Report on the 11th IUVSTA workshop ‘Auger Electron Spectroscopy: from Physics to Data’

SURFACE AND INTERFACE ANALYSIS, VOL. 26, 72È91 (1998)

Report on the 11th IUVSTA Workshop ‘AugerElectron Spectroscopy : from Physics to Data’

M.-G. Barthe� s-Labrousse*CNRS-CECM, 15 rue Georges Urbain, 94407 Vitry Cedex, France

KEYWORDS: AES; electron scattering ; data processing ; Auger imaging

GENERAL PRESENTATION

The International Union for Vacuum Science, Tech-nique and Applications (IUVSTA) has established aworkshop programme to provide a forum for intensescientiÐc debate and discussion between a small numberof experts in a focused scientiÐc Ðeld that falls withinthe area of activity of the ScientiÐc Divisions ofIUVSTA. In this frame, the 11th IUVSTA Workshopwas held in Saint-Pierre de Chartreuse (France), 1È5May 1995, on the topic “Auger Electron Spectroscopy :from Physics to DataÏ.

This workshop brought together experts in the Ðeldof Auger electron spectroscopy with a view of estab-lishing the present status of this spectroscopy in relationto the supply of analytical data on which reliance canbe placed by the user community. The 43 members ofthe workshop were drawn from 16 countries, of whichthere were eight from the UK, seven from the USA andsix from France. A full list of participants is given inTable 1, indicating the speakers, moderators and com-mittee members. The meeting was chaired by Dr J.-P.Langeron and Professor J. E. Castle, and local organiz-ation and registration was undertaken by the Socie� te�

du Vide.FrancÓ aiseThe scientiÐc programme was produced by the Chair-

men and especially by Dr M.-G. Barthe� s working withher programme committee. For reference, the pro-gramme is given in Table 2, where it can be seen thatthere were eight sessions of approximately 3 h each.Each session was addressed by an individual speakerwho had been asked to punctuate his talk with dis-cussion points. The sessions were chaired/moderated byappropriate members of the workshop group. Speakersprovided topic lists as a guide to the content of theirtalks (see Table 3).

Each speaker has, with input from the moderators,provided a written version of his or her presentation. Aclosing summary of the technical content of the meetingwas provided by Professor Castle acting as rapporteur.All this material is presented below. It is hoped that itwill be widely discussed in national forums and will

* Correspondence to : M.-G. Barthe� s-Labrousse, CNRS-CECM, 15rue Georges Urbain, 94407 Vitry Cedex, France. E-mail :barthes=glvt-cnrs.fr.

form the basis for a future international meeting inorder that this important topic will be regularlyupdated.

The meeting was held in a small hotel, the “Beau-SiteÏ,in a quiet village in the heart of the Chartreuse Moun-tains close to Grenoble, France. The weather andscenery were superb and it seemed likely that there wereno other visitors in the village than the participants tothe workshop. It is a tribute to the quality of themeeting that sessions, to the last, had 100% attendancewith eager and focused discussion throughout. The foodsupplied was exceptionally good, served at tables ofeight in a restaurant that also lent itself well to dis-cussion. The Socie� te� du Vide, represented byFrancÓ aiseBernard Dallery as the person responsible for the logis-tics of the meeting, made excellent arrangements fortransportation to the railway station and to the air-ports. Delegates were in the main accommodated in theconference hotel but a small overÑow group were in anadjacent hotel. A social occasion was provided by theincorporation into the programme of a visit to theChartreuse monastery and distillery.

PIERRE AUGER (1899–1993)

Pierre Auger was born in Paris in 1899. After his uni-versity studies and agre� gation, he joined, in 1922, theLaboratoire de Chimie Physique headed by Jean Perrin,to prepare a thesis on the photoelectric e†ect. A fewmonths later, he published the Ðrst description of thephenomenon that now bears his name.

It was more than 30 years before technologicalprogress changed Pierre AugerÏs discovery into apowerful means of studying atoms and solid surfaces.Even though there still remains a lot to be done, espe-cially from a theoretical point of view, Auger electronspectroscopy has now proved to be particularly fruitfulin many Ðelds like fundamental physics (with the studyof atoms, molecules, collision processes . . .) or funda-mental and applied surface science, and fortunatelyPierre Auger lived long enough to see the formidablescientiÐc and technological impact of his Ðrst discovery.

Novelty and diversity are the most suitable words tocharacterize Pierre AugerÏs whole life. The discovery ofradiationless transitions in the 1920s was followed bythe Ðrst experimental observation of spallation in the

CCC 0142È2421/98/010072È20 $17.50 Received 19 June 1997( 1998 John Wiley & Sons, Ltd. Accepted 5 August 1997

REPORT ON THE 11TH IUVSTA WORKSHOP 73

Table 1. List of participants at the 11th IUVSTA workshopaDr M. A. Baker (C) Ispra Italy

Dr Marie-Genevieve Barthe� s (C) Vitry France

Prof. Hugo Bender Leuven Belgium

Dr Hugh Bishop (M 7) Harwell UK

Dr Reuven Brenner Technion Israel

Prof. James E. Castle (C) Guildford UK

Prof. Jacques Cazaux (IS 7 ½M 5) Reims France

Dr Kenton D. Childs Eden Prairie USA

Dr Peter Coxon East Grinstead UK

Dr Peter Cumpson Teddington UK

Dr Alain Dubus Bruxelles Belgium

Prof. Mohamed El Gomati (M 3) York UK

Dr Stephen Gaarenstroom (IS 6) Warren USA

Dr Gyo� rgy Gergely Budapest Hungary

Dr Onno Gijzeman Utrecht The Netherlands

Prof. Keisuke Goto Nagoya Japan

Prof. John Grant (M 5) Dayton USA

Prof. Aleksander Jablonski (IS 3) Warsawa Poland

Dr Laszlo Ko� ve� r Debrecen Hungary

Dr Jean-Paul Langeron (C) Vitry France

Prof. Hans J. Mathieu Lausanne Switzerland

Dr Ralf Mu� ller Kaiserslautern Germany

Prof. Hans Oechsner (C ½M 8) Kaiserslautern Germany

Dr Eric Ollivier Suresnes France

Dr Cedric J. Powell (C ½IS 2) Gaithersburg USA

Prof. Martin Prutton (IS 8) York UK

Prof. David Ramaker Washington USA

Dr Monique Repoux Sophia-Antipolis France

Dr Nadia Roose Brussels Belgium

Prof. Jose� Sanz Madrid Spain

Dr Martin Seah (C ½IS 5) Teddington UK

Prof. Peter Sherwood (M 6) Manhattan USA

Prof. Ryuichi Shimizu (M 2) Osaka Japan

Prof. Daniel Spanjaard (M 1) Orsay France

Dr Robert Sporken Namur Belgium

Dr Hans Steffen (M 6) Dresden Germany

Dr Stephen M. Thurgate Murdoch Australia

Dr Tanaka Chigasaki Japan

Prof. Sven Tougaard (IS 4) Odensee Denmark

Dr Sergio Valeri Modena Italy

Prof. Peter Weightman (IS 1) Liverpool UK

Dr W. S. M. Werner (M 4) Wien Austria

Miss Hai Qing Zhou Arlington USA

Bernard Dallery Socie� te� FrancÓ aise du Vide

a C ¼Scientific Committee ; IS¼Invited Speaker ; M¼Moderator.

1930s, and of the giant cosmic ray showers in the 1940s.After World War II, Pierre Auger successfully partici-pated in the creation of the CEA (French AtomicEnergy Commission), the Science Department at

UNESCO (United Nations Education, Science andCultural Organization), CERN (European Organizationfor Nuclear Research), CNES (French National Centrefor Space Studies) and the European Space Agency . . .

Table 2. Programme for 11th IUVSTA workshop

Sessions Invited speaker Moderator

1. The Spectral Profiles of Auger Transitions P. Weightman (UK) D. Spanjaard (France)

2. Inelastic Electron Scattering in AES C. J. Powell (USA) R. Shimizu (Japan)

3. Elastic Scattering of Auger Electrons. Distribution of A. Jablonski (Poland) M. El Gomati (UK)

Inner-shell Ionizations

4. Escape of Electrons S. Tougaard (Denmark) W. S. M. Werner (Austria)

5. Instrumental Effects M. P. Seah (UK) J. Grant (USA)

J. Cazaux (France)

6. Data Processing S. Gaarenstroom (USA) P. Sherwood (USA)

J. A. Steffen (Germany)

7. Artefacts J. Cazaux (France) H. Bishop (UK)

8. Imaging M. Prutton (UK) H. Oechsner (Germany)

( 1998 John Wiley & Sons, Ltd. SURFACE AND INTERFACE ANALYSIS, VOL. 26, 72È91 (1998)

74 REPORT ON THE 11TH IUVSTA WORKSHOP

Table 3. Session contents

Session 1: The Spectral Profiles of Auger Transitions

The Auger spectra of free atoms:

profiles ; theory compared with experimentI Auger

energies, relaxation processesI Auger

intensitiesI Auger

structureI Multiplet

processesI Coster–Kronig

cascadesI Auger

distribution of Auger electronsI Angular

The Auger spectra of molecules and condensed matter :

shifts and extra-atomic relaxationI Chemical

parameter studiesI Auger

profiles as probes of local electronic structureI Auger

correlation effects ; Cini–Sawatzky theoryI Electron

processes and core peak lineshapes and lifetimesI Auger

Session 2: Inelastic Electron Scattering in AES

I Terminology

of inelastic mean free paths, effective attenuation lengths and mean escape depthsI Calculations

of inelastic mean free paths and effective attenuation lengthsI Measurements

and measurements of inner-shell ionization cross-sectionsI Calculations

issuesI Other

Session 3: Elastic Scattering of Auger Electrons. Distribution of Inner-shell Ionizations

outline of quantitative analysis with correction factors. Assumptions of quantitative AESI General

scattering cross-sections for medium-energy electronsI Elastic

scattering effects in AESI Elastic

depth distribution function (DDF)I Emission

depth distribution function. Key role of the depth distribution of inner-shell ionizations in AES and electron probeI Excitation

microanalysis (EPMA)

backscattering factor (BF) in AES. Overview of different definitions of the BF encountered in the literatureI The

of the backscattering factor on energy and the primary beam incidence angleI Dependence

BF for complex systemsI The

correction factors in AESI Remaining

Session 4: Escape of Electrons

diffraction, forward focusing and holographyI Channelling,

depth distribution function (homogeneous and non-homogeneous systems)I Emission

on peak intensity, background and inelastic tailsI Effects

for different surface morphologiesI Algorithms

Session 5: Instrumental Effects

I Introduction

typesI Spectrometer

positioning effectsI Specimen

systems (simple)I Detection

systemsI Multidetector

resolution and sensitivityI Spectral

of lens, electrodes and structures on spectral shape—internal scatteringI Effect

of spectrometer intensity scalesI Calibration

effectsI Angular

effectsI Spatial

spectrometersI Other

of instrumental effects in dataI Removal

data banksI Spectral

Session 6: Data Processing

Qualitative Analysis

and chemical state identificationI Element

systemsI Expert

I Databases

Lineshapes

vs. dN /dEI N(E)

I Smoothing

analyzer broadening correctionsI Energy

fitting, background suppression, energy loss and in-depth effectsI Background

SURFACE AND INTERFACE ANALYSIS, VOL. 26, 72È91 (1998) ( 1998 John Wiley & Sons, Ltd.


Table 3. Continued

Spectral Overlap

fittingI Curve

derivatives, divisions, deconvolutionsI Differences,

regression and factor analysis, additivity of spectraI Linear

Quantitative Analysis

standardless methodsI Automated

transferabilityI data

Depth Profiling

beam effects on depth resolutionI Ion

beam effects on concentration (preferential sputtering)I Ion

interface profilesI Modelling

Session 7: Artefacts

of the specimen during analysisI Modifications

field built up and its consequencesI Electric

I Others

Session 8: ImagingÔresolutionI Spatial

effectsI Topography

analysisI Image

He also wrote poetry (both in French and in English)and sculpted small objects.

Pierre Auger died in Paris, 24 December 1993.

THE SPECTRAL PROFILES OF AUGERTRANSITIONS

Auger spectra of free atoms

From a theoretical point of view the Auger processmust be considered as a one-step process. However, ifthe lifetime of the initial hole is sufficiently long, it canbe assumed, with a good approximation, that the cre-ation of the initial hole can be separated from the fol-lowing cascade. It is this point of view that is usuallyadopted in theoretical calculations.

These calculations are made, most of them, within thefollowing approximations :(1) Independent particle model.(2) Central Ðeld approximation.(3) Self-consistent Ðeld (SCF) approximation.The following questions can also arise :(1) Are relativistic calculations necessary?(2) Can a local Slater potential be used?(3) Is the frozen orbital approximation valid?With these approximations, the wavefunctions areformed from Slater determinants and the resolution ofthe Schro� dinger equation can be done within theHartreeÈFock, HartreeÈFockÈSlater, DiracÈFock orDiracÈFockÈSlater schemes. The resolution of theseequations in the ground state give :(1) the total energies ;(2) the eigenvalues of electrons ;(3) the wavefunctions of single electron orbitals.The above properties can also be found for excitedstates provided that they are orthogonal to the groundstate. This implies a “relaxationÏ of atomic wave func-

tions between di†erent states and the energy di†erenceare deduced from *SCF calculations.

Auger intensities. The atomic coreÈcoreÈcore Augerspectra exhibit multiplets as a consequence of LSJ spin-orbit coupling. The calculation of such a spectruminvolved the determination of these multiplets in theÐnal state and the transition probabilities from theinitial state to each component of them. The corre-sponding matrix elements can be evaluated, as statedabove, using atomic wavefunctions formed from Slaterdeterminants of single-electron orbitals found fromatomic structure calculations. Can then the frozenorbital approximation be used? The results obtainedusing this approximation for the KLL Auger transitionof Ne and Kr show non-negligible errors : D6 eV forNe and D19 eV for Kr. It is thus necessary to use*SCF calculations in which the atomic potentials “relaxÏbetween the initial and Ðnal states. In these conditionsgood agreement is obtained between theoretical andexperimental spectra for nearly all Auger kinetic ener-gies.

Auger spectra in solids

We will distinguish the coreÈcoreÈcore (CCC) Augertransition from the Auger transition involving thevalence electrons : coreÈcoreÈvalence (CCV) and coreÈvalenceÈvalence (CVV) transitions.

The CCC spectra. The main modiÐcations of CCCspectra involving the levels i, j and k due to the solidenvironment are the shifts (i, j, k) of the kinetic*Ekenergies of Auger electrons. Let us call the energy*Eb(i)shift of the initial level measured by XPS; can be*Ebsplit into two contributions

*Eb(i)\ *V [ *R



*V is the shift due to the environment on the initialstate and *R is the contribution due to the response ofthe local electronic structure to the creation of coreholes.

Similarly one can write, with good approximation

*Ek(i, j, k) D [*V ] 3*R

Wagner1 deÐnes an Auger parameter a in the Ðnal statesuch that

*a\ *Ek(i, j, k) ] *Eb(i)D 2*R

and Matthew2 deÐnes an Auger parameter b in theinitial state such that

*b\ *Ek(i, j, k)] 3*Eb(i)D 2*V

It is of considerable interest to separate the e†ects dueto the initial state from those due to the Ðnal state,because this gives insight into charge transfer and elec-tronic screening. Thomas and Weightman3 relate thecore atomic potential V to the atomic valence charge qby

V \ C] kq ] U

where C is the contribution due to the nucleus and thecore electrons, k is the change in the core potentialwhen one valence electron is removed and U is the con-tribution due to the chemical environment.

If we assume that k and q depend linearly on theoccupancy of core orbitals N, it can be shown that

*a\ *CqA dkdNB

]Ak [ 2

dkdNB dq

dN] dU

dND

The Ðrst term comes from the relaxation of theoccupied valence orbitals after the creation of the corehole ; the second term is due to the transfer of screeningcharge from the surrounding atom to the core-ionizedatom; and the last term comes from polarization of theenvironment after the creation of the core hole.

In a metal, dU/dN \ 0 and dq/dN \ 1, thus *(dq/dN)\ 0 and

*a\ *qdkdN

dk/dN being derived from calculations and *q deducedfrom experiments. The results obtained on the alloysAuZn and AuMg lead to Au~0.12Zn0.12 andAu~0.17Mg0.17, with an accuracy of D0.02 el.

However, a more careful analysis of k in the metaland in the atom leads to the conclusion that k should bealso a function of q

k(N, q) \ a ] bN ] dq

where b represents contraction of the valence shell whenthe atom is core ionized and d comes from dilation(contraction) of the valence shell, which follows the gain(loss) of valence charge.

A systematic analysis of V and k through the ele-ments of the Ðrst and second series of the PeriodicTable leads to a new deÐnition of electronegativity.

The CCV and CVV spectra in relation to local electronic struc-ture. T ransitions CCV involving no d electrons. Whenthe Auger process involves the conduction electrons of ametal, it is clear that the corresponding wavefunctions

di†er noticeably in the initial state and in the Ðnal state.Several theoreticians have been able to establish a “Ðnal-state ruleÏ. This rule states that :(1) the shape of the Auger line involving an s or a p

band is given by the local density of states in theÐnal state ;

(2) the intensity of this line is given by the local elec-tronic conÐguration in the initial state.

Calculations of the CCV Auger transitions in Na, Mgand Al using a spherical potential and a self-consistentdetermination of the local density of states are in goodagreement with experimental results.4

CV V T ransitions in transition metals. The Ðnal staterule does not apply when the electronic correlationsbetween the two holes in the Ðnal state cannot beneglected. This is the case when these two holes belongto the d band. If this band is completely Ðlled and if thecoulombic repulsion U between the holes is large com-pared to the bandwidth W , these two holes give rise toa bound state below the d band and the correspondingAuger spectrum is quasi-atomic, as shown by Cini5 andSawatzky.6 This is the case for cadmium. However, onecan vary the ratio U/W by alloying. For example, inpure Ag, W D 3.5 eV and the ratio U/W is not largeenough to give rise to a bound state. On the other hand,in the alloy AgMg, the width of the d band is reducedto D2 eV and the bound state appears.

Conclusion

The study of the Auger e†ect has now reached the pointwhere we have a good understanding of the Auger pro-Ðles in free atoms as well as their modiÐcations due to acrystalline environment. The local nature of the Augere†ect allows a determination of the local characteristicsof the electronic environment : charge transfer, screen-ing, electronic correlations.

References

1. C. D. Wagner, Farad.Discuss . Chem.Soc. 60, 291 (1975).2. S. D. Waddington, P. Weightman, J. A. D. Matthew and A. D.

L. Grossie, Phys.Rev.B 39, 10239 (1989).3. T. D. Thomas and P. Weightman, Phys. Rev. B 33, 5406

(1986).4. P. S. Fowles, J. E. Inglesfield and P. Weightman, J. Phys.

Condens.Matter 3, 641 (1991).5. M. Cini, Solid State Commun. 20, 605 (1976); 24, 681 (1977).6. G. A. Sawatzky, Phys.Rev. Lett . 39, 504 (1977).

INELASTIC ELECTRON SCATTERING IN AES

Terminology

The ASTM Committee E-42 on Surface Analysis hasrecently approved or is currently considering new deÐ-nitions for the following terms : inelastic mean free path ;emission depth distribution function (for a measuredsignal) ; average emission function decay length ; e†ec-tive attenuation length ; mean escape depth ; and infor-mation depth. The new deÐnitions are needed toaccount for the e†ects of elastic electron scattering. Thecurrent deÐnitions can be found in ASTM Document E



673, Standard Terminology Relating to SurfaceAnalysis.1

H. E. Bishop commented that inelastic mean freepaths, e†ective attenuation lengths and other relatedterms should be expressed in mass-thickness unitsrather than conventional length units. M. El Gomatisuggested that the information depth should be relatedto the signal/noise ratio in an experiment, although J.Cazaux pointed out that the information depth wouldthen depend on the data acquisition time. Cazaux alsorecommended that the symbol often used for the emis-sion depth distribution function (often given as /)should be changed to minimize possible confusion withthe '(oz) function commonly used in electron probemicroanalysis.

Calculations of inelastic mean free paths and e†ectiveattenuation lengths

Inelastic mean free paths (IMFPs) have been recentlypublished for 50È2000 eV electrons in 56 materials byTanuma et al.2 These IMFP calculations have beenmade using an algorithm due to Penn2 in which experi-mental optical data are used to give information on thedependence of the inelastic scattering probability onenergy loss, and theory is used to describe the depen-dence of the inelastic scattering probability on momen-tum transfer. The accuracy of the optical data for eachmaterial was checked using two sum rules.3 PennÏsalgorithm has uncertainties, estimated to be 10È20%,due to the neglect of vertex corrections, self-consistency,exchange and correlation, which are expected to be sys-tematic.

It has been possible to Ðt the calculated IMFPs foreach of the 56 materials to an empirical modiÐcation ofBetheÏs equation4 for electron inelastic scattering inmatter.2 This equation provides a physically reasonablebasis for describing the dependence of the calculatedIMFPs on electron energy and material parameters ; itis also superior to other empirical expressions that havebeen proposed earlier. The values of the four param-eters in the modiÐed Bethe equation for each materialhave been analysed to yield a predictive equation(currently designated TPP-2M) with which IMFPs canbe estimated for other materials.

Examples were given of the IMFP calculations forelements, inorganic compounds and organic com-pounds, the Ðts with the modiÐed Bethe equation andthe predictions of TPP-2M. Comparisons were alsomade with the widely used expressions for e†ectiveattenuation length (EAL) that were developed by Seahand Dench.5 While the IMFP and EAL values are gen-erally similar, systematic di†erences exist in the depen-dences on electron energy and material parameters ;further experiments are needed to test the IMFP calcu-lations. The IMFP calculations of Tanuma et al. are forbulk solids, and extensions to take account of surfaceexcitations (and interface excitations for thin-Ðlmstructures) would be desirable ; subsequent to the Work-shop, Chen and Yubero et al. have published papersaddressing this issue.6 Further clariÐcation is alsoneeded of the appropriate choice for the parameter Nv(the number of valence electrons per atom or molecule)in the TPP-2M formula.

Several authors have obtained EALs from calculateddepth distribution functions (DDFs) for Auger electronsignals. Jablonski7 found that the DDFs were essen-tially exponential in simulations for a cylindrical mirroranalyser (CMA) oriented with its axis normal to thespecimen surface. The EAL was found to be proportion-al to the IMFP for several specimens, and the pro-portionality constant ranged from 0.77 (for Ni) to 0.92(for Al). Werner8 showed that the DDFs could be non-exponential for other emission angles and developed anempirical expression giving the dependence of the DDFon the IMFP, the total mean free path (which includesthe contributions of elastic electron scattering), theemission angle and the depth of the emitting atoms.Ding and Shimizu9 have calculated DDFs in simula-tions of Auger electron emission from Cu and Au sur-faces that were detected with a CMA tilted at variousangles to the surface normal. They found that the meanescape depths were reduced from the correspondingIMFPs by about a factor of two due to elastic electronscattering and geometrical e†ects.

S. Tougaard commented that the dependence used inthe Penn algorithm of the electron energy-loss functionon momentum transfer (the Lindhard function) was rea-sonable for metals but that he had found a weakerdependence for insulators. He also indicated that “e†ec-tiveÏ IMFPs could depend on emission angle and depthof the emitting atom due to surface excitations (e.g.surface plasmons in free-electron-like solids). Wernerrecommended that the di†erential inverse mean freepath be used in discussions of inelastic scattering nearsurfaces (rather than the IMFP) to avoid difficultiesassociated with describing speciÐc excitation modes ator near surfaces.

Experimental tests

There are several techniques by which IMFPs andEALs can be measured. In addition, limited informationconcerning IMFPs can be obtained from measurementsof absolute and relative Auger electron yields.

The most well-known method for measuring EALs isthe thin-Ðlm overlayer method. Although this method isconceptually simple, there are many sources of experi-mental error (e.g. Ðlm morphology and uniformity,accurate measurement of Ðlm thickness, surface andinterface reconstruction, interface mixing, surfaceroughness, surface and interface excitations, couplingbetween extrinsic and intrinsic excitations, and angularanisotropies in electron transport in orderedmaterials).10 As a result, there can be large discrep-ancies (e.g. up to a factor of three) in the results of sup-posedly identical measurements.11 The potentially largeand often unspeciÐed systematic errors in many of thepublished EAL measurements make it difficult todevelop improved predictive EAL equations and tomake meaningful comparisons with calculated IMFPs.It should also be noted that angular anisotropies inelectron transport in ordered Ðlms can lead to appre-ciable deviations from the exponential dependence ofAuger intensities on Ðlm thickness that is expected fromsimple models.12

The IMFPs can be measured relatively simply byelastic peak spectroscopy.13 These measurements



depend on knowledge of di†erential cross-sections forelastic electron scattering.14 While relatively few IMFPmeasurements have been reported by this technique,there is often good agreement between experimentaland calculated IMFPs, although there are discrepanciesin the dependences on electron energy for a fewmaterials. Surface excitations are important in thesemeasurements,6 and it is likely that specimen roughnessshould be considered.

Yubero et al. have reported an analysis of measuredreÑection electron energy-loss spectra of for inci-ZrO2dent electron energies between 200 and 2000 eV.15 Thisanalysis led to determination of IMFPs that showedgood agreement with values obtained from TPP-2M.Similar measurements by Yubero et al.,6 published sincethe Workshop, indicate that the IMFPs for a givenenergy in Si and Fe decrease for more grazing angles ofelectron incidence and emission and are in generalagreement with a theoretical model developed by thesame group.

Another technique for measuring IMFPs is to deter-mine the decrease in electron transmission (at a selectedelectron energy) through thin Ðlms of increasing thick-ness.16 The observed attenuation in these experimentsdepends, however, on both inelastic scattering andelastic scattering outside the acceptance solid angle ofthe analyser. The results are also expected to depend onÐlm morphology and thickness, surface excitations,surface reconstruction and the substrate.

Measurements of absolute Auger electron yields canbe analysed to give IMFP data but peak intensitieshave to be measured carefully and inner-shell ionizationcross-sections and analyser parameters need to beknown with sufficient accuracy. Powell et al.17 havereported IMFPs for two elements from proton-excitedAuger electron yields, while Seah and Gilmore18 havemeasured electron-excited Auger electron yields frommost solid elements and have compared these yieldswith values expected from TPP-2M and the Gryzinskiand Casnati et al. formulae (see below) for inner-shellionization cross-sections. The agreement in the lattercomparisons is generally satisfactory, particularly ifTPP-2M is used excluding the 4f electrons for the rareearth metals.

Relative Auger electron yields from two elements inbinary compounds can be compared to give informa-tion on IMFPs.19 These measurements are based onexpected corrections to elemental sensitivity factors andrequire careful intensity measurements (including cor-rections for lineshape changes with changes of chemicalcomposition), corrections for the backscattering factor,and knowledge of the stoichiometry in the surfaceregion probed by the measurement. Surface roughnesscould also be a factor in these experiments.

There was extensive discussion on factors that coulda†ect IMFP and EAL measurements. H.-J. Mathieupointed out that surface roughness of a specimen wouldbe important, particularly for near-glancing emissionangles. G. Gergely stated that surface roughness e†ectswould be very small in his recent IMFP measurementsby elastic peak spectroscopy because the specimenswere rotated while being cleaned by ion sputtering. M.Prutton indicated that information on specimen rough-ness could be obtained from analysis of measurementsof backscattered electrons by four detectors on the

Auger electron microscope at the University of York. D.Spanjaard pointed out that a substrate could modifythe measured lineshape associated with intrinsic pro-cesses from an overlayer thin Ðlm. S. M. Thurgate sug-gested that specimen crystallinity could be important,even for polycrystalline specimens, in elastic peak spec-troscopy. W. S. M. Werner remarked that di†raction-type e†ects should be smeared out for the relativelylarge range of elastic scattering angles, and A. Jablonskiindicated that e†ects due to ordering should be impor-tant only for relatively small scattering angles. Gergelypointed out that the elastic peak intensity from a siliconsingle-crystal surface was reduced after amorphizationof the surface by ion bombardment. Thurgate com-mented that the amorphization depth would depend onthe ion bombardment parameters and that the resultinge†ects on the elastic peak intensity of an amorphouslayer on a crystalline substrate would depend on theelectron energy. Prutton draw attention to the pioneer-ing work of Aberdam and Baudoing,20 who demon-strated angular anisotropies in Auger electronintensities from single-crystal surfaces ; while such aniso-tropies lead to complexity in quantitative surfaceanalyses by Auger electron spectroscopy, they can yielduseful structural information.

Calculations and measurements of inner-shell ionizationcross sections

A simple equation introduced by Bethe4 has been exten-sively used for the analysis of inner-shell ionizationcross-sections.2 This equation is convenient because itsrange of validity can be easily assessed (from a so-calledFano plot of measured cross-sections) and becausecross-section data for di†erent elements can be simplyscaled and compared.

There are many measurements of K-shell ionizationcross-sections but relatively few for other shells. Cross-sections for K-shell and ionization Ðt theL23-shellBethe equation if the overvoltage ratio U (equal to theincident electron energy divided by the appropriatecore-electron binding energy) is greater than four.

Gryzinski22 has developed a classical description ofinner-shell ionization that has been widely used becauseof its analytical simplicity and general utility. Empiricalformulae for K-shell ionization have been published byCasnati et al.23 and by Jakoby et al.24 The latter twoformulae provide good descriptions of near-threshold(U \ 4) measurements of K-shell ionization cross-sections (as well as for U [ 4). The Gryzinski equationagrees reasonably well (generally better than 10%) withmeasured K-shell cross-sections for U [ 4 but there arelarger discrepancies for U \ 4.21 Smith et al.25 havereported delayed onsets in near-threshold measure-ments of the yields of Auger electrons fromN67O45O45Au, Pb and Bi as a function of incident energy. Thesedelayed onsets are associated with the large centrifugalbarrier in the potential for ionization from states of highangular momentum, and it is thus likely that formulaefound useful for describing the near-threshold ioniza-tion of K-shell electrons will be less accurate for ioniza-tion from subshells with higher angular momentum. Leeet al. have also reported deÐciencies in the Gryzinskiformula in simulations of Auger electron intensities for



ÐlmÈsubstrate systems.26 Further comparisons of pre-dicted and measured inner-shell ionization cross-sections are given in another review.27

Other issues

Jablonski and Powell28 have made recommendationsfor choosing the appropriate inelastic scattering param-eter for quantitative applications of Auger electronspectroscopy. The IMFP should be used for measure-ments of the surface composition (of a material that ishomogeneous over the information depth), while theEAL should be used for determining the thickness ofoverlayer Ðlms (as long as the DDF is close toexponential). The mean escape depth (obtained from theDDF) should be used to determine the average depth ofanalysis.

Summary

Inelastic electron scattering is important in Auger elec-tron spectroscopy for three main reasons. First, thesurface sensitivity of an experiment is determined by therelevant inelastic scattering cross-section or IMFP.Second, information on the inelastic scattering prob-ability as a function of energy loss (the so-called energy-loss function) is useful in deducing the intensity ofAuger electrons that reach the detector without inelasticscattering. Finally, inelastic scattering parameters areneeded to make the matrix corrections required inquantitative surface analyses.

The following needs are believed to be important forthe future :(1) While information is now available on the shapes of

the energy-loss functions for many materials, analgorithm is needed to correct for surface excitationsat di†erent Auger electron emission angles.

(2) Adequate data are now available for K-shell ioniza-tion cross-sections, but more experimental cross-sections are required for ionization from othershells.

(3) The IMFPs have been calculated in a consistentway for 56 materials, and the TPP-2M formulaappears to be useful for predicting IMFPs in othermaterials. These calculations, and the TPP-2Mformula, need to be evaluated by further measure-ments.

(4) Values of the IMFP, the EAL and the mean escapedepth are needed for di†erent analytical applicationsin Auger electron spectroscopy. Further experimen-tal tests need to be made of algorithms and simula-tions to assess their validity in describing the e†ectsof elastic electron scattering for di†erent materials,di†erent specimen morphologies and di†erentexperimental conÐgurations.

References

1. Annual Book of ASTM Standards , Vol. 3.06. ASTM, Phila-delphia (1995).

2. S. Tanuma, C. J. Powell and D. R. Penn, Surf . Interface Anal .21, 165 (1994) and references therein.

3. S. Tanuma, C. J. Powell and D. R. Penn, J. Electron Spectro-sc.Relat . Phenom. 62, 95 (1993).

4. H. Bethe,Ann. Phys. 5, 325 (1930).5. M. P. Seah and W. A. Dench, Surf . Interface Anal . 1, 2

(1979).6. Y. F. Chen, J. Vac. Sci . Technol . A 13, 2665 (1995); F.

Yubero, D. Fujita, B. Ramskov and S. Tougaard, Phys. Rev. B53, 9728 (1996).

7. A. Jablonski, Surf . Interface Anal . 15, 559 (1990).8. W. S. M. Werner, Surf . Interface Anal . 18, 217 (1992).9. Z.-J. Ding and R. Shimizu, Surf . Interface Anal . 23, 351

(1995).10. C. J. Powell, J . Electron Spectrosc . Relat . Phenom. 47, 197

(1988).11. C. J. Powell and M. P. Seah, J. Vac. Sci . Technol . A 8, 735

(1990).12. C. Koziol, G. Lilienkamp and E. Bauer, Phys. Rev. B 41, 3364

(1990); W. F. Egelhoff, Phys.Rev.B 30, 1052 (1984).13. B. Lesiak, A. Jablonski, Z. Prussak and P. Mrozek, Surf . Sci .

223, 213 (1989); H. Beilschmidt, I. S. Tilinin and W. S. M.Werner, Surf . Interface Anal . 22, 120 (1994).

14. A. Jablonski and S. Tougaard, Surf . Interface Anal . 22, 129(1994).

15. F. Yubero, J. M. Sanz, J. F. Trigo, E. Elizalde and S. Tougaard,Surf . Interface Anal . 22, 124 (1994).

16. R. J. Stein, Surf . Sci . 60, 436 (1976); C. Martin, E. T.Arakawa, T. A. Callcott and R. J. Warmack, J. Electron Spec-trosc.Relat . Phenom. 42, 171 (1987).

17. C. J. Powell, R. J. Stein, P. B. Needham, Jr. and T. J. Driscoll,Phys. Rev.B 16, 1370 (1977).

18. M. P. Seah and I. S. Gilmore, J. Vac. Sci . Technol . A 14, 1401(1996); M. P. Seah, personal communication.

19. For example : A. I. Zagorenko and V. I. Zaporozchenko, Surf .Interface Anal . 17, 237 (1991); T. Wirth, Surf . Interface Anal .18, 3 (1992).

20. For example : D. Aberdam, R. Baudoing, E. Blanc and C.Gaubert, Surf . Sci . 71, 279 (1978).

21. C. J. Powell, in Electron Impact Ionization, edited by T. D.Mark and G. H. Dunn, p. 198. Springer-Verlag, New York(1985); C. J. Powell, in Microbeam Analysis—1990, editedby J. R. Michael and P. Ingram, p. 13. San Francisco Press,San Francisco (1990).

22. M. Gryzinski, Phys.Rev. 138, A336 (1965).23. E. Casnati, A. Tartari and C. Baraldi, J . Phys. B 15, 155

(1982).24. C. Jakoby, H. Genz and A. Richter, J . Phys. Coll . 48, C9-487

(1987).25. D. M. Smith, T. E. Gallon and J. A. D. Matthew, J. Phys. B 7,

1255 (1974).26. C. L. Lee, H. Gong and C. K. Ong. Surf . Interface Anal . 21,

199 (1994).27. C. J. Powell, to be published.28. A. Jablonski and C. J. Powell, Surf . Interface Anal . 20, 771

(1993).

ELASTIC SCATTERING OF AUGERELECTRONS. DISTRIBUTION OFINNER-SHELL IONIZATIONS

The session started with a brief account of the currentstate of quantitative analysis in Auger electron spectros-copy with the use of correction factors. References tohistorical reviews over the last two decades (Refs 1È4,and references therein) were given. In all these formal-isms the userÏs main aim was how to obtain reliablemeasurements by AES to aid one in :(1) determination of the surface composition of the

surface under investigation ;(2) an estimate of the average depth of analyses ;(3) an estimate of the thickness of an analysed over-

layer.A simple mathematical formulation relating the

Auger electron intensity to the concentration of a givenelement was presented, making assumptions of instru-



mental geometries applicable to the most widely usedanalysis : CMA and CHA types. SpeciÐc to this expres-sion is the fact the elastic scattering of the signal (Auger)electrons by the host atoms was neglected. A corollaryof this formalism implied that the depth distributionfunction (DDF) had an experimental behaviour.

The meeting agreed that this is indeed an accurateaccount of the state of AES formalism, up to the mid-1980s.

The speaker then drew the audienceÏs attention torecent reports in the literature to show that the abovetreatment is not generally correct when including elasticscattering in the formalism, and that elastic scatteringe†ects in AES may have an important role in quantitat-ive AES.5h7

Elastic scattering cross-sections

A review of the elastic scattering cross-sections fol-lowed. A comprehensive data bank on an easily acces-sed computer program was shown at the meeting. Thissoftware package will be released by NIST.8 The pro-gramme, which is menu-driven, calculates di†erentialand total elastic scattering cross-sections of any of 96elements in the Periodic Table in the energy range 50È10 000 eV in steps of 1 eV and o†ers the user the choiceof two potential centres (MT, HF). The calculationincludes relativistic e†ects.

Calculated AES intensities, including elastic scat-tering e†ects, show a signal that is 10% smaller than thecase when elastic scattering was neglected. A new cor-rection factor to take account of the elastic scattering Qis suggested, where

Q\ IelInel

and is the Auger electron current that includes elasticIelscattering e†ects and is that neglecting elastic scat-Ineltering.A second correction term K relating the attenuation

length L to the inelastic mean free path j, based on theabove formalism, was also introduced, where

K \ L /jIt was reported that K depends on the experimentalconÐguration, i.e. emission angle, analyser solid angle,Auger transition and matrix ; K can vary by up to 30%(smaller than unity).

If the above corrections are taken into account, then

dIeldInel

\ QK

exp ([1/K)

Discussion followed this part of the presentation andattention was drawn to the similarity of the problemunder consideration with low-energy electron di†rac-tion (LEED) calculations. Such powerful codes that cannow be run on 80486 PCs have been available in thepublic domain for a number of years and summarisemany years of e†ort by the LEED community. Thesecodes, if to be used in AES, would have the furtheradvantage that the e†ects of the neighbouring hostatoms would also be taken into account.

The speaker agreed and noted that the present for-malism also shows some unreliability for the low-energyregime (\200 eV) and for low and high atomic numbersystems.

It was agreed that the relativistic e†ects would have asmall overall inÑuence on the Auger intensity, but theirinclusion may be desirable in some applications andthey are readily available in recent calculations of dp/dhand j.

The depth distribution function (DDF)

A presentation of the current state of the literature andunderstanding of the DDF followed. This term wasagreed by most to account for the escape probability ofthe signal electrons as a function of depth. Other deÐni-tions described the same phenomenon.9h11 However, inspite of the recent reports on the importance of includ-ing this factor in quantitative AES and for a largenumber of Monte Carlo calculations of this term fordi†erent systems, a lack of experimental veriÐcationappears to be the case. The speaker, in an attempt tostimulate interest and discussion, presented a suggestionof an experimental set-up but this was not elaboratedupon. The meeting agreed that this is an important areaof development that requires urgent attention.

Backscattering e†ects in AES

The second session was devoted to the e†ects of back-scattered electrons in quantitative AES. A review oftheoretical estimates and experimental measurements ofthe Auger backscattering factor r was made. It wasagreed that this factor was perhaps the most signiÐcantin the AES formalism and has therefore received themost extensive attention in comparison to the otherparameters. However, and in spite of the theoreticalactivity including both Monte Carlo and transporttheory, fewer direct experimental measurements havebeen reported.12h16 Furthermore, a prior knowledge ofthis factor by bulk analysis of the sample appears to benecessary for the AES formalism. It was suggested, andlater in the imaging session demonstrated, that thesimultaneous detection of a backscattered signal withthe Auger signal is an efficient way of including thise†ect in the AES formalism.17 Another suggestion, alsodemonstrated in the imaging session, was to referencethe Auger intensity to that collected simultaneously at arelatively higher energy than the Auger transition. Thechoice of this correction energy depends on the primaryincident beam energy and appears to change by, atmost, 20% across the Periodic Table, in comparison toa change in the Auger yield by the backscattered elec-trons amounting to 150%.

The speaker proposed a new term to deÐne the e†ectof backscattering in AES. The proposed deÐnition is totake account of both inelastic and elastic scattering ofelectrons with the solid and is referred to simply as thescattering factor. The audience felt that although thismay show a slight di†erence in some situations, itwould send a confused signal to the AES community atlarge and therefore should not be adopted. The speakeragreed.



Can the backscattering factor be less than unity?

Another point raised during this session was the case ofelastic scattering e†ects operating at given primaryenergies to cause ionization to be in a region where theovervoltage is \2. This has been reported to give riseto a backscattering factor of less than unity. Thespeaker Ñoated the idea of whether this should also betaken into consideration in the AES formalism. Thegeneral consensus was that such energies are seldomused by analysts for quantitative measurements andtherefore such a change may also cause confusion in thewider AES community.

References

1. L. E. Davis, N. C. MacDonald, P. W. Palmberg, G. E. Riachand R. E. Weber, in Handbook of Auger Electron Spectros-copy. Physical Electronics, Edina, MN (1972).

2. P. M. Hall and J. M. Moraboto, Surf . Sci . 83, 391 (1979).3. M. P. Seah, Scanning Electron Microsc . 1983, 521 (1983).4. D. Briggs and M. P. Seah, Quantitative Surface Analysis ,

Wiley Chichester (1993).5. A. Jablonski, Surf . Sci . 188, 164 (1987).6. A. Jablonski, Surf . Interface Anal . 15, 559 (1990).7. A. Jablonski, Surf . Sci . 364, 380 (1996).8. NIST Elastic Scattering Cross-section Database, Standard

Reference Data Program, Database 64. National Institute ofStandards and Technology, Gaithersburg, MD (1996).

9. A. Jablonski and H. Ebel, Surf . Interface Anal . 6, 21 (1984);11, 627 (1988).

10. W. H. Gries and W. S. M. Werner, Surf . Interface Anal . 16, 149(1990).

11. P. J. Cumpson, Surf . Interface Anal . 20, 727 (1993).12. T. E. Gallon, J. Phys.D 5, 822 (1972).13. M. M. El Gomati, Surf . Sci . 152/153, 877 (1985).14. D. Gramari and J. Cazaux, Surf . Sci . 136, 296 (1984).15. H. Bishop, J.Appl . Phys. 18, 703 (1967).16. S. Ichimura and R. Shimizu, Surf . Sci . 112, 386 (1981).17. I. Barkshire, M. Prutton, J. Greenwood and M. M. El Gomati,

Surf . Interface Anal . 20, 984 (1993).

ESCAPE OF ELECTRONS

Inelastic lineshape analysis

Quantitation of surface electron spectra requires accu-rate understanding on the details of the emissionprocess. In particular, the processes inÑuencing thehistory of a signal electron from its point of generationto its eventual escape from the surface must be properlyunderstood. If an appropriate model for the escapeprocess is at hand, one can then relate the measuredpeak intensities to the sampleÏs composition and depthproÐle. However, it was shown that the informationcontained in the peak intensity is not unique : identicalpeak intensities (for one given experimental geometry)may be recorded for a given signal electron line for twosamples with a di†erent depth proÐle of the consideredspecies. The extrinsic lineshape will generally be di†er-ent for the two peaks in such a case. This is easilyunderstood qualitatively : if a signal electron is gener-ated at a large depth, the path length travelled betweengeneration and escape will be long compared to a casewhen the source emission takes place closer to the

surface. Enhanced pathlengths imply a larger probabil-ity for energy loss and consequently signal electronsoriginating from larger depths will contribute more tothe inelastic tail (extrinsic features) of the peak thanthose emitted from shallower depths. Therefore, analysisof the experimental lineshape may yield useful informa-tion concerning the sampleÏs in-depth composition.

The basic assumption in the theoretical modeldescribing the signal electron emission is that energylosses and deÑections occur separately and that thecharacteristics of the emission process are energy inde-pendent in the vicinity of the no-loss peak. The latterassumption corresponds to the so-called quasi-elasticapproximation, which is justiÐed in the near-peakregion in AES and XPs. On the basis of this startingpoint, the escape process may be conceptually dividedinto an energy-loss mechanism and an emission mecha-nism. The energy-loss mechanism describes the energydistribution for a given history of the signal electron ;the emission mechanism describes the probability that agiven history takes place and eventually leads to escapefrom the surface and detection. The emission mecha-nism comprises the following phenomena :(1) depth proÐle/morphology ;(2) angular distribution of the source emission process ;(3) depth and energy-dependent elastic and inelastic

scattering characteristics ;(4) surface roughness ;(5) experimental geometry ;(6) coherent scattering (e†ects of crystallinity).

The loss mechanism requires understanding on theenergy distribution for a given electron history. Thisdistribution is determined by :(1) the loss function ;(2) the probability for a given energy loss after an indi-

vidual inelastic event ;(3) the inÑuence of surface excitations.

The history of escaping signal electrons may be mod-elled in terms of the distribution of pathlengths trav-elled and the corresponding energy loss experienced inthe course of the escape.1 At present this allows theescape process to be modelled for an arbitrary depthproÐle with due account of multiple inelastic scattering.For homogeneous samples it is also possible to accountfor multiple elastic scattering and the typical XPSanisotropy in the source emission process in this way.2

The required information on the loss mechanism maybe derived from optical data3,4 or reÑection5h7 electronenergy-loss spectra, but in the majority of cases whenthe inelastic electronÈsolid interaction is not dominatedby a single, narrow, free electron-like plasmon loss (asin Al), the universal cross-section proposed byTougaard8 is sufficiently accurate. The use of thisempirical single scattering loss function is recommendedwhenever narrow plasmon features are absent in theloss function of the considered material.

The model for the emission of signal electronsdescribed above has been incorporated into theQUASES software package.9,10 A large number ofexamples of successful applications of this software werepresented. The application areas included metallurgy,11semiconductor materials12 and organic overlayers onmetals.13

In the discussion the question was raised whether thepresented analysis of an oxide layer on a semiconductor



accounted for the di†erence in the single scattering lossprobability of the nearly free electron substrate and theinsulating overlayer. Although this was not the case, theachieved results were very satisfactory. The explanationwas that in the course of many inelastic events thedetails of the features of the single scattering loss prob-ability are sufficiently smeared out, and that for analysisof the spectral region in the immediate vicinity of thepeak the exact shape of the loss function is not veryimportant.

A further question addressed the amount of informa-tion that can be gained from inelastic peak shapeanalysis. This issue is presently being investigated. Thespeaker mentioned that if the spectral intensity can beaccurately measured in a 500 eV or larger energy range,Ðve or possibly as many as ten parameters concerningthe depth proÐle can, in principle, be extracted fromsuch analysis. In practice this is, however, hardly pos-sible because of, for example, interfering peaks, e†ects ofelastic scattering and the inÑuence of inaccuracies in theanalyser transmission function. Furthermore, the depthresolution and assessable depth scale are intrinsicallylimited. The exact limitations are also the subject offurther study. It was then remarked that the number ofparameters that can be retrieved depends heavily on thesignal/noise ratio. Preliminary results of another studyindicate that for realistic signal/noise ratios the informa-tion retrievable from a lineshape analysis at onegeometry is limited to less than three parameters.

It was furthermore noted that the neglect of elasticscattering and other phenomena of the emission processthat are not accounted for by the presented model mayhave a serious e†ect on the result of the evaluation. Thespeaker replied that the incorporation of a moredetailed emission mechanism in the model is underdevelopment. It was furthermore pointed out that analternative approach describing the energy/angular dis-tribution of signal electrons exists that is based on thepartial intensities of particles having experienced agiven number of inelastic collisions. This approach pre-sently incorporates e†ects such as surface roughness,depth dependence of the elastic and inelastic scatteringcharacteristics, anisotropy of source emission, etc.14

References

1. S. Tougaard and P. Sigmund, Phys.Rev.B 25, 4452 (1982).2. I. S. Tilinin, A. Jablonski and S. Tougaard, Phys. Rev. B 52,

5935 (1995).3. D. R. Penn, Phys.Rev.B 35, 482 (1985).4. E. D. Palik (Ed.), Handbook of Optical Constants of Solids .

Academic Press, New York (1985).5. S. Tougaard and I. Chorkendorff, Phys. Rev. B 35, 6570

(1987).6. S. Tougaard and J. Kraaer, Phys.Rev.B 43, 1651 (1991).7. F. Yubero and S. Tougaard, Phys.Rev.B 46, 2486 (1992).8. S. Tougaard, Surf . Interface Anal . 11, 453 (1988).9. S. Tougaard, The QUASES software package for quantitative

AES/XPS of surface nanostructures by inelastic peak shapeanalysis.

10. S. Tougaard and H. S. Hansen, Surf . Interface Anal . 14, 730(1989).

11. S. Tougaard and H. S. Hansen, Surf . Sci . 236, 271 (1990).12. M. Schleberger, D. Fujita, C. Scharfschwerdt and S. Tougaard,

J. Vac. Sci . Technol . B 13, 949 (1995).13. H. S. Hansen, S. Tougaard and H. Biebuyck, J. Electron Spec-

trosc,Relat . Phenom. 216, 141 (1992).14. W. S. M. Werner, Surf . Interface Anal . 23, 737 (1995).

Crystalline e†ects

A measurement of the energy and angular distributionof Auger electrons escaping from a crystalline solid maybe strongly inÑuenced by di†raction e†ects. The twomost important phenomena that have to be consideredin this connection are :1(1) channelling of the incoming (high energy) beam;(2) di†raction of the (medium- or low-energy) Auger

electrons, as well as secondaries and backscatteredprimaries.

Although the origin of both phenomena is the coher-ent scattering of an electron wave by the periodic arrayof scattering centres in the crystal, the most e†ectivetheoretical description of these phenomena is widely dif-ferent. Channelling of the high-energy primary beamcan be adequately described qualitatively in the frame-work of kinematic di†raction theory using an appropri-ate number of beams.2,3 Because forward focusingdominates the Auger electron di†raction (AED) patterndue to the strongly forward peaked shape of the elasticscattering cross-section,4,5 this phenomenon can bemodelled by single-scattering cluster calculations,6,7although quantitative agreement cannot generally beachieved in this way and more involved multiple scat-tering calculations are required.8

Both, channelling and di†raction for energies Z500eV probe real space rather than reciprocal space. This isagain due to the large forward scattering peak in theelastic cross-section at these energies. Therefore, thestrongest crystalline e†ects are generally expected alongdirections coinciding with crystallographic axes/planes.

Although di†raction e†ects in AES data may beemployed to gain structural information on a crystalsurface, it severely complicates quantitative analysis.Because the theoretical understanding on the energy/angular Auger electron distribution is not complete andappropriate models for di†raction are highly involved,the experimental procedure for quantitative analysisshould be devised in such a way as to minimize eventuale†ects caused by di†raction. According to the abovetheoretical summary, there exist several ways to accom-plish this, or at least relax the problems in connectionwith di†raction. First of all, experimental geometrieswhere the emission and/or incidence direction fallsalong a major crystallographic direction should beavoided. Secondly, increasing the opening angle of theanalyzer tends to smooth the di†raction pattern. Forexample, for an apparatus equipped with a CMA, ageometry where the specimenÏs surface is o† theanalyzer axis by 45¡ yields an intensity averaged overthe entire polar range, which is to be preferred over amere azimuthal average obtained for a perpendicularincidence geometry. Even if the analyzer opening angleis large and di†raction e†ects are thus suppressed, caremust be taken to avoid channelling of the primarybeam.

Many modern analytical problems deal with thinlayers on single crystalline substrates. Very often, thecontrast in these layers is not a†ected by crystallinee†ects (e.g. oxides on semiconductors, segregated layerson crystalline substrates, etc.) Then it is always advis-able to base the evaluation on the overlayer signal onlyand discard the information from the bulk lines, as thelatter is more likely a†ected by crystalline e†ects. A



spectral feature far o† the characteristic lines (and alsosufficiently far below the elastic peak) may be used as areference signal instead.

An entirely di†erent approach to alleviate the prob-lems with the crystallinity is amorphization of theanalyzed area, e.g. by ion bombardment. Very often,however, this may lead to other highly undesirable sidee†ects that are difficult to quantify, such as preferentialsputtering, recoil implantation, etc. Moreover, not allmaterials amorphize under ion bombardment.

Summarizing, it can be said that there exist nogeneral recommendations on how to avoid erroneousresults in quantiÐcation owing to crystalline e†ectsother than the rather rough guidelines discussed above.However, if the emission or excitation geometry can bevaried in the apparatus (or both simultaneously bytilting the sample), it should generally be possible toverify whether crystalline e†ects are likely to play a rolein a given experiment by repeating the measurement atone or more slightly di†erent geometries, because bothdi†raction and channelling e†ects exhibit very sharpfeatures in angular space.

References

1. H. E. Bishop, Surf . Interface Anal . 16, 118 (1990).2. G. R. Booker, A. M. B. Show, M. J. Wheelan and P. B. Hirsch,

Philos .Mag. 16, 1185 (1967).3. M. V. Gomoyunova, S. L. Dudarev and V. I. Pronin, Sov. Phys.

Solid State 30, 1562 (1988).4. J. B. Pendry, Low Energy Electron Diffraction. Oxford Aca-

demic, Oxford (1973).5. A. C. Yates, Comp.Phys.Commun. 2, 175 (1971).6. S. Kono, S. M. Goldberg, N. F. T. Hall and C. S. Fadley, Phys.

Rev.B 22, 6085 (1980).7. C. S. Fadley, Prog. Surf . Sci . 16, 275 (1984).8. J. J. Rehr and R. C. Albers, Phys.Rev.B 41, 8139 (1990).

INSTRUMENTAL EFFECTS

Spectrometer types1h3

The types of spectrometers that are used in AESinclude :(1) the retarding Ðeld analyser (RFA) ;(2) the 127¡ analyser ;(3) the cylindrical mirror analyser ;(4) the hemispherical analyser ;(5) the Shimadzu/Du Pont analyser ;(6) the Staib analyser ;(7) the Riber MAC2 analyser.Important considerations include working distancesand the types of surfaces to be studied (e.g. Ñat vs.rough), as well as the usual parameters of resolutionand sensitivity.

Manufacturers should use switches and not contin-uously variable settings at the front ends of instrumentsto improve reproducibility. The stability of the analyserpower supply used for AES needs, in general, to beimproved and the stability should be speciÐed. TheWorkshop conclusion was that manufacturers, ingeneral, need to improve the repeatability and stability ofmechanical and electronic controls.

Energy calibration

Copper, silver and gold are used for calibrating theenergy scale. The procedure for measuring the energy ofthe low-energy Cu doublet was described, resulting in avalue that is independent of instrument resolution.4

Corrections for changes in peak energy with energyresolution can be made using analytical functions.5Relativistic corrections can also be made when neces-sary.6 In measuring the energy of elastically back-scattered electrons, small corrections may need to bemade for recoil energy changes.7,8

The required stability of energy scale calibration inAES will be dealt with by ISO, (TC 201 on SurfaceChemical Analysis).

Specimen positioning e†ects

The measured electron kinetic energy varies, in general,with specimen position in relation to the analyser,4,9and also with the position on the specimen itself. Theextent of the variation depends on the shift in the speci-men position and the dispersive power of the spectro-meter. In this regard, hemispherical analysers tend to bemuch superior to cylindrical mirror analysers. The e†ectin hemispherical analysers may be further reduced byreducing the magnifying power of the analyser inputlens.4 The Workshop conclusion was that e†ective useof an instrument is only possible if the user understandsthe design principles of the spectrometer.

Detectors

For analogue detection, electron multipliers should notbe driven to give output currents [1% of the wall orstanding current.10 For pulse counting, a dead time andmultiplier gain loss above 1 Mc s~1 must be considered.The setting of the discriminator is also important. Thefront of the multiplier should be biased : if it is notbiased, problems at low kinetic energy may occur if themultiplier voltage is reduced to produce a count ratebelow 1 Mc s~1.11,12

With microchannel plates, half of the surface area islost, and it is not clear what happens when electrons hitthe Ñat regions between the channels.

Taking one long (time) scan vs. several fast scans wasdiscussed and, as long as the spectrometer voltagesupplies can scan adequately fast, it was agreed by theattendees that taking several fast scans was preferred.This averages out any beam current or sample-dependent drift. The Workshop conclusion was thatmultiple scans at a fast rate (determined by theelectronics) are preferred over a single scan at a slowrate.

The coating of the inner walls in deÑection-typespectrometers is important to reduce patch Ðelds andscattering of electrons. If there is scattering inside theanalyser, steps and peaks will appear when a spectrumis divided by the corresponding reference spectrum.13,14Internal scattering may be reduced in manufacture bysuitable Ðeld stops, the use of “electron dumpsÏ, theavoidance of retarding lenses and by care in the use ofgrids. In spectrometer use, very low pass energies orhigh retardation ratios may exacerbate the problem.



The Workshop conclusion was that in making a spectro-meter, scattering needs to be below the 1% level.

With multidetectors, problems may arise in the FRRmode, unless great care is taken with the software forassigning data to the spectral energy channels, as theindividual detectors will misalign with the energy chan-nels as the scan proceeds.15

Intensity scales

Each instrument can be calibrated by comparingspectra obtained from Cu, Ag and Au with standardreference spectra of these elements.12 These referencespectra have been determined by the National PhysicalLaboratory (UK), who will be providing a calibrationservice and software.

Several years ago, variations between instruments ofthe same model (from the same manufacturer) may havebeen a factor of three.17 Manufacturers have many sup-pliers of components and must put speciÐcations onthem to minimize such variations.

Electron beam current measurements should be madein a Faraday cup18 using a calibrated electrometer.Manufacturers, in general, need to improve the incidentelectron beam current stability. The Workshop conclu-sions were that manufacturers, in general, need toimprove the incident electron beam current stability, andthat in the near future it will be possible for manufac-turers to issue calibration certiÐcates with each instru-ment, providing a Ðgure for the spectrometerÏs efficiency,the scatter of calibration, the variation in absolute levels,scattering in the instrument and the overall uncertainty.

Such measurements enable all data to be comparableso that future data banks have universal applicability.This, therefore, provides the Ðrst step to e†ective databanking and to the direct use of theoretical calculationsof quantity.

Spatial e†ects

Various approaches to imaging, such as deÑectingfocused beams or using a pre-lens scanning system, werediscussed. Angular e†ects in Auger spectroscopy werealso discussed.

Data banks

There are several handbooks of Auger spectra avail-able.19h22 However, none of the spectrometers used toobtain such data were calibrated.

Auger data have now been obtained on a calibratedspectrometer and will be available in handbook form, aswell as in the VAMAS format,23 from NPL. Data forboth 5 and 10 keV electron beam energies wereobtained,24,25 over wide energy (“whole spectrumÏ) andnarrow energy (“individual peaksÏ) ranges. Di†erentialspectra were obtained from the direct spectra. Allspectra were obtained from sputtered elemental andeither cleaved or sputtered compound surfaces. Someoxides were studied additionally following in situ oxida-tion. Good consistency of the data, converted to anabsolute intensity scale of sr~1, was found with theoreti-

cal predictions also using the sr~1 scale with no inde-pendent parameters in either theory or experiment. Theworkshop conclusion was that a public-domain softwareprogram needs to be provided so that standard referencespectra can be output into ASCII format for incorpor-ation into other software.

Conclusions

The workshop conclusions from this session are sum-marized here :(1) E†ective use of an instrument is only possible if the

user understands the design principles of thespectrometer.

(2) Multiple scans at a fast rate (determined by theelectronics) is preferred over a single scan at a slowrate.

(3) In making a spectrometer, scattering needs to bebelow the 1% level.

(4) Manufacturers, in general, need to improve therepeatability and stability of mechanical and elec-tronic controls, and the stability of the incident elec-tron beam current.

(5) In the near future it will be possible for manufac-turers to issue calibration certiÐcates with eachinstrument, providing a Ðgure for the spectrometerÏsefficiency, the scatter of calibration, the variation inabsolute levels, scattering in the instrument and theoverall uncertainty.

(6) A public-domain software program needs to be pro-vided so that standard reference spectra can beoutput into ASCII format for incorporation intoother software.

(7) Auger lineshapes are important for chemical stateidentiÐcation and we need to have more on record.

(8) New opportunities arise from the World Wide Webin the distribution of reference spectra.

Future work should emphasize conclusions (4)È(8).

References

1. B. Wannberg, U. Gelius and K. Siegbahn, J. Phys. E 7, 149(1974).

2. R. C. G. Leckey, J. Electron Spectrosc . 43, 183 (1987).3. M. P. Seah, in Methods of Surface Analysis , edited by J. M.

Walls, p. 57. University Press, Cambridge (1989).4. M. P. Seah, G. C. Smith and M. T. Anthony, Surf . Interface

Anal . 15, 293 (1990).5. I. S. Gilmore and M. P. Seah, J. Electron Spectrosc . 83, 197

(1997).6. M. P. Seah and M. T. Anthony, J. Electron Spectrosc . 35, 145

(1985).7. N. E. Erickson and C. J. Powell, Phys. Rev. B 40, 7284

(1984).8. D. Laser and M. P. Seah, Phys.Rev.B 47, 9836 (1993).9. J. D. Geller, Appl . Surf . Sci . 18, 18 (1984).

10. M. P. Seah, J. Electron Spectrosc . 50, 137 (1990).11. M. P. Seah and G. C. Smith, Surf . Interface Anal . 15, 701

(1990).12. M. P. Seah, C. S. Lim and K. L. Tong, J. Electron Spectrosc .

48, 209 (1989).13. M. P. Seah, Surf . Interface Anal . 20, 865 (1993).14. M. P. Seah, Surf . Interface Anal . 20, 876 (1993).15. M. P. Seah and P. J. Cumpson, J. Electron Spectrosc. 61, 291

(1993).16. M. P. Seah, J. Electron Spectrosc . 71, 191 (1995).17. M. P. Seah and G. C. Smith. Surf . Interface Anal . 17, 855

(1991).



18. I. S. Gilmore and M. P. Seah, Surf . Interface Anal . 23, 248(1995).

19. L. E. Davis, N. C. MacDonald, P. W. Palmberg, G. E. Riachand R. E. Weber, Handbook of Auger Electron Spectroscopy,2nd Edn. Physical Electronics Industries, Eden Prairie, MN(1976).

20. G. E. McGuire, Auger Electron Spectroscopy ReferenceManual . Plenum Press, New York (1979).

21. Y. Shiokawa, T. Isida and Y. Hayashi, Auger Electron SpectraCatalogue: A Data Collection of Elements . Anelva, Tokyo(1979).

22. T. Sekine, Y. Nagasawa, M. Kudoh, Y. Sakai, A. S. Parkes, J.D. Geller, A. Mogami and K. Hirata, Handbook of Auger Elec-tron Spectroscopy. JEOL, Tokyo (1982).

23. W. A. Dench, L. B. Hazell and M. P. Seah, Surf . Interface Anal .13, 63 (1988).

24. M. P. Seah and I. S. Gilmore, J. Vac. Sci . Technol . A 14, 1401(1996).

25. M. P. Seah, I. S. Gilmore, P. J. Cumpson, J. P. Langeron andG. Lorang, ECASIA 95, Proc. European Conference on Appli-cations of Surface and Interface Analysis , edited by H. J.Mathieu, B. Reihl and D. Briggs, p. 607. Wiley, Chichester(1996).

DATA PROCESSING

Qualitative analysis

Three topics were raised :(1) Element and chemical state identiÐcation.(2) Expert systems. A number of systems have

emerged.1,2 In the discussion, questions such as theproblem of identifying peaks in isolation were con-sidered. The importance of examining patterns ofother peaks, the elimination of signal/noise e†ects,the problem of peak position uncertainty being aslarge as 10 eV in AES, the need to consider systemresponse function and the need to consider spectralshape were all raised as factors that complicate thecorrect development of expert systems.

(3) Databases. Recent advances were discussed, includ-ing :

(a) Availability of electronic databases : SurfaceScience Spectra (Auger spectra for 24 elements) ;NIST XPS database (Auger energies for 250compounds) ;NPL (Auger reference data of 61 elements) ;Surface Analysis Society of Japan, Spectra Data-base (75 Auger spectra for 25 elements) ; andhandbooks such as the PHI Handbook (4threvision).Several questions were raised in the discussion.Should the main interest be spectral shaperather than spectral peak position? There wasalso a concern that di†erent instrumentsinvolved the convolution of the experimentaldata with a di†erent instrument function leadingto a difficulty in establishing standards. Clearlythere is a need to establish the instrument func-tion, a practice currently being followed inSurface Science Spectra. The importance ofincluding calibration spectra with databasespectra was agreed upon. The e†ect that thesample state has upon spectral appearance alsoneeds to be considered. Thus spectra may be dis-

torted by sample charging and sample decompo-sition (during spectral collection) and may bedependent upon sample preparation. We alsoneed to make databases more interchangeable.

(b) ClassiÐcation by pattern recognition : e.g. studiesof the lineshape of Si LVV, O KLL and N KLLpeaks in air-exposed surfaces.3Si3N4(c) ClassiÐcation by neural networks : e.g. the use ofartiÐcial neural network approaches to the setsof di†erent Auger spectra (Fe, Au, Si, Sn andCu).4

(d) VAMAS data format. The VAMAs data format5provided a comprehensive data format thatallowed data from di†erent surface analyticalmethods to be transferred efficiently from onecomputer to another. It is clearly important thatdata analysis systems have a method for conver-sion of this type of format, although it is oftentoo detailed for routine data storage.

(e) Making more use of photon-excited Auger peaksin XPS studies. There is clearly potential herefor more information to be obtained from XPSstudies.

Lineshapes

Three topics were raised :(1) N(E) vs. dN(E)/dE. It was noted that the majority of

practicing analysts using AES rather than thosestudying AES itself do, in fact, currently use di†eren-tiating routines.6 Many agree that data should beinitially collected in N(E) mode, and then (ifrequired) converted to dN(E)/dE form by appropri-ate di†erentiating methods.

(2) Smoothing. It is important that the initial N(E) databe of high quality before di†erentiation, andsmoothing may be needed. Recommendations forsmoothing have been given.7,8

(3) Background Ðtting and suppression. The use of aninelastic loss (Tougaard9) background model ordeconvolution of the experimental data with elec-tron energy-loss experimental data (e.g. Ref. 10) wasemphasized.

Spectral overlap

The handling of overlapping peaks has been investi-gated by various methods in XPS studies, but is notoften used in AES. The reason for this is that this typeof analysis normally requires N(E) data, as well as datarecorded with high resolutionÈboth of these featuresare often absent from routine AES studies. A review wasgiven of a number of approaches that have been used inXPS. Approaches that can resolve spectral overlapinclude smoothing, spectral subtraction, deconvolution,curve Ðtting, multiple linear regression and factoranalysis. According to the validity of additivity of refer-ence spectra, it was pointed out that charging-inducedenergy shifts, ion and electron beam damage, and back-ground e†ects may impose some restrictions on linearmethods like factor analysis and multiple regression.For example, the secondary electron cascade duringdepth proÐling may change and cause substantial spec-tral changes that eventually necessitate background



removal.12,13 Thus the use of non-linear least-squarescurve Ðtting for depth proÐle analysis was mentioned.11Factor analysis methods have proved popular withAuger analysts : used in 3% of all recent Auger papers ;12% of Auger papers deal with composition or analysis.It is recommended that analysts spend more e†ort inexamining the residuals from factor analysis models toimprove their results.

In the discussion it was clear that precise curve Ðttingcould be very valuable in AES, e.g. in studies of phasetransitions. Nevertheless, it was recognized that it wasmuch more difficult to Ðt AES data in many cases dueto the presence of a multiplet structure (e.g. in the Ðttingof data from a mixture of copper metal, copper sulphideand copper oxide). It was also noted that an Augermultiplet structure may or may not change withchanges in chemical state.

Quantitative analysis

The discussion recognized :(1) the need for users to set up rules for the use of

expressions that relate the intensity of an AES peakto the amount of a material present ;

(2) the need for manufacturers to include details of theappropriate matrix factors to be used in the inten-sity expressions ;

(3) the need to determine the best method for obtainingthe intensity scale calibration in AES.

Future developments

It was recognized that expert systems and real-timedata processing would increasingly become a part ofnew instruments, which would also include data sharingby appropriate networking and distributive processing.In performing these operations it was essential thatusers and instrument manufacturers take properaccount of the basic physics of the processes. In particu-lar it was important to retain the original data in theappropriate manner.

Extra item: Positron Annihilation-induced AugerElectron Spectroscopy (PAES)

Ms Haiqing (Amy) Zhou of the University of Texas atArlington made a short presentation of data obtainedusing PAES. The analysis of the low-energy tail was dis-cussed. In the experiment an incoming positron is anni-hilated by an electron, leading to the emission of agamma ray together with a core hole. The core holeleads to Auger electron emission. In the discussion thatfollowed, the question of preferential trapping of thepositron on the surface was raised, and it was also sug-gested that better resolution could be obtained withproton-induced Auger electrons.

Extra item: standard free determination of detectionfactors in AES

Dr Ralf Muller presented data obtained with ProfessorHans Oechsner concerning standard free determination

of AES sensitivity factors from Auger images. Theapproach was shown to have advantages over anapproach using the standard relative sensitivity factorsgiven in Auger handbooks.14

References

1. S. Bumgarner, S. Hofmeister, D. Griffis and P. Russell, J . Vac.Sci . Technol ., A 8, 2221 (1990).

2. F. Bruninx and A. Van Eenbergen, Anal . Chim. Acta 133, 339(1981).

3. J. Zemek, T. Vystreil, B. Lesiak-Orlowska, A. Jablonski and A.Luches, Surf . Interface Anal . 21, 771 (1994).

4. M. N. Souza, C. Gatts and M. A. Figueira, Surf . Interface Anal .20, 1047 (1993).

5. W. A. Dench, I. B. Hazell and M. P. Seah, Surf . Interface Anal .13, 63 (1988).

6. M. P. Seah, M. T. Anthony and W. A. Dench, J. Phys. F 16,848 (1983).

7. M. P. Seah and W. A. Dench, J. Electron Spectrosc. 21, 771(1994).

8. M. P. A. Sherwood, in Practical Surface Analysis , 2nd Edn,edited by D. Briggs and M. P. Seah, pp. 555–586. Wiley, Chi-chester (1990).

9. S. Tougaard and P. Sigmund, Phys.Rev.B 25, 4452 (1982).10. V. Contini, C. Presilla and F. Sacchetti, Surf . Sci . 210, 520

(1989).11. I. Kojima, N. Fukumoti, M. Kurahashi and T. Kameyama, J.

Electron Spectrosc . 50, 53 (1990).12. J. P. Langeron, Surf . Interface Anal . 14, 381 (1989).13. W. S. M. Werner, I. S. Tilinin, H. Beilschmidt and M. Hayek,

Surf . Interface Anal . 21, 357 (1994).14. H. Oechsner, Vide,Couches Minces 271, 141 (1994).

ARTEFACTS

This workshop is concerned with how to obtain infor-mation on the composition and chemistry of a surfacefrom electron-excited Auger spectra. The informationcan only be as reliable as our input data. If we assumewe have carried out all the procedures recommended inother sessions to obtain reproducible data and takenaccount of orientation, shadowing and crystallographice†ects, what else can go wrong? We have considered thefollowing :(1) modiÐcation of spectrometer transmission by the

specimen ;(2) insulating specimens ;(3) beam-induced changes to the surface composition ;(4) inhomogeneity of specimen.

Spectrometer transmission

If the specimen modiÐes the magnetic or electrostaticÐeld between the specimen and the spectrometer, thee†ective spectrometer transmission may be signiÐcantlymodiÐed, particularly for low-energy electrons. The twomost common causes are ferromagnetic specimens thathave not been adequately degaussed and the chargingof insulating fragments on a surface in the neighbour-hood of the analysed area. The former e†ect may beavoided by taking proper degaussing precautions ; local-ized charging is much more troublesome and may resultin some areas of the specimen being inaccessible foranalysis.



Insulating specimens

The charging of insulating specimens is perhaps thegreatest limitation to electron-excited AES. Even estab-lishing conditions where the surface potential is stabil-ized so that a reasonable spectrum is obtained may stillleave ambiguities resulting from the complex nature ofthe charging phenomenon. The exact nature of chargingand charge neutralization is critically dependent bothon the composition and structure of the sample and onits environment within a particular instrument. Impu-rities and defects act as scattering centres for electronsin the conduction band and as trapping sites, both ofwhich strongly inÑuence the detailed charging behav-iour.

There are a number of strategies that may be used tominimize charging e†ects. However, data obtainedunder these conditions must be evaluated with care.Charging shifts may not be the same on all areas of thesample, leading to peak broadening in spectra and tothe possibility of missing an element completely inimages and linescans. For quantitative analysis it isimportant to check that the transmission function hasnot been changed signiÐcantly and that charges trappedbelow the surface have not modiÐed the backscatteringcontribution to the signal.

Beam-induced e†ects

As electron beams interact strongly with bound elec-trons, it is not surprising that electron irradiation maymodify the surface composition. Particular examples arethe loss of Ñuorine from the surface of lithium Ñuoride,oxygen loss from silicon dioxide and titanium dioxide,and the cracking of adsorbed hydrocarbons. The rate ofmodiÐcation is dose dependent. Damage may be negli-gible if the exciting beam is rastered over a large area,but the required information may be completely lostbefore a spectrum can be recorded if a Ðne probe isused. It is worth noting that similar e†ects occur to alesser extent in XPS and that they will become moresigniÐcant as the lateral resolution of XPS is reduced.

For Ðne probes, specimen heating may become sig-niÐcant on poor thermal conductors. In extreme cases,it is possible to melt the sample.

Electrical charge is normally carried in solids by elec-trons. However, in some materials ions are sufficientlymobile to move in electric Ðelds. In materials such asb-alumina and some glasses, alkali metal ions aremobile and the local concentration may be eitherenhanced or reduced depending on the direction ofcurrent Ñow and/or the local Ðeld gradient.

Beam penetration e†ects

Artefacts may result from specimen heterogeneitieseither laterally or in depth.

Topography. Large-scale topographic e†ects were dis-cussed in session 8 (see next section). However, in thevicinity of edges, electron scattering may lead to a con-tribution to the signal coming from surfaces other than

the point of incidence of the primary beam, leading toanomalies in linescans and images.

High lateral resolution. When the primary beam diameteris below 200 nm or so, the area from which the back-scattered contribution to the signal comes may be sig-niÐcantly greater than the area contributing to thedirect component. Although it is possible to determinequalitatively the composition of small features in the10È100 nm range, quantitative analysis is more difficultand detection limits will be inÑuenced by the composi-tion of the surrounding material.

Layered structures. The standard correction proceduresfor AES assume a uniform composition with depth. Fora thin overlayer, thicker than the Auger escape depthbut less than the electron penetration, the backscattercorrection based on the surface composition may not beappropriate, (see next section). If composition varieswithin the escape depth, quantitative analysis requiresgreat care. In the case of technological samples wherethe structure is not known and may be ill deÐned, quan-titative results should be heavily qualiÐed. A carefulexamination of the inelastic tail of the Auger peaks willoften give an indication of the near-surface distributionof the particular element.

Conclusions

As with any technique, AES is subject to artefacts. It isusually possible to anticipate the circumstances wherethey will occur from a knowledge of the nature of thesample. A careful operator should not be misled if he orshe takes full account of the evidence in the spectrumand the secondary electron images generated whenlocating areas of interest. Note that it is good practiceto be conÐdent of the composition and structure of thesubstrate before embarking on any surface analysis.

Useful References

1. C. G. Pantano and T. E. Madey, Appl . Surf . Sci . 7, 115 (1981).2. Special issue of J. Electron Spectrosc . 59, no. 1 (1992).

IMAGING

The presentation started with a compilation of thefactors determining the spatial resolution with AES.These are the proÐle I(r) of the probing electron beam,the composition proÐle C(r) and the so-called pointspread function S(r). The latter was discussed by someexamples from computer simulations, which traced thepaths of secondary electrons in the excited volume inthe sample surface. For an abrupt chemical edge on aÑat surface, an “edge resolutionÏ can be deÐned as theproduct 2Rb, where R is the Auger backscatteringfactor and b equals the FWHM of the primary beamproÐle. In CMA instruments with a Ðeld emission gunproviding a spot size below 30 nm at 4 keV, structuresof 100 nm were reported to become safely detect-(SiO2)able.



Starting from such results, the possibilities forimproving the spatial resolution were discussed subse-quently. An important issue will be to reduce theprimary beam energy whilst maintaining the beamcurrent. Deconvolution techniques with regard to I(r)and S(r) should also be considered as important steps inthat context. However, real surfaces are not only heter-ogeneous, but also rough. Recent studies on spatialresolution under di†erent topographical conditions andfor improved sidewall selectivity in the analysis of VLSIpatterns1 were shown to deliver respective informationand to demonstrate the importance of appropriatesample preparation. Using Au bars on Si as test struc-tures and a 30 keV beam of 4 nA, the Auger signal wasfound to originate from only 3% of the interactionvolume in the solid when a sampling width of 400 nm atthe surface was considered.2 An interesting preparationtechnique for Auger microanalysis was reported quiterecently : the surface area to be analysed is prepared as athin microlever by undercutting it through sputtererosion with a highly focused ion beam from a liquidmetal ion source.3 Then the deterioration of the lateralresolution from the interaction volume in the solid andfrom backscattered primaries is almost completelysupressed.

The speaker then concentrated on the progressachieved with the MULSAM (multispectral analyticalelectron microscope)4 developed in his own laboratory.Besides the energy-dispersive detection of Auger elec-trons with a hemispherical analyser, this instrumentcomprises capabilities for simultaneous detection of theexcited x-rays, and a four-quadrant detector for back-scattered electrons. Such potentialities make it possibleto exploit the correlations between the various signals.By a parallel image detection via the di†erent signals,the Auger image can be corrected with regard to thebackscattering e†ect and the surface topography.Making use of the information delivered by the back-scattered electron detectors, examples for topography-corrected images are achieved in which thebackscatter-excited signals of elements other than theimaged element are eliminated. Impressive examples forthe application to “realÏ surfaces, such as wear-loadedsteel samples, have been used to demonstrate theprogress achieved by the MULSAM.

As another topic, the so-called multivariate statisticshave been addressed in some detail. Multidimensionalintensity histograms are shown to deliver the distribu-tion of di†erent phases in a multidimensional framewhere the axes represent the concentrations of di†erentelements or speciÐc compounds. A technical layer struc-ture consisting of W on top of TiN with an additionalTi layer at the interface to an substrate has beenSiO2used to demonstrate the capabilities of this evaluationtechnique. As another example, the principal com-ponent images derived for a PtRh catalyst from topo-graphically corrected Auger images have beenpresented.

Finally, the needs and the future of imaging withAuger electrons have been stressed by the speaker. Hepointed out that there is certainly a demand for betterelectron columns producing highly focused low-energybeams. Better sample stages that enable a microscopicmotion of the sample under a Ðxed beam are anotherattractive task. Improved detectors with low signal

losses and di†erential Z-sensitivity might come up. Thepossibilities of parallel detection and the development ofmore sophisticated evaluation algorithms, including thepossibilities of multivariate statistics, will be anotherpart of the future of Auger microanalysis.

Several additional points were raised in the dis-cussion during and after Professor PruttonÏs talk. Thepossibilities resulting from the deconvolution of Augerimages with respect to I(r) and S(r) were addressed by J.Castle. Following his previous ideas, J. Cazaux indi-cated the possibility to detect single atoms in a sur-rounding matrix by repeated Auger excitation as theultimate limit for lateral resolution. H. Oechsner com-mented on electron beam-induced e†ects that must beconsidered with increasing electron current in a decreas-ing spot, and on the advantage of parallel work func-tion imaging with the so-called onset techniques.

References

1. W. Ho� sler, Surf . Interface Anal . 17, 543 (1991).2. A. Umbach et al ., Surf . Interface Anal . 14, 401 (1989).3. C. F. Hoener et al ., Surf . Interface Anal . 23, 83 (1995).4. M. Prutton et al ., Surf . Interface Anal . 17, 71 (1991).5. G. Bachmann et al ., Thin Solid Films 174, 149 (1989).

CLOSING SUMMARY

The following report is based on the text used for theÐnal summary made by J. Castle acting as rapporteur.In making this report the most signiÐcant points madeby all workshop contributors were considered and theindividual presentations and titles were remixed underthe following headings :(1) Solving spectrum for line positions.(2) Solving spectrum for background intensity.(3) Solving spectrum for line intensity.(4) Solving for composition.(5) Separation of overlapping peaks.(6) Obtaining distribution in depth.(7) Crystallographic and topographic features.(8) Spatial distribution.(9) Instrumental procedures.

(10) Data banks.(11) Artefacts.In reading this report it must be appreciated that theseare the impressions received after a very intensive week.They were put together on the Ðnal evening for a 20min summary and, as such, cannot be comprehensive. Ihope, however, that they give the Ñavour of the meetingand represent both presentations and discussions. It hasnot been possible to mention all the speakers by name,nor will I necessarily have picked on the points that theauthors themselves considered to be most important.This is the nature of a rapporteurÏs summary and Ihope it will be read in this spirit.

Progress in solving the spectrum for the peak positions

Recognition of elements present. The Ðrst step in interpre-tation of the electron spectrum is that of attributing theintensity observed in the form of peaks to the elements



that are present in the sample. Two approaches are pos-sible :(1) Calculation of exact position (Weightman). Pro-

fessor Weightman described the current state of theart in the direct estimation of electron kinetic ener-gies by SCF calculations using a two-step processfor electron emission. He considered that excellentprogress had been made for elements, metals andalloys and that peak positions could now reliably becalculated.

(2) Comparison with exact standard (Seah). Dr Seahassessed the position with regard to the availabilityof standards and means of comparison, pointing tothe NIST database, the database provided in thePHI Handbook and to the database being assembledby Professor Goto in Japan. A further database hasbeen assembled under the auspices of the EuropeanUnion by Seah, Langeron and Lorang. Seah alsoconsidered that excellent progress had been madefor metals and alloys and that elements could beidentiÐed by exact comparison with standards. Inthe course of the workshop, Dr Gaarenstroom madethe very relevant point that the ASTM-recommended sequence for making comparisonswas still fully relevant.

Recognition of compounds present. In recognizing com-pounds from the peaks present in the spectrum, thereare a multitude of problems. Because compounds arefrequently electrically insulating, there is the problem ofalignment of the spectra ; there was general agreementwith the view expressed by Gaarenstroom that this isnot yet solved. The problem of electrostatic chargingwas also addressed by Professor Cazaux and by DrBishop, but no generally applicable solution wasadopted by the participants. The other problems areconcerned with the extent of charge transfer in the com-pound and the relaxation contributions to the Ðnalstate, these were discussed by Weightman. Again, Seahreviewed the situation with regard to comparison withstandards. The two approaches given by Weightmanand Seah are :(1) Calculation of exact position (Weightman). Most

progress in the understanding of peak positions forcompounds had been made using XPS, because inthis case the initial and Ðnal state contributions canbe established by use of the Auger parameter (acomparison between the shift of the single-hole statefrom the photoelectron position and that of thedouble-hole state from the Auger electron). By thismeans, good progress had been made in estimatingcharge transfer in compounds and the basis for afundamental scale of electronegativity now exists.However, the Auger parameters *a and *b areirrelevant to the AES spectrum. Goto made theinteresting point that an equivalent source of infor-mation on relaxation is available from a comparisonof the Auger peak with the relevant EELS peak.Unfortunately this measurement was normally pre-cluded by both the available geometries of thespectrometer and the energy ranges that can bescanned in a typical instrument.

(2) Comparison with standards (Seah). The principalproblem for those wishing to make comparison with

standards was the lack of suitable records. Forexample, the NIST XPS database contains Augerenergies for about 250 compounds, the SurfaceAnalysis Society of Japan Spectra Database list 75and the AVS spectra include some 63 records. Thisis obviously a serious shortcoming. Also complicat-ing the situation was the fact that both shape andposition are important and that quality in industrialsamples may be lacking. In the former context it waspointed out that peak positions at low KE may bemeasured better by dN(E)/dE or still better byd2N(E)/dE2. In conclusion, it was agreed thatprogress was only poor to moderate in compoundrecognition.

Progress in solving for the background intensity

There was a feeling of real progress in understandingthe background intensity, although it was recognizedthat there are more components to this than in the ana-logous case for XPS. Consideration was given in thediscussion to the order in which components should beremoved. The stages recommended are :(1) Correct the shape using reference curves ; soon to be

available (NPL).(2) Remove backscattered primaries.(3) Remove Tougaard background at peaks.(4) Remove cascade background.

In a Ðnal interesting discussion it was shown bystudies of true peak shape by positron annihilation (DrAmy Zhou) and, separately, by electron coincidencespectroscopy (Dr Thurgate) that an intrinsic loss back-ground is associated with Auger electron peaks.

Progress in solving intensity

Having separated the background component, it is pos-sible to determine the intensity and, in principle, calcu-late the composition by basic principles. The followingsteps summarize, in the view of the rapporteur, the posi-tion reached during the week :(1) Assume constant electron excitation Ñux through

Auger emitting region, or use '(oz) function toobtain the number of ionizations as a function ofposition in the excited volume.

(2) Assume that wavefunctions remain constant to cal-culate the transition rate by an SCF approach(recognizing limitations, even for alloys and certain-ly for compounds).

(3) Determine electron transport to surface using jimfpand for preference calculated by TPP2 (noting reser-vations expressed below).

(4) Use matrix factor (Hall and Morbito) and JablonskiQ factor (elastic scattering factor) to correct inten-sities.

(5) Normalize to the intensity of the background at D2keV (Bishop).

(6) Use the backscattered intensity (Prutton) to obtainr, the backscattered coefficient. (No need for newdeÐnition of r.)

There were problems, still, in each part of the calcu-lation. The Gryzinski formula for the estimation ofinner-shell ionization cross-sections is acceptable forlow and medium atomic numbers (i.e. K shell



ionization) but may be progressively less accurate for Land M shell ionizations (Powell). There was also aproblem as to which electron transport coefficientshould be used. Powell summarized the situation,making recommendations as follows :(1) IMFP (inelastic mean free path) used for derivation

of surface composition. This was described byWerner as the characteristic length for the energytransfer process. The IMFP calculated using theTPP-2M formula appear to be the best that wascurrently available.

(2) Mean emission depth used for calculation of depth ofanalysisÈthe characteristic length for momentumtransfer (Werner).

(3) Electron attenuation length used for calculation ofoverlayer thickness. The widely used Seah andDench formula is based on experimental AL mea-surements performed prior to 1979 but needs updat-ing.

Solving for composition

Professor Jablonski reviewed the four routes to obtain-ing a composition from the intensities of the peakspresent in the spectrum:(1) From Ðrst principles. This has been included in the

summary of the interpretation of intensity, above.Jablonski paid special attention to the elastic scat-tering process and to the factor Q.

(2) By use of local standards. Dr Gaarenstroom pointedout that local standards of a closely related materialcan be used to obtain the fractional concentration

giving as long as the matrixXA ; XA \ IAMA/IArefcorrection factor for element A can be estab-MAlished.(3) By use of a locally determined sensitivity factor.(4) By use of a book sensitivity factor.

The last methods are used in 99% of analyses by AES(Gaarenstroom).

One problem in using local standards is that, in manycases, it presupposes knowledge (or presumedknowledge) of the elements present. Seah pointed outthat normalization of the instrument response functionwould often enable elements to be revealed by theirinÑuence on the background intensity.

In all methods a warning note needs to be soundedwhen single crystals are being examined. This oftenarises with semiconductors and even with metals andalloys, because the grain size is often large compared tothe lateral resolution of the analysis. The problem isthat of signal modulation according to take-o† angle,because of forward scattering (electron di†raction)e†ects.

Separation of overlapping peaks

Separation of overlapping peaks is particularly difficultin AES because of the complex and variable shape ofthe peaks. For example, Sherwood needs seven param-eters in his model, and the importance of the verycompound-sensitive CosterÈKronig transitions on peakshape was mentioned by many speakers. The followingmethods were discussed :

(1) Curve resolution by use of a Fourier transform. Pro-fessor Sherwood pointed out that there was no basisfor the use of this method when AES was carriedout in a high resolution mode. Its normal use inXPS was to remove the x-ray lineshape, which hadno equivalent in AES.

(2) The second derivative of the spectrum was useful forlocation of exact peak centres.

(3) Spectra subtraction using library peak shapes. Thereis an alignment problem and the background needsremoving Ðrst because this will often be di†erent forlibrary and experimental peaks.

(4) Fitting programs : too many components change,e.g. for Coster Kronig or spectator hole reasons,and, at the very least, an educated guess is needed tostart the process.

(5) Factor analysis can be successful and the literatureshows it to have a fast growing use. However, factoranalysis can return too many factors for this reason.

(6) Multiple regression methods (Muller ;Gaarenstroom) may be better.

(7) Smoothing (SawitskyÈGolay, 5 point) aids the eyebut original data should be used in Ðts.

A recommendation by Sherwood that original datashould always be included with spectra Ðts wasendorsed.

Distribution in depth

The following points arose form the various authors,particularly Professor Tougaard, who spoke on thistopic :(1) A universal curve is good for representing the loss

function of most elements (Tougaard).(2) A carbon universal curve for polymers is being

developed.(3) MassÈthickness (oz) is returned by most methods.(4) Methods using backgrounds are preferable to those

using intensities. In obtaining the depth distributionthere is a need to maximize the Ðtted range andminimise noise, or the number of genuine param-eters (concentrations and depth) rapidly decreases.

(5) A transport mean free path needs deÐning (seeabove).

Crystallographic and other structure/topographicinÑuences

The following points were worthy of note :(1) Channelling contrast : a substrate e†ect. Do not nor-

malize against a substrate element if channelling ispresent or suspected.

(2) Majority of technological samples una†ected.

Spatial distribution

(1) Resolution may be determined by the backscatteredtails.

(2) Apart from this e†ect, the limit could be the currentdensity limit for damageÈCazaux suggested thatthis could be one atom! Possibilities can be found in



deconvolution applied to the image, i.e. since

A(r) \ I(r)] C(r) ] S(r)

where I(r) is the beam shape and S(r) is the scattershape, then I and S can be deconvoluted.

(3) Scatter diagrams reveal phases.(4) Multivariant analysis can be useful.

Instrumental procedures

There is a need to calibrate our instrumentalparametersÈe.g. Channeltron life cycle, lens power,sample positionÈfor each FAT or FRR mode ofworking. Standard curves will assist in doing this.

Data banks of peak shape

(1) Their use increases understanding of instrument byoperators.

(2) New opportunities with World Wide Web.(3) NIST is assessing algorithms for peak Ðtting.

Artefacts

The following points arose :(1) The spectrum changes during acquisitionÈoften

because of charging and mass loss.(2) We cannot analyse most insulating materials, e.g.

there are only two papers on polymers. One reasonis the increase of dose with the square of lateraldimensionsÈthe critical dose was considered to beof the order 1 A cm~2. (It was also considered thatimaging XPS will meet the same problems as lateralresolution increases.)

(3) Electrostatic charging is best described by a two-layer model. Note that the local electrostatic Ðeldmay enhance mass di†usion. However, the predic-tion of Ðeld strength and direction is impossiblebecause it depends on trapping sites, etc.

(4) Revision of ASTM E983 is recommended.(5) Ionization damage through an Auger electron

cascade cannot be neutralized in an insulator.(6) Beam desorption of neutral species (from the lattice)

is important and the range of species emitted maybe wide.

(7) There is a need to reduce charging by, for example :a Ñood gun, a UV lamp with hl[ bandgap, use of a

low beam potential, windowing an area on thesurface or by extrapolating charging shift with time.

(8) Again, attention should be paid to linearization ofthe background.

FINAL COMMENT

One aim of the Workshop was to establish which physi-cal principles of AES were now so well established thatthey could be used as part of a rule-base for use in anExpert System designed to make the transition from“Physics to DataÏ in a semi-automatic manner. At theend of the Ðrst day, following the sessions of Weigh-tman and Powell, Castle attempted to illustrate howsome of the principles contained in the presentationsand the ensuing discussions could be encapsulated inrules. The following is taken from this presentation andhopefully presents a challenge for the future that willinspire further work along these lines.

Rules for identiÐcation of peak position

(1) If sample (s) \ free·atom, then run 2·step·(*SCF)calculation.

(2) If s \ metallic·element and not transition·element,then run 2·step·(*SCF) calculation.

(3) If s \ metallic·alloy then ? (look up table).(4) If s \ compound then ? (look up table).

Rules for interpretation of intensity

(5) If s \ metallic·element, then I\ *SCF·calc.(6) If Z\ 40, then identify Coster·Kronig.(7) If Z\ 40 and satellites, then Coster·Kronig \ yes.(8) If Z \ 40 and broadening, then Coster·

Kronig \ yes.(9) If Coster·Kronig \ yes, then Auger·intensity·ratio

unreliable.(10) If Auger·intensity·ratio \ unreliable, then quanti-

Ðcation \ unreliable.

Rules for interpretation of chemical state

(11) If Auger and photoelectron\ yes, then run *a.(12) If Auger and photoelectron\ yes, then run *b.


Documents

Report on the 11th IUVSTA workshop ‘Auger Electron Spectroscopy: from Physics to Data’