface states. At higher oxygen exposures new shoulders appeared on the higher energy side of Ti L2 -peak corresponding to the formation of Ti-oxides.

,3

We have studied [53] by SXAPS the Ti and Ni L2 3 levels in Ti-Ni alloys of atomic composition TixNil_x (x = 0, 0.3, 0.5, 0.7, 1.0). An important aspect of SXAPS should be pomted out in the present context. Since the matrix element governing the core hole creation involves a very short range wave function of the initial core electron state, the technique is expected to reveal a localized DOS. Since the spectra of different


L3 Clean Ti

450 460 470

Primary electron energy ( eV >

Figure 6. DAPS, AEAPS, and SXAPS spectra of %,3-levels for polycrystalline Ti. Arrow indicates the low energy structure associated with the main peak. The figure is reproduced by permission from ref. [51].


constituents are well separated in energy, the application of SXAPS is by no means limited to binary alloys. The changes in SXAPS spectral features and shifts in BE which accompany alloy formation will better characterize the alloys. The I3 spectra of Ti and Ni consist of intense main peaks which exhibit no secondary structure. The amplitude of both the Ti and Ni I3 peaks decreases with increasing concentration of the other metal. This decrease can be ascribed to the decrease in the DOS at EF upon alloying. The Ni spectra exhibit an increasing negative-going peak with an increase of Ti concentration. This has been interpreted as due to the hybridization of Ti and Ni bands. The BE as determined by the intersection of the extrapolated projection of the background and low energy slope of the positive going peak, is found to increase for the I3 levels of both Ti and Ni in the alloys. The chemical shifts for Ti increase with Ni concentration while the Ni shifts show an opposite trend. These shifts have been explained on the basis of charge transfer and the Fermi level changes accompanying alloy formation. The alloy TiO 7Ni0 3 shows a shift of about 0.9 eV for the Ti L3 peak. According to ‘the ‘impurity model, charge transfer is not expected to play a significant role at this level of Ni. The overall shift in BE of Ti is primarily attributed to the change in EF With increasing concentration of Ni, the charge transfer effect should become increasingly important, and the BE of Ti should increase. This is in agreement with the observed chemical shit of Ti in Ti-Ni alloys the magnitude of which increases with an enrichment of Ni content. Similar conclusions may be drawn from the observed chemical shifts of Ni. In TiO 3Nio 7, Ti atoms act as an impurity in the Ni metal. The Ni shift of 1.3 eV in the alloy should approximately correspond to the shift in the Fermi level of Ni. With increasing Ti concentration, Ni shifts decrease, which again are consistent with the contribution of charge transfer. These results are in agreement with the magnetic measurements on the Ti-Ni system [54,55].

The full width at half maximum (FWHM) of Ti in Ti-Ni alloys progressively decreases with increasing concentration of Ni while that of Ni increases with increasing Ti concentration. Fuggle et al. [56] have assigned the conduction band narrowing in aluminum-noble metal alloys to the increase in the interatomic distances of the similar kind of atoms in the alloy. This interpretation does not seem to apply in the present case. If it were so, then both Ti and Ni widths should have decreased with the increasing concentration of the other constituents. The measurements are, therefore, consistent with the predictions of the common-band model. Other SXAPS (57) and soft x-ray absorption [58] studies of Ti-Ni alloys also suggest the formation of a common band.

The geometric structure and chemical composition of surfaces and interfaces is of fundamental importance for the initial growth of films


and for the bonding of the film to the substrate. In recent years, modern surface analysis techniques, such as AES, XPS and UPS, have been extensively employed to gain considerable insight into the occupied part of the valence bands. Since the unoccupied part of the valence bands also plays an equally important role in determining the characteristics of surfaces, APS could be effectively used for such a study. In semiconductor technology, studies on the interaction of silicon with transition metals have gained significant interest recently. Because of their high electrical conductance, these silicides are used in integrated circuit technology as Schottky barriers, ohmic contacts, and low-resistivity interconnects [59,60]. Considerable theoretical and experimental work has been done to understand the electronic and stoichiometric properties of these silicides and the associated silicide-silicon interfaces. It is these properties which determine the electrical conductance of silicides. The hybridization of Si 3s and 3p electrons with transition metal 3d electrons plays a significant role in determining the bonding and band-structure properties of the s&ides. Among these silicides, the formation of titanium silicide is of considerable interest because its resistivity is the lowest of all refractory metal s&ides and because it is compatible with metal-oxide semiconductor processing techniques. Investigations of thin overlayers of Ti deposited on Si have shown that the Ti/Si interface remains unreacted at room temperature [61-631.

Vahakangas et al. [64] have studied the layer-by-layer growth of Ti on Si by APS. Thin films of Ti were evaporated onto the Si substrate at the rate of l/3 of a monolayer per minute. The Si wafers used in this study had (100) and (111) orientations and were cleaned by two different processes. In one case, the sample substrate was cleaned by annealing at > 850’ C for several minutes to remove the oxide layer. In the other, the Si surface was heated briefly at 450’ C, then sputtered lightly with argon ions. It was then annealed at 850’ C to restore surface order. DAPS spectra of the Ti 2p levels were recorded as Ti was deposited on clean Si surfaces. The 2p312 p eak showed a continuous shift to higher energy to a maximum of - 1 eV with increasing Ti coverage. Moreover, the spectra broadened and a shoulder on the low-energy side of the main peak appeared as the Ti concentration increased. UPS studies (see refs. 5 and 12 in [64]) have shown that there is no Fermi level shift associated with the deposition of Ti on Si (100) or Si (111) surfaces. Therefore, the observed shift represents a real change in the local DOS (LDOS) above EF. The low energy shoulder observed in the DAPS spectra is characteristic of polycrystalline Ti and has been attributed to the surface states at EF of Ti. After the deposition of the first layer of Ti on Si, the surface state becomes visible in the Ti DAPS spectrum. This state then broadens and the bulk LDOS behavior is seen in the spectra at low coverages (> 10 monolayer) of Ti on Si. They have also confirmed the


4 '0

PRIMARY ELECTRON ENERGY in eV

Figure 7. DAPS spectra of Ti 2p levels for - 6 layers of Ti deposited on Si and annealed at five different temperatures. At T > 4OO’C the low energy shoulder to the main peak disappears and the extremum of the negative peak shifts towards higher energy. The figure is reproduced by permission from ref. [66].

surface nature of the shoulder by raising the lower limit of the quasielastic scattering yield, thereby increasing the probing depth of DAPS. The shoulder was found to become less pronounced as the lower limit was increased. From these observations Vahakangas et al. concluded that the growth of Ti on the Si surface is dependent upon the surface cleaning procedure. They also concluded that room temperature deposition of very thin films of Ti on Si leads to the formation of Ti islands on Si and that


there is no intermixing of Ti and Si beyond the first layer of Ti. Their conclusions are, however, not in agreement with those of Loenen et al. [65] who concluded that the first 3-4 Ti layers intermix with Si (111) surface atoms and form a layer with the composition TiSi.

Idzerda et al. [66] extended the above study on the Ti-Si system by observing the reaction of thin films of Ti with the Si substrate as a function of temperature. After depositing 10 layers of Ti on Si, the substrate temperature was raised from room temperature to 850’ C at intervals of approximately 50’ C. At each temperature interval, the sample was annealed for 10 min. The Ti 2~312 and 2plj2 DAPS spectra observed in this system at different temperatures are shown in fig. 7. The spectra obtained at temp. < 250’ C exhibit a low-energy shoulder apart from the main peaks. As stated earlier, this shoulder is characteristic of polycrystalline Ti. At 250’ C, the shoulder is reduced while the position and the overall shape of the spectrum remains unchanged. This reduction is due to the disruption of the surface state because of the diffusion of Si into Ti. At higher temperatures (400’ and 650’ C) the low-energy shoulder completely disappears and the extremum of the negative peak in the spectrum shifts towards higher energy. This behavior has been interpreted as due to changes in the empty LDOS above EP as indicated by the silicide reaction. The conclusions arrived at in this study are in agreement with those obtained with AES. The DAPS spectra clearly demonstrate the appearance and disappearance of surface states. Thus the extent of the Ti-Si reaction can be inferred directly from looking at the shape of the spectra. This is the unique advantage of APS over other techniques. In other techniques such a direct conclusion is not possible from observing the shape of the spectra.

Nilsson et al. [67j have studied the unoccupied states of the valence band of single crystal of V,Si by APS. This compound falls into the class of Al5 superconducting compounds with T, = 17K. They have compared the DOS above EP obtained by Mattheis (see ref. 10 in [671) with the APS spectrum calculated from these DOS data. The agreement between these two is fair up to about 3 eV above EF’ Beyond this the experimental spectrum gets broader than the theoretrcal spectrum. The difference is attributed to the secondary structure in the APS spectrum arising from the two electrons entering the high DOS region. The V L3 APS spectrum observed in this compound displays the peaks characteristic of the core excitation features. They .have compared this experimental spectrum with that derived from theoretical calculations and have determined the BE for V L3 in this compound to be 511.9 eV. Agreement between these two spectra is good in the region of the first maximum. However, the negative peak in the experimental spectrum occurs at an energy higher than that expected from the theoretical spectrum. They also


performed XPS studies on V3Si and found that the BE of the V L3 level as determined by XPS is in agreement within 0.3 eV with that determined by APS.

Webb and Williams [68] have studied the transition-metal dichalcogenides (TiS2, TiSe2 and VSe2) by APS. These crystals are made up of layers consisting of hexagonal sheets of metal atoms sandwiched between sheets of chalcogen atoms. In the APS L2 3 spectra of Ti and V in these compounds prominent splitting is observed. This doublet structure is due to the crystal-field-splitting of the d-like conduction band. In these compounds the metal is octahedrally coordinated by chalcogen atoms. In this environment the metal d band is expected to split into two subbands as predicted by band-structure calculations. The APS spectra were deconvoluted (in the case of TiS2 and TiSe2) to give the conduction band DOS. These were found to be in very good agreement with theoretical calculations. The crystal-field splitting obtained by these measurements is found to be 2.1 eV which is again in good agreement with those obtained from x-ray and synchrotron radiation absorption data. This study demonstrates that the deconvoluted APS spectra provide valuable information about crystal-field splittings.

Adsorption

The surface electron spectroscopies are powerful experimental methods for investigating the sorption process. This section deals with some of the important applications of APS to the study of the adsorption phenomenon of gases on transition metals.

The SXAPS spectra of metals when exposed to oxygen showed [69] different oxygen 1s spectra for chemisorbed oxygen and for oxygen in the oxide. Chemisorption is characterized by one single peak in the oxygen 1s spectra and no change in the metal 2p spectra. The oxidation phase is identified by changes both in the oxygen 1s and in the metal 2p spectra. Thus chemisorption and oxidation phases can be distinguished with the help of SXAPS. Nyberg [29] has studied by SXAPS the reaction of oxygen with evaporated films of Ti, Cr, Fe and Ni. In all cases both oxygen 1s and metal 2p spectra were recorded for different oxygen exposures of the metals. The signal strength of the oxygen 1s spectra continues to increase with increase in exposure. In a plot of the magnitude of the first peak in these spectra as a function of exposure, a change in the slope is observed in the case of Ti, Cr, and Fe at about 10 L exposure of oxygen while for Ni the change occurs at an exposure of about 20 L. This change indicates the formation of the oxide phase because the electronic structure of the material changes when an oxide starts to grow on the surface. The metal


2p spectra remain unchanged during the chemisorption stage (exposures before the change), and undergo appreciable changes during the oxide formation. As the oxide starts to grow on the metal surface, the 2p spectra exhibit a contribution from both the chemisorption and oxide stages. The oxide stage contribution dominates with increasing oxygen exposure and at large exposures (- lo3 L) only the oxide phase is seen. AES measurements performed on the same systems show a similar trend. In some cases, AES spectra do not show any change beyond a certain exposure of oxygen while the APS spectra continue to show additional changes. Also, the DOS effect is suppressed in AES when using large modulation voltages for taking the first derivative. This AES detection scheme also makes it difficult to distinguish clearly between the chemisorption stage and the oxide formation stage by simply looking at the shape of the spectra. Andersson and Nyberg [26] have also studied the chemisorption of C, N and S on the transition metals Ti, Cr, Fe, and Ni. They have interpreted the spectra in terms of the substrate-adsorbate complex. The FWHM of the peaks has been found to correlate with the variation in the width of the unfilled portion of the substrate 3d band. APS, thus provides a simple means for studying the kinetics of the reaction of gases with 3d-transition metals. The important advantage of the technique is the possibility of distinguishing clearly between the chemisorption stage and the oxide formation stage -- a feature not directly available by other techniques.

Konishi et al. [70] have studied the penetration of oxygen and nitrogen atoms into Ti surfaces as a function of exposure to the gases by AEAPS and DAPS techniques. Their measurements show that the DAPS L3 negative peak heights saturate at an oxygen exposure of 80 L, whiie the AEAPS peak heights continue to increase up to an exposure of 100 L. The Ti L3 peak heights of both DAPS and AEAPS spectra saturate at a nitrogen exposure of about 10 L. As stated earlier DAPS has higher sensitivity in the neighborhood of the surface than AEAPS. Since the back-scattered electrons in DAPS travel at least twice the path corresponding to the penetration depth, the probing depth in DAPS spectra is estimated to be about half of that in the AEAPS spectra. The differences in the DAPS and AEAPS spectral measurements on the same solid specimen should, therefore, give information about the depth distribution of the diffused gas atoms into the surface. Konishi et al. from their DAPS and AEAPS measurements concluded that the saturation values of the diffusion depths of nitrogen atoms in Ti thin films are much smaller than those of oxygen atoms. They arrived at this conclusion from the fact that the Ti L3 negative peak height is a measure of the degree of overlap of Ti 3d and gas (02 and N2) 2p wave functions.

We have also studied La and LaH3 by AEAPS [71] with the view


7r La

820 830 840 880

ENERGY bV)

Figure 8. M4 +evels AEAPS spectra of La in pure metal and in LaH3. Comparison of the two spectra shows the absence of the shoulder A to the main peak and the presence of two additional peaks (C and D) on the low energy side. See ref.

[711*

that this type of study would be helpful in understanding the hydrogen absorption process in La and also in LaNi5. The M4 5-level spectra are shown in fig. 8. The spin-orbit splitting in the case of La is found to be 16.5 eV. In the pure metal each spin-orbit level is seen to give rise to two structures : a shoulder-like peak (A) followed by a more intense peak which is accompanied by an undershoot. The peak A reflects the 3d1 6

B) -+

3dg4f1 excitation while peak B corresponds to 3d1’ + e + 3dg4f2 excitation [72]. In XPS 3d-level spectra of La, a similar structure on the low energy side of the main peak is observed [73,74]. This has been interpreted as due to the lowering of the 4f level in the presence of a 3d hole. Based on the screening mechanism, the low energy structure is shown to correspond to the well screen d

tid 3dg4f1) hole, whereas the main

peak is due to the poorly screened (3d 4 ) final state [75]. As the energy of the incident electron is increased, excitation of a 3d electron becomes possible. Since the 4f-levels are localized in the core region, they are


sensitive to their mutual repulsion and to the attractive Coulomb potential of the 3d hole. As a result, the 3d hole pulls the empty 4f level down 1.2 eV below EF. The 3d excitation then becomes possible which gives rise to peak A in fig. 8. Increasing the incident electron energy further, both the incident and the core electrons may go into the strongly localized 4f levels which, due to the 4f-4f repulsion are raised above EF giving rise to peak B shown in fig. 8. From these measurements, we estimate the position of the 4f level in this excited configuration of pure La as 1.5 eV above EF

In the AEAPS spectra of La in LaH3, also shown in fig. 8, three peaks C, D and B’ are observed in the M5 region. Comparison of these structures with those in the metallic spectrum shows the absence of peak A and the existence of two additional peaks C and D on the low energy side of the main peak B’. One of these additional peaks is due to the pulling down of the 4f level in the presence of a 3d hole, and the other is due to the hydrogen induced band. In the case of insulating La compounds, the XPS spectra exhibit a satellite on the high BE side of the main peak. This has been interpreted [76] as the lowering of the 4f level to a few eV above the ligand derived band. However, no high BE satellite is observed in the AEAPS spectra for LaH3. This implies that the model for insulating La compounds is not applicable to LaH . Therefore, in the case of LaH3, the 4f levels are pulled down below the 2 ydrogen induced band. On the basis of this, peak C is then assigned to the transition of the 3d electron to the pulled down 4f level below the hydrogen induced band and peak D to the transition of 3d electron to the hydrogen induced band. The intense peak B’ occurs when both the incident and the core electrons undergo resonant transition to the strongly localized 4f levels above EF From these measurements, we estimate the position of 4f level in the present excited configuration of La in LaH3 to be 4.1 eV above EF. Thus, AEAPS spectra are useful in determining the position of the 4f levels relative to EF in rare earth metals and their hydrides. Such a determination of the position of 4f levels is extremely important in understanding the applications, such as superconductivity and permanent magnetism, of rare earths in modern technology.

Extended Appearance Potential Fine Structure

In recent years the extended x-ray absorption fine structure (EXAFS) associated with an x-ray absorption discontinuity has proven to be a reliable technique for obtaining structural information about bulk materials [77-791. The information obtained includes the interatomic distance, the coordination number, and the degree of disorder. In contrast to the bulk information available from the EXAFS data, extended appearance potential fine structure (EAPFS) gives information about the


0 4 8 12 18 20

DISTANCE &,

Figure 9. The magnitude of the optical Fourier transform of the extended fine structure above Ll appearance potential edge of polycrystalline Cr for 1 = 0. The prominent peak is at 2.33 + 0.1 A. The nearest neighbor atomic spacing 2.53 + 0.1 A is in good agreement with that for bulk values. The figure is reproduced by permission from ref. [87].

surface structure since it utilizes a low energy electron beam for excitation. The fine structure is obtained by recording the APS spectra as a function of primary electron energy [80]. The structure occurs on the high energy side of a particular core level appearance potential edge and extends over several hundred eV. This structure appears as a result of a spherical wave emanating from the central atom, modified by weak backscattering from neighboring atoms. Recent work has shown EAPFS to be analogous to EXAFS [81-831. Therefore, the analysis of EAPFS is carried out in much the same way as for EXAFS. After the background subtraction, the fine structure x(k) is plotted as a function of free particle momentum of the excited core electrons via the equation:

lik = [ 2m(E - Eo)]1/2.

Here E corresponds to the energy of the primary electron, and E, is the BE of the core level involved, Ii is h/2?r (h is Plan&s constant), k is Boltzmann’s constant, and m is the mass of the electron. The Fourier transform of x(k) yields a radial structure function:

s k max

F(r) = ti x(k) eq3(-2ikr) dr. (10) k min


The nearest neighbor distance of the surface constituents is easily obtained from the prominent peak position after appropriate corrections for phase shift and multiple scattering effects as in EXAFS [84]. These phase shifts are due the interference of the waves, scattered from the central atom and backscattered from the surrounding atoms. These depend on the angular momentum of the outgoing electron wave function. In EXAFS, photon excitation obeys the dipole selection rule. This greatly simplifies the determination of the angular momentum and hence of the appropriate phase shifts. In EAPFS, however, the major question in the analysis is the angular momentum of the two final state electrons. The knowledge of the phase shifts, which is a function of the angular momentum of the electrons, determines the accuracy with which the nearest neighbor distances can be determined. In their model calculations, Mehl et al. [85] have suggested a pseudodipole excitation rule for EAPFS, i.e., the data can be analyzed with the same phase shifts as would be used for EXAFS. Using this proposed model for the K edge of Al and L3 edge of Ti, they have determined the distance of the first shell from the origin within + 0.04 A for Ti and z!z 0.01 A for Al. This success encourages EAPFS to be used as a viable tool in surface measurements since theoretical phase shifts for all the elements have been calculated [86] and successfully employed in EXAFS analysis.

Konishi et al. [87] have measured the EAPFS spectra for polycrystalline Cr using AEAPS. The fine structure was recorded in the second derivative mode. In fig. 9 is shown the magnitude of the Fourier transform of the fine structure obtained by them above the Ll appearance potential edge of polycrystalline Cr. Using the phase shifts calculated by Teo and Lee [86], they determined, in the case of Cr, the nearest neighbor atomic spacing as 2.53 A (1 = 0), 2.54 A (1 = 1) and 2.54 A (1 = 2) in good agreement with bulk value (2.50 A).

Einstein et al. [88] have studied by EAPFS the oxidation of a Si (100) surface when exposed to air to form a saturated SiO layer. Because of the difficulty in observing DAPS spectrum due to di fraction effects, T they have used the SXAPS technique to obtain the fine structure. The fine structure associated with the oxygen K-edge as observed by them is shown in fig. 10. The spectrum shows a peak at 535 eV with some oscillatory structure on the high energy side of this peak. This structure extends up to 450 eV above the edge. The Fourier transform of this data, shown in fig. 11, is found to consist of two well-defined peaks corresponding to Si-0 and O-O nearest neighbor bonding. These peaks are shifted from the actual values by an amount equal to the correction due to the chemical environment. Taking into account the above correction, they have extracted parameters for nearest neighbor distances in Si02 and found


500 1000 Primary Energy (eV)

Figure 10. The fine structure associated with oxygen K-edge in Si02. The first derivative of X-ray yield is plotted as a functron of primary energy. The figure is reproduced by permission from ref. [88].

them to be in good agreement with those reported in the literature.

Terauchi et al. [89] have studied the fine structure associated with the Ga-Auger peak of a GaAs(001) surface. They have observed the AEAPS spectrum using a cylindrical mirror analyzer. Figure 12 shows the EAF’FS of Ga in GaAs as obtained by them. The oscillations on the high energy side are clearly visible.. The whole spectrum was divided into several regions. Each region was fitted with a third-order polynomial, the constraint upon the polynomials was to have equal slopes at the spline point. The fine structure x(k) was obtained with the help of eqn; (9). This function x(k) is shown in fig. 13. The Fourier transform of this function obtained with the help of eqn. (10) is shown in fig. 14. They have used a Harming window function [90] to minimize the effect of terminating k at high values. Since eqn. (10) involves both the real and imaginary parts, in fig. 14 is shown the imaginary part along with the magnitude of the


_I Si-0

6 8

Spacing (11

Figure 11. Optical Fourier transform of the data in Fig. 10. The peaks correspond to the near neighbors of oxygen in SiO2 The figure is reproduced by permission from ref. [88].

Fourier transform. The peaks in the transform correspond to different near neighbors of the central atom (here Ga) and appear to be shifted from the actual values by an amount determined by the phase shift. Applying this correction they have determined for the first time the near neighbor distances for the GaAs (001) surface to be 2.77 A. Based on these values they have constructed the surface structure of GaAs (001). This structure is consistent with those found by other studies (see refs. 10 and 11 in [SS]). These studies prove EAPFS to be a powerful and relatively simple technique for surface structure determination of semiconductors.

The major complication in EAF’FS is diffraction oscillations due


GaAs(OO1) Ga-Auger

a

zi lo-

2 1 1 I

1.18 1.34 1.50 1.66 1.82

E (keV)

Figure 12. Extended appearance potential fine structure observed from the Ga-Auger peak in GaAs (001) surface as a function of incident electron energy E. The figure is reproduced by permission from ref. [89].

5.6 7.4 9.0 10.6

k (l/b 12.2 13.6

Figure 13. The EAPFS function x(k) after the removal of the background from the data in fig. 12. The figure is reproduced by permission from ref. [89].


Figure 14. Extended appearance potential fine structure observed from the Ga-Auger peak in GaAs (001) surface as a function of incident electron energy E. The figure is reproduced by permission from ref. [89].

to elastically scattered electrons from ordered regions of the sample. This can be overcome by monitoring soft x-ray emission during the core de- excitation. However, for the soft x-ray levels, the probability for this SXAPS process is less than one percent. It is, therefore, important to detect the x-rays with high quantum efficiency in order to minimize the effects due to the scattering of incident electrons. High quantum efficiency has been achieved with a nude solid state surface barrier detector consisting of a < 111~ oriented Si crystal coated with a thin Al layer and cooled by liquid nitrogen [46]. Such a detector has been utilized by Einstein et al. [SS] in their study.

EAPFS has been found to achieve good signal-to-noise ratio up to at least 11 A-’ i.e., approximately 500 eV above the threshold, the upper limit of typical EAPFS measurements [91]. On the other hand, surface EXAFS (SEXAFS) signals are relatively weak to be resolved above about 8 A-' [92]. EAPFS has also been successful in providing structural information on the oxidation of Al [Sl] and Ni [93] where the LEED


patterns were found to be extinguished upon exposure of metal surfaces to oxygen. EAPFS is produced by backscattering from close neighbors of the excited atom and thus probes only very short-range order. It is, therefore, ideally suitable for systems lacking long-range order. Also, EAPFS can obtain fully adequate signal strength from a relatively thin adsorbate layer in contrast to EXAFS. Thus, it probes features distinctly characteristic of the surface region. EXAFS experiments require high intensity x-ray sources such as those available from rotating anode tubes or synchrotron facilities. These are not commonly available. EAPFS experiments use equipment which is commonly accessible in most surface science laboratories thus making it a widely available option for fine structure measurements.

CONCLUSIONS

In the previous sections we have discussed the interesting applications of APS to modern science and technology and highlighted its importance relative to other techniques available. In this section, we discuss the relative strengths and limitations of APS.

The intensity of a signal in an APS spectrum is determined by the transition of a core electron to the unoccupied states above E . The strength of the signal, therefore, depends upon the density o F these unoccupied states. For simple metals, the 3d transition metals, the rare earth metals, etc., having high density of unoccupied states at EF, this technique is particularly suitable for their study. Noble metals, such as, Cu and Au, having very low DOS, give a very weak signal in the spectrum. This is the reason why AI’S cannot be used as a common analytical tool. For the elements to which it is sensitive the spectra are much simpler and easier to interpret than those obtained from other techniques. Moreover, APS is a non-dispersive technique and requires only relatively simple and inexpensive instrumentation. These are the special advantages of APS over other techniques.

The one-electron theory discussed previously explains satisfactorily the features observed in the spectra of simple and 3d transition metals. The theory is valid for systems having continuous DOS above EF Discrepancies between theory and experiment were observed for rare earths, light elements and 4d transition metals. The breakdown of the theory for these materials is due to the inadequacy of the assumption that the incident and/or the excited core electrons occupy spatially extended states in the conduction band. For example, in the rare earth metals the 4f orbitals are localized making the excited core electron transition to one of these orbitals possible. More theoretical work, taking


into account the core-level widths, core-hole lifetime broadening, many- body and other effects contributing to the spectrum, is needed to provide a more plausible explanation for the AI’S spectra.

An interesting application of AI’S is the derivation of the DOS from the signal shape. Results for 3d and 5d metals have been found in excellent agreement with theoretical calculations. APS is found more suitable to the study of intermetallics, especially to multicomponent systems. Another important application of AI’S is the study of adsorption phenomenon. The AI’S investigation of the adsorption of oxygen on certain metal surfaces has shown that a change begins to appear in both the oxygen and metal spectra at the threshold of oxide formation. Thus, it is possible to distinguish the adsorption and oxide formation stages directly from the spectra. In other techniques, e. g., as in AES, such a direct result can not be obtained by simply examining the shape of the spectra.

The fine structure (known as extended appearance potential fine structure, EAPFS) occurs on the high energy side of a particular core level appearance potential edge and is analogous to the bulk phenomenon EXAFS for obtaining structural information. The analysis of EAPFS can be done on similar lines as with EXAFS. Unlike EXAFS, which requires high intensity sources such as those from synchrotron or rotating anode tubes, EAPFS makes use of an experimental set-up that can be easily fabricated in any surface science laboratory. Moreover, the structural information obtained from EAPFS pertains only to the surface. Thus EAPFS is a suitable alternative to EXAFS for fine structure measurements. The theoretical data available are suitable for the analysis of the K edge spectra of the elements. Additional theoretical data are necessary for the analysis of the L edge spectra. The diffraction oscillations due to elastically scattered electrons from single crystals pose a complication in EAPFS. This can be overcome by using SXAPS. However, to enhance the signal-to-noise ratio in SXAPS high quantum effkiency detectors may be used.

For surface studies it is necessary to prevent changes of the surface resulting from heating due to primary electrons. This means that low primary currents should be used. For greater signal-to-noise ratio SXAPS requires large current and hence can be used in cases where surface changes are not expected to occur due to heating of the sample. AEAF’S, on the other hand, uses lower primary currents. Also, the probability of the Auger process is approximately two orders of magnitude higher than the probability of x-ray emission in the O-2000 eV energy range of the analysis. Thus AEAPS is more sensitive and is commonly used for surface analysis. It must be noted that the structures observed in


SXAPS and AEAPS spectra may differ largely because of the core hole decay mechanisms following the excitation of the core electrons in these spectroscopies. In SXAPS X-ray emission is slow and core hole production and de-excitation are only weakly coupled. On the other hand, in AEAPS Auger decay is fast and the excitation and decay are strongly coupled. This may lead to some broadening of structure in AEAPS. Dose et al. [94] have observed in solid Ni the smearing of the threshold slope and structure in the AEAPS spectrum as compared to the SXAPS spectrum. APS is, however, not limited to solid metals only. With proper experimental arrangement it could be extended to the study of liquid metals, as has been demonstrated by Dose et al. [94].

The major problem in APS is the signal-to-noise ratio for the analysis of some elements. However, better sensitivity could be achieved, in the case of SXAPS, by using energy detectors with high quantum efficiency and by geometric arrangement of the different components of the spectrometer. The disadvantage of low signal-to-noise ratio does not limit the use of APS. It has been observed that APS has important advantages unique to this threshold spectroscopy as compared with other surface sensitive techniques, especially with respect to the most widely used AES [95]. In AES, the backscattering contribution to the signal intensity as a result of energetically scattered secondary electrons can introduce serious distortions which complicate quantitative analysis and microanalysis. These distortions do not exist at the excitation threshold and are, therefore, absent in SXAPS, AEAPS and DAPS, making these spectroscopies more adaptable for the study of surfaces.

At the present time, experience with this technique is very limited. Lack of sufficient theoretical data has restrained plausible interpretation of the relevant information obtained from the experimental data. The type of information provided by APS adequately compliments that obtained from other modern surface sensitive techniques, such as XI’S, UPS, AES, etc. However, theoretical and experimental work in this area is constantly expanding to understand more elucidly the different aspects of APS. Full exploration of APS will then allow thii spectroscopy to be accepted as a popular technique for better characterization of material surfaces.

Glossary of Symbols

AEAPS Auger Electron Appearance Potential Spectroscopy AES Auger Electron Spectroscopy APS Appearance Potential Spectroscopy BE Binding Energy BIS Bremsstrahlung Isochromat Spectroscopy

DAPS DOS

EF EAPFS EXAFS FWHM IPE LDOS LEED RPAPS

SEXAFS SXAPS UHV UPS XAS XPAPS XPS


Disappearance Potential Spectroscopy Density of States Fermi level, Fermi energy Extended Appearance Potential Fine Structure Extended X-ray Absorption Fine Structure Full Width at Half Maximum Inverse Photoemission Localized Density of States Low Energy Electron Diffraction Resonant Photoelectron Appearance Potential Spectroscopy Surface Extended X-ray Absorption Fine Structure Soft X-ray Appearance Potential Spectroscopy Ultrahigh Vacuum Ultraviolet Photoelectron Spectroscopy X-ray Absorption Spectroscopy X-ray Photoelectron Appearance Potential Spectroscopy X-ray Photoelectron Spectroscopy

ACKNOWLEDGMENTS

The authors gratefully acknowledge the financial support from the Robert A. Welch Foundation, the Texas Advanced Technology Research Program, and the American Chemical Society, PRF. The authors wish to thank Darrell Beauchamp, Rex M. Giddens, and Mona Towne for their assistance in the final phases of this work. Thanks are also due to the publishers/authors for their permission to use the figures in this article.

REFERENCES

1. D. R. Chopra and A. R. Chourasia, Scanning Micros. 2, 677-702 (1988).

2. R. L. Park and J. E. Houston, J. Vat. Sci. Technol. ll, l-18 (1974).

3. W. E. Harte, P. S. Szczepanek, and A. J. Leyendecker, J. Less Common Metals 93,189-200 (1983).

4. F. U. Hillebrecht, J. C. Fuggle, G. A. Sawatzky, M. Campagna, 0. Gunnarsson, and K. Schonhammer, Phys. Rev. B 30,1777-1787 (1984).

5. D. van der Marel, G. A. Sawatzky, and J. C. Fuggle, Solid State Commun. 50,47-50 (1984).


6. W. Speier, J. C. Fuggle, R. ZeIler, B. Ackermann, B., K. Szot, F. U. Hillebrecht, and M. Campagna, Phys. Rev. B 30,6921-6930 (1984).

7. J. B. Pendry, Phys. Rev. Lett. 45,1356-1358 (1980).

8. F. J. Himpsel, T. Fauster, J. Vat. Sci. Technol. A 2,815821(1984).

9. W. Drube, F. J. Himpsel, and R. Ludeke, J. Vat. Sci. Technol. B 5, 930-932 (1987).

10. L. V. Azaroff and D. M. Pease, in X-ray Spectroscopy, edited by L. V. Azaroff, (McGraw-Hill, New York, 1974) pp. 284-337.

11. J. E. Muher and J. W. Wilkins, Phys. Rev. B 29,4331-4348 (1984).

12. C. Mande and V. B. Sapre, in Advances in X-ray Spectroscopy, Edited by C. BonnelIe and C. Mande (Pergamon Press, New York, 1982) pp. 287-301.

13. R. L. Park, J. E. Houston, and D. G. Schreiner, Rev. Sci. Instrum. 41, 1810-1812 (1970).

14. R. L. Gerlach, J. E. Houston, and R. L. Park, Appl. Phys. Lett. 16, 179-181(1970).

15. J. E. Houston and R. L. Park, Phys. Rev. B 5,3808-3809 (1972).

16. J. Kirschner and P. Staib, Appl. Phys. 6,99-109 (1975).

17. J. Kanski and P. 0. Nilsson, Phys. Scripta 12,103-112 (1975).

18. S. Kato, R. Konishi, and S. Mogami, Jpn. JAppl. Phys. 18, 835-836 (1979).

19. Z. Hua, J. Zhuge, and X. Pan, Chin. J. Sci. Instrum. 3,10-17 (1982).

20. 0. W. Richardson and C. B. Bazzoni, PhiIos. Mag. 42, 1015-1019 (1921).

21. G. Shinoda, T. Suzuki, and S. Kato, Phys. Rev. 95,840-841(1954).

22. R. J. Liefeld, BuII. Amer. Phys. Sot. 12,562 (1967).


23. R. L. Park and J. E. Houston, Surf. Sci. 26,664-666 (1971).

24. R. J. Smith, M. Piacentini, J. L. Wolf, and D. W. Lynch, Phys. Rev. B 14,3419-3431(1976).

25. V. Dose and A. Haertl, Phys. Rev. Lett. 47,132-134 (1981).

26. S. Andersson and C. Nyberg, Surf. Sci. 52,489-504 (1975).

27. C. Nyberg, Surf. Sci. 82,165176 (1979).

28. G. Ertl and K. Wandelt, Z. Naturforsch. 29a, 768-772 (1974).

29. C. Nyberg, Surf. Sci. 52,1-9 (1975).

30. G. E. Laramore, Phys. Rev. B l&5254-5264 (1978).

31. B. Dev and H. Brinkman, Ned. Tijdsch. Vacuum. 8176-184 (1970).

32. J. E. Houston and R. L. Park, J. Chem. Phys. 5546014606 (1971).

33. A. M. Bradshaw, in Surface and Defect Properties of Solids (Chemical Society, London, 1974) Vol. 3, pp 153-183.

34. Y. Fukuda, W. T. Elam, and R. L. Park, Phys. Rev. B 16, 3322-3329 (1977).

35. V. Dose and T. Fauster, Appl. Phys. 20,299-303 (1979).

36. V. Dose, T. Fauster, and H. J. Gossman, J. Comp. Phys. 41, 34-50 (1981).

37. S. W. Schulz, K. T. Schleidcher, D. M. Ruck, and H. U. Chun, J. Vat. Sci. Technol. A 2,822~825 (1984).

38. D. Chopra, H. Babb, and R. BhaIIa, Phys. Rev. B 14, 5231-5236 (1976).

39. D. L. Grolemund and D. Chopra, IEEE TRans. Nucl. Sci. NS30,934- 936 (1983).

40. L. Papagno and R. Scarmozzino, Thin Solid Fiis 70,249-252 (1980).

41. M. Euler, Eur. J. Phys. 1,18-21(1980).


42. L. Eckertova and J. Pavluch, Czech. J. Phys. B 34,622-634 (1984).

43. J. Pavluch and L. Eckertova, Czech. J. Phys. B 35,630-642 (1985).

44. J. Kirschner and W. Losch, J. Vat. Sci. Technol. 14,1173-1179 (1977).

45. J. C. Tracy, J. Appl. Phys. 43,4164-4171(1972).

46. S. Andersson, H. Hammarqvist, and C. Nyberg, Rev. Sci. Instrum. 45, 877~881(1974).

47. R. N. Lee, Rev. Sci. Instrum. 48,1603-1609 (1977).

48. R. L. Park and J. E. Houston, J. Appl. Phys. 44,3810-3811(1973).

49. R. L. Park and J. E. Houston, J. Vat. Sci. Technol. 10,176-182 (1973).

50. R. L. Park and J. E. Houston, in Electron Spectroscopy edited by D. A. Shirley (North Holland, Amsterdam, 1971) pp 895-901.

51. K. Nishimori, H. Tokutaka, M. Kohno, and N. Ishihara, Jpn. J. Appl. Phys. 23, L366-L368 (1984).

52. P. J. Feibehuan, J. A. Appelbaum, and D. R. Hamann, Phys. Rev. B 20,1433-1443 (1979).

53. T. K. Hatwar and D. Chopra, Surf. Interface Anal. 7,93-96 (1985).

54. I. P. Gregory and D. E. Moody, J. Phys. F 5,36-44 (1975).

55. W. S. Ghan, K. Mitsouka, H. Miyajima, and S. Chikazumi, J. Phys. Sot. Jpn. 48822-829 (1980).

56. J. C. FuggIe, L. M. Watson, D. J. Fabian, and P. R. Norris, Solid State Commun. 13,507-510 (1973).

57. J. E. Houston and R. L. Park, J. Vat. Sci.Technol. 9,579-583 (1971).

58. J. R. CuthiII, A. J. McAIister, and M. L. Williams, J. Appl. Phys. 39, 2204-2208 (1968).

59. S. Zirinsky, W. Hammer, F. d’Heurle, and J. BagIin, Appl. Phys. Lett. 33,76-78 (1978).

60. S. P. Murarka, J. Vat. Sci. Technol. 17,775-792 (1980).


61. M. A. Taubenblatt and C. R. Helms, J. Appl. Phys. 53, 6308-6315 (1982).

62. R. Butz, G. W. Rubloff, T. Y. Tan, and P. S. Ho, Phys. Rev. B 30, 5421-5429 (1984).

63. W. Yang, H. Iwakura, H. Yagi, T. Kuroda, and S. Nakamura, Jpn. J. Appl. Phys. 23,1560-1567 (1984).

64. J. Vahakangas, Y. U. Idzerda, E. D. Williams, and R. L. Park, Phys. Rev. B 33,8716-8723 (1986).

65. E. J. van Loenen, A. E. M. J. Fischer, and J. F. van der Veen, Surf. Sci. X5,65-78 (1985).

66. Y. U. Idzerda, E. D. Williams, R. L. Park, and J. Vahakangas, Surf. Sci. 177, L1028-L1034 (1986).

67. P. 0. Nilsson, I. Curelaru, and T. Jarlborg T., Phys. Stat. Sol. b 79, 277-281(1977).

68. C. Webb and P. M. Wiiams, Phys. Rev. B l&2082-2086 (1975).

69. S. Andersson and Nyberg C., Solid State Commun. 15, 1145-1148 (1974).

70. R. Konishi, Y. Miyada, and H. Sasakura, Jpn. J. Appl. Phys. 24, 923- 927 (1985).

71. A. R. Chourasia and D. R. Chopra, J. Electron Spectrosc. Rel. Phenom. 43,233-241(1987).

72. G. Wendin and K. Nuroh, Phys. Rev. Lett. 39,48-51(1977).

73. J. Osterwalder, Z. Phys. B 61,113-128 (1985).

74. L. Schlapbach and H. R. Scherrer, Solid State Commun. 41, 893-897 (1982).

75. J. C. Fuggle, M. Campagna, Z. ZoInierek, R. Lasser, and A. Plateau, Phys. Rev. Lett. 45,1597-1600 (1980).

76. G. Crecelius, G. K. Wertheim, and D. N. E. Buchanan, Phys. Rev. B l&6519-6524 (1978).


77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

91.

92.

A. R. Chourasia, V. D. Chafekar, S. D. Deshpande, and C. Mande, Pramana 24,787-795 (1985).

P. A. Lee, P. H. Citrin, P. Eisenberger, and B. M. Kin&d, Rev. Mod. Phys. 53,769-806 (1981).

E. A. Stern, D. E. Sayers, and F. W. Lytle, Phys. Rev. B 11,4836-4846 (1975).

R. L. Park, Surf. Sci. 86,X%515 (1979).

M. L. den Boer, T. L. Einstein, W. T. Elam, R. L. Park, L. D. Roelogs, and G. E. Laramore, Phys. Rev. Lett. 44,496-500 (1980).

G. E. Laramore, Surf. Sci. 81,43-56 (1979).

G. E. Laramore, T. L. Einstein, L. D. Roelofs, and R. L. Park, Phys. Rev. B 21,2108-2121(1980).

E. A. Stern, Phys. Rev. B 10,3027-3037 (1974).

M. J. MehI, T. L. Einstein, and G. W. Bryant, J. Vat. Sci. Technol. A 2,862-863 (1984).

B. K. Teo and P. A. Lee, J. Am. Chem. Sot. 101,2815-2832 (1979).

R. Konishi, H. Tanigawa, and H. Sasakura, Jpn. J. Appl. Phys. 25, 1616-1617 (1986).

T. L. Einstein, M. L. den Boer, J. F. Morar, R. L. Park, and G. E. Laramore, J. Vat. Sci. Technol. l&490-491 (1981).

H. Terauchi, S. Sekimoto, N. Sano, H. Kato, and M. Nakayama, Appl. Phys. Lett. 46,148-149 (1985).

G. H. Via, J. H. Sinfelt, and F. W. Lytle, J. Chem. Phys. 71, 690-699 (1979).

J. F. Morar, and R. L. Park, J. Vat. Sci. Technol. A 1, 1043-1046 (1983).

J. Stohr, L. I. Johansson, S. Brennan, M. Hecht, and J. N. Miller, Phys. Rev. B 22,4052-4065 (1980).


93. M. L. den Boer, T. L. Einstein, W. T. Elam, R. L. Park, L. D. Roelofs, and G. E. Laramore, J. Vat. Sci. Technol. 17,59-62 (1980).

94. V. Dose, R. Drube, and A. Hartl, Solid State Commun. 57, 273-275 (1986).

95. J. Kirschner and W. Losch, J. Vat. Sci. Technol. 14,1173-1179 (1977).

Date post:	09-Jul-2016
Category:	Documents
Upload:	rimoal
View:	219 times
Download:	0 times

CHOPRA -Characterization of Semiconductor Surfaces by Appearence Potential Spectroscopy

Documents