JUNE 1983 Vol. 108 No. 1287 The Analyst Auger Techniques in Analytical Chemistry A Review J. C. Rivisre Materials Develo9ment Division AERE Harwell Oxfordshire OX11 ORA Summary of Contents Introduction Historical Auger effect Chemical information Surface specificity and vacuum requirement Electron energy analysis Electron sources Data acquisition and handling First principles Elemental standards Auger peaks Satellite peaks Ionisation loss peaks Secondary electron peak Analysis in depth Angular dependence of Auger intensity Erosion by ion bombardment Erosion by mechanical lapping Technique and instrumentation Quantification Spectral identification Chemical bonding information Applications Corrosion and oxidation Catalysis Reactions in the solid state Analyses using high spatial resolution Adhesion Analysis in depth Conclusions Advantages and disadvantages of AES Future developments Comparison with other surface analytical techniques References Keywords ; Review ; Auger electron spectroscopy ; surface analysis Introduction Historical Although the Auger effect was discovered1 as long ago as 1925 during experiments in a cloud chamber in which X-radiation was used as the ionising source the first report of stimulation of Auger emission from solids by incident electrons did not appear until 1953.This was by Lander,2 who can justly be regarded as the founder of the modern technique of Auger electron spectroscopy (AES). At that time however both lack of sensitivity and inadequate vacuum capability limited the development of the technique and the possibilities lay dormant 64 650 R I V I ~ R E AUGER TECHNIQUES IN Analyst Pol.108 until 1967 although in 1958 an isolated application was published by Powell et aZ.,3 who used it to monitor carbon contamination on tungsten. By 1967 production of ultra-high vacua had become routine and in that year two groups of workers Scheibner and Tharp4 and Palmberg6 showed that the sensitivity problem could also be reduced significantly by using the spherical-grids-and-screen arrangement already in use for LEED as a retarding-field energy analyser. In the same year it was demonstrated by Harris6 and by Weber and Peria' that visual enhance-ment of small or badly resolved Auger features could be achieved simply by electronic differen-tiation of the energy distribution N ( E ) of secondary electrons to give dN(E)/dE.Most Auger spectra are still recorded in this form. It soon became obvious that retarding-field analysers (RFA) suffered from poor signal to noise characteristics representing a serious limitation on the development of the technique, and attention was turned to the possibilities of dispersive analysers. In 1969 Palmberg et aZ.* showed that the cylindrical mirror analyser (CMA) had properties that made it emi-nently suitable for AES (see Technique and Instrumentation) and in a short space of time the CMA had almost completely superseded the RFA. Nowadays the only situations in which an RFA is used for AES are those in which the experimentalist wishes to perform LEED also.The CMA is available commercially in a variety of designs and where AES alone is to be performed is likely to be the accepted analyser for the foreseeable future. On the other hand, where both AES and X-ray photoelectron spectroscopy (XPS) are combined in the same instrument the preference is for a concentric hemispherical analyser (CHA) in which better energy resolution is obtainable. The basic technique of AES has not changed since 1967 in as much as differentiation of the spectra is still almost universally employed but of course there have been continual refinements and improvements to improve the quality of the spectra. The greatest changes since the early days have been in the interpretation of the spectra in the understanding of the physical pro-cesses involved in the production and ejection of an Auger electron and in the ever widening field of application of the technique.This review will endeavour to describe these and other aspects of AES with particular emphasis on those features of relevance to analytical chemistry. Auger Effect Inter-action with an incident electron creates a hole in a core atomic binding level that is causes ionisation ; the level shown ionised in the example is the K or Is. Relaxation of the atom back towards its ground state occurs by the filling of the core level with an electron from an outer level. The excess of energy represented by the difference in the binding energies can then cause emission either of a photon of characteristic energy or of an Auger electron. The two processes The Auger effect taking place in an isolated atom is shown schematically in Fig.1 (a). t (a) VAC v.0. either Fig. 1. (a) Schematic diagram of competing Auger and X-ray emission processes in an isolated atom following ionisation of a core level and (b) schematic diagram of competing Auger and X-ray emission processes in a solid following ionisation of a core level Jwae 1983 ANALYTICAL CHEMISTRY. A REVIEW 65 1 compete but for core level binding energies of less than about 2000 eV the probability of Auger emission is close to unity. As can be seen in Fig. 1 (a) the Auger electron also comes from an outer level which may or may not be the same as the de-exciting outer level. In the specific example shown the de-exciting level is the L and the emitting level the L2,3 and the Auger transition depicted would thus be designated the KL,L, according to convention.Where spin - orbit splitting cannot be resolved in any sub-shell of any particular atom e.g. as be-tween L and L3 in Fig. l ( a ) then the convention combines the levels together as for instance, L,,,. The simple notation based on X-ray spectroscopy is widely used but becomes inade-quate in those instances in which the two holes in the final doubly ionised state cannot decay independently but remain together sufficiently long for coupling between them to occur. This leads to fine structure appearing in the Auger spectrum interpretable in terms of final state spectroscopic terms. When atoms combine to form a solid atomic energy levels shift and broaden to form energy bands and in particular the outermost levels containing the valence electrons go to form the valence band.Auger transitions can take place just as readily in atoms in the condensed phase as in the gaseous phase and those transitions which involve electrons in the valence band are often the most intense in the spectrum. Such a transition involving an electron from the valence band of a solid is shown diagrammatically in Fig. l ( b ) . The designation of that transition would be KL,V; if both the de-exciting and emitted electrons had originated in the valence band it would have been designated KVV. Once an Auger electron has been ejected either from an atom or from the surface of a solid, its energy can be measured. Reference to Fig. l(a) shows that the energy is equal to where ABC is the Auger transition being considered where Ei are the binding energies of the levels i in the atom in a singly ionised state and where El’ is the binding energy of level i in a doubly ionised state.For Ei binding energies measured by X-ray or photoelectron methods can be used but for Ei’ it is necessary to consider the effect on the electrons in outer orbitals of the atom due to the presence of the additional hole or positive charge. The effect is to give the outgoing Auger electron some additional kinetic energy arising from the relaxation of the outer orbitals towards the hole in order to provide additional screening. The contribution of this atomic relaxation energy to the Auger energy was first pointed out by S h i r l e ~ ~ who showed that it could be large of the order of 10-20 eV.When the ionised atom is not isolated but forms part of a solid an additional relaxation energy appears called the extra-atomic relaxation energy; this is gained from the shift of electrons in immediately adjacent atoms towards the positive charge. Relaxation energies and therefore Auger energies can now be calculated with great accu-racy; the reader is referred to the compilation of such energies by Larkinslo as an example and also as a manual. It will be clear from the above that each element in the Periodic Table (except H He and atomic Li) has a unique Auger spectrum as no two elements have the same set of binding energies and that analysis of the energies of the Auger electrons therefore pro-vides a means of elemental analysis.Even when an element is in solid form or in combination with other elements in a solid the fact that the binding energy of the initially ionised core level is the dominant term in equation (1) allows unambiguous elemental identification. Chemical Information The combination of one atom with another causes changes in the electron density surround-ing the atom that is changes in the electron binding energies. These changes are most pro-found in the outermost or valence electrons but inner levels can be affected too to a lesser degree. Direct observation of these “chemical shifts” forms the basis of the technique of X-ray photoelectron spectroscopy (XPS) but in AES it should be obvious from Fig. 1 that chemical effects will be much harder to interpret as there is an additional ionisation in the final state of the atom.In general the shifts observed in AES have been used in a qualitative “fingerprinting” way without much effort to understand their precise origin. Some of these shifts can be very large greater than any seen in XPS; e.g. from Si to SiO reported values range from 10 to 15 eV from A1 to Al,03 -13 eV and from Mg to MgO -11 eV. Amongst th 652 RIVIERE AUGER TECHNIQUES IN Analyst Vol. 108 transition metals of the first series the shifts are lower between 1 and 4 eV. For a comprehen-sive tabulation of the shifts published in the literature the reader is referred to the paper by Madden .ll The other source of information on chemical changes in the vicinity of an atom comes from study of the shape of an Auger peak based on a transition in which either one or both electrons originate in the valence band.Again the changes observed in Auger peaks can and have been used in a purely “fingerprinting” way and indeed for carbon have been so used from the earliest days of AES. Among the observable spectral changes are the presence or absence of fine structure components and changes in their widths and relative intensities. However al-though the use of line-shape changes in such a qualitative way provides some chemical informa-tion more recent developments seek to derive fundamental knowledge about the local density of electronic states around individual surface atoms. This is achieved by unravelling or self-deconvoluting spectra in which both electrons originate in the valence band because of course each electron carries the same information about conditions in the band.Because the Auger process is atomic in nature the self-deconvolution then produces a density of states appropriate to the immediate surroundings of the ionised atom. Examples of the ways in which chemical information has been extracted from Auger spectra will be given later. Technique and Instrumentation Surface Specificity and Vacuum Requirement If the characteristic energy of an Auger electron in a solid is to be measured then the electron must be able to diffuse to and escape from the surface of the solid without losing energy by inelastic scattering. Although the range of primary energies used in AES is 5-20 keV the energy range of the AHger electrons used in the technique is normally 20-2000 eV.Fig. 2 shows how the kinetic energy of low-energy electrons varies with inelastic mean free path Le., the average distance travelled before losing energy by an inelastic collision ; the compilation is by Seah and Dench.12 It can be quickly appreciated that for an Auger electron of energy in the above range to be able to escape from the surface with its original characteristic energy it must originate within the outermost few atomic monolayers ; hence the highly surface-specific nature of AES. 100 -10 -/ x 1 -Fig. 2. Variation of inelastic mean free path of electrons in a solid with kinetic energy. The inelastic mean free path is related to the average escape depth of an electron ejected from a solid into the vacuum.The dashed lines are theo-retical relationships.’ Jzcne 1983 ANALYTICAL CHEMISTRY A REVIEW 653 As AES is so sensitive to the state of the surface and as most applications of AES involve treatment of the surface at some stage e.g. cleaning by heating or ion erosion or both reaction with introduced gases and vapours fracture etc. it follows that interference from the ambient must be eliminated. In practice this means reducing the total pressure to the level at which the build-up of contaminant layers on the surface by adsorption of gas molecules is so slow that there is enough time to carry out the AES analyses without such interference. Consideration of gas kinetics shows that AES like the other surface-specific techniques must be performed in ultra-high vacuum at pressures of the order of lo-* Pa ( Torr).To achieve such vacua requires accelerated outgassing by bakeout at -200 "C and hence the materials of construction are limited to those which will withstand such treatment and which do not themselves contribute anything to the residual gas background. Electron Energy Analysis As stated in the Introduction the first electron energy analyser to be used extensively for AES was the retarding field analyser (RFA) simply because the RFA used the existing electron optical arrangement for LEED and there were at the time a large number of LEED systems in existence. For those engaged in LEED measurement for its own sake the RFA is still con-venient for checking the cleanliness of a surface or for monitoring surface reactions but for all other applications energy analysis in AES is now performed in either of two types of analyser, the cylindrical mirror (CMA) or the concentric hemispherical (CHA).The former is the more popular for AES alone the latter for combined AES and XPS. It consists of two co-axial cylinders with annular entrance and exit apertures cut in the inner cylinder. If a deflecting (Le. negative) potential I' is applied to the outer cylinder while the inner cylinder is earthed and if the radii of the inner and outer cylinders are rl and yZ respectively then electrons emitted at the source on the sample surface with a kinetic energy E are refocused according to the equation A diagraml3 of a modern CMA and its associated circuitry is shown in Fig. 3. * (2) E/eV = 1.31 ln(rl/r2) .. . . for a critical emission angle of 42" 18'. At that angle the first-order aberration terms vanish and the device becomes second-order focusing. Obviously r2 must be large enough to allow free passage of electrons and it is convenient to make it about double yl. Typical dimensions for a CMA would be outside diameter 15 cm and length about 50 cm. UHV chamber Cylindrical mirror analyser Fig. 3. A modern cylindrical mirror analyser (CMA) and its associated circuitry. The electron gun is co-axial with the CMA. Electrons ejected from the irradiated point on the target positioned at the source of the CMA are re-focused at the electron-multiplier detector by a deflecting potential applied to the outer cylinder.1 654 RIVIBRE AUGER TECHNIQUES IN Analyst Vol.108 As dispersive analysers such as the CMA allow refocusing of only those electrons whose energies are in a narrow range whose width is dependent on the energy resolution of the indi-vidual instrument simply ramping the deflecting potential V through the energy range of interest produces an energy distribution EN(E). If the differential distribution EdN(E)/dE is required a small modulation of a few volts is superimposed on V and the collecting circuitry is then tuned to the fundamental frequency of the a.c. component of the collected current. Typical energy resolution to be found in a commercially available CMA would be 0.3% (Le., resolving power 330). The high luminosity and low time constant of the CMA enable it to be operated at high scanning speeds if required so that an AES spectrum could be displayed on a TV screen.Alternatively the incident electron beam generated from an internal electron gun on axis can be scanned rapidly over the specimen surface while the CMA is set to the kinetic energy of a chosen Auger peak thus producing a distributional map of a particular element over the surface. This variation of the technique is called scanning Auger microscopy (SAM). Fig. 4. Schematic diagram of Concentric hemispherical analyser (CHA). The radius of the inner spherical surface is y1 and that of the outer Y,. If the potential applied between the hemispheres is V then electrons entering the analyser at S with energy E are re-focused at F according to equation (3). The other commonly used dispersive analyser the CHA is shown diagrammatically in Fig.4. As the name indicates it consists of two concentric hemispheres whose included angle can be 180" as shown in Fig. 4 or 150". If the radii of the inner and outer hemispheres are y1 and y2, respectively and if the deflecting potential applied across the hemispheres is 'V then electrons entering the analyser with energy E are refocused according to the equation The CHA is often operated under constant resolution conditions that is the deflecting poten-tial V is fixed at some chosen value according to the required energy resolution and electrons approaching the entrance slit are retarded to that potential. Modern versions have an input lens system to decelerate and focus electrons on to the slit and thus improve the over-all sensitivity by accepting a greater input solid angle of electrons from the specimen.A typical size of CHA in common use would be of average radius about 10 cm and have an energy resolution of about 0.1% (ie. a resolving power of 1000). Modulation to produce a differential distribution is normally applied to the appropriate input lens electrodes and the effect of that practice is that the distribution is unfortunately not necessarily EdN(E)/dE but some com-plicated function of it. The CHA is in general more applicable to XPS where differentiation is not employed J w e 1983 ANALYTICAL CHEMISTRY A REVIEW 655 Electron Sources As the only basic physical requirement of a primary source of electrons for AES is that of ionisation of an atomic core level the energy conditions required are not stringent.Because of the shape of the variation of ionisation cross-section with energy,14 it is desirable that the primary energy be greater than about five times the binding energy of the core level but there are no restrictions on energy spread in the beam. Only if characteristic loss peaks also observable in the spectrum are to be studied should the spread be restricted to about 5 0 . 5 eV. The principal direction of development in recent years has been towards ever smaller irradiated areas i.e. electron spot sizes in order to achieve high spatial resolution. Electron optical design parameters such as space charge and lens aberrations inevitably push the development towards lower beam currents and higher beam voltages so that the current state-of-the-art instrument would operate typically at 1-10 nA and 20-30 keV as primary beam conditions.Under those conditions spot sizes of -50 nm can be obtained although it should be realised that the Auger resolution will be worse than that 100-200 nm owing to back-scattering effects that broaden the region from which Auger emission occurs compared with the irradiated region. As electrons can be deflected easily by electrostatic or electromagnetic fields it is relatively simple to raster the primary beam across a surface and use either the low-energy secondary electrons to produce a topographical image as in the scanning electron microscope or the Auger electrons at a selected energy to produce an elemental map. Much of the drive in development is towards smaller spot areas and the scanning mode in this context is proving invaluable in quality control in such areas as semiconductor device manufacture.Data Acquisition and Handling A typical spectrum of the number of electrons N(E) at a particular kinetic energy E ejected from a surface as a function of E is shown in the lower part of Fig. 5 for a boron specimen. 1 @--==- Ep= 1000 eV 175 167 I B Ep= 1000 eV A 0 200 400 600 800 1000 E ne rg y/eV Fig. 5. (A) Secondary electron distribution N(E) from a boron specimen at a nominal primary energy of 1000 eV. To the right are the elastic peak and associated plasmon loss peaks. To the far left is the steep slope leading up to the true secondary peak. At 167 eV is the boron KVV Auger peak. (B) Portions of the differential distri-bution dN(E)/dE in the regions of the elastic peak and the Auger peak.In the differential distribution the position of a peak is taken conventionally as that of the high energy minimum e.g. that for the boron Auger peak is at 175 eV 656 RIVIBRE AUGER TECHNIQUES IN Awalyst Vol. 108 The N(E) distribution as it is called contains two prominent peaks one at the primary energy due to elastically scattered electrons and the other at very low energy due to so-called “true” secondary electrons. Elsewhere there are minor features and in particular there is a peak at -180 eV due to electrons ejected by the boron KVV Auger transition. As can be seen not only is the Auger feature superimposed on a high background but in parts of the spectrum the background is also changing rather rapidly.For those reasons and also for visual enhance-ment of Auger features as they are not always as clear as the boron peak in Fig. 5 it is still normal to differentiate the spectrum electronically with respect to E to provide the dN(E)/dE spectrum [it would in practice be EdN(E)/dE when using a dispersive analyser]. The differentiated boron peak is shown in the upper part of Fig. 5. The energy resolution of a differentiated Auger peak will be a function of the way in which the differentiation is performed, i.e. of the sinusoidal modulating voltage applied to the outer cylinder of a CMA and if too great an amplitude of modulation is used in an effort to achieve greater sensitivity then the peak will suffer serious distortion.To a first approximation the difference between the maximum positive and negative excursions of an Auger peak in the differential distribution is taken as a quantity proportional to the concentration of the element giving rise to that peak. This approximation has been shown to be valid and useful in a large number of instances but as more knowledge is gained about the changes in Auger peak shapes as a result of changes in chemical environment so it is being realised that alteration of a peak shape can also alter the proportionality between differentiated peak height and concentration. The more appropriate quantity to measure is the area under an Auger peak in the undifferentiated or N(E) spectrum. Unfortunately this is not as easy as it sounds because of the problems of removing the inelastic background in a physically correct way before integration and of removing the effects of instrumental broaden-ing.The first problem in particular has not yet been solved in a satisfactory way and for most quantification purposes it is still the differentiated peak-to-peak height that is used. Most modern Auger spectrometers are now controlled by dedicated minicomputer systems. These will acquire spectra according to pre-programmed acquisition parameters such as energy region modulation voltage and counting statistics and store the acquired spectra for subsequent manipulation. Some systems will also allow a set of spot analysis positions to be chosen and stored by placing a cursor on features of interest in a scanning electron display of the specimen surface.Once the positions have been recorded then the primary beam returns to them at each subsequent acquisition without requiring re-positioning. Such a facility is useful when following changes in individual features as a function of surface treatment e.g., erosion by ion bombardment. Quantification First Principles Suppose the surface of a multi-component specimen contains i atomic species and that the current of Auger electrons of kinetic energy Ei resulting from the transition ABC in the ith species is measured. Then the relationship between that current I i and the density N i of atoms of the ith species within the analysed volume is where 1 = primary electron current; EA = binding energy of level A ionised in atom i; Ep = primary electron energy; u(Ep,EA) = ionisation cross-section of E at E,; h(E,) = inelastic mean free path for electrons of energy E i ; T(E,) = transmission of analyser at energy E,; D(Ei) = efficiency of detector at energy E,; P(ABC) = transition probability of ABC including contributions to ionisations of A from both direct and Coster - Kronig transitions from other levels; G = geometric factor; R = roughness factor; and y(EP,EA) = back-scattering factor taking account of additional ionisation of A due to energetic secondary electrons produced by Ep Jane 1983 ANALYTICAL CHEMISTRY A REVIEW 657 Under normal experimental conditions the primary energy and current would be held constant the geometry of the system would be fixed the detector efficiency would be sensibly constant as in most analysers the electrons are accelerated through a few hundred electronvolts into the multiplier and across the surface of any one specimen the degree of roughness is also likely to be constant.Equation (4) can thus be reduced to Ii(ABC) = Ko(Ep,E,) X (Ei)T(Ei)P(ABC)[l + r (E,,E,)]N . . * (5) where K includes the experimentally constant factors. Even in this expression there are still quantities that are not well known. There are various theoretical expressions that fit the cross-section CJ reasonably well the inelastic mean free path X is known fairly accurately as a function of energy from the work of Seah and Dench,12 the transmission function T is becoming better known for most commercial analysers but the probability P and the back-scattering factor r are much less well known.Calculations exist for both the latter quantities but their accuracy is doubtful. Approximations have always to be made in using equation (5) and the first-principles approach is rarely accurate to better than 560y0. Elemental Standards In this approach to quantification use is made of compilations of elemental Auger spectra recorded under well characterised conditions of energy resolution and of amplification and the assumption is made that the ratio of the intensity of an Auger peak from a given element in any situation to that of the same Auger peak from the pure element can be taken as proportional to the atomic fraction of that element. With the help of the compilations the proportionality constants can be regarded as relative elemental sensitivity factors Si so that the atomic per-centage concentration Ci of the ith element in an analysed volume containing j elements can be written as Equation (6) is used frequently and leads to an accuracy in quantification of about &20y0, but reference to equation (5) shows that there is total neglect of matrix effects in equation (6).In certain simple situations e.g. binary alloys the matrix effects can be included so that the accuracy can be improved to &loyo but in the more commonly encountered multi-component systems data on the modifications of inelastic mean free paths and back-scattering factors are inadequate. In either method of quantification the assumption is implicitly made that all elements are uniformly distributed within the analysed volume.Non-uniform distribution with depth cannot yet be treated although in certain instances e.g. where the surface is sufficiently smooth some information about depth distribution can be obtained by variation of take-off angle. See also the section on depth distribution by ion profiling. Spectral Identification Auger Peaks Prediction of the expected Auger energies from any element can be made using equation (1) ; until recently a semi-empirical approximation was used that was accurate enough for most purposes particularly in the earlier days of the technique but now the theory for the calcula-tion of Auger energies has become very exact and tables of calculated energies are available. The most comprehensive are those of Larkins.lo Obviously the number of possible Auger transitions increases progressively with atomic number owing to the proliferation of atomic energy levels but fortunately for AES there is a large variation in the probabilities of the transitions so that in practice within the energy range normally encompassed in the technique, only a few transitions are observed.Each region of the Periodic Table tends because of the similar electronic structure of atoms in that region to have a spectrum of characteristic appearance. For example in the first series of transition metals the three major LMM Auger peaks act as the “fingerprint”; their relative intensities and separations alter from scandium to zinc but they are unmistakable 658 R I V I ~ R E AUGER TECHNIQUES IN Analyst Vol. 108 Some of them are shown in Fig.6 (from Weber15) for the metals chromium manganese and iron. In the same spectra it is also possible to see the low-energy M2,3M4,5M4,5 Auger peaks. Another set of spectra based on MNN transitions characteristic of another region of the Periodic Table is shown in Fig. 7 (also from Weberls) for the elements silver cadmium, indium and antimony. In both Figures there are low-intensity features at energies below those of the principal peaks due to minor Auger transitions. Commercially produced hand-books have been available for some time in which Auger spectra in the energy range 0-2000 eV are displayed for nearly all elements; most elements have been cleaned ilz sitzc for the recording, but obviously most of the non-metallic elements would have to have been used in the form of compounds.A Ic Ni Ni Ni 0 200 400 600 800 1000 0 200 400 600 800 1 Electron energylev Electron energylev DO Fig. 6. Auger spectra from three of the first-series transition metals (A) chromium (B) manganese and (C) iron. The LMM triplet appearing for those metals in the energy region 450-700 eV is a characteristic “finger-print” for the series. The low-energy peak is due to the Ma,8M4 IM4 transition. l6 Fig. 7. Auger spectra from (A) silver (B) cadmium (C) indium and (D) antimony. The sharp doublet due to MNN transitions is very characteristic of these metals.16 It should be realised when attempting to identify Auger spectra that chemical effects are common. They take the form of shifts in peak energy and changes in peak shape generally both together and are strongest when the Auger transition is based on one or more electrons originating in the valence band of the solid.The first example of this and still one of the most striking is that of carbon in various states of combination; Fig. 8 shows the carbon KVV Auger spectra from graphite and from various carbides recorded by several workers.lG-l* Examples of large chemical shifts in particular for Mg A1 and Si have already been given in the Intro-duction. Unless the possible presence of such shifts is taken into account misinterpretation could result Jzcne 1983 ANALYTICAL CHEMISTRY A REVIEW 659 272 eV 1 273 eV Sic* Sic N isC T 1 271 eV 272 eV 240 260 280 300 240 260 280 En erg y I eV Fig. 8. Differences in the carbon KVV Auger peak shape in various chemical situations.(a) In silicon carbide and in graphite; the asterisk refers to the ion-bombarded surface. (b) In nickel carbide. (c) In titanium vanadium and chromium carbides.l6-l8 Satellite Peaks Any electron of sufficient energy moving within a solid can interact with the Fermi sea of electrons setting up collective oscillations (“waves”) within the sea. These oscillations have characteristic frequencies and so the interaction involves the loss of a characteristic amount of energy from the electron termed the plasmon loss. When the energy spectrum of electrons ejected from the surface is measured the plasmon loss peaks can be seen on the low-energy side of the principal peaks in the spectrum. Often multiple losses can be observed owing to the excitation of second third etc.harmonics of the fundamental oscillation although of course they decrease progressively in intensity rather quickly. All such losses are called “bulk” plasmon losses. Sometimes observable particularly when a low primary excitation energy is used is a loss peak at a smaller energy separation from the parent peak. This is the so-called “surface” plasmon loss arising from interaction with collective electronic oscillations in the layers of the solid near the surface; the presence of the surface i.e. the termination of the solid changes the characteristic frequency so that the energy of interaction is 1 / 4 2 that of such interaction in the bulk. Bulk plasmon losses are typically in the range 15-25 eV depending on the material so that the corresponding surface plasmons would be in the energy range 10.5-17.5 eV.Neither peak in the energy spectrum is more than 10% of its parent often less but clearly if the parent peak is one of the major features then there is the possibility of confusion of a loss peak with the Auger peak of a minor constituent. Occasionally a satellite peak appears at higher energy than the parent but is invariably small. Such peaks have been attributed in the past to plasmon gains but are now believed to be due to double ionisation of the atom 660 R I V I ~ R E AUGER TECHNIQUES IN Analyst VoZ. 108 Ionisation Loss Peaks As stated in the Introduction the initial stage in the process leading to Auger emission (or photon emission) in AES is the ionisation of a core level by a primary electron.The ionisation occurs by excitation of an electron from the core level to the first unoccupied state above the Fermi level. If there is a narrow band of such unoccupied states and a high probability of transition to it then at the threshold for ionisation a substantial number of electrons are lost from the primary current and a significant step loosely called a peak appears in the spectrum separated from the primary energy by the ionisation energy. Two such steps after differentia-tion are shown in Fig. 9 in which can also be seen bulk plasmon loss peaks associated in this instance with the elastic peak at the primary energy; the material is carbon-contaminated boron. The binding energies of the K shells of boron and carbon are 188 and 284 eV respec-tively and the corresponding ionisation loss peaks appear at 807 and 71 1 eV below the elastic peak.952 I I loss plasmon I I I” I 600 700 800 goo 1000 1100 En e rg yleV Fig. 9. Differential spectrum from a piece of carbon-contaminated boron in the energy region near the elastic peak. Primary energy was nomi-nally 1000 eV. The peaks a t 71 1 and 807 eV arise from losses from the primary energy owing to ionisation of the carbon and boron K shells a t 284 and 188 eV respectively. Near the elastic peak can be seen the peaks due to plasmon losses from the primary energy. Secondary Electron Peak All electrons in the energy distribution ejected from a surface except those elastically scattered are strictly secondary electrons but conventionally the term “secondary electrons” is taken to mean the so called t m e secondary electrons produced in the solid by a cascade pro-cess.The current of true secondaries peaks at very low energies of the order of a few electron-volts but the peak is very broad as can be seen in Fig. 5 and extends to 200-300 eV; it is the most prominent feature in the energy distribution at high primary electron energies while at low primary energies (<lo00 eV) it is the next most prominent after the elastic peak. The presence of the secondary electron peak is inconvenient in that Auger transitions of low energy are thus situated on a steeply sloping background June 1983 ANALYTICAL CHEMISTRY A REVIEW 661 Analysis in Depth Although the average escape depth of Auger electrons in the energy range used by AES is of the order of only a few atomic layers as can be seen from Fig.2 in practice it is often necessary to obtain information from greater depths. Compositional variations through corrosion-formed films through multi-layer structures on semiconductor devices and through surface layers altered by implantation to name but a few as a function of depth are needed in order to understand the mechanisms involved in the formation of the films etc. The methods that have been used to study such variations differ according to the thickness of the layer or film involved and will be discussed here in order of increasing thickness. Angular Dependence of Auger Intensity For an ideally smooth surface it is easy to show that if h is the inelastic mean free path for electrons of a particular kinetic energy and if the angle to the surface at which the energy analyser is placed (called the take-off angle) is 8 then the average escape depth is hsin8.In other words by varying 8 from 90 to O” it is possible in principle to vary the average escape depth from h to zero. In practice of course no surface is ideally smooth and it is normally impossible to use take-off angles lower than about 15”. Nevertheless there are special cases, e.g. very thin films of SiO on Si and segregated layers where the angular dependence method has proved useful. Where the information on compositional variation is required from depths greater than the inelastic mean free path i.e. in the overwhelming majority of instances then it is obvious that the surface must be removed in order to allow analysis by AES at the depths required.The removal to achieve the depth profile has been carried out either by erosion by ion bombardment or by mechanical lapping to produce a taper section. Its major advantage is that it is non-destructive. Erosion by Ion Bombardment A beam of positive ions almost invariably those of argon of energy between 500 and 5000 eV, and of current density between 5 and 50 pA crn-, is directed at the surface from an ion gun for a chosen length of time corresponding to the depth of erosion required. As data acquisition in AES is relatively fast and as the pressure of argon in the spectrometer chamber is generally between 10-7 and 10-5 Torr during ion bombardment it is perfectly feasible to carry out simultaneous bombardment and analysis thus achieving a continuous compositional profile down to the chosen depth.In practice the energy analyser is programmed to record the intensities at several selected energies corresponding to major Auger peaks of the elements whose profile is required in continuous succession during bombardment ; this is known as “multiplexing”. The other method of recording a compositional depth profile is to stop the bombardment at selected intermediate intervals and record the Auger spectra from the newly exposed surfaces. The advantages of the latter procedure are that the surface condition is not changing further during analysis that a more complete analysis can be carried out and that the possibility of synergistic effects due to simultaneous ion and electron irradiation is removed.However it is of course much slower than the multiplexing procedure. Profiling by ion bombardment as described above is a very widely used technique but its range of usefulness in terms of interpretable information is limited by the effects of the bom-bardment itself on the solid surface. Argon ions of a few kiloelectronvolts kinetic energy can penetrate several atomic layers below the surface before finally stopping during which they can transfer varying amounts of energy to many atoms; the more energetic of the atoms excited in this way can transfer energy to other atoms and so on in a cascade process. The result is that for each incident ion there might be several atoms in the solid displaced from their normal positions.This can result in the effects of atomic mixing and of “knock-on,” in which a certain species might not be removed completely from the surface but simply driven further into the solid. These effects cause a local smearing of the compositional depth distribution that is carried forward with the advancing bombardment front and is constant with depth once equilibrium has been reached. More serious is the effect of the statistical nature of the sputtering process. Atoms in the solid are not peeled off layer by layer but in a statistical sequence which means that if the average number of layers removed is N then statistically N i layers remain only partially sputtered so that the compositional information comes from man 662 RIVIBRE AUGER TECHNIQUES IN AutaZyst Vd.108 fractional layers rather than from an ideally flat surface. In this instance the smearing of the compositional depth information clearly becomes progressively worse with the amount of material removed and an effective limit is eventually set to the depth to which it is worth going before the information is lost. The problem has been considered in detail by Lea and Seah," and the very general conclusion can be drawn from calculations that for real situations in which the outer surface has a roughness of about 50 nm there is no point in attempting to use ion bombardment to depths greater than 2 pm where the constituent sputtering yields are similar or to much smaller depths if the sputtering yields differ substantially.Where a compositional profile is required to depths greater than about 2 pm the analysis will be less prone to errors due to artefacts if mechanical methods of surface removal are used, as described below. Erosion by Mechanical Lapping The principle involved in mechanical lapping is that of exposing the depth to be analysed by cutting across it at an acute angle to the surface. If the cut is made by lapping the surface at a constant angle then the result is the familiar taper section used with great success in electron probe microanalysis. Thus if the depth of interest is d and the taper angle is a the length of section to be analysed is dcosecct; a taper angle of 5" would therefore spread out the depth by a factor of about 11.5 and an angle of 1" by a factor of about 60.Of more general use in AES is the variation of mechanical lapping called ball-cratering. The principle is illustrated in Fig. 10 (from Walls et A highly polished stainless-steel ball is caused to rotate in contact with the specimen surface fine diamond paste being used as the lapping medium in the area of contact. If the specimen surface consists of a film or layer of a material A on a dissimiliar material B then lapping would continue until the interface had been crossed and B became visible. Alternatively if it is required simply to profile to a certain chosen depth then for a given ball diameter (usually 30 mm) it is easy to calculate from the geometry what the diameter of the resultant crater should be. Distance can of course be measured very accurately with a travelling microscope.The advantages of ball-cratering are that relatively small areas of the sample are eroded that it can be used on curved surfaces and that where there is an interface the taper angle can be made very small indeed. \ \ \ \ I / 4 R\ I I I I I \ i I 4 t d Coating b Fig. 10. Principle of depth profiling by ball-cratering. A highly polished ball of radius R is smeared with fine diamond paste and rotated against the surface to be cratered until an area of the substrate of diameter D is revealed. If D is the diameter of the crater then the thickness of the film or coating is given by C = (D2* - D1*)/SR).*O In all instances of erosion by mechanical lapping it is necessary to sputter briefly with argon ions to remove contamination after introduction into the vacuum system for analysis but such sputtering does not affect the depth resolution unduly.Not all materials are suitable for mechanical lapping e.g. brittle oxides may break away from an underlying metal substrate where the section is thinnest near the interface and it is also possible for soft materials to be smeared across the eroded section Jzcne 1983 ANALYTICAL CHEMISTRY A REVIEW 663 Chemical Bonding Information When discussing any technique in relation to analytical chemistry there are two questions to which it must be capable of providing a t least partial answers. These are (1) what is the elemental composition? and (2) what is the nature of the inter-elemental bonding? Up to now the answers obtained from AES have been overwhelmingly to the first question and any answers to the second have been obtained indirectly and incidentally e.g.by “matching” the atomic proportions of elements found on a surface to compounds suspected to be present from other evidence. In the various areas of application of AES discussed in the next section, therefore the analyses with a few exceptions could not be considered fully chemical on the above criteria as the only information produced has been compositional. However it is worth considering the exceptions because the ways in which some workers are starting to use Auger spectra to extract answers to the second question above are pointers to one of the directions of rapid expansion of the technique in the near future. The three properties of any peak in the secondary electron spectrum are intensity energetic position and shape.In AES just as in XPS the first two of these are used for the derivation of elemental composition the intensity for quantification and the energetic position for identi-fication. Again as in XPS it is not usually necessary to measure energetic position particu-larly accurately for mere elemental identification especially where an element has several Auger peaks in the energy range recorded but in both techniques additional chemical informa-tion can be obtained by more accurate measurement of the peak energy. An example of the changes observed in the silicon LMM and KLL Auger peaks in going from elemental silicon to SiO and Si,N is shown in Fig. 11 from the work of Holloway.21 The energy shift in the Si KLL LMM 9 eV Si T 3 I I 1618 I 191 I I I artifact 506 I I Fig.11. Variations in the silicon LMM and KLL Auger peak shapes between (A) elemental silicon (B) SiO and (C) Si,N,. The peaks a t 91 eV in the LMM spectrum and a t 161 1 eV in the KLL spectrum from SiO are due to reduc-tion effects by the incident electron beam.a 664 RIVIBRE AUGER TECHNIQUES IN Analyst Vol. 108 LMM peak is from 91 eV in silicon to 87 eV in Si,N4 and to 78 eV in SiO, with accompanying shifts in the same sense if not of the same magnitude in the KLL peak. (Note also in Fig. 11 one of the problems not infrequently encountered in AES namely the effect of the incident electron beam on the decomposition of compounds near their surfaces; the peak at 91 eV in the SiO LMM spectrum is due to elemental silicon produced by beam reduction.) As men-tioned in the Introduction there are by now many published figures for chemical shifts in Auger spectra and those observed during oxidation have been listed by Madden.1l Provided that the energy calibration of ones spectrum is always the same as that used in the listed observations then the figures can be used to determine the nature of the oxidation product at a surf ace.Although intelligent guesses can often be made about the chemistry of a surface from the derived elemental composition more direct chemical information should be available from the detailed analysis of Auger line shapes. Some of the most intense Auger transitions from solids are of the CVV type that is ionisation of a core level C followed by an Auger process in which both the de-exciting and the ejected electrons are from the valence band V.As either electron can originate from anywhere within the valence band the CVV Auger spectrum should in principle have the same shape as a self-convolution of the density of electronic states within the valence band. Thus a self-deconvolution of such an Auger spectrum might after the correct background has been subtracted and instrumental broadening removed be expected to pro-duce an energy distribution looking very like the valence density of states (DOS). The DOS so derived would not necessarily resemble that of the bulk material but would be related to the local DOS in the immediate neighbourhood of the ionised atom Where the atom is situated at or very near the surface as in AES then it would be the surface local DOS that would be observed with a distribution probably rather different from that in the bulk.Such informa-tion could be very valuable in studying the chemistry of surface reactions. Of course Auger transitions involving only one valence electron i.e. of the CCV type should have the DOS information directly reflected in their spectrum if the direct relationship held good but often their intensity is insufficient. The more or less direct relationship described above between the local DOS and the Auger line shape has been found to be true at the time of writing for only a few elements and some of their compounds but has been useful nevertheless. Those elements which include Li Be C, Al Si and some transition elements with valence bands less than half full are said to have “band-like” CVV Auger spectra because the spectral line shapes can be related to the valence band structure.The relationship is not necessarily one-to-one as Auger transition matrix elements vary across the valence band and indeed it is possible for different Auger transitions to give different apparent local DOS after line shape analysis. The differences can be used to derive information about the contributions of individual components of the valence band to 15 10 5 0 EnergyleV 15 10 5 0 Fig. 12. (a) Comparison of the local density of states derived from (1) analysis of the silicon L,L,,3V Auger spectrum with (2) the theoretical prediction.(b) Local density of states derived from the silicon L,,,VV Auger spectrum by (1) self-deconvolution and by (2) convolution square root.% Jane 1983 ANALYTICAL CHEMISTRY A REVIEW 665 the Auger process so that in fact not just the total local DOS can be derived but also a partial local DOS. This is well illustrated by the work of Brockman and on silicon to which most attention has been devoted for both theoretical and technological reasons. Fig. 12 compares the DOS extracted from analysis of the shapes of the L,L,,,V and L2,,VV Auger transitions in silicon; the full lines are the experimental results. The two derived DOS curves are different. That based on the L,L2,3V agrees well with the theoretical (dashed line) calculation of the DOS near a silicon surface in other words all the contributory terms s*s s*p, and p*p from the valence band are included but the L,,,VV-derived DOS is almost entirely p*p i.e.represents a partial local DOS. Clearly then analysis of the shapes of these two transitions during chemical reaction at a silicon surface can provide detailed information about the ways in which electrons in different regions of the local valence band are affected by what is happening at the surface. A good example of the application of Auger line shape analysis to the study of the differences in local or site-specific DOS has been given by Davis et aL2 for the compound semiconductor GeSe. In order to avoid the problems involved in self-deconvolution of a CVV Auger transi-tion they recorded the M,M,,,V transitions of Ge and Se in the compound in other words, CCV-type transitions even though they were admittedly much weaker than the CVV.Their results after integration and appropriate data processing are shown as the lower graphs (3) in Fig. 13 the solid line for Ge and the dashed line for Se with energies referred to the Fermi level. . . -20 -15 -10 -5 0 E - EV/eV Fig. 13. Differences in the local densities of states around germanium and selenium atoms in GeSe studied by line shape analysis of the X-ray excited MlM4,,V Auger transitions. (1) Valence band spectrum of GeSe obtained by un-monochromatised XPS dotted line before solid line after data processing. (2) Same using monochromatised X-rays. (3) Solid line, local DOS from germanium M,M4,,V spectrum and dashed line local DOS from selenium M1M4, V spectrum.3 666 R I V I ~ R E AUGER TECHNIQUES IN Analyst Vol.108 For comparison the valence band spectrum for GeSe obtained by XPS is shown in the upper part of Fig. 13 as graph (l) the solid line being the result after data processing. XPS of course provides a general picture of the DOS in the near-surface region without being able to distinguish variations in DOS at individual sites. The significant result is that the Auger line shapes at (1) in Fig. 13 reflecting the local DOS each show only some of the features in the XPS valence band spectrum. Hence the Ge shape shows only two of the peaks and similarly the Se shape only two but not the same two. In other words the CCV Auger transitions in GeSe involve electrons only in those states that overlap with the site of the core hole and the DOS derived from their line shapes relates to the immediate vicinity of the ionised atom.Additional information can be derived from the relative magnitudes in each instance of the low-energy bonding p peaks and the higher energy non-bonding s peaks in comparison with the XPS result. At the opposite extreme from those elements that exhibit “band-like” CVV Auger transitions is the group of elements in which the CVV peaks are said to be of “quasi-atomic” character. This term has been applied because the observed line structure is much too narrow to have any possible direct relationship with the band structure the peak widths being closer to those observed from free atoms than from atoms making up a solid.The narrowness arises from distortions imposed on the spectra by the interaction of the two holes left behind after the ejection of the Auger electron (see the section Auger Effect in the Introduction) ; in the group of elements in question the two holes can be localised near the ionised atom long enough for their interaction to be significant. Hence for these elements it seems unlikely that any chemical information can be obtained from analysis of the CVV line shapes although such analysis has given much information about the Auger process itself. Between the extremes of purely band-like and purely quasi-atomic character lies a largely unchartered region in which there may be many elements whose CVV Auger spectra contain components of both. Moreover it seems possible that surface reaction involving transfer of electronic charge in one direction or another may alter the local DOS sufficiently for a change in character towards either extreme to be observable and therefore usable for chemical informa-tion.As yet there is insufficient work in this area for confirmation or otherwise of this possi-bility. One area in which the use of AES as a chemical tool looks definitely promising is that of the Auger spectra of gas-phase molecules. There are many papers in which the Auger spectra of molecules adsorbed on solid surfaces have been reported but only recently has attention been turned to molecules in the gas phase. Obviously there are some experimental problems as a sufficiently high pressure of the gas must be maintained at the focus of the energy analyser, while at the same time the pressure in the analyser itself must be low but they seem to have been largely solved.The Auger data can and have been used in two ways firstly as a “fingerprint” of the molecule as the spectra can show large variations between neighbouring molecules in a series and secondly by comparison of the gas-phase and condensed (adsorbed) molecular Auger spectra to derive information about the local chemical environment at the surface via the spectral changes. The first of the ways is exemplified in Fig. 14 (from Rye et in which the carbon KVV Auger spectra from the three hydrocarbons methane, ethylene and acetylene in the gas phase are shown. The carbon hybridisation is of course changing markedly from one to another of these three and this is reflected in the spectra.From the same paper24 is taken Fig. 15 as an example of the second way of using Auger spectra of molecules. Here the gas-phase and condensed (on nickel) phase oxygen KVV spectra are superimposed for molecules in the series H,O CH,OH and (CH,),O with naturally energy shifts as appropriate to take account of transition to the solid state. It can be seen that the two spectra almost coincide for (CH,),O but differ increasingly from CH,OH to H20. This reflects directly the nature of the intermolecular forces in the condensed layer or in other words the local density of electronic charge around each atom in the molecule. In (CH,),O for instance the forces are mostly of the weak Van der Waals type Le. with little movement of charge whereas in the other two molecules hydrogen bonding between the condensed particles becomes increasingly important.The changes in the degree of hybridisation of the oxygen orbitals as a result of condensation are so significant that the condensed phase spectrum starts to look very much like a solid-state valence band rather than that of a molecule. Clearly there is much useful chemical information to be obtained from measurements such as these J m e 1983 ANALYTICAL CHEMISTRY A REVIEW 667 G z Gas phase C(KVV) 180 220 260 3 E I ect ro n ene rg yIeV I I 1 470 51 0 t Electron energylev i0 Fig. 14. Carbon KVV Auger spectra from (A) Fig. 15. Comparison of the oxygen KVV Auger methane (B) ethylene and (C) acetylene in the gas spectra from several molecules in (1) the gas phase phase.In these three molecules the carbon hybri- [(A) water (B) methanol and (C) dimethyl ether] disation is drastically different and the differences and (2) condensed multi-layer phase on nickel at are revealed in the carbon Auger spectra empha- 110 K. The departure of the condensed phase sising the sensitivity of the Auger process to the spectra from the gas phase spectra reflects the local electronic environment.** degree of charge transfer between the condensed molecules very little for dimethyl ether and con-siderable for water.24 Applications Most of the more recent applications of AES have been those in which the high spatial resolution capability of the technique has been a necessary requirement i.e. applications in the metallurgical and semiconductor fields.Nevertheless the total number of applications of all types is now so great that there is no difficulty in selecting examples of a more chemical nature. These will be discussed under the headings of the various fields from which they have been chosen. Corrosion and Oxidation In general because information about the detailed chemical nature of corrosion films in terms of elemental oxidation states is always required the preferred technique for surface analysis in corrosion has been XPS rather than AES. Many workers believe too that the high spatial resolution obtainable in AES is wasted in corrosion studies but it is likely that more situations arise than are realised in which that capability could be useful. I t was found to be very useful for example by Lumsden et ~ 1 .~ 5 in their study of the susceptibility of iron to pitting in solutions containing both chloride and phosphate ions. Whereas aggressive ions such as chlorides accelerate the general dissolution of an iron surface that occurs in the presence of water inorganic inhibitors such as phosphates can passivate the iron if added in adequate concentration. When both types of additive are present together the form of attack of the iron is intermediate between general dissolution and passivation and pitting of the surface can result 668 RIVIBRE AUGER TECHNIQUES IN Analyst Vol. 108 The iron sample was held at a constant potential in a solution of disodium hydrogen orthophosphate and potassium chloride and then scratched.Immediately after scratching faceted pits of the type shown in Fig. 16(a) appeared where dissolution occurred along crystallographic planes. The pits soon became hemispherical as in Fig. 16(b) and were then surrounded increasingly by reaction products eventually becoming completely covered [Fig. 16(c) and (41. Using an incident electron beam of diameter -1 pm Auger spectra could be obtained from the inside walls of the pits and from the corrosion product covering a pit. The films covering the walls were found to be either of iron and oxygen only or of the composition corresponding to the Auger spectrum of Fig. 17(a) ; in the latter instance the film contained the ionic species from the solution. The Auger spectrum from the pit cover Fig. 17(b) showed two prominent phosphorus peaks whose energies were those of phosphorus in a phosphate.The only other peaks were those of oxygen and iron suggesting that the corrosion product was iron( 111) phosphate. Further confirma-tion comes from the fact that the Auger peak heights are in approximately the right relative magnitudes for FePO,. Fig. 16 (from Lumsden et al.25) shows a sequence during pit growth. - 3 Y 3 !2 W 0 500 1000 0 500 1000 EnergyleV Fig. 17. (a) AES spectrum from the inside wall of a faceted pit showing residues from the ionic (b) AES spectrum from the reaction The positions of the phosphorus peaks indicate a phosphate, species in solution (Fe; 0.1 N Na,HPO + 0.001 N KCl). product covering a pit as in Fig. 16(d). suggesting that the material is iron(II1) pho~phate.,~ From its inception AES has been used very extensively in studies of the earliest stages of reaction between gases and solids.Its high surface specificity and its ability to determine both the nature of elements present on a surface and their approximate concentrations have made it an obvious technique to use in such studies although it is now realised that care must be taken to avoid artefacts that may be introduced by the incident electron beam itself. Of all the gas - solid reactions studied those between oxygen and the first series of transition metals form probably the largest single section both for technological reasons and because they are intrinsically interesting. Typical of some of the careful measurements made of that type is the work of Benndorf et aL26 on the chemisorption of oxygen on and the initial oxidation of the nickel (1 10) surface.Starting with an initially atomically clean surface they followed the changes in the nickel low-energy MVV spectra and high-energy LVV spectra and in the oxygen KLL spectrum as a function of oxygen exposure. The work function of the surface was monitored at the same time. Fig. 18 shows some of the changes they observed in the nickel spectra after oxidation at room temperature; the M,,,VV Auger peak has shifted split and decreased and the L,VV spectra have also decreased in intensity substantially. There is also additional structure below the M,,,VV peak at -30 eV after oxidation. When the oxygen KLL peak intensity (measured as the peak-to-peak height) was plotted against exposure along with the variations in work function and energetic position of the oxy-gen peak the result shown in Fig.19 was obtained. Three distinct stages in the room-temperature reaction could be distinguished. The initial stage was that of dissociativ Fig. 16. Scanning electron micrographs showing different stages in the formation and growth of a pit In the earliest stages the pit is of crystallographic Corrosion products build up in a circle and its subsequent burial under reaction products. morphology with faceted sides but soon becomes hemispherical. around it and eventually cover the pit completely.2 Jzcne 1983 ANALYTICAL CHEMISTRY A REVIEW 669 Isr I 20 60 100 700 800 900 Kinetic energyIeV Fig. 18. Nickel MVV and LMM Auger spectra from a nickel (110) surface before and after oxidation a t room temperature in ca.Torr of oxygen. The M,,,VV peak shifts splits and decreases and the LMM spectra also decreases in intensity. There is also additional structure a t 30eV. below the M,,VV peak after oxidation.,6 chemisorption of oxygen molecules marked by rapid linear increases in the oxygen Auger intensity and in the work function and an equally rapid decrease in the oxygen peak position. In the second stage the rate of uptake of oxygen decreases and passes through an inflection, while both the work function and the oxygen kinetic energies reach a plateau; the interpreta-tion here is that a rearrangement of the chemisorbed layer is occurring with oxygen being incorporated in the nickel sub-surface layers.Finally with further oxygen exposure the third stage is reached in which the rate of oxygen uptake increases again and then saturates the work I I I I 10 30 50 70 90 Oxygen exposure11 Fig. 19. The variations in oxygen KLL peak intensity and energy and The unit of Torrs. Oxidation was in the work function as a nickel (110) surface is oxidised. exposure is the Langmuir which is equal to carried out at room temperature and 5 x lo-* Tom of oxygen.* 670 RIVIBRE AUGER TECHNIQUES IN Analyst VoL 108 function decreases to below that of the clean surface and the oxygen Auger kinetic energy also decreases sharply and then becomes constant. This stage is said to be due to the growth of islands of nickel oxide over the surface. From the oxygen KLL intensities and from the attenuation of the nickel L,VV and M,,,VV intensities the authors were able to calculate the limiting oxide thicknesses at temperatures from 300 to 670 K; these varied from -3.5 layers of NiO at 300 K to -10 layers at 670 K.The broadening of the M,,,VV peak on oxidation is due to alterations in the local electron densities of states at the surface as the d band of nickel is broadened and pushed to lower energies on oxidation and oxygen-induced 3p states appear. The additional structure below the M,,,VV peak is due to an interatomic or “cross,” transition involving the nickel M,, levels and the oxygen L and L,, levels. Catalysis There is not a large amount of published work to be found in which AES has been applied directly to the study of commercial catalysts.The reasons are probably three-fold firstly, that AES does not provide all the chemical information required secondly that on supported catalysts the effect of the incident electron beam might be catastrophic and thirdly that more commercial secrecy surrounds the specification and performance of catalysts than is found in any other field. Nevertheless many interesting basic studies of model catalyst systems have been carried out using AES usually in combination with other surface analytical techniques. Generally these studies have taken the form of experiments in which two or more pure gases are adsorbed either simultaneously or sequentially on the surfaces either of evaporated films or of single crystals of compositions corresponding to those of materials of known catalytic activity.Changes in surface composition during reaction and the nature of the reaction products are monitored. The relationship between such studies and the real situation tends to be regarded as tenuous, but there are instances where it is demonstrably not so. One of these has been studied by Cros et at?.,,’ and it originated in an earlier observation that silicon atoms that had diffused through a thick gold layer on silicon could be oxidised at surprisingly low temperatures below 400 “C. Their experiment consisted in evaporating small amounts of gold on to a clean silicon surface produced by cleavage in vacuum and then observing the effects of oxidation at various temperatures by AES inter alia. The basic result at room temperature is shown in Fig.20. The uppermost spectrum (A) is of the clean silicon surface typified by the intense LVV peak at 92 eV. The next spectrum (B) is of the same surface now covered by four monolayers of gold. The large complex peak at 69 eV is due to the gold NVV Auger transition while the presence of the silicon LVV peak and its splitting into two peaks at 90 and 94 eV show that silicon has diffused to the surface and alloyed with the gold. Exposure of the gold-covered silicon to oxygen at 0.2 Torr for 3 h at room temperature then produced spectrum (C) in which the gold NVV peak is reduced in intensity but otherwise unchanged but in which the elemental silicon LVV peak has been reduced considerably and a new silicon peak at 78 eV has appeared. Comparison of (C) with (D) which is the spectrum obtained from SiO grown at 900 “C shows that the 78 eV peak is indeed due to the oxide.However the comparison of (C) with (E) is the most interesting for the latter is the spectrum obtained from a clean gold-free surface of silicon (k spectrum A) exposed to exactly the same oxidising conditions as for (C). Apart from a reduction in the intensity of the elemental 92 eV peak a weak feature appears at 84 eV ascribable to the formation of SiO (where x ml) but there is no suggestion of the character-istic SiO peak at 78 eV. Thus the presence of the gold on the silicon surface enhances con-siderably the oxidation of the silicon in that SiO is formed where it is not formed in the absence of gold but the gold itself is unaffected. In other words the gold is acting as a cata-lyst for the oxidation.By extending their experiment to higher temperatures and with the help of depth profiling, the authors were able to show that the crucial factor was indeed the gold - silicon alloy formed at the surface. When present in a sufficient concentration of gold atoms silicon atoms adopt different hybridisation states which results in the disappearance of the normally strong covalent silicon bonds and stabilisation in a quasi-metallic state. In this condition the Si-0, tetrahedra can grow easily as their free energy of formation is greater than that of the gold -silicon alloy and thus an SiO layer is produced June 1983 ANALYTICAL CHEMISTRY A REVIEW 671 Reactions in the Solid State In addition to the study of gas - solid and liquid - solid reactions AES has been used to observe the effects of reactions between solids and the thermal decomposition of solids.Often such experiments are easier to perform than those involving gases or liquids and in general the potentially disturbing effects of the incident electron beam do not intrude to the same extent. In several technologically important fields such as device fabrication and the design of therm-ionic electron sources information about the composition and thickness of surface films pro-duced by solid - solid reactions as well as the nature of the interface between the film and the substrate is vital. One of the requirements in the fabrication of certain types of integrated circuits is the ability to produce thin films of metal on a semiconductor in which the metal - semiconductor interface has the correct electronic properties.This is known as Schottky barrier formation. It has been found that a relatively simple and attractive way of generating interfaces of the correct properties is to deposit thin films of those metals which will react easily with the clean semi-conductor substrate. One such system that has been the subject of much study is that of 84 v I I I I 60 70 80 90 Electron energytev AES spectra recorded before and after oxidation of clean and of gold-covered silicon surfaces. (A) Clean silicon surface with characteristic LVV Auger peak a t 92eV. (B) The same surface covered by four monolayers of gold. The new peak near 69eV is due to the gold NVV Auger transition while the splitting of the silicon LVV peak into two peaks a t 90 and 94 eV indicates alloying of silicon with the gold.(C) Exposure of the gold-covered surface to 0.2 Torr of oxygen for 3 h a t room temperature. A new peak a t 78eV has appeared. (D) SiO grown on silicon at 900 OC showing that the 78eV peak is due to the oxide. (E). Exposure of gold-free silicon surface as in (A) to the same oxidising conditions as in (C). The 92 eV peak is reduced slightly and a weak peak appears a t 84eV due to SiO, but no 78-eV peak appears.97 Fig. 20. - Electron energylev Fig. 21. Auger spectra from a silicon surface taken a t succes-sive stages of deposition of palladium a t room temperature : (A) clean (B) 20 A evaporated ; (C) 60A evaporated; and (D) 120A evaporated.After a deposition of only 20 A the silicon L,,3VV spectrum has changed from the single domi-nant peak at 92 eV to four peaks a t 81 86 91 and 98eV in the same positions as found for bulk Pd,Si. No changes are observed in the palladium MNN spectra.2 672 RIVIBRE AUGER TECHNIQUES IN Analyst Vol. 108 palladium on silicon. When Okada et aL28 deposited palladium on to silicon at room tempera-ture after cleaning the silicon surface by the well tried method of ion bombardment and annealing they observed a sequence of Auger spectra of which a selection is shown in Fig. 21. It can be seen that although there was no change in the shape of the palladium MNN peak at 330 eV as the thickness of the palladium increased that of the silicon LVV spectrum changed drastically.After only a few gngstroms of palladium the 92 eV peak characteristic of clean silicon had disappeared and was replaced by four peaks at 81 86,91 and 95 eV. With increase in the thickness of the palladium deposit to over 100 A the silicon peaks vanished altogether, and only the palladium spectrum was left. The complex four-peak silicon spectrum observed at the beginning of deposition was compared with the spectrum from the alloy Pd,Si and found to be identical. Thus even at room temperature palladium will alloy spontaneously with a clean silicon sur-face. However the authors were able to extract more information from the spectra by realis-ing that the silicon spectra in the very early stages of palladium deposition were in fact com-posite spectra being made up of a superposition of the elemental (it?.covalent) silicon spectrum and of the alloy (i.e. metallic) spectrum. Using the standard spectra from silicon and from Pd2Si in linear combination they were able to synthesise silicon spectra to match exactly the observed spectra as shown in Fig. 22. When therefore a film of palladium on silicon of thickness about 300 A was depth profiled by ion bombardment the authors could determine at each stage the proportions of covalent and of silicide silicon in the film with the result given in Fig. 23. The profile consists of three regions unreacted palladium palladium silicide and the silicon substrate. In consideration of the Schottky barrier formation it is thus necessary to take into account two interfaces that between palladium and the silicide and that between the silicide and silicon.Measu /J r/ d Pd Synthesised Pd2Si Fig. 22. Synthesis of silicon spectra at various stages of palla-dium deposition on silicon from standard L,,,VV spectra of silicon in elemental silicon and in Pd,Si. Comparison with the observed spectra on the left then allowed the proportions of covalent and of silicide silicon to be determined at each stage of deposition.28 100 s d 5 m 50 0 100 200 300 400 ! 10 Fig. 23. Profile across the various Depthh inteaaces in the silicon - palladium system (A) covalent silicon (B) silicide and (C) palladium. The proportions of covalent silicon and of silicide silicon were determined as illustrated in Fig. 22. Three regions can be distinguished un-reacted palladium palladium silicide and the silicon substrate so that the system thus contains two interfaces one between palladium and the silicide and the other between the silicide and In the manufacture of thermionic electron emitters for vacuum tubes which over the years has progressed from being a black art to having a sound scientific basis oxides of the alkaline earth metals are used in varying proportions and the efficient operation of the emitter depends on one or more of the oxides being reduced at the surface of the device to free metal.The point is that a monolayer of free alkaline earth metal has a low work function lower than that of most other materials except some of the alkali metals which of course means that it is a copious source of electrons.The operation to produce the thin film of free metal is calle Jzcne 1983 ANALYTICAL CHEMISTRY A REVIEW 673 activation. One of the methods used to maintain a supply of free atoms during the life of the emitter is to include in the recipe a reducing agent as the supply must balance the loss by evaporation if the device is to have an acceptable working life. The surface interactions between typically employed alkaline earth oxides and reducing agents have been studied by Verhoeven and Van Doveren,29 and Fig. 24 shows the results from one such interaction that between barium oxide and magnesium. The upper spectrum is that from pure barium oxide prepared by complete oxidation of a thick barium film and shows peaks at 52 and 65 eV characteristic of the oxide.The three lower spectra were recorded following increasing amounts of deposition of magnesium metal. As the thickness of magnesium increases the size of the magnesium LVV peak at 44 eV of course also increases but the interesting changes are those taking place in the barium spectrum for even after a few mono-layers of magnesium a substantial peak due to metallic barium at 71 eV appears. The v 65 52 '44 44 I I 50 1 oc Electron energylev Fig. 24. Surface reaction at room temperature between barium oxide and a magnesium film deposited on it. (A) Barium Auger peaks from a completely oxidised barium film, showing peak positions a t 52 and 66eV characteristic of the oxide. Lower spectra the result of deposi-ting increasing amounts of mag-nesium on the barium oxide.(B) BaO after 6 s Mg deposition. (C) BaO after 12 s Mg deposition. (D) BaO after 112 s Mg deposition. The peak at 71 eV is due to metallic barium and that a t 44eV to mag-nesiu m .a* 50 100 Y 0 Deposition time/s Fig. 25. Peak-to-peak heights of various Auger signals from a barium oxide film plotted as a function of time of deposition of magnesium a t a constant deposition rate (A) Mg; (B) 0; (C) Ba; and (D) BaO. An equilibrium situation develops in which there is always free barium a t the surface.2 674 RIVIBRE AUGER TECHNIQUES IN Analyst VoE. 108 intensities of the original barium oxide peaks quickly become too small to measure. The changes as a function of deposition time (it?. thickness) of the magnesium are summarised in Fig.25 where it can be seen that an equilibrium situation develops at room temperature in which there is always a substantial amount of free barium at the surface. Notice that Auger effects due to oxidised magnesium (at -34 eV) were very small or absent; MgO would be expected to be the other product of the reaction between Mg and BaO. The interpretation is either that the reduction is taking place below the outermost layers or that the barium is segregating to the surface at once. Analyses Using High Spatial Resolution As described in the section on Technique and Instrumentation one of the most useful variations of AES is scanning Auger microscopy (SAM) in which the incident electron beam is rastered over a chosen area of the sample allowing production either of topographical images using the low-energy secondary electrons or of elemental maps using selected Auger energies.The optimum spatial resolution obtainable in SAM is currently about 0.2 pm but in many applications such as those discussed below such high spatial resolution is not always necessary. Cast iron often contains nodules of diameter 10-30pm as shown in the SEM image of Fig. 26(a) (from Joshi30) which is that of a cast iron surface polished and then cleaned in vacuo by ion bombardment. Each nodule is surrounded by an annular region of contrast between those of the nodule and of the matrix. The two SAM images in Fig. 26(b) and (c) are those of carbon and of iron respectively. Clearly the nodule itself must be very rich in carbon and indeed a complete Auger spectrum from the centre of the nodule showed that apart from the carbon peak there was only a very small oxygen peak.The shape of the carbon peak was typical of that of graphite. Interestingly the annular region corresponds to a denudation of carbon from the cast iron matrix presumably caused by diffusion of carbon to the growing graphite nucleus. Over the rest of the surface iron is fairly uniformly distributed and the Auger spectrum from a point in the matrix between the nodules and their haloes showed that the carbon was in the form partly of graphite and partly of carbide indicating a mixture of iron carbide or pearlite and finely divided graphite. Solar cells for energy conversion should be of high efficiency and reliablity have a long life and be cheap to produce.Costs can be reduced by using polycrystalline rather than single crystal materials but then it is found that the presence and the properties of grain boundaries become the limiting factors in the performance and reliability of the cells. The electrical characteristics and thus the eventual efficiency will depend on the chemistry of the grain boundaries i.e. on the nature and amount of impurities that segregate there and on any separate phases that might form. Kazmerski31 has studied the grain boundary chemistry by a variety of techniques including SAM. In order to expose the grain boundary in such a way that the analysis would be unambiguous a piece of polycrystalline large-grain cast silicon was fractured inside the vacuum system in ultra-high vacuum conditions and the fracture surface analysed immediately by SAM.In that way interference from atmospheric and other con-tamination was avoided. Fig. 27 (from Kazmerski’s paper31) compares the features seen in the topographical image of the fracture surface at the top with elemental maps for Ni Al C and 0. Clearly the particle labelled 1 is mostly A1 and 0 probably alumina while particle number 2 is a mixture of Ni A1 and 0. On the other hand the unlabelled particle in the top right-hand corner of the topographical image is mostly carbon and indeed the Auger spectrum revealed that it was in the form of graphite. When the fracture path passed thyough a silicon grain rather than along a grain boundary Auger analysis revealed no elements present apart from silicon itself.It seems that the casting process had concentrated impurities present in the bulk at very low levels at the grain boundaries with subsequent mutual interaction and the precipitation of particles of separate phases. Analyses such as these when combined with electrical measurements allowed identification of those impurities whose action was the most detrimental to the performance of the cell. The ability of AES to provide elemental maps of high spatial resolution has been particularly valuable in the study of processes occurring during the activation of thermionic cathodes, which as already mentioned are complex multi-component devices. For these it has been found that so-called conventional AES in which spot analyses are performed with a relatively large beam size does not provide much useful information as the area analysed is too large an Fig 26.AES analysis of nodules in a cast iron surface. The nodule appears black in the secondary electron image in (a) and is surrounded by a grey annular halo. The SAM images in (b) and (c) are those of carbon and of iron respectively. The nodule is almost pure carbon, probably graphite while the surrounding halo corresponds to a carbon-denuded region. Between the nodules and their haloes is a matrix of iron carbide.30 Fig. 27. SAM analysis of the surface of a grain boundary in polycrystalline cast silicon. At the top a secondary electron image of the surface showing particles embedded in the boundary surface. The elemental maps for Ni Si C and 0 below show that particle 1 is probably alumina and particle 2 probably nickel aluminate.The unlabelled particle in the top right hand corner is mostly carbon probably graphite.31 [to face P. 67 Fig. 28. Elemental maps by SAM analysis at mom temperature of an impregnated thermionic cathode before activation. (a) Absorbed current image; (b) tungsten (1 736 eV); (c) barium (584 eV); (d) oxygen (604 ev); (e) calcium (292 eV) ; (f) sulphur (150 ev) ; (g) carbon (278 ev) ; and (h) osmium (1 850 eV).32 Fig. 29. Elemental maps pregnated thermionic cathode, while the cathode temperature Barium (584 eV) ; (b) oxygen (d) sulphur (149 eV) ; (e) osmium (274 eV) ; and (g) distributio J m e 1983 ANALYTICAL CHEMISTRY A REVIEW 675 a spatially averaged analysis results.For instance in their study of impregnated commercial cathodes Jones et found no correlation between the elemental surface concentrations measured with a beam size of 100 pm and the emission performance of eight different cathode surfaces. However when they turned to SAM with a beam size of less than 1 pm and com-pared the cathodes before and after activation significant correlations both between the distributions of various elements and between the electron emitting areas and the distributions of certain elements were found. Fig. 28 from their paper shows the distribution of seven ele-ments of interest and the adsorbed electron current image (k the negative of the emitted electron current) for an unactivated cathode. In this instance the cathode consisted of a sintered porous tungsten matrix impregnated with a mixture of BaO CaO and A1203 and then sputtercoated with about 0.5 pm of an 0 s - Ru alloy.There is clear correlation between the distributions of tungsten barium oxygen calcium and sulphur the rest of the surface being occupied by carbon. The osmium seems uniformly distributed. After activation and while the cathode was held at 1 145 "C the elemental maps in Fig. 29 were recorded; also shown is the electron emission distribution on the same scale. The correlations have now changed; tungsten and ruthenium are found associated together with sulphur and their areas correspond to the regions of low electron emission. The barium and oxygen distributions correlate and those elements occupy the regions not occupied by the sulphur.In addition the barium and oxygen regions correspond to the areas of high electron emission. The osmium is still more or less uniformly distributed while carbon has disappeared. Hence of the impregnating materi-als BaO is the only one found at the surface after activation and the efficiency of the cathode is directly related to the amount of BaO present. Also clear from Fig. 29 is that sulphur is an effective cathode poison so that the efficiency will also depend on the distribution and amount of sulphur still present after activation. As there is no basic limitation to the temperature at which analysis by AES may be carried out high-resolution elemental maps such as those in Figs. 28 and 29 can be recorded continuously if required at all stages of cathode activation and operat ion.Adhesion Clean metal surfaces when brought into contact adhere to each other very strongly because of the interatomic bonds formed between them. The presence of other elements at the surfaces before contact either as native oxides as deliberately introduced thin films or as airborne contamination is bound to have an effect on the nature and strength of the adhesion and such an effect will not necessarily be deleterious. For example Hartweck and Grabke33 measured the force of adhesion between two iron surfaces that were first cleaned and then covered with fractions of a monolayer of various pure gases and found that at certain coverages the ad-hesion was considerably enhanced. This result is in contrast to that observed when much N 'E 2000 ; 1200 E G 0 .-v) m w-W 2 400 LL 0.2 0.6 1.0 1.4 ( a ) Monolayers 0.2 0.6 1 .o 1.4 1.8 I I I I I ( b) 0.1 0.2 0.3 0.4 0.5 0.6 0.2 0.4 0.6 0.8 1 .o Nitrogen in the boundary AN/& Carbon in the boundary Ac/AF, Fig.30. Variation of the force of adhesion between two pure iron samples with the coverages of (a) nitrogen and of (b) carbon. The coverages are expressed as relative in terms of the ratio of the peak-to-peak height of the principal Auger peak of the adsorbent to the peak-to-peak height of a principal Auger peak of iron.3 676 RIVII~RE AUGER TECHNIQUES IN Analyst Vol. 108 thicker adsorption or reaction films are present when adhesion is definitely reduced. Their results for adsorbed films of nitrogen and of carbon are shown in Fig.30; the coverages of nitrogen and carbon were measured by AES before contact was made between the iron sur-faces in an adhesion testing apparatus built in a UHV chamber. In both instances a maximum in adhesion occurs at a coverage of about 1.0 mondayer the force of adhesion at that point being about three times that for the bare metal. Similar results were found for other adsorbed species at fractional coverage and the authors concluded that the iron - non-metal - iron interatomic bonds formed at the interfaces were obviously much stronger than the iron - iron bonds formed by clean surfaces. At much higher coverages where the formation of three-dimensional compounds is possible it is logical that the cohesion should be weak as the bonds between metals and compounds are generally weaker than between the metals themselves, Analysis in Depth The combination of AES with surface erosion by ion bombardment or occasionally by mechanical lapping as described previously has been used very extensively to obtain mainly qualitative information about the variation in composition with depth; qualitative because the relationship between ion dose and the depth of erosion is usually known only approxi-mately and because the ion beam itself can cause changes in the composition.Nevertheless, the combination has been able to provide information not obtainable in any other way. A few examples will give an indication of how the AES depth analysis has been used. I Sb I 0 10 20 30 40 50 t/m i n Fig. 31. (a) Auger spectrum of an AlSb surface in the “as received” state ie.with the native air-formed oxide film present and the contaminant species silicon chlorine and carbon. (b) Auger spectrum of the same surface after ion bombardment has removed the contaminants but not the oxide. (c) Depth profile through the oxidised AlSb surface showing the five zones identified by the Roman numerials. Region (A) corresponds to the oxide film (B) to the oxide - semiconductor interface and (C) to the bulk emi icon duct or.^ Juvte 1983 ANALYTICAL CHEMISTRY A REVIEW 677 The oxide film formed in dry air on the compound semiconductor AlSb has been studied by Guglielmacci et the material has potential use in optoelectronics particularly in the photo-voltaic conversion of solar energy and the properties of its native oxide film are important.Fig. 31 shows the results of ion bombardment of the surface of the AlSb. The contaminant species silicon chlorine and carbon seen in the as-received state disappeared after bombard-ment light enough to remove only a few atom layers as demonstrated by a comparison of spectra A and B. The Auger peaks corresponding to the remaining elements aluminium, antimony and oxygen were then monitored as a function of bombardment time until the oxy-gen had also disappeared. The authors identified five zones in the film of characteristic composition zone I in which the surface contamination was removed; zone 11 the bulk of the oxide film of constant composition; zone 111 an interface region in which the antimony signal decreased and the antimony to aluminium ratio increased significantly over that in either the oxide or the substrate; zone IV another interface region in which both antimony and alumin-ium signals increased as the oxygen signal fell; and zone V the substrate AlSb.The depth removed to the end of zone IV was estimated to be 750A The presence of an excess of antimony in the interface region of zone I11 would probably modify the electrical charac-teristics of the surface. who were interested in knowing what were the effects of different surface treatments on the compositions of films formed on a 50 + 50 Fe - Ni alloy. Their treatments included ultrasonic cleaning in methanol boiling in hydrogen peroxide and subjecting to an oxygen radiofrequency plasma. Depth profiling in each instance gave the results shown in Fig.32. The nickel-to-iron ratios Oxide films of a different sort were examined by Wittberg et (a) ( b) Ni Ni Fe Fe C C 0 0 I I I 1 I I 0 10 20 0 10 20 I I I I I 1 I 0 10 20 30 40 50 60 Sputtering tirne/rnin Fig. 32. Depth profiles of the surface films formed on a iron - nickel 50 + 50 alloy after being cleaned in (a) methanol (b) hydrogen peroxide and (G) oxygen plasma. The oxide films on the peroxide- and oxygen plasma-treated surfaces are much thicker than on the alloy simply cleaned in methanol. However, the levels of carbon contamination are much lower than after the latter treatment.3 678 RIVIBRE AUGER TECHNIQUES IN Analyst Vol. 108 could be measured and it was found that after a conventional ultrasonic cleaning treatment in methanol the ratio was the same as that of the bulk alloy but that after the other two treat-ments there was either a surface enrichment of iron or a surface depletion of nickel.The oxide films on the peroxide and plasma treated surfaces were thicker than on the ultrasonically cleaned surface being particularly thick after the oxygen plasma treatment as expected. On the other hand the levels of carbon both at the surface and in the bulk were significantly lower after the plasma treatment. Measurements such as these can provide a guide to the surface treatment that should be used to produce a surface condition of the desired properties. In the field of ceramic fabrication the stability of surface composition under severe environ-mental conditions can be a problem in some applications e.g.in the electrical distribution industry and AES with profiling has been used to study progressive changes of ceramic surfaces subjected to such conditions. exposed alumina con-taining a variety of impurity elements to saturated steam (266 "C) at a pressure of 5.3 x lo6 Pa, and carried out depth analyses at regular intervals on specimens removed after various times. The principal result they found was a progressive segregation of calcium (present as an impur-ity at 0.1% m/m in the bulk) to the surface of the alumina. As can be seen from the calcium profiles in Fig. 33 the calcium concentration at the surface increased more or less monotonic-ally with time up to an exposure of 12 d beyond which its concentration at the surface did not increase further.At that point the calcium level at the surface was about two orders of magnitude greater than that in the bulk. Fig. 33 also gives an indication of the observation made by other techniques that after the time at which the surface saturates in calcium the calcium-rich region starts to extend into the bulk and eventually was found some 10 pm from the surface. In parallel work the authors established a correlation between the enhancement of calcium concentration at the surface and the reduction in flexural strength of the alumina. Clearly then commercial-grade alumina used for fabrication of components for electrical and other applications must not be used in conditions where one or more of the impurities added for mechanical strength reasons is likely to segregate to the surface in significant amounts and be lost from the bulk.Another area in which depth profiling with AES has proved most useful is that of ion im-plantation. The latter technique is being applied more and more to the fabrication of certain For example Sinharoy et 14 I I I -12 I I I 0 20 40 60 Sputtering time/min Fig. 33. Depth profiles for calcium into the surfaces of commercial alumina specimens (Coors AD-99 alumina) treated in high-pressure saturated steam (750 lb in-2) for different periods of time (A) Unexposed; (B) I d ; (C) 2 d ; (D) 3 d ; (E) 6 d ; and (F) 12d. The amount of segregated calcium increased with time of exposure up to about 12 d. After that the calcium-rich region starts to extend further into the bulk of the material.36 1019 0 0.04 0.080 0.04 O.Ot Depth/pm Fig.34. Depth profiles into surfaces of GaAs that have been implanted (a) with tellurium (120 keV 10l6 cm-2) and (b) with cadmium (120 keV 10le cme8). The observed profile (1) in (a) corre-sponds well to (2) the theoretically expected profile although the maximum is at a lower concentration. The observed cadmium profile (3) however does not bear much resemblance to (4) the theo-retical profile. The reason is suggested to be diffusion of cadmium towards the surface owing to a temperature rise caused by the implantati~n.~ Jame 1983 ANALYTICAL CHEMISTRY A REVIEW 679 solid-state devices but its usefulness is severely limited without reliable information about the distribution of the implanted species with depth and also the depth a t which its concentration has reached a maximum; in semiconductor jargon the dopant profile.Such profiles of tellurium and cadmium implanted into GaAs were examined by Park et aZ.37 The implants were carried out at 120 keV and a t fluences from 1015 to l O l 6 cm-2 at toom temperature. Tellurium and cadmium are used to alter the local electronic characteristics of the semi-conductor and it is important therefore to know where they are in the surface after implanta-tion. Depth profiling for the two implants gave the results shown in Fig. 34 where the Auger profiles are compared with the dopant profiles predicted from theory. For tellurium the pro-files are similar in shape although the concentration at the maximum is lower than expected, but the cadmium Auger profile is different from the theoretical profile.The maximum is flattened and there is a tail of concentration towards the surface. Indeed some cadmium was detected at the external surface before Auger profiling started. The explanation for such a profile is that cadmium has probably diffused towards the surface during implantation as the current density was high and some local heating may have occurred. From observations of this type can be deduced the implantation conditions needed to produce a locally altered sur-face layer of the depth and dimensions required for the fabrication of the device. In the above example for instance it is clear that implantation of cadmium to achieve the correct profile should be carried out either a t a lower current density or with efficient cooling of the host material .Conclusions Advantages and Disadvantages of AES No one analytical technique whether its application is in the field of surface analysis or in any other field is ever perfect and AES is no exception. The content of this review so far has concentrated on a description of the mode of operation of the technique and of some of the ways in which it has been useful and little mention has yet been made of the drawbacks. It is only fair then to those who might be thinking of applying AES to their own problems to provide a more balanced picture by setting the advantages and disadvantages against each other and they will be mentioned again later when a brief comparison is made of AES with other surf ace analytical techniques.The advantages should have become apparent from the previous sections but for complete-ness are summarised as follows: 1. high spatial resolution ; 2. fast data acquisition; 3. sensitivity to light elements; 4. relatively uncomplicated spectra ; and 5. continuously variable primary energy. The first two of these allow elemental mapping to be performed of which examples have been given. The third represents an advantage over other electron-probe techniques such as electron-probe microanalysis (EPMA) where analysis of the light elements is always difficult. The fourth advantage is again a relative one particularly in comparison with the complemen-tary technique of X-ray photoelectron spectroscopy (XPS) in which the spectra contain both photoelectron and Auger peaks.The fifth is useful if either the sensitivity for a particular element is to be maximised by choosing a primary energy at the maximum of the ionisation cross-section or if a charging problem is energy dependent or if loss spectra are required to be highly surface-specific which can be realised by reducing the primary energy to a low value. Among the inherent disadvantages the two principal ones stem directly from the use of an incident electron beam as the ionising source and they are electron-beam damage and surface charging. Even in the early days of AES when the irradiated area was large of the order of 1 mm diameter and the current density was relatively low typically A cm-2 effects on the surface composition arising from prolonged electron irradiation were observed.With the development of high spatial resolution the electron spot size has been reduced progressively to less than 1000 A but the beam current although also reduced has not been so in proportion, and as a result the current density has leaped to -lo2 A cm-2 for “conventional” AES and to -lo8 A cm-2 for high-resolution AES. Under these conditions the number and variety o 680 RIVI~RE AUGER TECHNIQUES IN Analyst Vol. 108 observations of disturbance of the surface by the incident beam have proliferated. These disturbances can take the forms of electron-induced desorption electron-assisted adsorption, electron-induced or -enhanced surface diffusion and surface decomposition. Some of the most severe effects have been observed with glasses.For example in Fig. 35 (from the work of Ohuchi et aZ.)35 are plotted the decays of the sodium Auger signal for two glass films of different thicknesses and for bulk glass as a function of the time of irradiation at room temperature by an electron beam. A cm-2 must be regarded as mild by today’s standards. The authors were able to show that the rate of decay i.e. of surface decomposition increased with increasing current density but decreased at higher electron energies and at lower temperatures. The irradiation conditions 3 keV and a current density of 5 x 0 1 3 5 7 9 Beam impingement timehin 1 Fig. 35. Decay with time of electron irradiation of the sodium Auger signal from the surfaces of glass films of different thicknesses (A) 1000 A (B) 2000 A and of !C).bulk glass at room temperature.The incident current density was 6 x 10-3A cm-2.3B During ion profiling of glasses there may also be synergistic effects between the ion and electron beams. When Ahn et aL39 carried out simultaneous ion erosion and Auger analysis of a soda-lime glass for a time long enough to have produced a crater of depth about 0.3 pm and then mapped the topography of the crater with a profilometer they obtained the result shown in Fig. 36. The diameter of the ion-eroded area was 3.2 mm and the electron beam being of much smaller diameter (-0.25 mm) was located in the centre of the crater. The profilometer traces show that the bottom of the crater is reasonably flat except where the electron beam has been impinging where there is an additional small crater of depth 600 A.The irradiation conditions chosen were such that for the same length of time the electron beam by itself did not remove measurable amounts of material. Thus both beams are required to be impinging on the surface together for the effect to be produced. The authors attributed the effect to electron-stimulated desorption of surface oxygen leading to enhanced sputtering in the electron irradiated area. The accumulation of surface charge during electron irradiation of insulating or semi-insulating surfaces can also be a problem in AES. It arises of course from the inability of the material to provide an electron current to balance the loss of electrons from the surface by secondary emission so that the surface becomes positively charged.The effect occurs at primary electron energies above that energy at which the secondary emission yield becomes greater than one. For most insulators or semi-insulators this threshold energy is only a few hundred electronvolts much lower than the primary energies normally used in AES. By the same token if the charging surface is flooded with electrons from an auxiliary source at energies well below the threshold e.g. 10-20 eV the excess positive charge at the surface can to a greater or lesser extent be neutralised. This is known as charge neutralisation and the neutralising source as a “flood gun.” Severe charging in AES can produce shifts in the entire spectrum by as much as several hundred electronvolts e.g. for powdered insulating materials Jane 1983 ANALYTICAL CHEMISTRY A REVIEW 681 1 m m Horizontal scale I I Depth scale 1 prn I Fig.36. Profilometer traces across the circular crater of diameter 3.2 mm formed in the surface of a soda-lime glass by ion erosion. In the centre of the ion-eroded area is another crater of additional depth, formed where both ion and electron beams have impinged together. Without the ion beam present the electron beam under the same energy and current density conditions and for the same length of time did not cause any measurable removal of material from the glass.30 and the shift can vary in a random manner following treatment of the material by for instance, ion bombardment. Use of a flood gun for charge neutralisation will not necessarily return the spectrum to the right position but it will stabilise its shift so that spectra recorded after successive surface treatments are at least comparable.The other present disadvantages of AES are not inherent but are the subjects of considerable research and are therefore likely to be minimised progressively. One is the difficulty of quantification and the other is the relative (compared with XPS) absence of chemical informa-tion. Most Auger spectra in AES are still recorded in the differential energy mode in which the raw measure of an elemental concentration is taken from the peak-to-peak height. This is a reasonable approximation as long as the undifferentiated peak width is not changing from one chemical situation to another but unfortunately there is now plenty of evidence that the width does change as does the associated fine structure.Thus the correct measure to use is the area under a peak in the undifferentiated N(E) distribution but here again there are substantial unresolved problems particularly with regard to the treatment of the inelastic tail that follows each peak. The exact amount of the tail to be included in the peak area is still a matter of argument. Further no general quantification treatment has yet been formulated that takes account of concentration inhomogeneities near the surface ; most treatments assume a homogeneous region within the analysed volume. Chemical information is implicit in many Auger peaks in particular those in which one or more electrons arise in the valence band of the solid but because of the very nature of th 682 RIVIERE AUGER TECHNIQUES IN Analyst Vol.108 Auger process the information has been largely inaccessible. In a few instances the changes in the spectra with chemical state are sufficiently dramatic for the various forms of the spectrum to be usable in an empirical “fingerprinting” way but it is only recently that improved physical interpretation coupled with the colossal advances in computing capability has allowed the correct unravelling of the peak shapes. Compared with the technique of XPS in which chemical information is available directly and in general comprehensibly AES still has a long way to go. On the other hand it does seem that the type of chemical information that AES will provide will be rather different from and complementary to that provided by XPS.In the same way as for the advantages the disadvantages of AES may thus be summarised as follows : 1. disturbance of the surface by the incident electrons; 2. surface charging due to the incident electrons; 3. difficulties in quantification; and 4. little direct chemical information as yet. Future Developments To a large extent the foreseeable developments in AES will be those intended to reduce or eliminate the disadvantages listed above. It seems unlikely given the current knowledge of the physics of electron - solid interactions that the potentially damaging effects of the electron beam can ever actually be removed but at least they can be understood and allowed for if that knowledge is extended beyond its present limits which is certainly one direction of develop-ment.In the same direction that of better physical understanding will go the development of quantification of AES as many of the poorly known quantities that enter the quantifica-tion procedure are of physical origin such as the back-scattering factor as a function of energy and atomic number and the inelastic mean free path. Theoretical calculation of these quantities has benefited enormously from the rapidly increasing capacity of computers over the last few years and as a result the models set up to simulate electron paths and electron interactions with atoms and with other electrons are becoming more sophisticated and realistic. Similarly there is no doubt that the amount of effort devoted to extracting chemical informa-tion will expand rapidly in the near future now that several workers have shown the way forward both experimentally and theoretically.On the technical side there will undoubtedly be further moves towards ever higher spatial resolution although a more pressing need in this connection is reduction of the incident current density that goes into the highly focused beams and that can only be achieved use-fully if the sensitivity of detection is increased. Current density reduction without added sensitivity simply means that recording times become unacceptably long. With the use of computers to control data acquisition store the data and then carry out data processing it is not now necessary to record Auger spectra in the differential mode by the conventional modula-tion techniques because spectra acquired in the N(E) mode can be differentiated by the computer if required.There will therefore be a general trend towards initial acquisition in the undifferentiated mode and that in itself could lead to greater sensitivity. Not all the developments of a technique can be foreseen and it is always possible that some major step forward will appear from an unexpected quarter. Comparison with Other Surface Analytical Techniques In every analytical field one can find several techniques in use because invariably no one technique can provide all the information required and they are used therefore in a comple-mentary fashion. This is certainly true of surface analysis and it is becoming the exception rather than the rule to find surface analytical equipment devoted to a single technique.Probably the most common combination as they can both use the same electron energy analyser is that of AES and XPS. In several ways their advantages and disadvantages balance each other. Both provide elemental compositional information at about the same ultimate sensitivity but XPS provides direct chemical information as well. Against that AES can analyse with high spatial resolution whereas at the moment XPS is restricted to an average analysis of a relatively large area. Again AES is a fast technique while XPS is slow but there is always the worry about electron-beam damage in AES such effects being virtually absent in XPS. As it is possible to switch from one technique to the other quickl June 1983 ANALYTICAL CHEMISTRY A REVIEW 683 tion 1.2. 3. 4. 6. 6. 7. 8. 9. 10. 11. 12. 13. 14. 16. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. and without altering the sample position the combination is an obvious and very useful one. The other techniques frequently employed in the same system as AES or XPS or even with both are static secondary ion mass spectrometry (SSIMS) and ion scattering spectroscopy (ISS) . Both use ion probes SSIMS of argon and ISS of helium in the low-energy range (1-4 keV). As their names suggest SSIMS is concerned with the mass analysis of secondary ions positive or negative and ISS with the energy analysis of elastically scattered primary ions. For SSIMS therefore the detector is a mass spectrometer usually of the quadrupole type and for ISS an energy analyser.In fact it has been shown to be possible to use the same energy analyser for ISS as for AES and XPS by reversing the polarities of the deflecting potentials as the energies of positive ions rather than of electrons are analysed. Each of these ion probe techniques has its advantages and disadvantages compared with AES and XPS. SSIMS has much higher sensitivity to some elements by some orders of magnitude can analyse for hydro-gen and hydrogen-containing species and in principle is capable of establishing the nature of surface compounds. On the other hand it is inherently destructive in that the surface has to be removed to be analysed and still very difficult to quantify because of matrix-dependent effects.ISS is the most surface-specific of the surface analytical techniques and quantification is straightforward but there is no possibility of obtaining chemical information while indi-vidual identification and analysis of heavy elements when there are several present together, is difficult. Ideally a surface analysis system should contain all the techniques whose elemental and chemical information is complementary so that as complete a picture as possible is obtained. In practice it is found that such multi-technique systems tend to be counter-productive in that, when more than two or three separate techniques are associated at least one of them is significantly under-used. I t is necessary to decide in advance for any particular problem or range of problems that combination of just two or three techniques will maximise the informa-required and concentrate on making optimum use of those.References Auger P. J . Phys. Rad. 1926 6 206. Lander J. J . Phys. Rev. 1953 91 1382. Powell C. J . Robins J . L. and Swan J . B. Phys. Rev. 1958 110 657. Scheibner E. J . and Tharp L. N. Surface Sci. 1967 8 247. Palmberg P. W. J . Appl. Phys. 1967 38 3137. Harris L. A. GEC Res. Dev. Rep. No. 67C201 (1967); J . Appl. Phys. 1968 39 1419. Weber R. E. and Peria W. T. J . Appl. Phys. 1967 38 4355. Palmberg P. W. Bohn G. K. and Tracy J. C. Appl. Phys. Lett. 1969 15 254. Shirley D. A. Chem. Phys. Lett. 1972 17 312. Larkins F. P. A t . Data Nucl. Data Tables 1977 20 311. Madden H. H. J . Vac. Sci. Technol. 1981 18 677. Seah M. P. and Dench W. A. Surf. Interface Anal. 1979 1 2 . Landolt D. and Mathieu H. J Oberfluche 1980 21 8. Powell C. J . Rev. Mod. Phys. 1976 48 33. Weber R. E. J . Cryst. Growth 1972 17 352. Kny E. J . Vac. Sci. Technol. 1980 17 658. Kleefeld J. and Levenson L. L. Thin Solid Films 1979 64 389. Smith M. A. and Levenson L. L. Phys. Rev. B 1977 16 1365. Lea C. and Seah M. P. Thin Solid Films 1981 75 67. Walls J. M. Hall D. D. and Sykes D. E. Surf. Interface Anal. 1979 1 204. Holloway P. H. Surf. Sci. 1976 54 506. Brockman R. H. and Russell G. J. Phys. Rev. B 1980 22 6302. Davis G. D. Viljoen P. E. and Lagally M. G. J . Electron Spectrosc. 1980 21 135. Rye R. R. Madey T. E. Houston J . E. and Holloway P. H. J . Chem. Phys. 1978 69 1504. Lumsden J. B. Stocker P. J. and Tsai S. C. Appl. Surf. Sci. 1981 7 347. Benndorf C. Egert B. Nobl C. Seidel H. and Thieme F. Surf. Sci. 1980 92 636. Cros A. Derrien J. and Salvan F. Surf. Sci. 1981 110 471. Okada S. Oura K. Hanawa T. and Satoh K. Surf. Sci. 1980 97 88. Verhoeven J. and van Doveren H. Thin Solid Films 1981 77 367. Joshi A. in Strauss B. M. and Cullen W. H. Editors “Fractography in Fdilure Analysis,” ASTM Kazmerski L. L. Appl. Surf. Sci. 1981 7 55. Jones D. McNeely D. and Swanson L. W. Appl. Surf. Sci. 1979 2 232. Hartweck W. and Grabke H. J . Acta Metall. 1981 29 1237. Guglielmacci J . M. Charfi F. and Joullie A. Thin Solid Films 1981 76 69. STP 645 American Society for Testing arid Materials Philadelphia 1978 p. 275 684 RIVI~RE 35. 36. 37. 38. 39. Wittberg T. N. Hoenigman J. R. Moddeman W. E. and Salerno R. L. Appl. Surf. Sci. 1980, Sinharoy S. Levenson L. L. and Day D. E. J. Vac. Sci. Technol. 1979 16 603. Park Y. S. Theis W. M. and Grant J. T. Aflpl. Surf. Sci. 1980 4 445. Ohuchi F. Ogino M. Holloway P. H. and Pantano C. G. Surf. Interface Anal. 1980 2 86. Ahn J. Perleberg C. R. Wilcox D. L. Coburn J. W. and Winters H. F. J. Appl. Phys. 1976, Received December Sth 1982 Accepted December 29th 1982 4 531. 46 4581