PHOTOELECTRON EXPERIMENTS ON SEMICONDUCTOR HETEROSTRUCrURESUSING SYNCHROTRON RADIATIONbyKEVIN MICHAEL COLBOWB.Sc. (Honors), Simon Fraser University, 1986M.Sc., University of British Columbia, 1988A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PH[LOSOPHYinTHE FACULTY OF GRADUATh STUDIESDepartment of PhysicsWe accept this thesis as conformingto the required standard..E..:.SkSTHE UNIVERSITY OF BRITISH COLUMBIADECEMBER 1991© Kevin Michael Colbow, 1991National Litxaryo( CanadaBibliothêque nationaledu CanadaCanadian Theses Service Service des theses canadiennesOttawa. CanadaKIA 0N4The author has granted an irrevocable nonexclusive hcence allowing the National Libraryof Canada to reproduce, loan, dstiibute or sellcopies of his/her thesis by any means and inany form or format, maFing this thesis availableto interested persons.The author retains ownership of the copyrightin his/her thesis. Neither the thesis norsubstantial extracts from it may be printed orotherwise reproduced without his/her permission.L’auteur a accordé une licence irrevocable etnon exclusive permettant a Ia Bibliothèquenatiönale du Canada de reproduire, prêter,distribuer ou vendre des copies de sa thesede quelque maniére et sous quelque formeque ce soit pour mettre des exemplaires decette these a Ia disposition des personnesintéressées.L’auteur conserve Ia propnété du droit d’auteurqul protege sa these. Ni Ia these ni des extraitssubstantiels de celle-ci ne doiverit êtreimprimés ou autremerit reproduits sans sonautorisatiori.11+anaaaISBN 0-315-75397-8In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department of r/17J(—-The University of British ColumbiaVancouver, CanadaDate / ,27 /92DE-6 (2/88)11ABSTRACTPhotoelectron spectroscopy using synchrotron radiation has been used tostudy a variety of semiconductor heterostructures. The electronic properties andcomposition of the interfaces between selected wide band gap materials and GaAshave been explored. Three types of wide band gap materials were considerednamely chalcogen compounds, alkaline-earth fluorides, and rare-earth trifluorides.Prior to the growth of thin surface layers, the semiconductor substrates hadto be atomically clean. The final step of the cleaning process normally involvesdesorbing the surface oxide layer. The effect of surface roughening was observedfor the first time as a result of optically monitoring the oxide desorption in GaAs.High resolution photoemission spectroscopy of clean GaAs (100) surfacestreated with H2S revealed that these surfaces are completely terminated by a GaSspecies which is stable in air or water. The H2S treated surfaces are better definedin terms of the interfacial chemical bonding than the recently proposed chemicaltreatments involving Na2S or (NH4)2S.The interface between ZuSe and GaAs (100) was studied to determine theinterfacial composition as a function of annealing temperature. From an analysisof the chemical shifts and relative intensities of the atomic core levels, we obtainedthe first direct measurements of the interfacial composition. With increasingannealing temperature, an interfacial Ga2Se3 layer forms with loss of Zn and As atthe interface. The first direct valence-band photoemission measurements fortheZnSe/GaAs (100) interface yielded a band offset of 1.25 ± 0.07 eV.The formation of the interfaces between the alkaline-earth fluorides (CaF2,SrF2, BaF2) and GaAs (100) have also been studied by photoemission. Thin films ofthese fluorides were deposited on clean GaAs (100), and subsequently annealed. ForCaF2, by 550°C, a monolayer of Ca was found to react with As at the interface with an111associated loss of approximately one Ga and one F by evaporation per formulaunit ofCaF2. The loss of Ga and F at the interface had not been reported previously. Thethickness of the interfacial layer terminates at one monolayer for therange ofannealing conditions studied. For both the SrF2/GaAs and BaF2/G As interfaces,results supporting cation-As bonding accompanied by the loss of F andGa wereobtained. The interfacial valence-band offsets for CaF2, SrF2, and BaF2 on GaAs werefound to range between 7 eV and 8.5 eV.X-ray absorption spectroscopy and deexcitation electron spectroscopy (DES)were used to investigate the nature of the electronic transitions near the Ca L edgein CaF2, the Ba M and N edges in BaF2 and the F K edges in CaF2, SrF2 and BaF2. Thecation absorption edges were interpreted in terms of atomic multiplet theory andthe bound excited states were identified. The binding energies for these coreexcitons were estimated for the first time. By resonantly exciting these bound statesand measuring the resulting deexcitation spectra (DES) the nature of theexcited-state electrons was determined. The presence of sharp spectatorlines nearthe Auger region of the spectrum showed which absorption transitionshad boundfinal states. The participator lines, which appear as an enhancement of the normalphotoeniission lines, provide information on the atomic location of thebound-stateelectrons.Thin films of the rare-earth trifluorides (LaF3, NdF3, TmF3) grown on cleanSi (111) surfaces were also studied. By resonant excitation of the giant 4d—> 4 ftransition, the photoemission from the cation 4f orbitals was enhancedand thusdistinguished from the photoemission signal of the overlapping F 2p valence band.The valence-band offset was found to be the same for the three heterojunctions.The lack of any interfacial electron transfer into the unoccupied 4f orbitals wasreported here for the first time and is attributed to the on-site Coulombrepulsionenergy U.ivTABLE OF CONTENTSABSTRACT iiLIST OF TABLES viiLIST OFFIGUPES viiiACKNOWLEDGMENTS xiii1. INTRODUCTION 11.1 CHALCOGENS ON Ill-V SEMICONDUCTORS 11.1.1 Surface Passivation of GaAs with Sulfur 11.1.2 ZnSe on GaAs (100) and InP (111) 21.2 SEMICONDUCTOR INTERFACES WiTH FLUORIDES 31.2.1 Alkaline-Earth Fluorides on GaAs (100) 31.2.2 Rare-Earth Trifluorides on Si (111) 72. EXPERIMENTAL 92.1 SAMPLE PREPARATION 92.1.1 Substrate Cleaning 92.1.2 Optical Monitoring 122.2 GROWTH OF THIN SURFACE LAYERS 142.2.1 Surface Passivation with Sulfur 142.2.2 Solid Phase Epitaxy 152.2.2.1 ZnSe Deposition 152.2.2.2 Fluoride Deposition 162.2.3 Temperature Measurement 172.3 PHOTOELECTRON SPECTROSCOPY 212.3.1 Photoemission Spectroscopy 262.3.1.1 Chemical Shifts 262.3.1.2 Film Thickness Determination 29V2.3.1.3 Deexcitation Electron Spectroscopy .332.3.2 X-ray Absorption Spectroscopy 342.3.2.1 Total Electron Yield Measurement 352.3.2.2 Microchannel Plate Detectors 353. RESULTS AND DISCUSSION 373.1 CHALCOGENIDES 373.1.1 H2S Treatment of GaAs (100) 373.1.2 ZnSe on GaAs (100) 453.1.2.1 Interfacial Bonding 453.1.2.2 Valence-Band Offsets 523.2 ALKALINE-EARTH FLUORIDES 583.2.1 Interfaces 583.2.1.1 Formation of the CaF2/G As (100) Interface 583.2.1.2 SrF2 and BaF2 Interfaces with GaAs 673.2.1.3 Band Offsets at Insulator/Semiconductor Interfaces 733.2.2 X-ray Absorption 853.2.2.1 Atomic Multiplet Theory 853.2.2.2 The Calcium L Edge in CaF2 903.2.2.3 The Barium M and N Edges in BaF2 943.2.2.4 The Fluorine K Edges 1043.2.3 Deexcitation Electron Spectroscopy 1073.2.3.1 The Calcium L Edge in CaF2 1103.2.3.2 The Barium M and N Edges in BaF2 1133.2.3.3 The Fluorine K Edges 1163.3 RARE-EARTH TRIFLUORIDES 1213.3.1 Resonant Photoemission 1213.3.2 Band Offsets 126vi4. SUMMARY AND CONCLUSIONS .131REFERENCES 135APPENDIX I Microchaunnel Plate Detector 143APPENDIX II Chemical Shifts From Electronegativity Values 144viiLIST OF TABLESTable PageI. Comparison of valence-band offset values determined in this study 57with those predicted by the LCAO model of Harrison for somesemiconductor heteroj unctions.II. Summary of the atomic calculation for the Ca2 2p— 3d absorption 93transition.III. Summary of the atomic calculation for the Ba2 3d—4f absorption 96trans i ti o a.IV. Summary of the atomic calculation for the Ba2 4c1—*4f absorption 101transition.V. Summary of the binding energies of the various excitons for the 109alkaline-earth fluorides determined in this study.viiiLIST OF FIGURESFigure Page1. Unit cells for the fluorite and zincblende structures. 42. Energy band gap versus lattice constant for Ill-V semicouducting 6materials.3. Photoemission spectra during the cleaning of GaAs showing the 11As 3d and Ga 3d core levels and the valence-band region.4. Intensity of the diffuse reflection from a polished GaAs wafer 13during a linear temperature ramp.5. Experimental apparatus for the thermocouple/pyrometer 19temperature measurement.6, Temperature as measured by the thermocouple compared with the 20temperature measured by the optical pyrometer.7. Optical configuration of the Ul beamline. 228. Top view of the analysis chamber at the Ui beamline. 249. The three types of photoelectron spectroscopy experiments 25performed in this study.10. Photoemission spectra of GaAs (100) following various surface 38treatments including H2S.11. Photoemission spectra of the As 3d core level of GaAs (100) for the 41(a) clean, (b) H2S exposed and (c) annealed surface.12. Photoemission spectra of the Ga 3d core level of GaAs (100) for the 42(a) clean, (b) H2S exposed and (c) annealed surface.ix13. Photoemission spectra for a sample consisting of 14 A of ZnSe on 46GaAs (100) for a series of annealing temperatures from 410°C to590°C measured with a photon energy of 105.2 eV.14. Peak intensity for each of the species (Se, Zn, Ga and As) relative to 48the total intensity of all four species as a function of annealingtemperature.15. Thicknesses of the surface ZnSe layer, the interfacial Ga2Se3 layer 49and the total overlayer as a function of annealing temperature forthe sample of Fig. 13 as determined from the fits in Fig. 14.16. Photoemission spectra of the Se 3d and Ga 3d core levels for the 51series of annealing temperatures for the sample of Fig. 13.17. Photoemission spectra for the valence-band region of 7 A and 14 A 53of ZnSe on GaAs.18. Photoemission spectrum for the valence-band region of 2-5 A of 55ZnSe on InP.19. Photoemission spectra for 1.1 ML of CaF2 on GaAs for a series of 59annealing temperatures from 460°C to 590°C measured with aphoton energy of 83.6 eV.20. Fitted As 3d and Ca 3p core-level spectra for a series of annealing 60temperatures for the sample of Fig. 19.21. Ratios of the areas of the F 2p to Ca 3p and the Ga 3d to As 3d 62photoemission peaks as a function of annealing temperature forthe sample of Fig. 19.22. Photoemission spectra for a series of CaF2 overlayers (a) 6.1 ML, 63(b) 2.2 ML, (c) 1.4 ML and Cd) 0.48 ML on GaAs annealed to 570°C.23. Proposed structural model for the CaF2/G As (100) interface. 65x24. Photoemission spectra for 2.3 ML of SrF2 on GaAs for a series of 68annealing temperatures from 370°C to 610°C measured with aphoton energy of 83.6 eV.25. Photoemission spectra of the Sr 3d core level and the As 3d, Sr 4s 70spectral region for a series of annealing temperatures for thesample of Fig. 24.26. Ratios of the areas of the Sr 4p to F 2p and Ga 3d to As 3d 71photoemission peaks as a function of annealing temperature forthe sample of Fig. 24.27. Photoemission spectra for 3.8 ML of BaF2 on GaAs for a series of 72annealing temperatures from 470°C to 620°C measured with aphoton energy of 83.6eV.28. Ba 4d core-level spectra for a series of annealing temperatures for 74the sample of Fig. 27 measured with a photon energy of 160.9 eV.29. Fitted As 3d and Ba 4d core-level spectra for the sample of Fig. 27 75after annealing at 620°C.30. Photoemission spectrum of the valence-band region for 6 ML of 77CaF2 on GaAs measured with a photon energy of 83.6 eV.31. Band alignment and Fermi level for 6 ML of CaP2 on GaAs. 7832. Valence-band offset as a function of CaF2 overlayer thickness for 79the CaF2/G As interface.33. Fermi level position relative to the GaAs valence-band maximum as 81a function of CaF2 overlayer thickness for the CaF2/G As interface.34. Valence-band offset as a function of overlayer thickness for the 83alkaline -earth fluoride/GaAs interfaces.35. Fermi level position as a function of overlayer thickness for the 84alkaline -earth fluoride/GaAs interfaces.xi36. Total electron yield measurement of the absorption of CaF2 at the Ca 92L edge with 160 meV resolution37• Total electron yield measurement of the absorption of BaF2 at the 95Ba Medge.38. Total electron yield measurement of the absorption in the vicinity 98of the Ba N edge for BaF2 showing the giant 4d—+ 4f resonancetransition.39 Total electron yield measurement of the absorption of BaF2 at the iooBa N edge.40. Total electron yield measurement of the absorption of CaF2 at the F 105K edge.41. Total electron yield measurement of the absorption of BaF2 at the F 106K edge.42. Total electron yield measurement of the absorption of SrF2 at the F 108K edge43. Photoemission spectra with 500 meV overall resolution produced 111by excitation at different energies in the Ca L edge of CaF2corresponding to points A-E in Fig. 36.44. Photoemission spectra produced by excitation at different energies 114in the Ba M edge of BaF2 corresponding to points A-B in Fig. 37.45. Photoemission spectra produced by excitation at different energies 117in the Ba N edge of BaF2 corresponding to points A-E in Fig. 39.46. Photoemission spectra produced by excitation at different energies 118in the F K edge of CaF2 corresponding to points A-D in Fig. 40.47. Photoemission spectra produced by excitation at different energies 120in the F K edge of BaF2 corresponding to points A-D in Fig. 41.xii48. Photoemission spectra of the valence-band region of NdF3 on Si 122taken at resonant and non-resonant photon energies. The insetshows the absorption spectrum of NdF3 in the vicinity of the NdN edge.49. Photoemission spectra of the valence-band region of TmF3 on Si 124taken at resonant and non-resonant photon energies. The insetshows the absorption spectrum of TmF3 in the vicinity of the NdN edge.50. Photoemission spectra of the valence-band region of LaF3 on Si 125taken at resonant and non-resonant photon energies. The insetshows the absorption spectrum of LaP3 in the vicinity of the NdN edge.51. Valence-band spectra for the three rare-earth trifluoride/silicon 128interfaces used to determine the valence-band offsets shown inFig. 52.52. Measured band alignments for the rare-earth trifluoride/Si (111) 129interfaces showing the relative positions of the F 2p valence bandand the cation 4f levels.xiiiACKNOWLEDGMENTSI would like to express sincere gratitude to my supervisor, Dr. Tom Tiedje forhis guidance throughout the course of this project. He provided a stimulatingworking atmosphere.I would like to also thank Dr. Wolfgang Eberhardt, Dr. Dale Sondericker, Dr.Zugen Fu and the Ui beamline support staff for their assistance at the NationalSynchrotron Light Source, Brookhaven National Laboratory, which is supported bythe U.S. Department of Energy. Dr. Jeff Dahn, Yuan Gao and Duncan Rogers alsocontributed to the sometimes exhausting chore of collecting data and for that I amalso grateful.Finally, the author acknowledges financial support from the Natural Sciencesand Engineering Research Council in the form of a 1967 postgraduate scholarshipand from the University of British Columbia for a Killam predoctoral scholarship.11. INTRODUCTIONThe heteroepitaxy of compounds on both elemental and compoundsemiconductors poses many interesting questions about interface formation. One ofthe most important of these questions is the nature of the chemical bonding at theinterface and the related atomic arrangement. The alignment of the energy bandsof the overlayer relative to the substrate is another property that characterizesthese interfaces. It is the goal of this study to explore the electronic properties andcomposition of the interfaces between selected wide band gap materials and GaAs,using synchrotron radiation based photoelectron spectroscopies. Three classes ofwide band gap overlayer materials will be considered, namely chalcogencompounds, alkaline-earth fluorides, and rare-earth trifluorides. (see next page)For semiconductor device applications it is important to be able to fabricateheterojunctions with a low density of defect states, located in the band gap of thematerial. Oxidation of GaAs surfaces does not yield device quality interfaces, sincethe density of interface states in the band gap is high.1 Currently, device qualityinterfaces of GaAs can be produced by growing epitaxial films of AlGaAs on GaAs.For these epitaxial structures the defect state density is reduced to levels acceptablefor device applications. In optoelectronics technology, it would be useful to havethe same quality interface but between GaAs and an insulator or at least a largerband gap material than AIGaAs.241.1 CHALCOGENS ON Ill-V SEMICONDUCTORS1.1.1 Surface Passivation of GaAs with SulfurVarious new wet chemical treatments involving sulfur compounds have beenreported that produce passivated surfaces of GaAs with a decreased carrierrecombination velocity and increased luminescence efficiency compared to earlierIIBfIB IVB VB VIB VIlESIPSFCa ZnGa AsSeSr InBaLa1AFigure Caption. Periodic table showing the relationship between the compoundsused throughout this work. The semiconductor substrates consisted of the 111-Vmaterials (GaAs and InP) and silicon. The wide band gap overlayer materialsconsisted of the chalcogen (group VIB) compounds (H2S and ZnSe), thealkaline-earth (group hA) fluorides (CaF2, SrF2 and BaF2), and the rare-earthtrifluorides (LaP3, NdF3 and TmF3),GROUPIAH hAlilA2passivation techniques.5’6 However the composition of these surfaces was notrevealed by these studies. Therefore, initially, photoemission measurements onGaAs (100) wafers passivated using Na2S7’8 or (NH4)2S7’9 chemical treatments asprescribed in the literature5’6 were performed. These high-resolution core-levelstudies enables one to resolve the chemical shifts induced by a sulfur or any oxygenatom chemically bound to either a Ga or an As atom at the surface. In combinationwith valence band photoemission studies, a microscopic picture of the atomic andelectronic structure of these passivated surfaces can be obtained.In these photoemission studies7’8 it was found that even a small amount ofsulfur on the the surface was effective in improving the electronic quality of thesurface. As a result, the possibility of developing better passivation techniquesresulting in surfaces completely terminated with sulfur with superior electronicproperties has been considered. One such alternative technique is a treatment withgaseous H2S, and the results of these experiments will be reported here andcompared with previous work.1021.1.2 ZnSe on GaAs (100) and InP (111)Considerable efforts are also being devoted to the characterization of theGaAs/ZnSe system and other heterostructures formed from Ill-V and Il-VIsemiconductor materials due to their potential application in optoelectronic devices.The lattice mismatch of ZnSe (a = 5.6676 A) and GaAs (a = 5.6537 A), both of whichhave the cubic zincblende structure, is only 0.27% and the two semiconductors alsohave closely matched thermal expansion coefficients. Epitaxial ZnSe layers canmodify the electronic properties of the GaAs surface and the wide band gap of ZnSe(2.67 eV at room temperature compared to 1.42 eV for GaAs) makes it potentiallyuseful in optoelectronic devices which operate in the blue portion of the visible3spectrum. Optical nonlinearities and integrated optical waveguiding in the visibleare also expected to be achievable with GaAs/ZnSe heterostructures)Epitaxial growth of ZnSe has been achieved on GaAs (100)13-17 and(110). 17-19 ZnSe films deposited by molecular beam epitaxy (MBE)1316 and bycongruent evaporation from ZnSe powder1749 have been studied using a variety ofspectroscopic techniques. In particular Li et al. 16 have shown by transmissionelectron microscopy (TEM) that the structure of the interfacial layer is consistentwith the formation of Ga2.Also Tu and Kahn’7 have suggested the formation of agallium selenide interfacial layer at the ZnSe/GaAs heteroepitaxial interface andshown that after extensive post-deposition annealing a surface ZnSe layercompletely converts to Ga2Se3. Here the results of a photoemission study of theZnSe/GaAs (100) interface are described, in which the first direct measurements ofthe interfacial composition are obtained. Also, the interfacial band offsets for thissystem as well as the analogous ZnSe/InP system have been determined fromphotoemission measurements of the valence band density of states for the first time.1.2 SEMICONDUCTOR INTERFACES WITH FLUORIDES1.2.1 Alkaline-Earth Fluorides on GaAs (100)The three alkaline-earth fluorides, CaF2, SrF2 and BaF2 and their alloys, canbe lattice-matched to a broad range of Ill-V compounds. These fluorides have thefluorite structure2° which consists of the alkaline-earth cation located at the centerof eight F anions situated at the corners of a surrounding cube. Conversely, each Fanion is surrounded by a tetrahedron of alkaline-earth cations. The symmetry iscubic with four molecules per unit cell as shown in Fig. 1(a), where CaF2 is used asan example. The Ill-V compounds considered here have the zincblende structure.The structure of GaAs is shown in Fig 1(b), where the unit cell also contains fourmolecules. This crystal structure is face-centered cubic, where each Ga atom has4(a)•Ca 0F(b)•Ga QAsFigure 1. Unit cells for the (a) fluorite (CaF2) and (b) zincblende (GaAs) structures.5about it a tetrahedron of As atoms and similarly each As atom has about it atetrahedron of Ga atoms. The difference between these two crystal structures is theoccupation of tetrahedral sites in the fcc framework of the Ca and Ga atomsrespectively. In the unit cell of the fluorite structure, all eight of the tetrahedralsites are occupied by F ions, whereas in the unit cell of GaAs, As atoms occupy onlyfour of these tetrahedral sites and the other four are vacant. One would then expectthat epitaxial growth of one material onto the other would be possible if the cubiclattice constants were similar.Figure 2 shows the energy band gap versus the lattice constant for the cubicunit cell for a number of semiconducting and insulating materials. The linesjoining the binary compounds give the ternary energy gap and lattice constant.The lines joining alkaline-earth fluorides give the energy gap and lattice constantfor a solid solution of two of the fluorides. No solid solution between CaF2 and BaF2can be made. As shown, the alkaline-earth fluorides and their alloys can belattice-matched to a broad range of Ill-V compounds including GaAs, luP and InAs.The growth of epitaxial fluorides on semiconductor substrates has beenexplored extensively in recent years.24’32 Epitaxially grown group hAfluorides on elemental2t23’930 and compound semiconductors2428 arepotentially useful as epitaxial insulators in multi-level integrated device structures,in electro-optic devices, and as strain-relieving buffer layers for heteroepitaxialgrowth of semiconductors on lattice-mismatched substrates. The electronicproperties and thermal stability of these ionic-covalent interfaces are crucial indetermining the suitability of these heterostructures for device applications.Together with CaF2 on Si, CaF2 on GaAs is a useful model system for studies of theionic-covalent interface. In addition CaF2 on GaAs can be regarded as a prototypefor an alkaline-earth fluoride/Ill-V semiconductor interface.612100 5.7 5.9 6.3LATTICE CONSTANT (A)Figure 2. Energy band gap versus cubic lattice constant for the alkaline-earthfluorides and a selection of semiconducting materials. The lines joining thecompounds give the energy gaps and lattice constants for the alloys.5.3 5.5 6.17Calcium fluoride is a wide band gap (12.1 eV33) dielectric, nearlylattice-matched to GaAs (3.5% lattice mismatch). Epitaxial growth of CaF2 has beenachieved on GaAs (100), for GaAs substrates cleaned in vacuum by thermaldesorption of the oxide24 and in ultrahigh vacuum (UHV) using the technique ofion-bombardment and annealing.25 Recently, the CaF2/G As (100), (110) and (111)interfaces formed at room temperature and subsequently annealed have beenstudied by photoemission spectroscopy.26 CaF2 films deposited by MBE onGaAs (111)27 substrates at 500°C and on GaAs (100)28 substrates at high (580°C,620°C) and low (420°C, 500°C) substrate temperatures have also been studied byphotoemission spectroscopy and electron diffraction. Here the results of the earlierwork2628 are extended with a systematic study of the effects of thermal annealingon the interfacial composition and bonding for CaF2 deposited on GaAs (100) at roomtemperature and subsequently annealed. New results on the interfacialcomposition, Fermi level, and band alignment as a function of overlayer thicknessand annealing temperature are reported. In our experiments the substrates werecleaned in UHV using a repetitive Ar sputter-anneal process. High resolutionphotoemission spectroscopy was used to determine the electronic structure andbonding at the CaF2/GaAs (100) interface as a function of the annealingtemperature. Similar experiments were also performed for the SrF2/GaAs andBaF2/G As systems. The chemical bonding and valence-band offsets for these threesystems are then compared.1.2.2 Rare-Earth Trifluorides on Si (111)The rare-earth trifluorides with the hexagonal LaF3 structure are anotherclass of fluorides that can be grown epitaxially on semiconductor substrates.4’312The hexagonal tysonite structure of the lighter rare-earth trifluorides is well suitedfor epitaxy on Si (111) because the Si (111) surface has hexagonal symmetry with8an equivalent lattice parameter of a = 3.840 A which is 8% mismatched from thelattice parameter of LaF3 (a = 4.148 A) and only 2% from TmF3 (a = 3.905 A).20 Inaddition LaF3 can be doped with luminescent rare-earth ions which might open uppossibilities for novel silicon-based optoelectronic devices.32In the present work photoemission and x-ray absorption spectroscopy areused to measure the interfacial energy level alignment in heterojunctions of the La,Nd and Tm trifluorides on Si (111). Determination of the band alignment for theserare-earth trifluorides is complicated by the overlap between the photoemissionsignal from the partially filled 4f orbitals with the photoemission signal from theF 2p valence band. By resonant excitation of the giant 4d—. 4f transition, present inthe first row rare-earth trifluorides including NdF3 and TmF3 it is shown that it ispossible to greatly enhance the photoemission signal from the 4f electrons andthereby distinguish it from the photoemission signal from the overlapping F 2plevels. This allows one to determine the position of the cation 4f levels and the F 2pvalence band separately and thus determine the alignment of these energy levelsrelative to the conduction and valence bands of the silicon at the interface.92. EXPERIMENTALIn this section, the experimental techniques and analysis used throughoutthis work will be discussed. The first two sections deal with the experimental detailsconcerning sample preparation, thin film deposition and annealing. The thirdsection will describe the Ui beamline at the National Synchrotron Light Source(NSLS), Brookhaven National Laboratory where the photoelectron experimentswere performed. The three types of photoelectron experiments will be comparedand some of the fundamental principles of analysis associated with these techniqueswill be discussed.2.1 SAMPLE PREPARATIONThe GaAs samples used throughout this work were semi-insulating liquidencapsulated Czochralski (LEC)-type wafers with a (100) surface orientationsupplied by Johnson Matthey Electronics. In certain experiments, indiumphosphide (InP) and silicon (Si) were used as substrates, and both these types ofwafers had a (111) surface orientation.Prior to chemical treatments in the laboratory or thin film deposition inultrahigh vacuum (UHV), it is necessary to carefully clean on an atomic scale thesemiconductor surfaces so as to remove primarily carbon and oxygen contaminants.In this section, these cleaning procedures and the techniques used to analyze theclean surfaces are described.2.1.1 Substrate CleaningPrior to the sulfide treatments of the clean GaAs surfaces, the substrates werefirst solvent washed ultrasonically for 5 mm with acetone and methanol and thenrinsed with deionized water. The substrates were then chemically etched in a10concentrated H2S04:30% H20 : H20 mixture (1:1:50 by volume). The solution wasgently agitated for 2 miii and then the samples were rinsed with deionized waterand finally blown dry with nitrogen and kept under an inverted funnel in flowingnitrogen for a few minutes prior to the chemical sulfide treatments.Samples used as substrates for H2S treatments or thin film deposition werenormally loaded into the UHV chamber as received. The substrates were thencleaned by thermal desorption to remove the surface oxides. A repetitivelow-energy (500 eV) Ar+ or Ne+ion sputter-anneal process was normallynecessary to remove the residual carbon that was detected in photoemission. Thesurface oxides desorb in two steps: the arsenic oxide evaporates first (in the400-500°C range) followed at higher temperature by the gallium oxide, whichdesorbs over a narrow temperature interval, typically above 5 80°C. The galliumoxide desorption peak as measured with a quadrupole mass spectrometer is relativelysharp and is used as a reference point for substrate temperature calibration. Thiscalibration technique will be discussed in a later section.Figure 3 shows a series of photoemission spectra taken at a photon energy of84.3 eV showing the As 3d and Ga 3d core levels and the valence-band region for aGaAs sample during the cleaning procedure. The top spectra are the As 3d andGa 3d core levels for a GaAs sample as received. The middle spectrum is for thesample after it had been annealed to approximately 450°C, where the majority of thearsenic oxide and carbon contaminants were removed. The bottom spectrum is forthe clean GaAs sample after it had been sputtered and annealed to approximately5 80°C. At this annealing temperature all the remaining surface oxide and carbonhad been removed. In the top spectra, significant contributions to thephotoemission intensity of the As 3d and Ga 3d core levels are observed on thelow-kinetic-energy side of the corresponding core levels in the spectrum for theclean sample. This contribution is due to As-O bonding and Ga-O bonding on the1180Figure 3. Series of photoemission spectra taken at a photon energy of 84.3 eVduring the cleaning of a GaAs sample, (a) as received, (b) after annealing to 450°C(arsenic oxide removed), and (c) after sputtering and annealing to 580°C (clean).The large peak in spectrum (a) is the chemically shifted component of the As 3dcore level associated with As-O bonds.30 40 50 60 70KINETIC ENERGY (eV)12GaAs surface. After annealing at 450°C, only the gallium oxide remains asillustrated by the broad Ga 3d peak in the middle spectrum. The valence-bandregion is also altered during sample cleaning. As the surface oxides are removed byannealing, the 0 2p contribution to the valence-band region of the spectrumdecreases as illustrated in Fig. 3. The clean GaAs sample has a flat valence bandwith no additional features.Similarly, prior to the deposition of the rare-earth trifluorides on silicon, thesubstrates were cleaned by the sputter-anneal process, until no residual carbon wasdetectable in photoemission and the Si 2p photoemission peak showed no chemicalshift associated with a surface oxide. The sample was annealed at approximately600°C for one minute after each sputtering cycle.2.1.2 Optical MonitoringIn desorbing the oxides from the GaAs wafers, a faint haze, which is difficultto detect in room light, becomes visible under intense illumination. This haze wasthought to be caused by surface roughening during the oxide desorption process.The formation of the haze was monitored optically by measuring the diffusereflection of a He-Ne laser beam from the front surface of the polished wafer.Figure 4 shows the intensity of the diffuse reflection as a function of annealingtemperature during a linear temperature ramp. The sharp rise in the diffusereflectivity coincides exactly with the peak in the Ga20 mass spectrometer signalcorresponding to the oxide desorption.From an investigation of the oxide desorption for varying oxide thicknesses,it has been shown35 by Van Buuren et at. of this lab that the desorptiontemperature increases linearly with oxide thickness up to approximately 20 A. Thiswas interpreted in terms of a thin film instability. One example of an appropriateinstability is the breakup of the homogeneous oxide film into islands. Such an1310Cl)>->I0w-JLL4wwCl)UU0540 660Figure 4. Intensity of the diffuse reflection of a He-Ne laser beam from the frontsurface of a polished GaAs wafer during a linear temperature ramp in which a 17 Aoxide is desorbed.560 580 600 620 640TEMPERATURE (°C)14inhomogeneous oxide desorption process is also consistent with the observedsurface roughening.The diffuse reflectivity measurements are interpreted as evidence that thesurface roughens on a length scale comparable to the wavelength of light duringthe oxide evaporation. Once the oxide desorption is complete the diffuse reflectivitycontinues to increase with time as shown in Fig. 4. This continuing change isinterpreted as progressive faceting of the surface into a lower energy structurefollowing the disruption associated with the oxide removal. These results highlightthe desirability of surface oxide layers which evaporate at low temperatures in GaAswafer preparation for subsequent growth. The low-temperature oxides minimizethe surface roughening and the need for growth of a surface-smoothing layer.However, in our experiments no surface-smoothing layers were grown prior toheteroepitaxy.2.2 GROWTH OF THIN SURFACE LAYERS2.2.1 Surface Passivation with SulfurTwo types of sulfide treatments were performed on the clean GaAs surfaces.The first involved chemical treatment using solutions of either Na2S or (NH4)2S andthe second was a treatment with H2S in the gas phase. Although the results of theH2S treated samples are reported here, comparisons are also made to chemicallytreated samples prepared by our group and others)°2 For this reason, theexperimental details of both types of treatments are described. The Na2S chemicaltreatment consisted of the addition of a drop (—0.025 mL) of sodium sulfide solution(25% by weight of Na2S.9H0 in H20) evenly onto the surface. After drying thesolution under flowing nitrogen, the surface was then rinsed briefly with waterand dried again under flowing nitrogen. The (NH4)2S treatment consisted ofsoaking the wafer in a saturated (NH4)2S solution for up to 15 mm with periodic15agitation. The sample was then rinsed and dried as described above. The sampleswere then loaded into the UHV system approximately 30 mm after the treatment.The H2S treatments were performed in a separate UHV deposition chamber.The clean substrates were transferred into this chamber and exposed to9 x i0 Torr of H2S for 15 mm. The gas was introduced through a leak valve, andthe valve to the turbo-molecular pump was left only partially open during the gasexposure time. The FI2S was activated by a hot filament, in order to produce areactive sulfur species. After exposure, the valve to the turbo-molecular pumpvalve was re-opened so that the pressure would return to the 10-8 Torr range. Thesample was then transferred to the analysis chamber where the sample annealingand photoelectron experiments were performed.2.2.2 Solid Phase EpitaxyThin films of ZnSe, the alkaline-earth fluorides (CaF2, SrF2, BaF2) and therare-earth trifluorides (LaF3, NdF3, TmF3) were deposited in UHV onto cleansubstrates by evaporation of the solids from an evaporation vessel. The evaporationrate could be adjusted and controlled by the power input to the resistively heatedvessel.2.2.2.1 ZnSe DepositionZnSe was evaporated at a rate of 0.3-0.5 A/s from a tantalum evaporationvessel containing the powder, in a background pressure of iO8 Torr. A quartzcrystal thickness monitor was used for relative thickness measurements duringdeposition and for determining the deposition rate. The directionality of theevaporation source and the geometry of the crystal monitor relative to the samplesubstrate prevented us from obtaining reliable absolute thickness measurementsfrom the crystal monitor. The substrates were near room temperature during16deposition, heated only by the thermal radiation from the evaporator. The presenceof elemental Zn was detected in photoemission after deposition, but it evaporatedwhen the sample was annealed at the lowest temperature, approximately 350°C. Theratio of the Ga 3d to Zn 3d photoemission peak was found to decrease exponentiallywith increasing film thickness as determined from the thickness monitor for filmsannealed at 370°C. This indicates that uniform ZnSe films were formed, a resultconsistent with the so-called Frank-van der Merwe (FM) growth mode, where thefilm grows by the successive addition of two-dimensional monolayers. In contrast,the Volmer-Weber (VW) mode is one where growth initiates from three-dimensionalislands. After deposition the ZnSe films were annealed at progressively highertemperatures for 1 mm intervals in a range from 370°C to 600°C.2.2.2.2 Fluoride DepositionThe alkaline-earth fluorides (CaF2, SrF2 and BaF2) and the rare-earthtrifluorides (LaF3, NcIF3 and TmF3) were deposited in a similar manner to ZnSe. Allof the fluorides evaporate stoichiometrically as molecules. The evaporation rate wastypically 0.2-0.6 A/s from a tubular evaporation vessel made of tantalum,containing either the powder or small crystals in a similar background pressure of1 Torr. The substrates were again near room temperature. Initially, for thedeposition of CaF2, a thoroughly outgassed boron nitride crucible was used as anevaporation vessel to prevent the incorporation of Ta into the deposited films, aneffect that had been reported earlier.23 The films grown with our evaporationsources, however, showed no evidence of Ta contamination. These vessels werecylindrical tubes of 0.002 inch Ta, 2-3 cm long and approximately 1 cm wide, withboth ends flattened and a small hole (3-5 mm in diameter) near the middle of thetube section. The ends were fastened to the copper posts of the electricalfeedthrough, and the tube was filled with the fluoride material. Approximately 20 A17(200 W) were required to achieve a stable evaporation rate. This design is thoughtto be more efficient in heating the fluoride material than conventional evaporationboats. As a result, the temperature of the evaporator was kept low enough toprevent Ta contamination of the fluoride films during evaporation. The thicknessmonitor was again used for relative thickness measurements during deposition andfor monitoring the evaporation rate. After deposition, the films were annealed atprogressively higher temperatures for I mm intervals in a range from 400°C to650°C.The ratio of the Ga 3d to the F 2p photoemission peak area was found todecrease exponentially with CaF2 overlayer thickness in a manner consistent with auniform film (see equation (8)), for both unannealed films as well as for filmsannealed at 460°C. This indicates that as in the ZnSe case, uniform CaF2 films wereformed and that the Volmer-Weber three-dimensional island growth reported withhigh temperature substrates28 is not present for growth on room temperaturesubstrates. In contrast, for the rare-earth trifluorides there was no evidence thatsupported layer by layer growth of the films at substrate temperatures rangingfrom 370°C to 460°C.2.2.3 Temperature MeasurementOne-square-centimeter pieces of the GaAs wafers were mounted in Ta foilbaskets spot-welded to Ta wires which could be heated resistively. Thesample/basket temperature was monitored with an IRCON infrared pyrometer,sensitive to 2 im radiation, which was calibrated by the temperatures at which theGa and As oxides were desorbed as observed with a quadrupole mass spectrometer,assumed to be 580°C and 450°C respectively.35’6 The pyrometer was also calibratedby a type-K (chromel-alumel) foil thermocouple placed in contact with the frontsurface of a similar sample.18A schematic of the experimental apparatus is shown in Fig. 5(a) and thedetails of the sample holder are shown in Fig. 5(b). A small vacuum system wasconstructed consisting of a UHV window providing a line-of-sight to the samplesurface for the optical pyrometer, and a thermocouple and electrical feedthroughfor measuring and adjusting the sample temperature. A portable pumping stationequipped with a mechanical roughing pump and a turbo-molecular pump was usedto evacuate the system.Figure 5(b) shows the sample, which consisted of a thin film (.—30 A) of CaF2on GaAs measuring 15 x 10 mm and suspended in a basket made from 0.002 inch Ta.The basket was screwed down to the copper sample holder with Ta wires spot-weldedto the basket which could be heated resistively. Approximately 10 A (40 W) wererequired to raise the sample temperature to 600°C in this arrangement. The type-K(chromel-alumel) thermocouple foil, 0.0005 inches thick, was held down with aninsulating MACOR block. The thermocouple point made contact with the GaAssurface 5 mm beyond the MACOR and 1 mm from the sample edge. A small Ta clipwas used to keep the thermocouple point in contact with the GaAs sample. Thecurrent to the heating wires was varied, and once the temperature had equilibrated,it was measured with the thermocouple and the pyrometer.Figure 6 shows the sample temperature as measured by the thermocoupleversus the temperature reading of the optical pyrometer. The limits on the range ofcalibrated temperatures resulted from two effects. Firstly, the pyrometer was onlysensitive to temperatures above approximately 350°C. Secondly, a sampletemperature of 600°C is near the upper limit of the type-K thermocouple sensitivityrange. The solid line in Fig. 6 shows the variation of the sample temperature withthe pyrometer reading. The sample temperature was determined self-consistentlyfrom the pyrometer reading by calculating the sample emissivity from Planck’sradiation law and by assuming the desorption temperatures for the Ga and As oxides19(a) VACUUM SYSTEM TYPE-KTHERMOCOUPLE(b) COPPER SAMPLE HOLDERFigure 5, Experimental apparatus for the thermocouple/pyrometer temperaturemeasurement. Also shown is the sample mounting technique used in theexperiments described in this thesis.STAINLESSROO4UHV WINDOWRESISTIVEHEATERTa HEATING WIRESTHERMOCOUPLEJUNCTIONPYROMEtER MLORSFOT BLOcK4HEATERWIRES20700600C)0wD500waUi400300400 1000Figure 6. Sample temperature as measured by the type-K thermocouple (points)compared with the temperature measured by the optical pyrometer (solid line). Thepyrometer was calibrated with the As and Ga oxide desorption temperatures.500 600 700 800 900PYROMETER READING (°C)21were 580°C and 450°C, respectively.35’6 The agreement between the sampletemperatures determined by the thermocouple and by the desorption of the oxideswere within 10°C over the temperature range of 450-600°C and within 20°C fortemperatures down to 350°C.2.3 PHOTOELECTRON SPECTROSCOPYThe photoelectron experiments were conducted on the Ui beamline at theNSLS, Brookhaven. The Ui beamline was designed to cover an energy range from 25to 1300 eV using an extended range grasshopper (ERG) monochromator,37 and forthis energy range the line is maintained under UHV conditions to avoid carbonbuild-up on the optical elements and for compatibility with the electron storagering. The optical layout of the beamline is shown in Fig. 7. Light from the storagering is horizontally collected by a bent flat mirror MO with its focus located at theexit slit S2 of the monochromator. The mirror MO collects an angle of 10 mrad. Aset of premonochromator slits S,, and Sh aperture the beam vertically andhorizontally in order to reduce stray light in the forward monochromator section.The ERG monochromator contains a bent elliptical mirror Ml, a flat mirror M2, anentrance slit Si, three interchangeable spherical gratings Gi, G2, G3 and an exit slitS2. The combination of the flat mirror and entrance slit are optically equivalent toa bilateral slit with a range from 10 to 300 J.tm while the exit slit is a bilateral devicewith a range from 10 to 1000 urn. The resolution of the monochromator is afunction of the aperture of these slits. The individual optical components arecomputer controlled and once the appropriate parameters are determined throughsoftware, the entire ERG assembly (Ml, M2, S2, Gi, G2, G3), located on an air bearing,translates to the correct position.The samples were introduced into the spectrometer through a sampletransfer line and transported through a small deposition chamber to the main22M3-M 2— — ——• ISAMPLE t’z:. ._POSITIONS2 i,X1 SiI MO SDI..RDESIDE VIEW[ = R• — — —...--MOM2ShG SiSAMPLEPOSITIONTOP VIEWFigure 7. Optical configuration of the Ui beamline (after Sansone et al., ref. 37).23analysis chamber as schematically represented in Fig. 8. New samples wereintroduced into the system through the load lock, and stored in a five-shelf sampleholder. Once the sample was on the manipulator arm in the analysis chamber, itwas rotated into position for: (a) cleaning with the sputtering gun, (b) oxidedesorption by annealing and monitoring with the mass spectrometer, (c) absorptionmeasurements with the home-built microchannel plate (MCP) detector, and (d)photoemission measurements with the hemispherical electron energy analyzer(100 mm mean radius, VSW-HAC 100). The combined resolution of themonochromator and electron energy analyzer for photoemission varied dependingon the specific experimental setup. The resolution was estimated from theexperimental width of the Fermi edge in copper.The three types of photoelectron experiments that were performed areschematically represented in Fig. 9. The first is photoemission spectroscopy (PES),where a core hole is created by an incident photon and a photoelectron is emittedfrom the sample and energy analyzed. The energy of the incident radiation is fixedand the electron energy analyzer is swept through the various electron kineticenergies to obtain a spectrum. The second is x-ray absorption spectroscopy (XAS),where the energy of the incident radiation is varied over an energy rangecorresponding to a core-to-empty-state transition as shown in Fig. 9(b). A MCPdetector collects the electrons from the resulting nonradiative decay processeswithout regard to their energy and thus measures the absorption. Deexcitationelectron spectroscopy (DES) is the same as normal photoemission however thephoton energy is specifically tuned to resonantly excite a core-to-bound-statetransition. The photoelectrons are then energy analyzed as in photoemission. Itwill be important to distinguish between two of the decay processes as shown in thebottom of Fig. 9. The participator process has a single core hole final state, the sameas normal photoemission. The. spectator process has a 2 core hole, one bound24DEPOSITION ANALYSISLEAKEVAPORATOR VALVEOPTICALPYROMETERFigure 8. Top view of the analysis chamber, deposition chamber and sampleLOADTRANSFER ARM LOCKSPLJTERINGSAMPLESTOREANALYZERVALVESTHICKNESSMONITORMCPDETECTORMASSSPEC.BEAMLINE TOSYNCHROTRONtransfer line at the Ui beamline, NSLS.25PI-croNCORE HOLEW1y d’,’J(,W.1.Figure 9. The three types of photoelectron spectroscopy experiments performed: (a)photoemission and (b) absorption and (c) deexcitation of a bound state.(a) PHOTOEMISSIONVBe ANALYZERVAC. LEVEL(b) ABSORPTIONPHOTON(c) DEEXCITATION/-.-SPECTATORN4* .PARTICIPATOR26electron final state and is similar in energy to the normal two core hole final stateof the normal Auger transition.2.3.1 Photoemission SpectroscopyPhotoelectron spectroscopy is growing in popularity as a general analyticaltool, specifically in the area of surface analysis. It allows the investigation ofelectronic structure, providing a picture of molecular orbitals for gas-phasespecies, valence-band density of states, and core-level electron binding energies forsolids. The characteristic electron energies allow elemental analysis as well aschemical state identification. In addition, photoelectron spectroscopy probes onlythe surface region of solids.If a solid sample is irradiated with monochromatic photons of frequency v,the photons may be absorbed resulting in the emission of electrons with kineticenergy EK defined byhv=EB+EK+ (1)where EB is the ionization energy or binding energy of the electron in the materialand 1 is the spectrometer work function. The binding energy of any core level isdefined to be the energy separation between that core level and the Fermi level ofthe sample. For samples able to exchange electrons with an electron spectrometer,the Fermi levels of the sample and spectrometer will coincide. As a result, thespectrometer work function can be determined by obtaining the photoemissionspectrum of a material with known binding energies. The instrument is adjusteduntil the binding energies agree with the known values.2.3.1.1 Chemical ShiftsThe energy levels of the core electrons are characteristic for a givenelement. These core levels can therefore be utilized for an elemental analysis of the27material being studied. However, the exact value of the binding energy measureddepends on the chemical environment of that element, due to the so-called chemicalshifts. A complete understanding of the chemical shifts requires solutions of thetotal wave functions for the neutral species and the singly ionized counterpart,corresponding to a vacancy in a given inner shell orbital. That is, the chemicalshift iE1 for an orbital ni in a free atom A relative to the same orbital in aparticular molecule or compound AB is=E1(AB)- AE1(A) = [Tj(AB)- T(AB)] - [Tj(A) - T(A)J (2)where T and T are the total energies, respectively, for the neutral species and forthe ionized species having a single vacancy in the ni orbital. Hartree-Fockcalculations have been limited to simple molecules and the use of a modifiedHartree-Fock solution for solids has been limited. Fortunately, a much simplermodel can be used to explain and predict chemical shifts.The simple model is based on the idea that the core electrons are affected bythe potential of the valence shell, which changes depending on the specificchemical environment of the atom. If one envisions the core electrons as beinginside a hollow electrostatic sphere, then the potential seen by each of the coreelectrons due to the valence shell will be qe/r, where qe and r are the chargeand radius of the valence shell. The chemical shift would then beqe2 qe2 qye2AE1= r(AB)-(A) = A( r (3)V VIn the last part of equation (3), the radius of the valence shell, rv is assumed to beequal for the atom and compound and is approximated by the covalent radius, r.Before approaching the problem of calculating the valence shell potential inorder to evaluate the observed core-binding energy shifts, it is necessary to takeinto account the total electrostatic environment, which includes not only thevalence shell but all the other atoms in the solid. One may regard each atom, other28than the atom under study, as having a net charge qe, which will act as a pointcharge located at the center of the atom. The core electron in atom i will thus seethe effective potential of all the other electrons asV= E (qe/R1) (4)where is the distance between atoms i and j. The binding energy shift nowbecomes= tX(qe2/r) - (5)where AV = eV(AB). Since the change in the charge of the valence shell isessentially equivalent to the change in the net charge, calculations on the netcharge can be applied equally to q,, and qj• In a solid or crystal the calculation of Vmust be carried out in such a way that the sum will converge after taking a finitenumber of terms. Such calculations are similar to Madelung potential calculations.In correlating experimental binding energy shifts with theory it has beencommon practice to ignore AV. However, this is unjustified since V is usually ofthe same order as Aqe2/r and in some instances is larger. The only reason suchneglect still yields a reasonable correlation is that V also often changes linearlywith q,,, particularly if one compares compounds of similar structure like the XYzincblende structure where equation (5) can be written asAq 9x. 2 i 1 2R )e =qx( - R)e (6)where r is the covalent radius of atom X, R is the distance between atoms X and Y,and t q x and z q y are the net charge of atoms X and Y. In these cases equation (6)has been successful in correlating a multitude of shifts in small molecules as well asin a number of solids with charges tq, derived from the ionicity of bonds accordingto Pauling’s electronegativity values. This will be demonstrated in our analysis ofthe ZnSeIGaAs interface.29This chemical shift model fails in many instances to describe the correlationbetween Aq and AE. It fails to account for the energies involved in the drasticelectronic rearrangement of the system upon ionization of one atom. Thisrearrangement involves a flow of negative charge towards the hole created in thephotoemission process in order to screen the suddenly appearing positive charge.The screening lowers the energy of the hole state left behind and therefore lowersthe measured binding energy as well. This binding energy defect is commonlyreferred to as the relaxation energy ER. The magnitude of ER is expected to differ forthe same atom in different systems and therefore a term AER has to be added toequation (5) in order to describe the chemical shifts correctly.2.3.1.2 Film Thickness DeterminationOne of the most important features of PES is the ability to quantitativelydetermine the composition of surfaces. The number of emitted electrons of a givenkinetic energy is a function of the number of atoms of a given type on the surface,however, the measured signal depends on many factors. Equation (7) describes thesignal strength I observed in PES due to electrons originating from atoms of type iwith energy ej that have not been scattered inelastically:Ii =01la(E)D(e (7)where I = x-ray photon fluxi = density of atoms of type ia1 = photoexcitation probability(j) = escape depth of an excited electron with energy ejD(e) = the fraction of electrons detected by the analyzer.All the parameters must be accurately known to obtain good quantitative results.This however has not been accomplished because of difficulties involved inevaluating the parameters. Nevertheless, measured intensity ratios can be used to30derive the required quantities. The determination of the thickness of the depositedfilms is of great importance in this work. It is therefore necessary to relate thephotoemission intensity ratio, R = ‘f’Ts where I and I represent thephotoemission intensities of a core level from the overlayer film (f) and substrate(s) respectively, to the thickness d of the overlayer. Assuming that the contributionto the photoemission intensity from buried layers falls off exponentially with depthbelow the sample surface with characteristic length ), the ratio R is given byri -j (8)where K is a constant expressing the relative sensitivity of the spectral features ofthe overlayer film to the substrate. In arriving at equation (8), two approximationsare made concerning the electron escape depth ).. It is assumed that it remainsconstant over the kinetic energy range of the photoelectrons being considered,which is typically 50-70 eV but may extend to 40-100 eV, and that it is equal for theoverlayer film and substrate. An analytic expression for K can be obtained fromequation (7) by considering the photoemission intensity ratio of two core levels,one from a bare substrate and one from an infinitely thick overlayer film (i.e. d>>2.).K—If/F0— 11faff(ef)D(ef) 9—15/F05— 1o5A8(e)D(e (F0f and F05 are the respective incident photon fluxes on the bare substrate and theoverlayer film. The parameters 11, a, and are either known or can be calculated.The analyzer-detection efficiency, D(e) is usually not known38 and thus K isdifficult to calculate. However, K can be experimentally determined by the ratio ofIf/F0 divided by IIF05. The values of K, together with the measured peak intensityratio R and the assumed value of can then be used to determine the overlayer filmthickness d from equation (8).31For CaF2 overlayers on GaAs, the photoemission intensity ratio R wasdetermined from the intensity of either the Ca 3p or F 2p levels to the intensity ofthe Ga 3d level. Four measurements on clean GaAs substrates and two measurementson thick (>50 A) CaF2 films were averaged to obtain K. The resulting value for K was1.2 for the Ca 3p/Ga 3d photoemission intensity ratio. The value of the analogous Kfor the F 2pIGa 3d intensity ratio was found to be 4.7. The experimentaluncertainty in the determination of K, contributes a 20% uncertainty to the filmthickness. These values of K, together with the measured peak intensity ratio R andthe assumed value for 2 were then used to determine the CaF2 film thickness d fromequation (8).As discussed earlier, the electron escape depth is required before thethickness can be determined. The electron escape depth (ED) is related to two otherquantities namely the electron inelastic mean free path (IMFP) and the electronattenuation length (AL). These three terms are often used interchangeably buteach has a separate meaning.39 The IMFP is defined as the average distance that anelectron with a given energy travels between successive inelastic collisions, andcan be obtained from theory and certain types of experiments. The AL is defined inthe same way as the IMFP however it is derived from a particular model in whichelastic scattering is assumed to be insignificant. With this model, electrons areassumed to be scattered only inelastically and predominantly in the forwarddirection. For example, the AL can be obtained from overlayer-film experimentswhere photoelectrons reaching the surface are assumed to have travelled alongstraight-line paths from their point of generation. The ED is defined as the distancenormal to the surface at which the probability of an electron escaping withoutsignificant energy loss due to inelastic scattering processes drops to l/e of itsoriginal value. The ED is the product of the AL and the cosine of the angle definedby the analyzer direction and the surface normal.32The electron inelastic mean free paths (IMFP) for three alkali halides (LiF,KC1 and NaC1) have recently been determined40 by an algorithm developed byPenn41 where experimental optical data were used to give information on theinelastic scattering probability as a function of energy loss and theory was used todescribe the dependence of the scattering probability on momentum transfer. Inthe kinetic energy range of 50-100 eV the IMFP reaches a minimum, and variesbetween 6-8 A for the three alkali halides. The IMFP is systematically larger thanthe escape depth by about l530%42 because the net electron path is longer than theaverage distance between inelastic collisions due to elastic scattering. Thealkaline-earth fluorides (CaF2, SrF2, BaF2) are assumed to have a similar escapedepth to the above mentioned alkali halides, namely = 6 A, and the energydependence of ? in the 50-70 eV range of these experiments has been neglected.It is important to note that other groups have used alternative methods fordetermining the escape depth. For example, Himpsel et at.21 determined 2. = 5 Aby looking at a monolayer of Cl on the Si (111) 1 x 1 surface. The escape depthswere determined from the relative ratio of the photoemission peaks correspondingto bulk Si and to the surface layer of Si that bonds to Cl. Olmstead et at.22 used= 10-12 A determined by Battye et al.43 for the alkali fluorides using asemi-empirical theory based on a tight-binding model. Yamada et at.26 determinedtheir escape depths by extrapolating to lower kinetic energy the earlier work ofPenn44 in the 200-2000 eV range. The recent work40’2 was found most suitable indetermining the escape depths for the alkaline-earth fluorides in the 50-70 eVkinetic energy range of these experiments. According to these authors thetheoretical calculations are now more accurate than the experimental values.For the rare-earth trifluorides deposited on Si, the thickness of the annealedfilms was estimated from the intensity of the Si 2p photoemission peak relative tothe La 4d, Nd 5p or Tm 5p photoemission from the overlayer, however, there was33no evidence that supported layer by layer growth of the films. In this analysis itwas assumed that the escape depth was also 6 A in agreement with the otherfluorides. The layer thicknesses determined in this way for these films aresomewhat more uncertain but are still thought to be accurate to better than a factorof two. If the rare-earth trifluorides do not grow in a layer by layer way, thisprocedure underestimates the overlayer thickness.For ZnSe deposited on GaAs, the same method for determining the thickness ofthe films was used. The electron inelastic mean free path (IMFP) for ZnS has alsobeen calculated recently42 for the kinetic energy range of 50-100 eV and found tovary between 5-7 A. In the thickness determination we assume that ZnSe has thesame escape depth as ZuS, namely ? = 5 A, and the energy dependence of in the45-90 eV range of these experiments is neglected. The assumption on ? is critical indetermining the film thickness as it is directly proportional to the assumed valuefor .2.3.1.3 Deexcitation Electron SpectroscopyAs mentioned earlier, deexcitation electron spectroscopy (DES) is the same asnormal photoemission however the photon energy is specifically tuned toresonantly excite a core-to-bound-state transition. A deexcitation spectrum cancontain information that aids in the interpretation of both normal photoemissionspectra and x-ray absorption spectra in the vicinity of absorption edges. Thenonradiative decay processes that result from the core excited states (or coreexcitons) are conveniently studied with this technique. As shown earlier inFig. 9(c), one such decay process, the spectator process has a two core hole, onebound electron final state, which is similar in energy to the two core hole final stateof normal Auger decay. As a result, a spectator decay process in a deexcitationspectrum appears as additional spectral features in the vicinity of the normal Auger34lines and is also known as resonant Auger emission. The existence of thesespectator lines indicate the presence of bound electron states in the correspondingx-ray absorption spectra. This technique has widespread applicability in thespectroscopy of atoms and molecules in the gas phase,45 but only recently has itbeen applied to solids and surfaces.7Another decay process known as the participator process, is also shown inFig. 9(c) The participator process, results in a single core-hole final state, the sameas normal photoemission. In the deexcitation spectrum, this process appears as anenhancement in intensity of the normal photoemission line, thus this process isalso known as resonant photoemission. The existence of these participator lines indeexcitation spectra allows one to determine the atomic location of the boundelectron states.2,3.2 X-ray Absorption SpectroscopyThe importance of x-ray absorption spectroscopy as a spectroscopic tool willincrease as the resolution and intensity of the tuneable sources that are available inthe vacuum ultra-violet and x-ray region improves.46 The limitations in the energyrange and intensity of standard x-ray tubes, used to obtain the first absorptionspectra, were overcome by the use of synchrotron radiation sources for soft x-raystudies in the 1960s and early 1970s. In the past few years experimental progress inthe field of soft-x-ray absorption has improved the attainable resolution to itspresent best value of 30 meV at 300 eV.47The structure observed above x-ray absorption edges is attributed to twodifferent origins. The near-edge structure (XANES) within the immediate vicinityof the edge is interpreted in terms of the density of states of empty valence statesavailable for the excited electron, while the extended absorption fine structure(EXAFS) arises from the interference of the electron, represented now by a plane35wave, with the waves backscattered from the neighboring atoms. In the edgeregion of insulators bound excited states appear below the continuum thresholdover a range of 5 to 10 eV. These bound excited states will be studied for thealkaline-earth fluorides.2.3.2.1 Total Electron Yield MeasurementOur method for measuring the absorption of these solid thin films is by a totalelectron yield method. In total electron yield measurements, all electrons emittedfrom the sample are collected without regard to energy. The total yield from asample is composed of Auger electrons and photoelectrons, and inelasticallyscattered (or secondary) electrons. Only a small fraction of the photoelectrons andAuger electrons escape into vacuum without suffering inelastic scattering. As aresult the total yield is dominated by inelastically scattered electrons.Experimentally, total yield measurements are especially simple and consist ofcollecting the electrons from the sample by an electron multiplier or microchannelplate detector. The detector is often biased depending on the specific experimentalconditions, either positively or negatively up to 200 V, in order to obtain a bettersignal to noise ratio.2.3.2.2 Microchannel Plate DetectorsIn our measurements of the absorption by total electron yield amicrochannel plate (MCP) detector was used to count the emitted electrons.Microchannel plates are photoelectric devices consisting of a parallel array ofmicroscopic channel electron multipliers capable of ion, electron, ultra-violetphoton, and soft x-ray detection and signal amplification. A MCP begins as aspecially formulated lead glass tube and solid core assembly that is drawn and fusedto form a monofiber. A bundle of monofibers, stacked in a hexagonal array, is36drawn to form a multifiber. Multifibers are then stacked and fused to form a billetor boule. The fused billet is sliced into wafers at a specified bias angle, and these areground and polished to an optical finish. The individual wafers are chemicallyprocessed to remove the solid core, leaving a “honeycomb” structure of millions oftiny holes. Each hole functions as a single channel electron multiplier, relativelyindependent of adjacent channels.For normal operation, a bias voltage of up to 1000 V is applied across the MCP(output positive with respect to input). The bias current flowing through theresistive layer supplies electrons necessary to continue the secondary emissionprocess. When a photon or charged particle is incident at the input of a channel,secondary electrons are generated and accelerated down the channel toward theoutput end. When secondaries strike the channel wall, en route, additionalelectrons are generated, This process is continuously repeated until an output pulseof up to 10 electrons is generated for a straight MCP. If two or more MCPs areoperated in series, a single input event will generate a pulse of or moreelectrons. Output signals are typically collected by a metal anode. Appendix Ishows a schematic diagram of the MCP detector assembly used and the MCP biasingcircuit.373. RESULTS AND DISCUSSION3.1 CHALCOGENIDES3.1.1 H2S Treatment of GaAs (100)As discussed in section 1.1.1, various new wet chemical treatments involvingsulfur compounds have been reported to produce passivated surfaces of GaAs, that issurfaces with a decreased carrier recombination velocity and increasedluminescence efficiency compared to earlier passivation techniques.5’6 Howeverthe composition of these surfaces was not revealed by these studies. Photoemissionexperiments on GaAs (100) wafers passivated using Na2S and (NH4)2S chemicaltreatments have been performed by our group and others. In these photoemissionstudies7’8 it was found that even a small amount of sulfur on the the surface waseffective in improving the electronic quality of the surface. An alternativepassivation technique to these wet chemical treatments, is a treatment with gaseousH2S, and the results of these experiments will be reported here.In section 2.1.1, the cleaning of the GaAs surfaces was described, and the H2Streatment of the clean surfaces was described in section 2.2.1. Photoemissionspectra were taken prior to and immediately after the sulfur treatment, and alsoafter annealing the treated sample, in order to determine the nature of the chemicalbonding at the interface. Spectra were also taken for the treated sample followingexposure to air and water in order to determine the stability of the S-passivatedsurface.Figure 10 shows a series of photoemission spectra of GaAs (100) excited withphotons of energy hv 330 eV following the various surface treatment steps. Thebottom spectrum (a) is for the clean GaAs (100) wafer. All the features can beidentified as photoemission peaks of the accessible GaAs core levels or as Augerpeaks, as labeled in the figure. The photoemission spectrum (b) in Fig. 10 is for the38I I I I IwsAs As3s 3s3p 3p3°KINETIC ENERGY (eV)Figure 10. Photoemission spectra of GaAs (100) following various successivesurface treatments measured with a photon energy of approximately 330 eV.Spectrum (a) is for the clean surface, (b) is for the clean surface after exposure to“activated” H2S at 9 x 10 Torr for 15 mis, (c) is for the exposed surface annealedat 400°C for 3 mis, (d) is for the sample after it was removed from the UHV systemand rinsed with deionized water, and (e) is after the sample was annealed againunder UHV for 2 mm at 500°C.39GaAs sample after the H2S treatment. As a result of the sulfur treatment, the S 2pand 2s photoemission peaks and the sulfur LMM Auger peak can be identified. Inaddition to these spectral features associated with sulfur, some C is photoemissionwas also observed as indicated in the figure, and is thought to be the result ofimpurities in the deposition chamber. The surface treatment also resulted in a lossof As, which is evident from the relative change in photoemission intensitybetween the Ga and As spectral features in this spectrum.Heating this surface to 400°C for 3 mm results in spectrum (c). The carbonresidue that was deposited during the H2S treatment has been largely removed, andaccordingly all the Ga, As and S spectral features appear stronger. This sequence oftreatments yields the S passivated surface, but before examining the nature of thechemical bonding at the surface, if any, one can show that the sulfur is not lostwhen exposed to air or water. To show this, the sample was removed from the UHVsystem into air, then rinsed in deionized water and finally blown dry with flowingnitrogen. After reintroduction into the UHV system, the photoemission spectrumwas measured and is shown in Fig. 10(d). The largest change observed is again thelarge increase in the C is photoemission peak. The carbon containing speciesappears to be only loosely bound to the surface, because heating the sample to 400°Cfor 2 mm results in the removal of most of this carbon deposit as illustrated inspectrum (e) of Fig. 10. Furthermore, a carbon desorption peak was observed withthe mass spectrometer during the annealing process. Spectra (c) and (e) in Fig. 10are almost identical, the only difference being a larger amount of carbon waspresent on the surface in spectrum (e). As discussed, this carbon species waslargely physisorbed to the surface, demonstrating that the sulfur passivated surfaceis stable to the loss of sulfur, as the surface experienced no major chemical changesupon exposure to air or water.40In order to characterize the nature of the chemical bonding of the H2 Streated surface , high resolution photoemission spectra of the As 3d and Ga 3d corelevels were obtained for the various treatment steps. These spectra are shown inFigs. 11 and 12 and were measured with a photon energy of 84.3 eV. Figures 11(a)and 12(a) show the As 3d and Ga 3d core levels of the clean GaAs (100) surface.These spectra are identical to other previously published results for the (100)surface of GaAs.48After the H2S exposure as described in section 2.2.1, the intensities of theAs 3d and Ga 3d photoemission peaks were reduced, due to the coverage of thesurface with sulfur and the carbon contaminants. In addition, the relative intensityof the Ga 3d peak to that of the As 3d peak indicates that As has been lost relative toGa at the surface possibly due to an exchange reaction between S and As atomsduring the H2S exposure. These spectra are shown in Figs. 11(b) and 12(b). TheAs 3d core level also exhibits a component chemically shifted by 1.5-2 eV to largerbinding energies measured relative to the center of the spin-orbit split bulkphotoemission peak. This chemical shift is characteristic of an AsS species, asshown previously.8 An As oxide species would exhibit a chemical shift ofapproximately 3 eV. The chemical shift observed for the Ga 3d core level was onlyapproximately 0.6 eV, which is again typical for a sulfide species.8 Oxygen atomsreacting with Ga shift the 3d level by approximately 1.3 eV to larger bindingenergies. The chemical shifts observed on both the As 3d and Ga 3d core levelsindicate the As-S bonding and Ga-S bonding are present at the interface afterexposure to H2S. In addition to the chemically shifted core-level species, a shift ofthe bulk components was also observed indicating that the surface Fermi level wasshifted upwards in the band gap by approximately 0.15 eV following this treatmentstep. This may indicate the removal of some of the surface states through substratebonding to sulfur.41I ‘ ‘ I I I I I I•I I I I44(C) ::4 44 4-,w4 +•4 +• 4VA.bI. : 4.4%b4-(b) lb4b%.*444 •40 +•+ 4 +— +b- —>- 4U,4.a) r-444C—I i I I I I I I I I I I I(a)/.44 —4 44..4.I i i i I i , i i I , i i I32.5 35. 37.5 40KINETIC ENERGY (EV)Figure 11. Photoemission spectra of the As 3d core level of GaAs (100) for the(a) clean, (b) H2S exposed and (c) annealed surface. Details of these preparationsare identical to the ones given for Fig. 10(a), 10(b) and 10(c). The photon energywas 84.3 eV. The arrow in spectrum (b) represents the 1.5-2 eV chemical shiftcharacteristic of an AsS species.42J I I T(c)• / 4I• 4• +* 44 44 +•4•1 4-• +•4•+I I I I I I(b)• $C — 4+.0 • +I_— 4> 4 40) /+C 4-I j(a) A-14I• +* :• 44• 444• 4• +—•1 +I57.5 60. 62.5KINETIC ENERGY (EV)Figure 12. Photoemission spectra of the Ga 3d core level of GaAs (100) for the(a) clean, (b) H2S exposed and (c) annealed surface. Details of these preparationsare identical to the ones given for Fig. 10(a), (b) and (c). The photon energy was84.3 eV.43Next the sample was heated to 400°C for 3 mm, and the photoemission spectrafollowing this treatment are shown in Figs. 11(c) and 12(c). Again the bulkphotoemission peaks move slightly to higher binding energy, which indicates ashift of the surface Fermi level. The Fermi level has now shifted by 0.5 eVcompared to the clean surface, where it was measured at approximately 0.35 eVabove the valence-band maximum (VBM). This puts the Fermi level of thepassivated surface at 0.85 eV above the VBM, only slightly above midgap, which canbe interpreted as a fairly well passivated surface because one would expect thesurface states of the clean surface to be located below midgap and thus pinning theFermi level in this region. However, since the Fermi level of the S treated surface isabove midgap one can interpret this as resulting from the removal of the surfacestates through chemical bonding with sulfur at the surface.Upon this annealing step, the As 3d core level now exhibits only a bulkcomponent, as shown in spectrum (c) of Fig. 11. The chemically shifted componentobserved prior to annealing that was attributed to As-S bonds is no longer evident.In addition the As 3d core level is sharper than the one observed for the cleanGaAs (100) surface. The existence of a surface core level shifted componentbroadens the photoemission signal of the clean surface, as reported earlier.48 Thusone can conclude that the H2S treated annealed GaAs (100) surface has no orextremely few As atoms present at the surface. In contrast to this observed changein the As 3d core level, the photoemission peak from the Ga 3d core level showsvery little change upon the annealing of the H2S treated surface. The exceptionbeing the global shift due to the change in the Fermi level position and a generalincrease in intensity due to the removal of the carbon contaminants. Therefore,one can conclude that the H2S treated GaAs surface is dominated by Ga-S bonds andthat very little As is present at the interface.44The interaction between H2S and GaAs surfaces with different orientation hasbeen studied previously.102 J. Massies and co-workers’°’ observed that with thesubstrate held at 700 K an exchange reaction occurs between As and S atoms at theGaAs (100) surface upon exposure to (unactivated) H2S. This exchange reaction wasobserved for single monolayer coverages. The surface structure was observed intheir experiments to change from a centered (2 x 8) reconstructed clean surface8 1to a well defined (2 x 1) reconstructed S saturated surface. In our study, a decreasein the As/Ga ratio at the interface of the H2 S treated sample indicates a similarexchange reaction may occur for sulfur layers that are several monolayers thick.Massies et al. also observe that a room temperature H2S adsorption phase can betransformed into the stable (2 x 1) phase upon subsequent annealing. This agreeswith our results in that upon annealing of our H2S treated sample, the surface didnot lose any sulfur. Ranke et al. 12 report weak interactions between H2S adsorbedat 150 K and various surface orientations of GaAs, but they also report a 0.8 eV shiftof the Ga 3d level for high temperature (700 K) and high exposures of the (111)surface, which they attribute to the formation of a GaS2 species. Our results showthat there is Ga-S bonding at the interface in agreement with Ranke et al.’ sfindings.From the high-resolution photoemission studies of the clean GaAs (100)surfaces treated with activated H2S at room temperature, followed by annealing at400°C, we conclude that the surface passivates through removal of the As atoms andformation of a GaS chemical species. The Fermi level at the passivated surface islocated slightly above midgap at 0.85 eV above the valence-band maximum. Whileno comparative data has yet been obtained to evaluate the carrier recombinationrate at this passivated surface, the photoemission results indicate that the H2Streated GaAs (100) surface is better characterized in terms of the interfacialbonding than the ones produced by Na2S and (NH4)2S chemical treatments.5’68453.1.2 ZnSe on GaAs (100)3.1.2.1 Interfacial BondingHere the results of a photoemission study of the ZnSe/GaAs (100) interfaceare described. A thin ZnSe film was deposited on GaAs and then annealed insequential steps in order to determine the properties of the interface as a functionof annealing. The first direct measurements of the interfacial composition wereobtained and the thickness of the interfacial layer as a function of annealingtemperature was determined.Figure 13 shows photoemission spectra measured at hv = 105.2 eV for aseries of anneals increasing in temperature for approximately 14 A of ZnSe onGaAs (100), showing the Se 3d, As 3d, Ga 3d and Zn 3d photoemission lines. Thethickness of the deposited films was estimated from the ratio of the integratedintensities of the photoemission peaks from the ZnSe overlayer to those of the GaAssubstrate as discussed in section 2.3.1.2. The ratios of the Ga 3d to As 3d and theSe 3d to Zn 3d photoemission peaks were observed to increase with increasingannealing temperature. This is attributed to the formation of an interfacial galliumselenide layer, where Se atoms substitute for As atoms and Zn and As atomsevaporate.Selenium forms two stable compounds with gallium which arethermodynamically more stable than GaAs, namely GaSe and Ga2Se3.49 In order toquantitatively determine the deposited film thickness and composition as a functionof annealing temperature, a model is used where an interfacial gallium selenidelayer forms between the ZnSe overlayer and the GaAs substrate, with an escapedepth equal to that of ZnSe. This model assumes that the overlayer, interfacial layerand substrate are all laterally homogeneous. The thicknesses were obtained bydetermining the best fit for the ratios of the photoemission peak intensity for eachof the individual atomic species to the total intensity as a function of annealing46U)0C)>-I—Cl)zLUIz40 100Figure 13. Photoemission spectra for a sample consisting of 14 A of ZnSe onGaAs (100) for a series of annealing temperatures from 410°C to 590°C measuredwith a photon energy of 105.2 eV. The change in the character of the noise invarious sections is due to changes in the scan rate during acquisition of the spectra.50 60 70 80 90KINETIC ENERGY (eV)47temperature. The total intensity is given by the sum of the photoemission peakintensities of the Zn 3d, Ga 3d, As 3d and Se 3d core levels. The two fittingparameters were the overlayer thickness and the interfacial layer thickness. Thesolid lines in Fig. 14 represent the best fit to the relative peak area for each of the3d core levels for the sample in Fig. 13.The relative cross sections, a of the Zn 3d, Ga 3d, Se 3d and As 3dphotoemission peaks were determined from the absolute peak intensities for cleanGaAs and a thick layer of ZnSe. Two GaAs measurements and two ZuSemeasurements, taken on different samples, were averaged to obtain these relativecross sections. The resulting values, normalized relative to 0As = 1.0, are az 1.21,0Ga = 1.48 and 0Se = 1.07. The corresponding values, determined from the atomicsubshell photoionization cross sections calculated by Yeh and Lindau,50 are0As = = 1.14, GGa = 1.17 and °Se = 0.79. The small discrepancy between theexperimental and theoretical values for 0Ga and GSe can most likely be attributed tothe surfaces of the respective samples being rich in these two species afterannealing. Since the analogous surface and interface enrichment is expected to bepresent for the ZnSe/GaAs samples, the experimentally determined values for a areused in the model.The best fit to the data was found to result from an interfacial galliumselenide layer consisting of only the Ga2Se3 component, represented in Fig. 14 bythe solid lines. The dotted lines show the fit for a GaSe interfacial layer. Clearly theGa2Se3 interfacial layer fits the data in Fig. 14 better. These results are consistentwith those of Li et al. 1 6 who inferred, from transmission electron microscopy(TEM) studies of ZnSe grown by MBE on GaAs (100), that the structure of theinterfacial layer is consistent with the formation of Ga2Se3. Ga2Se3 has azincblende structure like GaAs and ZnSe, however one third of the cation sites arevacant. Figure 15 shows the thicknesses of the surface ZuSe layer, the interfacial480.60.50.4o 0.3LU>I.LU0.10350 600Figure 14. Peak intensity for each of the species (Se, Zn, Ga and As) relative to thetotal intensity of all four species as a function of annealing temperature. The solidlines represent a fit to the data corresponding to an interfacial Ga2Se3 layer and thedotted lines correspond to a GaSe interfacial layer. In the fits, the total overlayerthickness was fixed and the thickness of the interfacial layer as a fraction of thetotal was adjusted to match the temperature dependence of the data. The relativecross sections of the 3d photoemission peaks were determined experimentally, asdiscussed in the text.400 450 500 550TEMPERATURE (°C)4915 iSRg11100io 000< 0C’) 0 • ZnSe• 0•o 0 23o00 Total Overlayer......••.0 II iii iliii350 400 450 500 550 600TEMPERATURE (°C)Figure 15. Average thicknesses of the surface ZnSe layer, the interfacial Ga2Se3layer and the total overlayer as a function of annealing temperature for the sampleof Fig. 13 as determined from the fits in Fig. 14.50Ga2Se3 layer and the total overlayer as a function of annealing temperaturedetermined from the best fit model of Fig. 14. The ZnSe thickness decreases withannealing temperature until at 590°C no ZuSe remains, while, the interfacial Ga2Se3layer grows in thickness until 500°C, where it has approximately the same thicknessas the originally deposited ZnSe. This result was the first direct determination of thecomposition of the interfacial layer and its formation as a function oftemperature.51 These results compare with the results of Tu and Kahn,17 whosuggested that an interfacial Ga2Se3 layer formed, based on their Auger electronspectroscopy measurements which showed that after extensive post-depositionannealing a surface ZnSe layer completely converts to Ga2Se3.The chemical shifts in the Se and Ga 3d core levels also support the aboveinterpretation. Figure 16 shows these core levels for the series of annealingtemperatures for the sample of Fig. 13. The Se 3d level exhibits a chemical shifttowards higher binding energy of 0.52 ± 0.05 eV, taking into account a 0.1 eVincrease in the Fermi level. The expected chemical shift has been calculated forthis core level from Pauling’s electronegativity values using the method outlined byCarlson.52 Details of this calculation are presented in Appendix II. According tothis model, a chemical shift of the Se 3d core level due to simple cation exchange ofZn and Ga is expected to be 0.41 eV. Increasing the Se/Ga atomic ratio to 1.5 shouldincrease the chemical shift to 0.63 eV. The predicted Ga2Se3 chemical shift(0.63 eV) agrees reasonably well with the measured value (0.52 ± 0.05 eV). Apossible explanation for the shoulder on the high-kinetic-energy side of the Se 3dcore level at the higher annealing temperatures is surface Se-Se bonds.From the calculations of Appendix II, the predicted chemical shifts towardshigher binding energy of Ga in GaSe and Ga2Se3 relative to Ga in GaAs are 0.47 eVand 0.83 eV, respectively. Again taking into account the 0.1 eV Fermi level shift,the observed Ga 3d core-level shift in Fig. 16 is only 0.37 ± 0.05 eV. Figure 1551jFigure 16. Photoemission spectra of the Se 3d and Ga 3d core levels for the series ofannealing temperatures from Fig. 13, showing the chemical shifts to higherbinding energy as a result of the conversion of the ZnSe layer to Ga2Se3.44 45 46 47 48 79 80 81 82KINETIC ENERGY (eV)52shows that even at the lowest annealing temperature (determined by the lowtemperature limit of the optical pyrometer), approximately half of the ZnSeoverlayer has converted to Ga2Se3, and therefore the relatively broad Ga 3d corelevel at the lowest annealing temperature in Fig. 16 is already dominated by Gabonded to Se. Therefore, the shift of the Ga 3d core level at higher annealingtemperatures corresponds to an increase in Ga-Se bonding relative to As-Ga-Se andAs-Ga bonding as the interfacial Ga2Se3 layer gets thicker. The observed chemicalshift for the Ga 3d core level at the highest annealing temperature is approximately1.0 eV when compared to clean GaAs. In conclusion, the evidence for a Ga2Se3interfacial layer is further supported by an analysis of the chemical shifts of theSe 3d and Ga 3d core levels.3.1.2.2 Valence-Band OffsetsWhen two different semiconductors are placed together forming an abruptheterojunction, an important question is how the bands line up at the interface,Figure 17 shows the valence-band region taken at a photon energy of 46.8 eV fortwo samples with different thickness ZnSe overlayers deposited on GaAs (100)annealed at a low temperature (370°C). Note that the valence-band edge for both theZuSe and the underlying GaAs are visible in the sample with the 7 A overlayer. Theband offset can be determined by fitting the data with parabolic band edges for thedensity of states of the ZnSe film and the GaAs substrate. Assuming that thephotoemission intensity is proportional to the density of states and taking intoaccount the exponential decrease in electron escape probability with depth, theenergy dependence of the photoemission intensity is given by,1(E) = Dz(E)e + D(E)(1-e’) (10)where d is the thickness of the deposited ZnSe film, ? is the escape depth, and D(E)and DG(E) are the densities of states of ZnSe and GaAs respectively.53I>-200H(I,zwHz100Figure 17. Photoemission spectra taken atregion for 7 A (top) and 14 A (bottom) of ZnSeshows the measured band alignment andtheoretical fit shown as a solid line in the tophv = 46.8 eV for the valence-bandon GaAs, annealed to 370°C. The insetFermi level, determined from thespectrum.4003000-5 -4 -3 -2 -1 0 1ENERGY RELATIVE TO VBM (eV)2 354The solid line in Fig. 17 represents the sum of 1(E) and a linear backgroundfrom the secondary electrons due to second order radiation from themonochromator convoluted with a Gaussian function of width 0.29 eV, equal to ourexperimental resolution. The best fit to the data yielded a band offset of= 1.25 ± 0.07 eV, and is the first direct photoemission measurement of thevalence-band offset for the ZnSe/GaAs (100) interface.51 This is somewhat largerthan the 1.10 ± 0.03 eV18 determined for the ZnSe/GaAs (110) interface by x-rayphotoelectron spectroscopy of the 3d core levels and the valence-band region. Theinset of Fig. 17 shows the resulting band alignment and Fermi level position for thissample. The Fermi level is approximately 0.4 eV above the GaAs valence-bandmaximum and upon further annealing shifts up towards the conduction-bandminimum, presumably due to n-type doping of the GaAs substrate by Se.In a similar manner to the band offset determination for the ZnSe/GaAs (100)interface, the valence-band offset for the ZnSeIIuP (111) interface was found to be1.35 ± 0.15 eV. This value was obtained from the photoemission spectrum of thevalence-band region taken at a photon energy of 46.8 eV for a thin film (2-5 A) ofZnSe on InP annealed to 350°C as shown in Fig. 18.If the semiconductors are lattice matched as they are in ZnSe and GaAs forexample, it is possible to calculate or estimate the band offsets. Harrison’s LinearCombination of Atomic Orbitals (LCAO) theory53 is a simple yet fairly successfultheory for the prediction of band lineups in semiconductors. The LCAO descriptionof the electronic structure gives the atomic energy levels in terms of atomic termvalues which are measured from the vacuum level. The term value is the negativeof the energy required to remove the electron from the atom and place it at rest aninfinite distance away. Although the absolute values of the valence-band maximumenergy (E) calculated from the LCAO theory do not correspond numerically with55200C’,150D0C)>-HCl)zwi—100z50-5Figure 18. Photoemission spectra taken at hv = 46.8 eV for the valence-bandregion for a thin film (2-5 A) of ZnSe on InP, annealed to 350°C. The inset showsthe measured band alignment and Fermi level, determined from the theoretical fit,shown as a solid line.-4 -3 -2 -1 0 1 2 3ENERGY RELATIVE TO VBM (eV)• 56observed photothresholds, the relative values which represent the variation frommaterial to material, do correspond fairly well. Table 1(a) lists the magnitudes of E,and the band gaps for a few semiconducting materials. The predicted discontinuityin the valence-band maximum is obtained by simply subtracting the correspondingvalues of E. Table 1(b) summarizes the valence-band offsets (AE) determined inthis study and compares them with values determined from Harrison’s LCAOmodel.53 The values of zE do not agree well with those predicted by Harrison,suggesting that a more rigorous model is needed.57Table I(a) Valence-band edge E and band gap for some semiconducting materials.Material E (eV) Eg (eV)GaAs 9.53 1.42InP 9.64 1.35Si 9.50 1.1ZnSe 10.58 2.67(b) Comparison of Harrison’s valence-band offsets to experiment.Heterojunction Valence-Band Offset, (eV)Harrison53 This Study Other Expt.ZnSe/GaAs 1.05 1.25 ± 0.07 1.10 ± 0.0318ZnSe/InP 0.94 1.35 ± 0.15583.2 ALKALINE-EARTH FLUORIDES3.2.1 Interfaces3.2.1.1 Formation of the CaF2/G As (100) InterfaceFigure 19 shows photoemission spectra measured at liv = 83.6 eV for a seriesof anneals increasing in temperature for 1.1 monolayer (ML) of CaF2 deposited atroom temperature on GaAs (100), showing the As 3d, F 2s, Ca 3p, Ga 3d and F 2pphotoemission lines. (1 ML is defined as a CaF2 triplelayer which is 3.15 A thick).Low temperature annealing results in ordering of the CaF2 film evidenced by thesharpness of the Ca 3p photoemission features of CaF2. After annealing at 550°C theCa 3p level exhibits a component chemically shifted by 1.8 eV to lower bindingenergy. This chemically shifted component becomes more prominent at higherannealing temperatures. After annealing at 590°C only this portion of the Ca 3plevel remains.The As 3d and Ca 3p core levels for the series of annealing temperatures forthe sample of Fig. 19 were fitted in order to determine the nature of the chemicalbonding at the interface. The core-level spectra were fitted with two spin-orbitdoublets of the Lorentzian lineshape convoluted with a Gaussian approximating theexperimental resolution. Experimentally determined values48 of the branchingratio (1.53) and the spin-orbit splitting (0.69 eV) were used for the As 3d core level.The deviation of the branching ratio from its statistical value of 1.5 is attributed tothe steep increase in cross section of the 3d core level in the 50-100 eV photonenergy range.50 The spin-orbit splitting of the Ca 3p core level is very small, andassumed to be zero.33 The results of these fits are shown in Fig. 20. For the As 3dcore level the fitted component at higher binding energy corresponds to As in bulkGaAs, while the component chemically shifted to lower binding energy correspondsto As at the interface with CaF2. The intrinsic surface core-level shift, as modifiedby the CaF2 overlayer, will contribute to the chemically shifted component in the As59U)C0C)>-F-Cl)zLUFz35KINETIC ENERGY (eV)Figure 19. Photoemission spectra for 1.1 ML (3.5 A) of CaF2 on GaAs (100) for aseries of annealing temperatures from 520°C to 590°C measured with a photonenergy of 83.6 eV.45 55 65 75a CD CD CD C., CD CD CI-I CD p a a CD CD C CD CD — CD C) a I-i C., CD CD C., C CDC’CINTENSITY(counts/s)z Eli C.) m z m G)01 0)61core level at the lowest annealing temperature in Fig. 20. The larger chemical shiftat higher temperatures can be attributed to chemical bonding between As and Ca.The Ca causes a shift to lower binding energy (approximately 0.4 eV), because itdonates more charge to the adjacent As atoms than the Ga does due to the lowerelectronegativity of Ca relative to Ga. One expects a related chemical shift in theCa 3p core level towards lower binding energy due to the lower electrouegativity ofAs relative to ‘-F. As shown in Fig. 20 the Ca 3p does exhibit the appropriate signchemical shift. This chemical shift is larger (approximately 1.8 eV) reflecting thelarger difference in electronegativity between As and F as compared to Ca and Ga.There was no apparent evidence of a chemical shift in the Ga 3d photoemissionpeak.The relative intensity of the chemically shifted photoemission componentsincreases with higher annealing temperature. At the same time a relative loss of Gaand F was observed as demonstrated by Fig. 21 where the ratios of the area of theF 2p to Ca 3p photoemission peaks, and the Ga 3d to As 3d photoemission peaks, areshown as a function of temperature. The decrease of Ga and F relative to As and Cacan be attributed to the decomposition of CaF2 at the interface, followed by theformation of a volatile GaF compound which is lost by evaporation.54 This effectcan not be explained by the formation of CaF2 islands on the substrate as this wouldnot change the Ga 3d/As 3d ratio or the F 2p/Ca 3p ratio.Figure 22 shows photoemission spectra measured at liv = 83.6 eV for a seriesof CaF2 samples varying in thickness all annealed to 570°C. The photoemissionfeatures due to the substrate attenuate as the thickness of the CaF2 overlayersincreases. Also, the ratio of the Ca 3p interface component to the Ca 3p bulkcomponent decreases as the CaF2 overlayer becomes thicker, as is shown in Fig. 22.These results support the hypothesis that the CaF2 layers grow uniformly on thesurface.62450 500 550 600Figure 21. Ratios of the areas of the F 2p to Ca 3p and the Ga 3d to As 3dphotoemission peaks as a function of annealing temperature for the sample ofFig. 19.I I I I0.(U00.c..JUI I I I—%_S65432-S A (F 2p) I A (Ca 3p)1.510.500VCr)(U--0-- A (Ga 3d) / A (As 3d)I . I I I i I ITEMPERATURE (°C)U)04-’0C.)>-F(0zu-IFz35KINETIC ENERGY (eV)63Figure 22. Photoemission spectra for a series of CaF2 overlayers (a) 6.1 ML,(b) 2.2 ML, (c) 1.4 ML and (d) 0.48 ML on GaAs (100) annealed to 570°C with a45 55 65 75photon energy of 83.6 eV.64The core level measurements can be used to formulate a structural model forthe bonding of the CaF2 to the GaAs (100) surface. The polar (100) surface of GaAsmay consist of all anions or all cations, however surfaces cleaned by ion sputteringand annealing (550°C) are expected to be Ga terminated.55 Following deposition ofCaF2 and annealing, the results show a depletion of Ga relative to As, as discussedabove, when the interface is compared with the clean surface. This depletion isconsistent with the top monolayer of Ga from the Ga-terminated surface beingreplaced by Ca, and the Ga lost by evaporation, when the CaF2/G As interface isformed. Along with the loss of Ga is a loss of approximately one half of the fluorine,which may suggest that a volatile gallium fluoride compound is formed. In thismodel a CaF interfacial layer is left behind, with the Ca bonded to As on one side andto F on the other. This depletion of Ga and F at the interface54 has not been reportedpreviously. It is likely that a similar depletion of Ga and F will occur during MI3Egrowth of CaF2 on GaAs at elevated temperatures. The interface model shown inFig. 23, which is similar to one proposed earlier by Yamada et al.26’8 is consistentwith these experimental observations. In this interface structure Ca substitutes forGa and each Ca atom is bonded to two As atoms on the semiconductor side and four Fions on the CaF2 side.These results differ from CaF2 on Si, in that the interfacial reaction takesplace at lower temperature (550°C vs 700°C3) and the Ca bonds selectively to one ofthe components (As) in the compound semiconductor. There is now generalagreement3° that the CaF2/Si consists of direct Si-Ca bonds with no interveningfluorine layer, and in agreement with our results it has since been hypothesizedthat S1F4 is lost during the interface formation.Yamada et al.26 interpret the Ca 3p spectrum measured with a lower energyresolution on GaAs (100) (0.4 eV compared with 0.25 eV in the present work) asconsisting of three overlapping chemically shifted components; Ca-i, Ca-2 and Ca-3.65QAs SGa . Ca QF(100)AFigure 23. Proposed structural model for the CaF2/G As (100) interface.66The core-level shifts were -2.3, -1.1 and 0 eV respectively relative to bulk CaF2. Intheir interpretation Ca-i corresponds to Ca atoms in the GaAs lattice occupying theGa position, Ca-2 to Ca atoms existing in an abrupt CaF2/G As interface (GaAs,As-terminated; CaF2, Ca-terminated) and Ca-3 to Ca atoms in the CaF2 lattice. Ourresults, at a somewhat better energy. resolution (0.25 eV), show only two distinctoxidation states of Ca corresponding to the Ca-i and Ca-3 of Yamada et al.26 even forrelatively thick CaF2 films under extreme annealing conditions. The interfaciallayer does not grow beyond a monolayer even when 20 A (-‘6 ML) thick films areannealed up to 650°C. The first oxidation state is interpreted as being Ca atoms in theCaF2 lattice and the second being Ca atoms occupying the surface Ga positions of theGaAs (Ga-terminated) substrate.Only for the submonolayer film at the highest annealing temperature (590°Cspectrum in Fig. 19), does one observe a chemical shift in the F 2s spectrum,suggesting the existence of F in an alternative local bonding configuration. This Fis interpreted as being in the Ca-rich environment of the CaF surface layer.Furthermore, F vacancies may exist in this CaF layer. In either case, this results in achemical shift in the F 2s spectrum towards lower binding energy due to-. the moreelectron-rich environment for the F ions due to their proximity to incompletelyionized Ca ions. A related chemical shift is observed in the F 2p spectrum, shown inFig. 19, where the photoemission spectrum for the F 2p valence band shifts tohigher kinetic energy and becomes narrower with increasing annealingtemperature. The reduction in the width of the valence band is consistent with thesmaller number of neighbors for the F as the film thickness and F mole fraction isreduced with annealing. The absence of a bulk CaF2 signal in the Ca 3p spectrum atthe highest annealing temperature supports the idea that the Ca forms aninterfacial compound in which it is bonded to both F and As. The peak in Fig. 19,which lies 3-4 eV higher in kinetic energy than the F 2p peak and becomes more67prominent with annealing, is interpreted as a bonding orbital/valence baudassociated with the interfacial CaAs compound.3.2.1.2 SrF2 and BaF2 Interfaces with GaAsExperiments similar to the ones for the CaF2/G As interface were performedwith SrF2 and BaF2. The data for the formation of the SrF2/GaAs and BaF2/G Asinterfaces is more difficult to interpret because of the relatively small cross sectionsof the Sr and Ba related spectral features as compared to Ca, and because of theiroverlap with the spectral features of the GaAs substrate. However, from ourunderstanding of the CaF2/G As interface, certain similarities in these systems canbe identified and thus conclusions can be drawn concerning the interfaceformation.SrF2/GaAs InterfaceFigure 24 shows photoemission spectra measured at a photon energy of83.6 eV for a series of anneals for 2.3 ML (1 ML = 3.34 A) of SrF2 deposited at roomtemperature on GaAs (100) showing the As 3d, Sr 4s, F 2s, Sr 4p, Ga 3d and F 2pphotoemission features. Unlike the CaF2/G As system, the cation core levels fromSrF2, namely Sr 4p and Sr 4s, are difficult to measure because of their relativelysmall cross sections and their overlap with the Ga 3d and As 3d core levels.Although there are no clear indications of any chemical shifts from these spectra, itis apparent that the photoemission intensity of the spectral features associated withthe GaAs substrate are clearly enhanced relative to those of the SrF2 overlayerindicating a decrease in the overlayer thickness with annealing temperatureprobably due to some re-evaporation of the SrF2 or the formation of islands.In order to determine the presence of any chemical shifts and thus thenature of the chemical bonding, and because of difficulties in obtaining this10KINETIC ENERGY (eV)6840 50 60 7030 80Figure 24. Photoemission spectra for 2.3 ML of SrF2 on GaAs (100) for a series ofannealing temperatures from 370°C to 6 10°C measured with a photon energy of83.6 eV.69information from the Sr 4p and Sr 4s core levels, the Sr 3d core level was alsomeasured for the same sample of Fig. 24 but with a photon energy of 172.7 eV. Thespin-orbit splitting of this core level is 1.8 eV. These spectra together with ablow-up of the As 3d and Sr 4s core-level spectra from Fig. 24 are shown in Fig. 25.A chemical shift on the As 3d is observed at approximately 590°C and a overlayerthickness of 1.2 ML. This chemical shift is accompanied by a similar one on theSr 3d core level of approximately 1.1 eV towards lower binding energy. As in thecase of CaF2 no chemical shift was observed on the Ga 3d photoemission peak.As further evidence of the similarity between the CaF2/G As and SrF2/GaAsinterfaces a plot similar to Fig. 21 showing the ratios of the areas of the F 2p toSr 4p and the Ga 3d to As 3d photoemission peaks is shown for SrF2/GaAs in Fig. 26.A decrease in both of these ratios indicates the amounts of F and Ga relative to Asand Sr is reduced and in analogy to the CaF2/G As system can be attributed to thedecomposition of SrF2 at the interface and the formation of Sr-As bonds.BaF2/GaAs InterfaceFigure 27 shows photoemission spectra taken at a photon energy of 83.6 eVfor 3.8 ML (1 ML = 3.58 A) of BaF2 deposited at room temperature showing theAs 3d, Ba 5s, F 2s, Ga 3d, Ba 5p and F 2p photoemission features. In these spectra aBa Auger peak is also present in the 45-50 eV kinetic energy range whichcontributes to the broad spectral feature in this region of the spectra. Like the SrF2case, the photoemission features of the substrate are clearly enhanced relative tothose of the overlayer at higher annealing temperatures indicating the possiblegrowth of islands or the re-evaporation of the BaF2. At the highest annealingtemperature there appears to be a chemically shifted component on the Ba 5p corelevel.70Figure 25. Photoemission spectra of the Sr 3d core level and the As 3d, Sr 4sspectral region for a series of annealing temperatures for the sample of Fig. 24.The Sr 3d core level was measured with a photon energy of 172.7 eV.29 30 31 32 33 34 35 36 37 38 39 40 41KINETIC ENERGY (eV)71‘cr18C,)-16C’4JLi..‘—14V1.5 c’C’)Cr)1 0I222021210400 450 500 550 600TEMPERATURE (°C)0.5650Figure 26. Ratios of the areas of the Sr 4p to F 2p and the Ga 3d to As 3dphotoemission peaks as a function of annealing temperature for the sample ofFig. 24.72Cl)C0C.)>-I—C’.)zwIz30 80KINETIC ENERGY (eV)Figure 27. Photoemission spectra for 3.8 ML of BaF2 on GaAs for a series ofannealing temperatures from 470°C to 620°C measured with a photon energy of83.6 eV.40 50 60 7073The Ba 4d core level was also measured for this sample over the same rangeof annealing temperatures at a photon energy of 160.9 eV and these spectra areshown in Fig. 28. The spin-orbit splitting of the Ba 4d core level was measured tobe 2.7 eV with a branching ratio of 1.2. The deviation of the branching ratio fromits statistical value of 1.5 is attributed to the steep decrease in cross section of the 4dcore level in the 100-170 eV photon energy range.5° At the highest annealingtemperature a chemical shift of approximately 1.2 eV toward lower binding energywas observed. In order to determine the nature of the interfacial bonding, one mustalso look at the substrate core levels. Figure 29 shows fitted Ba 4d and As 3d corelevels for the sample of Fig. 27 after it had been annealed at 620°C. Two spin-orbitsplit doublets were used to fit the core levels as was done for the CaF2/G As system.The experimentally determined values of the spin-orbit splitting and the branchingratio were used for the Ba 4d core level whereas the values of Ludeke48 were usedfor the As 3d core level. The fitted component of the As 3d core level at higherbinding energy corresponds to As in bulk GaAs, while the component chemicallyshifted to lower binding energy corresponds to As at the interface with BaF2. As inthe CaF2 case, the cation (Ba) causes a shift to lower binding energy because itdonates more charge to the adjacent As atoms than the Ga does due to the lowerelectronegativity of Ba relative to Ga. A related chemical shift is expected in theBa 4d core level towards lower binding energy due to the lower electronegativity ofAs relative to F. Both core levels exhibit these expected chemical shifts thatnaturally lead to an interpretation of Ba-As bonding at the interface.3.2.1.3 Band Offsets at Insulator/Semiconductor InterfacesThe electronic properties of the interface are best described by the energyoffset between the valence bands of the two materials making up theheterojunction. By expanding the vertical scale of spectra such as those in Fig. 19,CoCl)0C.)>-I(I)zwFz60 61 62 63 64 65 66KINETIC ENERGY (eV)74Figure 28. Photoemission spectra of the Ba 4d core level for a series of annealingtemperatures for the sample of Fig. 27 measured with a photon energy of 160.9 eV.67 68C0C.)>-I—Cl)zwIz7536 37 38 39 40 41 60 61 62 63 64 65 66 67 68KINETIC ENERGY (eV)Figure 29. Fitted As 3d and Ba 4d core-level spectra for the sample of Fig. 27 afterannealing at 620°C.76one can determine the valence-band offset between the overlayer and the substrate,as shown in Fig. 30 for the CaF2/G As interface. The CaF2 and GaAs valence-bandedges were determined by a simple linear extrapolation of the high-kinetic-energyside of the respective valence-band photoemission spectra, to the background. Thismethod is simpler than the one dealing with valence-band edges which areseparated by a relatively small amount. As was discussed earlier for the ZnSeinterfaces with GaAs and InP, a model using parabolic band edges was used todetermine the valence-band offsets in that case. The band alignment and Fermilevel in Fig. 31 were measured on a 6 ML sample of CaF2. The band gaps of the GaAsand CaF2 are taken to be 1.4 and 12.1 eV33 respectively. The 8.5 eV band offset forthis sample compares with 8.8 eV reported earlier for 2.5 ML of CaF2 on Si (1ll).23Figure 32 shows that the valence-band offset, AE decreases by more than1.5 eV as the thickness of the CaF2 is reduced, These results are consistent with thevalence-band offset of 7.7 ± 0.3 eV reported earlier for 1-2 ML of CaF2 deposited atroom temperature on GaAs (1 10).26 Note that within the accuracy of themeasurements the band offset in Fig. 32 seems to depend only on the film thicknessand not on the annealing temperature. The lack of dependence on annealingtemperature is not an obvious result in view of the interfacial compound formationwhich takes place with annealing as discussed above, Two physical mechanismsmight be expected to contribute to the observed change in the band offset withthickness. First, one might expect the structure of the overlayer to evolve as thelayer thickness increases through the first few monolayers due to the 3.5% latticemismatch between GaAs and CaF2. Thus as the structure and/or strain in theoverlayer changes with thickness it is not unreasonable to expect a strain-inducedchange in the band offset. This effect is difficult to quantify without more detailedinformation about the structure of the CaF2 overlayer as a function thickness.77(I)0C)>-F(1)zwFz64 84Figure 30. Photoemission spectrum of the valence-band region for 6 ML of CaF2 onGaAs measured with a photon energy of 83.6 eV, showing the method ofdetermining the valence-band offset.68 72 76 80KINETIC ENERGY (eV)78GaAs1.4 eV2.2—FERMITO.leV8.5 eVCBMIN.LEVELVB MAX.Figure 31. Band alignment and Fermi level for 6 ML of CaF2 on GaAs.9Cl)Uu- 80Uzth750zLU6.5079Figure 32. Valence-band offset, as a function of CaF2 overlayer thickness forthe CaF2/G As interface. The solid line is a simple quantum tunneling model asdiscussed in the text assuming a band offset of 8.5 eV for a thick CaF2 overlayer andan effective mass of O.Sme.1 2 3 4 5 6 7ThICKNESS (M L)80A second effect which will cause the measured band offset to change withthickness is quantum mechanical tunneling of valence-band electrons from theGaAs into the band gap of the CaF2. These electrons, which tunnel into theforbidden gap of the CaF2 contribute to the density of states just above the top of thevalence band of the CaF2. For very thin layers of CaF2, this density of states makesthe valence-band offset at the interface appear to be smaller, when measured inphotoemission, which is sensitive to the density of states. The solid line in Fig. 32shows an estimate for the size of this effect. In this model, the band offset is reducedfrom its value for very thick layers to the energy at which a tunneling electronfrom the GaAs valence band is able to tunnel through the potential barrier into theband gap of the CaF2 with a transmission coefficient of lie. The band offset as afunction of overlayer thickness d is then given by= AEo -2/(2m*d) (11)The effective mass m* for the CaF2 valence band was estimated to be O.Sme from atight binding model for the valence band in which the width of the valence band isassumed to be 4.0 eV as inferred from the width of the F 2p photoemission line forbulk CaF2. The calculated thickness dependence of the band offset in Fig. 32 with= 8.5 eV is qualitatively similar to the data. In view of the structural changesthat might be expected with changes in layer thickness as discussed above and theidealized nature of the model, it is surprising that the simple quantum tunnelingmodel fits the data as well as it does.As shown in Fig. 33 the surface Fermi level moves towards the top of the GaAsvalence band over approximately the same CaP2 thickness range as the change inthe valence-band offset shown in Fig. 32. This dependence on film thickness issimilar to the dependence on film thickness observed for the Fermi level in CaF2 onsilicon.23 In the silicon case the movement of the Fermi level towards the valenceband was interpreted as being due to the formation of interface states associated810.80.70.6>-00.4w0.3wLL0.20.100JD THICKNESS (M L)Figure 33. Fermi level position, EF relative to the GaAs valence-band maximum as afunction of CaF2 overlayer thickness for the CaF2/G As interface. The slopedsegment is a linear least-squares fit to the data for submonolayer thicknesses. Thehorizontal segment represents the position of the Fermi level for thicknessesgreater than 1 ML that would be consistent with the interpretation discussed in thetext.1 2 3 4 5 6 782with Ca-Si bonds, that lie below the top of the valence band of the semiconductor. Inthe GaAs case, one can interpret the downward movement of the Fermi level towardsthe valence-band maximum as being due to the formation of interface statesassociated with Ca-As bonds. In this interpretation, one would expect the downwardshift of the Fermi level to be completed at a coverage of 1 ML at which point itwould remain fixed in position relative to the valence-band maximum. In Fig. 33, alinear least-squares fit to the data for coverages up to 1 ML was determined, and theline of zero slope is consistent with the Fermi level remaining constant for highercoverages. Like the valence-band offset the Fermi level depends only on the filmthickness and not independently on the annealing temperature as shown in Fig. 33.Since annealing has been shown to effect the interfacial composition, which inturn would be expected to influence the Fermi level position, this result may meanthat the Fermi level position is determined not by interface states at all, but by bulkstates or fluorine vacancies in the CaF2 overlayer. The Fermi level extrapolatescontinuously to the measured clean-surface Fermi level position near midgap, at0.5-0.7 eV above the top of the GaAs valence band.Using the same method as previously, the valence-band offsets and Fermilevel positions were determined as a function of overlayer thickness for theSrF2/GaAs and BaF2/G As interfaces and these results were compared with theresults obtained for the CaF2/G As interface. Figure 34 shows the valence-bandoffset as a function of overlayer thickness for the three alkaline-earth fluorides.All the data points in this figure are from interfaces that had been annealed to400°C and some as high as 600°C. Within the accuracy of these measurements theband offset seems to depend only on the film thickness and not on the annealingtemperature. Figure 35 shows the Fermi level position as a function of overlayerthickness for the three alkaline-earth fluorides. Unlike CaF2, the Fermi level forSrF2 and BaF2 stays near midgap for coverages up to 4 ML.839 I0 08.5 a00on-LLoo°0o 0LL0• 075 o4je%.• SrF2Uo- L:JBaF2z6.560I I I I I I I I I ITH(CKNESS (ML)Figure 34. Compilation of the results of the valence-band offset as a function ofoverlayer thickness for the alkaline-earth fluoride/GaAs interfaces.841 I I I I I I I I I I I I.0.8-_S •SrF2>ci)•UBaF2UIUILii c9U 0000 000.2- QQ Q• Oq0 I I I i I I0 2 4 6 8 10THICKNESS (ML)Figure 35. Compilation of the results of the Fermi level position as a function ofoverlayer thickness for the alkaline-earth fluoride/GaAs interfaces.853.2.2 X-ray AbsorptionIn conjunction with the experiments on the alkaline-earth fluoridelGaAsinterfaces, we have also studied aspects of the electronic structure of the overlayerfilms with x-ray absorption spectroscopy. In this section we measure some of theabsorption edges for the alkaline-earth fluorides (CaF2, SrF2, BaF2) by determiningthe total electron yield as a function of photon energy. The cation absorption edgespectra are interpreted in terms of an atomic multiplet theory where the effects ofthe crystal field in the solid are initially neglected. We also determine the bindingenergies of the electron excited states associated with the core-to-bound-stateabsorption transitions for both the cation edges and the fluorine K edges.3.2.2.1 Atomic Multiplet TheoryThe starting point for understanding the physics and thus interpreting thespectral features in x-ray absorption spectra of solids is an atomic theory where themultiplet splittings of the absorption final state are calculated. Many of theabsorption edges to be discussed here are quite adequately interpreted in terms ofthe atomic theory. Recently there has been work47 which treats the mostprominent effect of the solid state, namely the crystal field. It is the crystal fieldwhich breaks the spherical symmetry around the atom. In interpreting theabsorption spectra of the alkaline-earth fluorides, the effects of the crystal fieldwill only be dealt with qualitatively. The absorption transitions that will be ofinterest are the cation absorption edges 2p6—* 2p53d1 (Ledge), 3d10—* 3d94f1(Medge) and 4d10—*4d9f (N edge).Atomic multiplet theory can be found in the standard textbooks.56 TheHamiltonian for the final state in the 2p6—* 2p53d1 absorption transition can bewritten asH = Hay + L.S(p) + L.S(d) + g(i,j) (12)86Hay consists of the kinetic energy term and the interaction with the nuclei. It givesthe average energy of the multiplet and does not contribute to the multipletsplittings. The splittings are caused by the spin-orbit couplings, L.S, for the 2p and3d electron and by the Coulomb repulsion term, g(i,j). This two-electron operatorcan be expressed in terms of spherical harmonics,56 which requires its divisioninto radial and angular parts. The radial part, RK(1112;l3 4) is divided into directCoulomb terms, FK(l112;lil2), and exchange terms, GK(ljl2;l1). The index K is forthe infinite series that results from expressing the l/r part of g(i,j) in terms ofLegendre polynomials. The angular part of g(i,j) results in the selection ruleswhich gives the possible K values. For the direct Coulomb term, no odd K values areallowed, and the maximum K value is two times the minimal 1 value. For the p5d1multiplet this means that F° and F2 are the only non-zero terms, while for a d9f1multiplet, which results from a d—> f absorption transition, F4 is also non-zero. The Kvalues in the exchange term equal 111-121, 11-2÷2k1, ..., 11+12. For the p5d1 multipletthis results in & and G3 being the only non-zero terms. Thus one can evaluate thetwo-electron operator g(i,j) for the p5d1 multiplet in terms of four interactions: F°,F2, & and G3. F° only contributes to the average energy and is taken into 11avThese terms can be calculated using the self-consistent Hartree-Fock method andthe results of these calculations are expressed in terms of the so-called Slaterintegrals F0, F2, G1 and G3. To avoid the occurrence of fractional coefficients thesequantities are related through numerical constants,56F2 G335’ 15’ 245for a p5d1 configuration and,F2 G- G3-105’ 35’ 35’ 315for a d9f1 configuration.87Zaanen et al.57 have calculated the optical spectrum in terms of the atomicmultiplet theory for the 2p63d0_2p51absorption transition in calcium. We haveextended Zaanen’s pd calculation applicable to Ca to a df calculation applicable to thed10f°—*d9 transitions of the Ba M and N edges,Consider the Ba2 atom, which has filled 3d and 4d shells and an empty 4fshell in the ground state. The ground state of the atom represented in LS couplingand written in spectroscopic notation is is. The general symbol in this notation isgiven by 2S+lLJ where S, L, and J are the respective angular momentum quantumnumbers for the given electron configuration. The only optically allowedtransitions from the ground state are to the final states. Considering only3d—.4f transitions the electronic configuration of the final states is d9f1 which, inLS coupling, splits into the terms H, 1H, 3G, 1G, 3F, 1F, 3D, 1D, 3P, and because ofthe d-f Coulomb and exchange interactions. If the Coulomb and exchangeinteractions were neglected, one would expect an absorption spectrum consisting oftwo peaks separated in energy by the spin-orbit splitting of the d core level with 3:2intensity ratio corresponding to the statistical occupation of the 2j+l substates of thed core level. Including the Coulomb and exchange interactions causes a largeredistribution of intensities and therefore strong deviations from the 3:2 intensityratio that would be expected if only the spin-orbit coupling were considered. If onthe other hand the spin-orbit coupling were negligible one would see a single linein the absorption spectrum corresponding to the ‘P final state.Because of the large spin-orbit coupling of the d hole it is more appropriateto work in a j,j representation. The transformation matrix from ES to jj couplinginvolving the state is given by88r a(512,7/2) 1 -Ii ‘[fl 2’1i b(3D1)[ a(512,512) j = 4z= 4 3[ .4 b(3P1) (13)a(312,5/2) s[jj b(1P)The matrix elements of the g(i,j) operator of equation (12) must bedetermined in terms of the j,j basis. Condon and Shortley56 has shown that the termenergies are given as(3DIg(i,j)1)=-F°-6F2+99F4( 3P I g(i,j) I 3P ) = - F° - 24F - 66F41PIg(i,j)1)=-F°-24F-664+70GUsing the transformation matrix of equation (13) and expressing the Coulomb andexchange interactions in terms of the Slater integrals F2, F4 and G1, the interactionmatrix elements ( I g(i,j) I J’ ,j ) are150 297 40 72f 132’[ 4[ 18”[T 33’[T 4[fl3675F24851j-Gi - 3675F24851+72[ 13 2’Ji 4’Ji 552 66 2 7 2’1Th 1 32Tl 2’ITE-3675F2485101-3675F4851fG1 3675F÷425-18 33[ 4Thi 72JT 1 32JT 2JT 84 28- 3675F24851+ G1 3675F+2425- G -F2t-jG1+Although the f-electron spin-orbit coupling is negligible and is neglected thespin-orbit coupling of the d hole is not. This coupling is accounted for by adding aterm , which equals the spin-orbit splitting of the d level, to the( 3/2,5/2 I g I 3/2,5/2 ) matrix element as shown above. Strictly speaking, thecoupling is accounted for by adding the matrix (id’if I L.S(d) I JdJf ) to the oneabove, however, the only element in this matrix which contributes to the splitting89of the multiplet is the ( 3/2,5/2 I L.S(d) I 3/2,5/2 ) element. The eigenstates of thed9f1 final configuration are written asIi> = x 15/2,7/2> + 13 15/2,5/2) i’j 1312,512) (14)and using equation (13) the optical spectrum is given by= f2f cc-+ q14 I2 6(0—Ei) (15)where E1 are the eigenvalues determined from the ( Jd’Jf I g(i,j) I d’’Jf’ >interaction matrix. A spectrum can be determined if values for the Slater integralsare known, or alternatively one can fit the experimental spectrum and determineempirical values for the Slater integrals.The results of the earlier work of Zaanen et al. are summarized below, sincethey will be useful for interpreting the Ca L edge spectrum measured for CaF2. Theanalogous expressions for the LS to jj coupling transformation matrix involving theP state and the Coulomb and exchange interaction matrix for the p5 d 1configuration arer a(3/2,512) 1 qj b(3D1)[ a(312,312) j = 2 4 -[ b(3P1) (16)a(1/2,3/2) [i -[ ITh b(1P)andi. LI-MF2 + l 2 - jGi j2 +751 2 - G1 - +lIF2-4F2 + j01 1I•F2 - Gl +90The eigenstates of the p5d1 final configuration are written asIi) = x 13/2,5/2) ÷ lj 3/2,3/2) + y 11/2,3/2> (17)and the optical spectrum is given byP(0) = I3’[ xj - + ‘J10yl26(w—Ej) (18)Atomic Hartree-Fock values for the Slater integrals are available for the Ca2+2p—*3d transition. For the Ba2+ 3d—.*4f and 4d—+4f transitions, however, no calculatedvalues appear in the literature. For the Ba2+ transitions, values for theisoelectronic La3+ species will be used as a starting point to fit the spectra andva1us for the Slater integrals will be determined from the best fit. -3.2.2.2 The Calcium L Edge in CaF2The absorption in CaF2 near the calcium L edge (and the fluorine K edge)have been studied extensively in thin films of CaF2 deposited on silicon.23’58 Themeasurements reported here were carried out on thin films of CaF2 evaporated onGaAs, as well as powder samples pressed into indium foil.In Fig. 36, total electron yield measurements of the absorption spectrum ofCaF2 in the vicinity of the Ca L edge for both a thin film sample and a powdersample are shown. The absorption is assumed to be proportional to the total electronyield. As discussed earlier, the total electron yield was measured with amicrochannel plate detector placed close to the sample with a retarding potential of140 V. The Ca Ledge in CaF2 is dominated by the Ca2 2p—*3d, core-to-bound-statetransition.47’59 All of the spectral features in Fig. 36, which are better resolved inthe powder sample, can be interpreted with reference to the spin-orbit splitting ofthe 2p core hole, the Coulomb and exchange interaction between the 3d electron andthe 2p core hole and the crystal field splitting of the 3d electron due to the presenceof the eight neighboring fluorine ions.47 The result of the atomic calculation of the91absorption spectrum for Ca2 is shown in Fig. 36 and summarized in Table II. Theoverall position of the calculated spectrum has been shifted to best fit the data. Thecalculation does not account for all of the spectral features because it does notinclude the crystal-field interaction,4‘We are now going to discuss the relation between the atomic-like absorptiontransitions and the energy levels of the CaF2 crystal. The binding energy of the 3delectron resulting from the absorption transition relative to the conduction-bandedge of the CaF2 can be estimated. This binding energy can be estimated bycalculating the difference between the energy of the absorption transition, and theenergy spacing between the Ca 2p core level where the excited electron originatedfrom and the bottom of the conduction band. The latter quantity is given by theenergy spacing between the Ca 2p core level and the top of the valence band asmeasured in photoemission, plus the optical band gap of CaF2 (12.1 eV). Thebinding energy of the Ca2P3/2 core level relative to the valence-band maximum is341.0 eV in this CaF2 sample. This results in a binding energy of the 3d electronrelative to the bottom of the conduction band of 3.6 eV in the final state of theabsorption transition C in Fig. 36, if the core hole is treated as a pure 2P3/2 state.The 3d electron in the final state of the transition E has nearly the same bindingenergy, since the difference in the absorption energy is almost entirely due to thespin-orbit splitting of the 2p core hole. In calculating the binding energy of thebound states, the core hole for the lower energy transition (C) in Fig. 36 has beentreated as a pure 2P3/2 state and the core hole for the higher energy transition (E)has been treated as a pure 2Pl/2 As shown in Table II, this approximation issupported by the u-coupling percentages, where transition C is associated with a92.2% pure 2P3/2 core hole and transition E is associated with a 90.1% pure 2P1/2core hole.92Cl)4-,0C)>-FC’)zLUFz345 347 349 351 353 355PHOTON ENERGY (eV)Figure 36. Total electron yield measurements of the absorption of a CaF2 thin filmand powder sample at the Ca L edge. Also shown is a calculated atomic spectrum(dashed line) in which the crystal field has been neglected. The letters indicate theexcitation energies used to obtain the spectra in Fig. 43.93Table IISummary of the atomic calculation for the Ca2+ 2p—> 3d absorption transition.Peak Energies (eV) Relative u-coupling LS-couplingIntensity Percentage Percentage- 347.2 1 71.9% p312d 7.5%26.0% p312d5 91.7% 3P2.1% p112d3 0.8%C 349.4 41 28.0% p312d 61.7% 3D64.2% p312d5 2.8%7.8% p112d3 35•5%E 352.7 74 0.1% p312d 30.8% 3D9.8% p312d5 5.5%90.1% p112d3 63.7%Parameters (eV):Ca2 (2p,3d): F2 = 3.79, G1 = 2.51, d = 3.5 (from ref. 47)94Even though the bottom of the conduction band in CaF2 has Ca 4scharacter,6° the absorption spectrum can be interpreted without reference to theCa 4s orbital. A possible explanation is that the relatively weak Ca 2p6—>2p54s1transition overlaps -with the 3d transitions, and cannot be observed separatelywithout a quantitative analysis of the lineshapes.3.2.2.3 The Barium M and N Edges in BaF2Barium M EdgeFigure 37 shows total electron yield measurements of the absorptionspectrum of BaF2 in the vicinity of the Ba M edge for a thin film and a powdersample. Unlike the L edge spectra of CaF2, the spectral features for the M edge ofthe thin film and powder samples of BaF2 are equally well resolved. This spectrumis dominated by the Ba 3c1—* 4f, core-to-bound-state transition, and can beinterpreted in terms of the spin-orbit splitting of the 3d core hole and the Coulomband exchange interaction between the 4f electron and the 3d core hole. The BaM edge absorption spectrum was fit with an atomic model for Ba2+, and the resultsof the fit are shown in Fig. 37 and summarized in Table III. The Slater integrals F2,F4 and G1 were used as parameters and a 3d spin-orbit splitting of 15.2 eV was usedin the fit to both the position and relative intensity of the spectral lines.6’ TheSlater integrals of La3+ were used as initial values and only G1 had to besignificantly adjusted to improve the fit to the Ba2 spectrum. As shown in Fig. 37the calculated spectrum accounts for all the spectral features. No additional linesresulting from the crystal field were observed which indicates that it is smaller forthe Ba M edge in BaF2 than the Ca L edge in CaF2. The crystal field has less of aneffect on the localized 4f orbitals of Ba compared to the 3d orbitals of CaSimilar to the CaF2 case, the binding energy of the 4f electron relative to theconduction-band edge can be estimated from the energy of the respective95Cl)D00>-I.Cl)zwFz775 780 785 790 795 800 805 810PHOTON ENERGY (eV)Figure 37. Total electron yield measurements of the absorptionand a powder sample at the Ba M edge. Also shown is a calculated fit to the spectrum(dotted line) in which the crystal field has been neglected. The letters indicate theexcitation energies used to obtain the spectra in Fig. 44.of a BaF2 thin film96Table IIISummary of the atomic calculation for the Ba2+ 3d—> 4f absorption transition.Peak Energies (eV) Relative u-coupling LS-couplingIntensity Percentage Percentage- 779.4 1 13.6% d512f7 21.2% 3D86.2% d512f 77.9%0.2% d312f5 0.9%A 782.7 48 83.7% d512f7 48.7% 3D13.8% d512f 8.6%2.5% d312f5 42.7%B 797.7 63 2.7% d512f7 30.1% 3D0.04% d512f 13.5%97.3% d312f5 56.4%Parameters (eV):La3 (3d,4f): F2 = 5.985, F4 = 2.63 3, G1 = 3.990 (from ref. 62)Ba2 (3d,4f): F2 = 6.0, F4 = 2.6, G1 = 3.0, d = 15.297absorption transitions, and the energy spacing between the Ba 3d core level andthe top of the valence band plus the optical band gap of BaF2 (11.0 eV33). Thebinding energies of the Ba 3d512 and 3d2 levels relative to the valence-bandmaximum are 775.6 eV and 790.8 eV respectively. This results in binding energiesof the 4f electron relative to the bottom of the conduction band of 3.9 eV and 3.7 eVfor the 3d512 and 3d12 core holes, which have been treated as pure states. The 4felectron in the final state of these transitions, labeled A and B in Fig. 37, havealmost the same binding energies, since the difference in the absorption energy isalmost entirely due to the spin-orbit splitting of the 3d core hole. As shown inTable III, the approximation of treating the core holes as pure 3d5,2 and 3d12 statesrespectively is supported by the jj-coupling percentages, where transition A isassociated with a 97.5% pure 3d5/2 core hole and transition B is associated with a97.3% pure 3d2 core hole.The width of the peak, labeled B in Fig. 37 is clearly larger than peak A. Thisadditional width can be attributed to the bubbling up of the 3d/2 core hole and aresulting band to band transition. This process is possible because the spin-orbitsplitting of the 3d core level (15.2 eV) is larger than the optical band gap (11.0 eV)of BaF2. An electron can decay from the 3d512 level in to the 3d/2 level, and theresulting energy can produce an electron transition from the valence band to theconduction band. The small peak located at a photon energy of 789 eV is interpretedas a transition from the lower energy 3d5,2 level to a resonance in the conductionband possibly of Ba 6s character.Barium N EdgeFigure 38 shows the total electron yield measurement of the absorptionspectrum of BaF2 in the vicinity of the Ba N edge. A giant absorption resonanceeffect is observed, which is attributed to collective excitations associated with the98(I)4.-IC0C)>-I—C/)zwIz90Figure 38. Total electron yield measurement of the absorption in the vicinity of theBa N edge of a powder sample of BaF2 showing the giant 4d— 4f absorption100 110 120 130PHOTON ENERGY (eV)resonance.994d—4 4f, ef transitions. Before discussing the origin of this giant absorption featurein more detail, it should be noted that there are spectral features present in the90-95 eV range which can be interpreted in terms of the multiplet splitting of the4d9f1 excited state arising from the atomic 4d— 4f core-to-bound-state transition.Figure 39 shows the absorption spectrum of BaF2 in the 87-97 eV photon energyrange. Similar to the previous analysis, this region of the Ba N edge spectrum wasfitted with the atomic model which includes the effects of the spin-orbit splitting ofthe 4d core hole and the Coulomb and exchange interactions between the 4f electronand the 4d core hole. The results of this fit are shown in Fig. 39 and are summarizedin Table IV.The Slater integrals F2 and F4 were used as parameters while G1 and a 4dspin-orbit splitting of 2.7 eV were used as constants to fit both the position andrelative intensity of the spectral lines. The value of G1 was obtained fromLucatorto et al.64 while the values of F2 and F4 for La, obtained from Sugar63 wereused as initial values for the fit. Only small deviations in F2 and F4 from the valuesfor La were found to provide the best fit to the Ba data.In contrast to the 3d absorption edge, where the 3d spin-orbit splitting islarger than the Coulomb and exchange interactions of the 4d electron and the 3dcore hole, one finds that in the 4d photoabsorption the Coulomb and exchangeinteractions are much larger than the spin-orbit interaction. As a result themultiplet structure of the Ba2 3d94f1 configuration is essentially determined by the3d spin-orbit coupling. The purity of the levels determined from the percentagecompositions at the 3d edge was shown earlier (see Table III) to be high in jjcoupling and therefore one can designate the excited states in this coupling. Asshown in Table IV, for the 4d photoabsorption the purity in jj coupling is low andthe levels have almost pure LS-coupling character, Therefore peaks A and C in(I)C00>-I—C/)zLUIzthe crystal field has been neglected.used to obtain the spectra in Fig. 45.The letters indicate the excitation energies10088 89 90 91 92 93 94 95PHOTON ENERGY (eV)96Figure 39. Total electron yield measurement of the absorption at the Ba N edge for aB aF2 thin film. Also shown is a calculated fit to the spectrum (dotted line) in which101Table IVSummary of the atomic calculation for the Ba2+ 4d—*4f absorption transition.Peak Energies (eV) Relative jj-coupling LS-couplingIntensity Percentage PercentageA 90.1 1 20.2% d512f7 6.3% 3D73.5% d,f51 93.5%6.2% d312f5 0.2%C 93.6 4 30.5% d512f7 93.3%23.7% d512f 6.2% 3P45.8% d312f5 0.5%- 109.8 710 49.2% d512f7 0.4% 3D2.8% d512f 0.3%48.0% d312f5 99.3%Parameters (eV):La (4d,41): F2 = 10.2, F4 = 6.5, G1 = 9.42 (from ref. 63)Ba2 (4d,4f): F2 = 10.8, F4 = 6.1, G1 = 9564, d = 2.7102Fig. 39 are identified as transitions from the ground state to terms and 3D of the4d9f1 configuration.As observed in Fig. 39, the and 3D components of the multiplet structure inthe Ba N edge exhibit asymmetric lineshapes. The asymmetry is considered to comefrom a small amount of interference between excitations such as4d105s2p6—*4d9f--4 105s2pd,4d105sp—÷4d105s2ped.In the Fano theory for this interference effect, the absorption line profile isexpressed by the formu1a65:I(o) oc (q + E)21(e2 + 1), and E (Iwo - E0)/T,where E0 is the idealized resonance energy which pertains to a discrete excited state,P is the line width associated with the mixing between the discrete and continuumstates, and q is the asymmetry parameter. The P and q values for the 3P and 3Dcomponents of the Ba N edge multiplet have been estimated through a best fitprocedure of the lineshape. The results of this fit yielded q values of -9 eV and-5 eV and P values of 0.18 eV and 0.19 eV for the 3P and 3D componentsrespectively.The binding energy of the 4f electron relative to the conduction-band edgecan again be estimated, this time from the energy of the respective absorptiontransitions, and the energy spacing between the Ba 4d core level and the top of thevalence band, plus the optical band gap. As shown in the table above, it is stillpossible to identify the excitonic state for peak A in terms of its jj character, since93.7% of this state is associated with a Ba 4d512 core hole. The binding energies ofthe Ba 4d512 and 4d312 core levels relative to the valence-band maximum are84.0 eV and 86.7 eV respectively. These measurements result in a binding energyfor the 4f electron relative to the bottom of the conduction band of 4.9 eV when thiselectron is bound to a 4d512 core hole. The corresponding value for the binding103energy of the excitomc transition labeled C in Fig. 41 is 4.1 eV. However, as shownin Table IV, this state has approximately equal character from the two 4d core holes,and thus the binding energy of the bound excited state is not well described in termsof jj coupling. Representing this exciton in LS coupling as a 3D excited stateconfiguration with a 4.1 eV binding energy is more accurateIn the atomic multiplet calculation used to identify the and 3D componentsof the Ba 4d—, 4f absorption, a third component approximately 16 eV above the 3Dcomponent was identified as the 1P component. The calculated : 3D :intensity ratio is 710 : 4 : 1. Clearly the 1P component is not observed as a sharpsingle structure but rather as a broad band. Wendin66 pointed out that the giantcontinuum structure can be interpreted as a collective excitation of 4d, 5s, and 5pelectrons. Strong collective effects result from transitions between orbitals whichare localized in the same region of space, that is for transitions within a main shell.The condition for collective effects in atomic subshells is that they belong to openmain shells so that there exists transitions to essentially empty subshells within thesame main shell, leading to a very large polarizability.There has been a considerable amount of work on giant resonances in atomsand solids.67 A simple classical interpretation of this resonance is briefly describedhere. In the photoabsorption process a photon excites the atom which induces adeformation in the charge distribution of the electrons, mainly those electrons inthe outer most shells. The resulting polarization of the atom causes the charges toredistribute. The net effect of this continual charge redistribution is similar to theoscillations observed in metals and given by the plasma frequency. In our case, theoscillation modes are confined to the atom and the associated frequencies would beexpected to be larger than those found in metals.1043.2.2.4 The Fluorine K EdgesCalcium FluorideAbsorption spectra of CaP2 in the vicinity of the F K edge are shown inFig. 40 for a thin film and a powder sample. This spectrum, like Fig. 36, is obtainedin a total electron yield measurement, with a retarding potential of 250 V. Althoughthe F K-edge absorption has been measured earlier,58 it is not understood to thesame extent as the Ca L-edge absorption. The F K edge is more complicated in that itinvolves an interatomic transition in which the final state electron is likely to beprimarily on the Ca ion.Following a similar procedure to that used for the Ca L core excitons, thebinding energy of the electron in the state corresponding to the lowest energytransition (A) in Fig. 40 was found to be 1.7 eV relative to the CaF2 conduction-bandedge. The second peak, B in Fig. 40, involves a transition to a state that lies abovethe bottom of the conduction band. Since the lower part of the conduction band inCaF2 is made up of Ca orbitals, with Ca 4s at the bottom and Ca 3d higher up,6° thelowest energy transitions in Fig. 40, must involve interatomic transitions. In thiscase the high degree of localization of the F is core hole is likely to have a smallereffect on the binding energy of the F core exciton relative to the valence exciton,than it would if the excited electron were on the same atom, as it is in the Ca L edge.In fact the valence exciton, which consists of a hole in the F 2p valence band andan electron in the Ca-like conduction band, has a binding energy of 1.1 eV,33 quiteclose to the binding energy of the F ls exciton.Barium FluorideFigure 41 shows absorption spectra in the vicinity of the F K edge for a thinfilm and powder sample of BaF2, As expected these spectra are qualitatively similarto the P K edge spectra obtained for CaF2, and thus a similar interpretation of aI105685 695 705 715 725PHOTON ENERGY (eV)Figure 40. Total electron yield measurement of the absorption at the F K edge for aCaF2 thin film and powder sample. The letters indicate the photon energies used toobtain the photoelectron spectra in Fig. 46.i106685 695 705 715 725PHOTON ENERGY (eV)Figure 41. Total electron yield measurement of the absorption at the F K edge for aBaF2 thin film and powder sample. The letters indicate the photon energies used toobtain the photoelectron spectra in Fig. 47.107transition in which the final state electron is primarily on the cation (Ba) isappropriate. Following a similar procedure to the ones used earlier, the bindingenergy of the electron in the state corresponding to the lowest energy transition(A) in Fig. 41 was found to be 1.3 eV relative to the BaF2 conduction-band edge.Strontium FluorideFigure 42 shows absorption spectra in the vicinity of the F K edge for a thinfilm and powder sample of SrF2. Again these spectra are qualitatively similar to theF K edge spectra of both CaF2 and BaF2. The spectral features at higher photonenergies are not as well defined for SrF2 as they are for CaF2 or BaF2. The bindingenergy of the electron in the state corresponding to the lowest energy transition inFig. 42 was found to be 1.7 eV relative to the SrF2 conduction-band - edge.A number of electron bound states have been identified in the discussion ofthe absorption edges of CaF2, SrF2 and BaF2. The estimated binding energies relativeto the conduction-band maxima for these bound states are summarized in Table V.3.2.3 Deexcitation Electron SpectroscopyX-ray absorption spectra in the vicinity of absorption edges are known to berich in chemical and structural information if they can be interpreted. Inconjunction with the experiments on the alkaline-earth fluoride/GaAs interfaces,we showed, in section 3.2.2, that aspects of the electronic structure of low lyingexcited states of the overlayer films can be studied with x-ray absorptionspectroscopy. In this section we show, with reference to CaF2 and BaF2, thatdeexcitation spectra which result from resonant excitation of individual absorptionlines provide new information that is useful in the interpretation of absorptionspectra in solids.IPHOTON ENERGY (eV)Figure 42. Total electron yield measurement of the absorption at the F K edge for a108685 695 705 715 725SrF2 thin film and powder sample.109Table VSummary of the binding energies of the various excitons for the alkaline-earthfluorides determined in this study.Core Hole CaF2 SrF2 BaF2This work Other This work This workFluorineis 1.7 1.7 1,32p 1.1Cation2P3/2 3.62P1/2 3.63d512 3,93d,2 3,74d52 4,91103.2.3.1 The Calcium L Edge in CaF2In Fig. 43 photoemission spectra for CaF2 excited at photon energiescorresponding to points A, B, C, D and E in Fig. 36 are shown. In this experiment thesample is excited with monochromatic radiation and the energy distribution of thephotoelectrons is measured. In addition to the Ca 3s, 3p and F 2s core levels thesephotoemission spectra also show the Ga 3d core level from the substrate. The GaAsis important because it provides the Ga 3d photoemission line which acts as aninternal intensity reference for the CaF2 deexcitation spectra. Also shown inFig. 43 is the Auger spectrum for CaF2 excited by broad spectrum radiation centeredat 400eV.The decay of the core-to-bound-state transitions occurs predominantly (>99%)via a radiationless Auger-type two-electron process. If the electron which wasexcited into the bound state in the original absorption event, takes part in thetransition, the transition is called participator decay. If on the other hand theinitially excited electron does not take part in the decay but rather stays around as aspectator, it is referred to as a spectator decay.45 The participator decay leads tosingle hole final states, which are degenerate in energy with first orderphotoemission final states. The spectator decay leads to two-hole, one-electron finalstates.The Ca 3s and 3p photoemission lines in spectra C and E in Fig. 43 are clearlyenhanced relative to the F, As and Ga photoemission lines. This enhancement isattributed to a participator process. Since the final state in the participator processis identical to a photoemission final state, the participator lines are degenerate withnormal photoemission lines. The F 2s photoemission line shows no participatorenhancement, as can be seen with reference to the Ga photoemission line. Thisconfirms that the excitation is localized at the Ca site.111Cl)•1-CD00>-I—C,)zwIz240 260 280 300 320 340KINETIC ENERGY (eV)Figure 43, Photoemission spectra produced by excitation at different energies in theCa L edge, corresponding to the points marked with letters in Fig. 36. The spectraare spaced vertically according to the excitation energy, and are normalized to theGa 3d photoemission peak. The three vertical bars below spectrum C correspond tothe relative energy of the 1S, 1D, and 3P configurations of Ca V (3s23p4).I I I I I I I I I I I I I I IEDC2.71.4349.58.57.9IIBA112The structure in spectra C and E in Fig. 43, in the vicinity of the main peak inthe CaF2 Auger spectrum, is associated with a spectator process. In this process, thebound 3d electron remains present to screen the two 3p core holes in the final state.This screening lowers the final state energy so that more kinetic energy isavailable to the emitted electron. This explains why the spectator lines are on thehigh-kinetic-energy side of the normal Auger emission. The kinetic energy of thesharp spectator lines in spectra C and E differ in energy by approximately 3 eV,resulting from the 3.2 eV difference in the excitation energy. In spectrum E thehigher energy 3=1/2 2p5 state is excited while in spectrum C, it is the lower energy3=3/2 state that is primarily excited.The broader spectral features in spectra C and F, at 283 eV and 285 eVrespectively, can be attributed to a shake-off process in which the 3d spectator isionized into the conduction band during the Auger decay. The peak is broadenedcompared with the spectator lines by the continuum of final states accessible to theionized spectator. Also the transition in which the final state is a bare,four-times-ionized Ca ion, may have a larger phonon broadening than thetransition in which the bound 3d spectator electron is present to screen the two 3pcore holes in the final state. The normal Auger spectrum is broadened in addition,by the spin-orbit splitting of the 2p core hole since the normal Auger is asuperposition of spectra due to the J=1/2 and J=312 core hole decays. With thesebroadening mechanisms removed or reduced as they are in the spectator Augerdecay, the experimental data suggests that the Coulomb and exchange splitting ofthe two 3p core holes can be resolved. Neglecting the spin-orbit splitting, the 3p4configuration should have three energy levels, corresponding to S, 1D andstates. Taking the spacing of these three lines from the 3p4 configuration of atomicCa V,68 one is able to match three spectral features in the spectator spectrum, asindicated next to spectrum C in Fig. 43. In making this comparison it is assumed113that the only effect of the 3d spectator electron is to shift the overall spectrum (thespectator model) and accordingly the energy scale has been shifted to match theobserved spectrum. Additional contributions to the low kinetic energy part of thespectator spectrum could also come from “shake-up” transitions in which thespectator electron is left in a higher energy bound state such as a 4d level after thetransition.69’7°The two lines between 260 eV and 275 eV kinetic energy in Fig. 43correspond to Auger decay processes with a Ca 3s1p5 final state. The sharpness ofthese lines in spectrum C in particular suggests some spectator contribution inthese Auger decays as well. The apparent broadening of these lines in spectrum Ecan be attributed to the decay of the two 2p angular momentum substates, since bothare excited in spectrum E.3.2.3.2 The Barium M and N Edges in BaF2Barium M EdgeFigure 44 shows spectra of the M4,5N Auger region for BaF2 excited atphoton energies corresponding to the transitions, labeled A and B in Fig. 37. TheAuger spectrum for BaF2 excited with broad spectrum radiation centered at 840 eVis also shown. This normal Auger spectrum is dominated by the transition wherethe initial state consists of one core hole in the 3d level of the Ba2 ion and the finalstate has two core holes in the 4d level. The 12.6 eV splitting between the two broadpeaks is mostly due to the difference in energy of the two angular momentum statesof the 3d level that is excited. Besides the phonon broadening of a four times ionizedBa V species, the peaks are further broadened by the Coulomb and exchangesplitting of the 4d8 final state configurations. These splittings are not resolved inthe normal Auger spectrum.I114550 560 570 580 590 600 610 620KINETIC ENERGY (eV)Figure 44. Photoemission spectra of the M45N,5Auger region for BaF2 excitedat photon energies corresponding to points A and B in Fig. 37.115In spectrum A, the lower-binding-energy Ba 3d5/2 level which is excited inspectrum A, and only the features of the M5N4,Auger transition are observed.The structure observed at 5 84-604 eV in spectrum A in Fig. 44 is associated with aspectator process similar to the one observed for CaF2 near the LMM Augertransition. In this process, the bound 4f electron remains present to screen the two4d core holes in the final state. This screening lowers the final-state energy so thatmore kinetic energy is available to the emitted electron, explaining why thespectator lines are on the high-kinetic-energy side of the normal Auger emission.The broader spectral feature in spectrum A at 574 eV can be attributed to ashake-off process in which the 4f spectator is ionized into the conduction bandduring the Auger decay. The same effect was observed for CaF2 in conjunction withthe spectator associated with the LMM Auger decay. The peak is broadenedcompared with spectator lines by the continuum of final states accessible to theionized spectator. To determine the origin of the sharp spectator lines one can drawupon previous work71 on the M4,5N Auger spectrum of atomic Ba wherethree main spectral features are also observed. These features are associated withthe 3d512 core hole initial state. The energy splitting between the normal Augerpeak and the strongest of the three spectator lines is 16.2 eV, which provides anestimate for the Coulomb screening energy for the spectator electron.It is the higher binding energy Ba 3d2 state which is excited in spectrum B.However, the band gap of BaF2 (11.0 eV) is smaller than the spin-orbit splitting ofthe 3d levels (15.2 eV) which provides a mechanism for the 3d2 core hole to decayinto a 3d52 core hole. Therefore, one would expect to see a superposition of spectradue to the J=3f2 and J=512 core-hole decays as one obtains with broad bandnonresonant radiation. This is what is observed in Fig. 44. The spectator linesassociated with the 3d512 core hole are now centered around 605 eV.116Barium N EdgeShown in Fig. 45 are photoemission spectra for BaF2 excited at photonenergies corresponding to points A-E in Fig. 39. Also shown in Fig. 45 is theconventional Auger spectrum for BaF2 excited by broad spectrum radiation centeredat 115 eV. The photoemission lines of the Ba 5s, F 2s and the spin-orbit split Ba 5pcore levels are labeled in the figure. In the nonresonant cases, illustrated inspectra B, D and E, the photoemission intensity ratio of the P3/2 to the P1/2component of the Ba 5p core level is close to the statistical value of two. In theresonant cases, illustrated in spectra A and C, the intensity of the P1/2 component ofthe Ba5p core level is clearly enhanced relative to the P3/2 component as well asthe other photoemission features in the spectra. This enhancement can beinterpreted as a participator process where the electron that made the transition inthe original absorption event is also involved in the nonradiative decay of theexcited state. Spectrum C corresponds to an absorption final state which in jjcoupling corresponds to 45.8%d3/2f5+ 30.5% d5/2f7÷ 23.7%d512f. Theorigin of this differential enhancement of one of the J substates over the otherreflects a higher probability of the decay channel leaving a core hole in the 5P1/2state than leaving one in the 5P3/2 state, however the origin of this effect is notunderstood.3.2.3.3 The Fluorine K EdgesCalcium FluorideIn light of the spectator-participator processes observed in connection withthe Ca core-hole decay one can look for similar features at the F K edge. The excitoninterpretation for the first peak (A in Fig. 40) is supported by the deexcitationspectrum shown in Fig. 46 obtained by direct excitation of this absorptionresonance. All the spectra in Fig. 46 show a clear KLL Auger band near 650 eV117Figure 45. Photoemission spectra for BaF2 at photon energies corresponding toI I I I I I•I I I I I I I I I I I I I I I I I IAuger95.94.93.91.590.Co-.-.CD00>-F-Cl)zwFz20 40I I I I30 50A60KINETIC ENERGY (eV)70 80points A-E in Fig. 39.118C,)4-’D0C.)>-FC,)zU]I—z610 620 630 640 650KINETIC ENERGY (eV)Figure 46. Photoemission spectra for CaF2 obtained by excitation at differentenergies in the F K edge as indicated by the letters in Fig. 40. The spectra arespaced vertically according to the excitation energy and are normalized to the samemaximum peak height.‘ I ‘ I 1 I J I— —‘ IAugerD .6Ca3pCB-5A.1L I I I660 670119kinetic energy. However only the first spectrum, A, corresponding to the firstabsorption peak in Fig. 40, shows a distinct spectator line as would be expected ifthis were the only transition that formed a bound state. The next peak, B in Fig. 40,is believed to be a peak in the density of states corresponding to the bottom of theconduction band.72 The higher energy spectral features are band structure related.In Fig. 46 and in other similar data there is no evidence for enhancement in the Fphotoemission lines that could be attributed to participator processes. This result isconsistent with the interatomic nature of the F K edge. The hole on the F ion has asmall overlap with the excited electron on the neighboring Ca ions and hence theprobability of a participator decay is small. The other small features in Fig. 46between 620 eV and 640 eV are F Auger processes in which the final state has theF 2s1p5 configuration rather than 2sp4.Barium FluorideThe exciton interpretation for the first peak (A in Fig. 41) is supported by thedeexcitation spectra of Fig. 47 obtained by direct excitation of this absorptionresonance. As in Fig. 46, all the spectra show a clear KLL Auger band near 650 eVkinetic energy. The first spectrum, A, corresponding to the first absorption peakclearly shows some additional intensity on the high-kinetic-energy side of the mainAuger band that one might associate with a spectator process. Although not as wellresolved as in CaF2, the structure is consistent with the interpretation of the firstabsorption peak in Fig. 41 corresponding to a core-to-bound-state transition.Cl)-I-’CD00>-HCl)zwHzKINETIC ENERGY (eV)120Figure 47. Photoemission spectra for BaF2 obtained by excitation at different610 620 630 640 650 660 670energies in the F K edge as indicated by the letters in Fig. 41.1213.3 RARE-EARTH TRIFLUORIDES3.3.1 Resonant PhotoemissionDetermination of the band alignment for the rare-earth trifluorides onsilicon is complicated by the overlap between the photoemission signal from thepartially filled 4f orbitals with the photoemission signal from the F 2p valenceband. It is possible, however, to greatly enhance the photoemission signal from the4f electrons by resonantly exciting the giant 4d—* 4f atomic transition and therebydistinguish it from the photoemission signal from the overlapping F 2p levels.As discussed earlier, the giant absorption resonance that exists at the BaN edge is also present for the rare-earth elements. This broad feature above the 4dionization threshold was attributed to a collective excitation. One of the decaychannels for the deexcitation of this excited state involves a process which leaves acore hole in the cation 4f levels, identical to the final state in the normalphotoemission process, which enhances the resulting cation 4f photoemission. Inthis manner the cation 4f levels can be enhanced relative to the F 2p levels.Figure 48 shows photoemission spectra in the valence-band region for a thinfilm of NdF3 on Si (111), approximately 20 A thick annealed to 460°C. Thephotoemission spectra were taken at two different photon energies correspondingto resonant (A) and nonresonant (B) excitation. These spectra were normalized tothe radiation flux incident on the sample so as to correctly represent the relativephotoemission intensity. The spectrum (A-B) was obtained by subtracting thenouresonant spectrum from the resonant one. In taking this difference, acorrection factor that describes the decrease in photoionization cross section of theF 2p level with increasing photon energy5° was included in the nonresonantspectrum to take into account the relatively small change in the F 2p cross sectionin going from the nonresonant to the resonant excitation energy. The inset ofU)4-.C0C)>-I—C’)zwI—z0122BINDING ENERGY (eV)Figure. 48. Photoemission spectra of the valence-band region of NdF3 on Si taken attwo different photon energies corresponding to resonant (A) and nonresonant (B)excitation. The difference spectrum (A-B) is also shown. The inset shows theabsorption spectrum of NdF3 in the region corresponding to the 4d edge.-15 -10 -5123Fig. 48 shows the absorption spectrum of the same NdF3 film in theregioncorresponding to the Nd 4d edge, which agrees well with previous measurementsalso obtained from thin film samples.34 The vertical lines represent the photonenergies used to obtain the valence-band spectra. The photoemission spectrum isclearly different in the resonant case and this is attributed to the enhancement ofthe signal from the 4f electrons on the Nd3 (4f3) cations. The difference spectrum(A-B) shows that the full width at half maximum (FWHM) of the 4f levels isapproximately 2 eV, and that they are located on the low-binding-energy side of theF 2p level just above the valence-band maximum.Photoemission spectra of the valence-band region for TmF3 taken at resonant(A) and nonresonant (B) photon energies analogous to the data in Fig. 48, areshown in Fig. 49. The excitation energies are shown on the absorption spectrum inthe inset for a 10 A film of TmF3 on Si, annealed at 410°C. This absorption spectrumagrees well with previous results.34 The difference spectrum (A-B), obtained in thesame manner as for NdF3, shows that the resonantly enhanced partof thephotoemission spectrum now extends above and below the F 2p valenceband. Thebroad feature in the difference spectrum is attributed to the filled 4f orbitalson theTm3 (4f12) cation.The interpretation of the NdF3 and TmF3 resonant photoemission spectraissupported by similar results on LaF3, the analogous material with no 4f electrons.Figure 50 shows photoemission spectra of the valence-band region forapproximately 10 A of LaF3 on Si (111) annealed to 580°C, taken with resonant (A)and nonresonant (B) excitation. The inset in Fig. 50 shows the location of theexcitation energies on the absorption spectrum of the same film annealed at alowertemperature (460°C). The difference spectrum (A-B) in Fig. 50 was obtained bysubtracting the nonresonant spectrum from the resonant one, with therelativeintensity adjusted so as to minimize the difference. The photon flux was notI-20124BINDING ENERGY (eV)Figure. 49. Photoeniission spectra of the valence-band region of TmF3 on Si taken attwo different photon energies corresponding to resonant (A) and nonresonant (B)excitation. The difference spectrum (A-B) is also shown. The inset shows theabsorption spectrum of TmF3 in the region corresponding to the 4d edge.-15 -10 -5 0:‘0CD0C)>-I—Cl)zwI—z0125BINDING ENERGY (eV)Figure. 50. Photoemission spectra of the valence-band region of LaF3 on Si taken attwo different photon energies corresponding to resonant (A) and nonresonant (B)excitation. The difference spectrum (A-B) is also shown. The inset shows theabsorption spectrum of LaF3 in the region corresponding to the 4d edge.-15 -10 -5126recorded making normalization to the incident flux impossible for spectra (A) and(B). The difference spectrum, like the resonantly excited spectrum (B), shows nonew features from 4f electrons, as one would expect because of the absence of felectrons in LaF3. The spectral feature on the low-binding-energy side of the F 2pband is a surface or interface state that is also observed in other thin film fluorideson semiconductors. For CaF2 deposited on GaAs, it was explained as an CaAs bondingorbital, the possibility exists that this feature arises from an analogous La-Sibonding orbital.In a previous x-ray photoelectron spectroscopy study73 of the valence-bandregion of the rare-earth trifluorides, the photoemission spectrum of the 4felectrons was interpreted in terms of the excited state distribution of the 4felectrons for the element immediately to the left in the periodic table. A prominentsplitting in the photoemission spectrum was identified as an exchange splittingassociated with whether the photoelectron is removed from a state with a spinparallel or anti-parallel to the majority spin of the 4f electrons. This splitting wasshown to be particularly strong for the rare-earths whose 4f orbitals are just overhalf-filled due to the strong tendency for the spins of the 4f electrons to align tomaximize the exchange energy lowering. The approximately 5 eV splittingbetween the two broad peaks observed in the 4f photoemission for TmF3 (spectrumA-B in Fig. 49) can be interpreted as this exchange interaction. In other words ifthe total spin of the electrons in the 4f ground state of Tm3+ is S = I then after thephotoemission event the final state will either have S = 1/2 or S = 3/2, with thelatter state having a lower energy than the former due to the exchange interaction.3.3.2 Band OffsetsNow that one can distinguish the F 2p valence band from the localized 4forbitals of the rare-earth cation in photoemission, the alignment of these energy127levels relative to the valence band of silicon at the silicon interface can bedetermined. Figure 51 shows the valence-band spectra used to infer the bandalignments for the three rare-earth trifluoride/silicon interfaces considered here.The F 2p band edges can be readily identified in the nonresonant spectra shown inFig. 51. For the LaP3 and TmF3 samples, the overlayers were thin enough to identifythe Si valence-band maximum in the same spectra. The valence-band edges of thefluoride overlayers and the Si substrate are determined by a simple linearextrapolation of the high-kinetic-energy side of the respective valence-bandphotoemission spectra to the background. The valence-band offsets weredetermined to be 6.75 ± 0.15 eV for LaF3 and 7.0 ± 0.2 eV for TmF3. Both sampleswere approximately 10 A thick and annealed to 410°C. For the thicker NdF3 sample(20 A thick annealed to 460°C) the Si valence-band photoemission signal is weakand its energy can only be determined to an energy equal to the Si band gap. Inthis case the valence-band offset was 7.0 ± 0.6 eV. The valence-band offsetsdescribed above for the rare-earth trifluorides can be compared with values7.3-8.3 eV determined for the CaF2/Si (111) interface by Olmstead et al.29 Theresults of these valence-band offset measurements and the position of the occupied4f orbitals are summarized schematically in Fig. 52. In this figure all therare-earth trifluorides have been assumed to have the same 10 eV band gap.7Within the experimental uncertainty of these measurements, the band offsets arealso the same for the three rare-earth trifluoride/silicon heterojunctions studiedhere. Fig. 52 also shows that the binding energy of the 4f electrons increases withincreasing atomic number and that the width of the distribution of 4f levelsincreases with increasing number of f electrons.For NdF3 and TmF3, it is apparent that there is a mechanism by which the Sivalence-band electrons are prevented from occupying the empty 4f levels. Thislack of electron transfer can be explained by an on-site Coulomb repulsion energy.128-8 -4BINDING ENERGY (eV)Figure. 51. Valence-band spectra for the three rare-earth trifluoride/siliconinterfaces used to determine the valence-band offsets shown in Fig. 52.-16 -12 0129(a) (b) (c)LaF3 Si NdF3 Si TmF3Figure. 52. Measured band alignments for the rare-earth trifluoride/Si (111)interfaces showing the relative positions of the F 2p valence band and the cation 4flevels.Si6.75 ± 0.15GB MINVB MAX7.0 ±0.6 7.0 ± 0.2130This repulsion is described by a parameter U, which is the Coulomb energy requiredto put an extra 4f electron on one of the cations. By comparing the measured bandoffsets to the values of the on-site Coulomb repulsion parameter U = 6.5 eVmeasured for Nd and Tm metals,75 one can see that the lowest lying state with anextra 4f electron lies above the Si valence-band maximum for both NdF3 and TmF3.This prevents any charge transfer from the Si valence band to the rare-earthcation at the interface even though there are many excited states of the 4f electronsthat lie below the occupied silicon valence band states.76 These results suggest thatthe lowest energy level with one extra 4f electron lies close to the silicon band gapbut the measurements, and probably also the estimate for U, are not accurateenough to determine whether the state lies above or below the bottom of the siliconconduction band. If the state is below the bottom of the silicon conduction bandthen it should be possible to populate it electrically by charge injection from thesilicon.1314. SUMMARY AND CONCLUSIONSPhotoelectron experiments using synchrotron radiation have beenperformed on a variety of semiconductor heterostructures. The electronicproperties and composition of the interfaces between selected wide band gapmaterials and GaAs have been studied. Three types of wide band gap materials wereconsidered namely chalcogen compounds, alkaline-earth fluorides, and rare-earthtrifluorides. Prior to the growth of thin surface layers, the semiconductorsubstrates had to be cleaned on an atomic scale. The final step of the cleaningprocess normally involved desorbing the surface oxide layer. The effect of surfaceroughening was observed for the first time as a result of optically monitoring theoxide desorption in GaAs by measuring the diffuse reflectivity of the sample surfacewith a He-Ne laser and a silicon diode detector.Clean GaAs wafers were treated with H2S to study interfacial bonding, a studymotivated by the search for suitable surface passivating layers, for GaAs-basedsemiconductor device applications. The high resolution photoemission studies ofthe clean GaAs (100) surfaces treated with activated H2S at room temperature,followed by annealing at 400°C, showed that these surfaces are completelyterminated by a GaS chemical species. This surface layer, which forms throughthe removal of As atoms at the interface, experienced no major chemical changesupon exposure to air or water. The Fermi level at the passivated surface is locatedslightly above midgap at 0.85 eV above the valence-band maximum. While nocomparative data has yet been obtained to evaluate the carrier recombination rateat this surface, the photoemission results indicate that the H2S treated GaAs (100)surface is characterized in terms of the interfacial bonding better than surfacesproduced by recent Na2S or (NH4)2S chemical treatments.5’68 Recent papers7132have appeared in the literature demonstrating the continued interest in GaAssurface passivation with sulfur.Photoemission spectroscopy was used to determine the valence-band offsetand composition at the ZnSeIGaAs (100) interface. On the basis of an analysis ofchemical shifts and relative intensities of the individual atomic core levels, weobtained the first direct measurements of the interfacial composition.51 Withincreasing annealing temperature, an interfacial Ga2Se3 layer forms above about400°C with loss of Zn and As at the interface. The determination of the interfaciallayer being Ga2Se3 confirmed the suggestions from earlier work.16’7 Thethickness of the interfacial layer as a function of annealing temperature was alsodetermined. Recent studies on the electronic properties of GaAs/ZnSeheterojunctions, show improved interface properties for ZaSe grown on Ga-richsurfaces,15 presumably because of the stable Ga-Se interface layer which forms.One could speculate that ZnSe/GaAs interfaces with high electronic quality, haveinterfacial Ga2Se3 layers, which passivate the interface in an analogous way to thesulfur passivation discussed earlier. The first direct valence-band photoemissionmeasurements for the ZnSe/GaAs interface yielded a band offset of 1.25 ± 0.07 eV.5Similar measurements for the ZnSe/InP (111) interface yielded a band offset of1.35 ± 0.07 eV.The interfacial bonding, energy band alignment and Fermi level positionwere measured for the alkaline-earth fluoride/GaAs interfaces as a function ofannealing temperature and overlayer thickness by high resolution photoemissionspectroscopy. Thin films of these fluorides were deposited in UHV by evaporationon clean GaAs (100), and subsequently annealed in steps, in the range 370-620°C. At550°C for deposited CaF2 layers less than a few monolayers thick, a monolayer of Careacts with As at the interface with an associated loss of approximately one Ga andone F by evaporation per formula unit of CaF2 at the interface. This was the first133report of Ga and F loss at the interface.54 The interfacial layer does not growbeyond a monolayer even when thicker films are annealed up to 650°C. Theseresults are based on an analysis of the intensity and chemical shifts of the shallowcore levels in Ga, As, Ca and F. An interfacial structure model in which Ca replacesthe first layer of Ga was found to be consistent with the data. For both the SrF2!GaAsand BaF2/G As interfaces, results supporting cation-As bonding accompanied by theloss of Ga and F were obtained. Of further interest in this area would be toaccurately lattice match the alkaline-earth fluorides to the semiconductor byevaporating alloys of the type CaSri ..F2 and SrB aiF2. This would most likely beachieved by a two source evaporation method in an MBE system.The CaF2/G As band offset was found to increase with layer thickness to amaximum of 8.5 eV for thick CaF2 layers. The band offset variation for SrF2 andB aF2 followed a similar trend. The Fermi level moves to the top of the GaAs valenceband for thick CaF2 overlayers in an analogous way to the Fermi level in CaF2 onsilicon.23 Whatever their origin, the presence of states that pull the Fermi leveldown to the top of the valence band suggests that CaF2 is not a promising materialfor high quality electronic heterojunctions such as would be required forGaAs/CaF2 based, MIS type devices,78 unless the states which pull the Fermi leveldown can be eliminated.X-ray absorption spectroscopy and deexcitation electron spectroscopy (DES)were used to investigate the nature of the electronic transitions near the Ca L edgein CaP2, the Ba M and N edges in BaF2 and the F K edge in CaF2, SrF2 and BaF2. Thecation absorption edges were interpreted in terms of atomic multiplet theory andbound excited states were identified for these cation edges as well as the F K edges ofCaF2 and BaF2. The binding energies for these core excitons were estimated for thefirst time.79 By resonantly exciting these bound states and measuring the resultingdeexcitation spectra (DES) the nature of the excited-state electrons was determined.134The presence of sharp spectator lines near the Auger region of the spectrumshowed which absorption transitions had bound final states. The participator lines,which appear as an enhancement of the normal photoemission lines, provideinformation on the atomic location of the bound-state electrons. Thus thecombination of high resolution absorption and deexcitation spectra have beenshown to be a powerful tool to investigate the nature of core electron excited statesin ionic compounds. For the Ba N edge, the giant absorption resonance located10-20 eV above the ionization threshold was described classically in terms of aplasma-like collective effect.Thin films of the rare-earth trifluorides (LaF3, NdF3, TmF3) grown on cleanSi (111) surfaces were also studied. The interfacial energy level alignments forthese heterostructures were determined. By resonant excitation of the giant 4d—* 4 ftransition in NIF3 and TmF3, the photoemission signal of the localized 4f electronswas distinguished from the photoemission signal of the overlapping F 2p levels.For NdF3 and TmF3, a mechanism by which the Si valence band electrons areprevented from occupying the empty 4f levels was described for the first time.8 0This lack of electron transfer was explained by an on-site Coulomb repulsionenergy. This repulsion is described by a parameter U, which is the Coulomb energyrequired to put an extra 4f electron on one of the cations. By comparing themeasured band offsets to the values of the on-site Coulomb repulsion parameterU = 6.5 eV measured for Nd and Tm metals,75 we found that the lowest lying statewith an extra 4f electron lies above the Si valence-band maximum for both NdF3 andT m F.This prevents any charge transfer from the Si valence band to therare-earth cation at the interface even though there are many excited states of the4f electrons that lie below the occupied silicon valence band states.7 6135REFERENCES1. W.E. Spicer, P.W. Chye, P.R. Skeath, C.Y. Su, and 1. Linclau, J. Vac. Sci. Technol. 16,1422 (1979).2. S. Sinharoy, P.G. McMullin, J. Greggi Jr., and Y.F. Lin, I. Appi. Phys. 62, 875 (1987);R.A. Hoffman, S. Sinharoy, and R.F.C. Farrow, Appi. Phys. Lett. 47, 1068 (1985).3. J.M. Phillips, L.C. Feldman, J.M. Gibson, and M.L. McDonald, Thin Solid Films 107,217 (1983).4. S. Sinharoy, R.A. Hoffman, J.H. Rieger, W.J. Takei, and R.F.C. Farrow, J. Vac. Sci.Technol. B 3, 722 (1985).5. C.J. Sandroff, R.N. Nottenburg, J.-C. Bischoff, and R. Bhat, Appi. Phys. Lett. 51, 33(1987).6. M.S. Carpenter, M.R. Melloch, M.S. Lundstrom, and S.P. Tobin, Appi. Phys. Lett. 52,2157 (1988).7. C.J. Sandroff, M.S. Hegde, L.A. Farrow, CC. Chang, and J.P. Harbison, Appl. Phys.Lett. 54, 362 (1989).8. T. Tiedje, P.C. Wong, K.A.R. Mitchell, W. Eberhardt, Z. Fu, and D. Sondericker, SolidState Commun. 70, 355 (1989).9. B.A. Cowans, Z. Dardas, W.N. Delgass, M.S. Carpenter, and M.R. Melloch, Appl. Phys.Lett. 54, 365 (1989).10. 1. Massies, J. Chaplart, M Laviron, and N.T. Linh, Appi. Phys. Lett. 38, 693 (1981).11. J. Massies, F. Dézaly, and N.T. Linh, J. Vac. Sci. Technol. 17, 1134 (1980).13612. W. Ranke, J. Finster, and H.J. Kuhr, Surface Sd. 187, 112 (1987).13. D.J. Olego, Phys. Rev. B 39, 12743 (1989).14. M.C. Tamargo, J.L. de Miguel, D.M. Hwang, and H.H. Farrell, J. Vac. Sci. Technol. B6, 784 (1988).15. J. Qiu, Q.-D. Qian, R.L. Gunshor, M. Kobayashi, D.R. Menke, D. Li, and N. Otsuka,Appi. Phys. Lett. 56, 1272 (1990).16. D. Li, J.M. Gonsalves, N. Otsuka, J. Qiu, M. Kobayashi, and R.L. Gunshor, Appi. Phys.Lett. 57, 449• (1990).17. D.-W. Tu and A. Kahn, J. Vac. Sci. Technol. A 3, 922 (1985).18. S.P. Kowalczyk, E.A. Kraut, J.R; Waidrop, and R.W. Grant, J. Vac. Sci. Technol. 21,482 (1982); K. Mazuruk, M. Benzaquen, D. Walsh, B. Makuc, H. Jayatirtha, and H.Aharoni, Can. J. Phys. 67, 339 (1989).19. D.-W. Tu and A. Kahn, 3. Vac. Sci. Technol. A 2, 511 (1984).20. R.W.G. Wyckoff, Crystal Structures 2nd ed. Vol. 1 & 2. (John Wiley & Sons, NewYork, 1963 & 1964).21. F.J. Himpsel, F.U. Hillebrecht, G. Hughes, J.L. Jordan, U.O. Karisson, F.R. McFeely,J.F. Morar, and D. Rieger, Appi. Phys. LeLt. 48, 596 (1986).22. M.A. Olmstead, R.1.G. Uhrberg, R.D. Bringans, and R.Z. Bachrach, I. Vac. Sci.Technol. B 4, 1123 (1986).13723. D. Rieger, F.J. Himpsel, U.O. Karisson, F.R. McFeely, J.F. Morar, and J.A. Yarmoff,Phys. Rev. B 34, 7295 (1986).24. S. Siskos, C. Fontaine, and A. Munoz-Yague, J. Appi. Phys. 56, 1642 (1984).25. S. Sinharoy, R.A. Hoffman, J.H. Rieger, R.F.C. Farrow, and A.J. Noreika, J. Vac, Sci.Technol. A 3, 842 (1985).26. Y. Yamada, M. Oshima, S. Maeyama, T. Kawamura, and T. Miyahara, Appi. Surf. Sd.33134, 1073 (1988); D. Mao, K. Young, A. Kahn, R. Zanoni, J. McKinley, and G.Margaritondo, Phys. Rev. B 39, 12735 (1989).27. Y. Yamada, M. Oshima, T. Waho, T. Kawamura, S. Maeyama, and T. Miyahara,Japan. J. of Appi. Phys. 27, L1196 (1988); Y. Yamada Maruo, M. Oshima, T. Waho, T.Kawamura, S. Maeyama, and T. Miyahara, Appi. Surf. Sci. 41/42, 647 (1989).28. Y. Yamada Maruo, M. Oshima, T. Waho, and T. Kawamura, Japan. J. Appi. Phys. 28,L299 (1989); T. Waho, F. Yanagawa, and Y. Yamada, J. Crys. Growth, 95, 415 (1989); Y.Yamada Maruo, M. Oshima, T. Waho, T. Kawamura, S. Maeyama, and T. Miyahara,Appi. Surf. Sci. 41/42, 647 (1989).29, M.A. Olmstead, R.I.G. Uhrberg, R.D. Bringans, and R.Z. Bachrach, Phys. Rev. B 35,7526 (1987).30. M.A. Olmstead and R.D. Bringans, Phys. Rev. B 41, 8420 (1990).31. P.G. McMullin and S. Sinharoy, J. Vac. Sci. Technol. A 6, 1367 (1988).32. H.D. Muller, J. Schneider, H. Lüth, and R. Strumpler, Appl. Phys. Lett. 57, 2422(1990).33. G.W. Rubloff, Phys. Rev. B 5, 662 (1972).13834. C.G. Olson and D.W. Lynch, J. Opt. Soc. Am. 72, 88 (1982).35, T. Van Buuren, M.K. Weilmeier, I. Athwal, K.M. Colbow, J.A. MacKenzie, T. Tiedje,P.C. Wong, and K.A.R. Mitchell, Appi. Phys. Lett. 59, 464 (1991).36. A.J. SpringThorpe and P. Mandeville, J. Vac. Sci. Technol. B 6, 754 (1988).37. M. Sansone, R. Hewitt, W. Eberhardt, and D. Sondericker, Nuci. Instrum. MethodsA266, 422 (1988).38. V. Leon and J. Carrazza, Rev. Tech. INTEVEP 9, 81 (1989).39. C.J. Powell, J. Electron Spectrosc. 47, 197 (1988).40. C.J. Powell (private communication),41. D.R. Penn, Phys. Rev. B 35, 482 (1987).42. S. Tanuma, C.J. Powell, and D.R. Penn, J. Vac. Sci. Technol. A 8, 2213 (1990).43. F.L. Battye, J. Liesegang, R.C.G. Leckey, and J.G. Jenldn, Phys. Rev. B 13, 2646(1976).44. D.R. Penn, I. Electron Spectrose. 9, 29 (1976).45. W. Eberhardt, E.W. Plummer, C.T. Chen, and W.K. Ford, Aust. J. Phys. 39, 853,(1986).46. F. Sette, B. Sinkovic, Y.J. Ma, and C.T. Chen, Phys. Rev. B 39, 11125 (1989).13947. F.M.F. de Groot, J.C. Fuggle, B.T. Thole, and G.A. Sawatzky, Phys. Rev. B 41, 928(1990); F.J. Himpsel, U.O. Karisson, A.B. McLean, L.J. Terminello, F.M.F. de Groot, M.Abbate, J.C. Fuggle, J.A. Yarmoff, B.T. Thole, and G.A. Sawatzky, Phys. Rev. B 43, 6899(1991).48. R. Ludeke, T.C. Chiang, and D.E. Eastman, Physica 117B-11SB, 819 (1983).49. K.C. Mills, Thermodynamic Data for Inorganic Suiphides, Selenides and Tellurides(Butterworth, London, 1974).50. J.J. Yeh and I. Lindau, At. Data Nucl. Data Tables 32, 1 (1985).51. K.M. Colbow, Y. Gao, T. Tiedje, J.R. Dahn, and W.E. Eberhardt, J. Vac. Sci. Technol. A9, 2614(1991).52. T.A. Carison, Photoelectron and Auger Spectroscopy (Plenum Press, New York,1975).53. W.A. Harrison, Electronic Structure and the Properties of Solids: The Physics ofthe Chemical Bond (Dover Publications, New York, 1989).54. K.M. Colbow, T. Tiedje, D. Rogers, and W.E. Eberhardt, Phys. Rev. B 43, 9672 (1991).55. J. Massies, P. Devoldre, and N.T. Linh, 3. Vac. Sci. Technol. 15, 1353 (1978).56. E.U. Condon and G.H. Shortley, The Theory of Atomic Spectra (CambridgeUniversity Press, Cambridge, 1959).57. J. Zaanen, G.A. Sawatzky, J. Fink, W. Speier, and J.C. Fuggle, Phys. Rev. B 32, 4905(1985).14058. F.J. Himpsel, U.O. Karisson, J.F. Morar, D. Rieger, and J.A. Yarmoff, Phys. Rev. Lett.56, 1497 (1986); C.T. Chen and F. Sette, Phys. Rev. Lett. 60, 160 (1988); F.J. Himpsel,U.O. Karisson, J.F. Morar, D. Rieger, and J.A. Yarnioff, Phys. Rev. Lett. 60, 161 (1988).59. J. Barth, F. Gericen, and C. Kunz, Phys. Rev. B 28, 3608 (1983).60. R.A. Heaton and C.C. Lin, Phys. Rev. B 22, 3629 (1980).61. M. Cardona and L. Ley, Eds., Photoemission in Solids I: General Principles(Springer-Verlag, Berlin, 1978).62. P. Motais, E. Belin, and C. Bonnelle, Phys. Rev. B 30, 4399 (1984).63. J. Sugar, Phys. Rev. B 5, 1785 (1972).64. T.B. Lucatorto, T.J. Mcllrath, J. Sugar, and S.M. Younger, Phys. Rev. Lett. 47, 1124(1981),65. A. Kotam and Y. Toyozawa, Synchrotron Radiation, ed. C. Kunz (Springer-Verlag,Heidelberg, Berlin, New York, 1979).66. G. Wendin, Vacuum UV Radiation Physics, ed. C. Kunz, R. Haensel, and E.E. Koch(Vieweg/Pergamon, 1974) p. 225.67. J.P. Connerade, J.M. Esteva, and R.C. Karnatak, Eds., Giant Resonances in Atoms,Molecules, and Solids (NATO ASI Series B, No.151, Plenum Press, New York, 1987).68. E. Biémont, Phys. Scripta 34, 318, (1986).69. H. Aksela, S. Aksela, H. Pulkkinen, G.M. Bancroft, and K.H. Tan, Phys. Rev. A 37,1798 (1988),14170. M. Meyer, E. v. Raven, M. Richter, B. Sonntag, R.D. Cowan, and J.E. Hansen, Phys.Rev. A 39, 4319 (1989).71. M. Kellokumpu and H. Aksela, Phys. Rev. A 31, 777 (1985).72. Y. Gao (private communication).73. G.K. Wertheim, A. Rosencwaig, R.L. Cohen, and H.J. Guggenheim, Phys. Rev. Lett.27, 505 (1971).74. C.G. Olson, M. Piacentini, and D.W. Lynch, Phys. Rev. B 18, 5740 (1978).75. J.F. Herbst, R.E. Watson, and J.W. Wilkins, Phys. Rev. B 17, 3089 (1978).76. G.H. Dieke, Spectra and Energy Levels of Rare Earth Ions in Crystals, (John Wiley& Sons, New York, 1968).77. M.S. Carpenter, M.R. Melloch, B.A. Cowans, Z. Dardas, and W.N. Delgass, J. Vac. Sci.Technol. B 7, 845 (1989); C.J. Sandroff, M.S. Hegde, and C.C. Chang, J. Vac. Sci.Technol. B 7, 841 (1989); J. Shin, K.M. Geib, C.W. Wilmsen, and Z. Lilliental-Weber, J.Vac. Sci. Technol. A 8, 1894 (1989).78. T. Waho and F. Yanagawa, IEEE Electron Device Letters 9, 548 (1988).79. T. Tiedje, K.M. Colbow, D. Rogers, and W.E. Eberhardt, Phys. Rev. Lett. 65, 1243(1990).80. K.M. Colbow, Y. Gao, T. Tiedje, J.R. Dahn, J.N. Reimers, and S. Cramm, J. Vac. Sci.Technol. A (July/Aug 1992, to be published).14281. For an explanation on surface reconstructions see: D.P. Woodruff and T.A.Decichar, Modern Techniques of Surface Science, (Cambridge University Press,Cambridge, 1986).APPENDIX I Microchannnel Plate Detector300 V470 K1430.005 tFSIGNAL(TO RATEMETER)Schematic diagram of the microchannel plate detector and the accompanyingbiasing circuit.MCPI II II II II I I• I II II I II I II I II I II I II I II IMCPCOLLECTORI I —•I•II IOur2 Mc470 K 5 x 470 K 5 x 470 K 470 K1 2GRQU’JD + H.V.144APPENDIX II Chemical Shifts From Electronegativity ValuesIn section 2.3.1.1 it was demonstrated that the chemical shift is closely relatedto the electron density surrounding the atom. It is reasonable to then try to relatethe core-binding energies to some chemical criterion reflecting this density. Theuse of a calculated charge based on Pauling’s scale of electronegativity has beenoutlined by Carlson.52 From equation (6), it is apparent that a determination of thenet charge allows one to subsequently predict the magnitude of the chemical shift.This calculation is applied to the analysis of the ZnSeIGaAs interface.In a covalent solid such as the zincblende structure of GaAs and ZnSe, thecalculated charge on each atom is determined by the partial ionic character ascontributed from each of the neighboring atoms. This value is derived from theelectronegativity scale for elements and is equal toX A - XB r n ,ci 2iq = {1 - e1V”\XA - XB .i} (19)‘ZA XB’where XA and XB are the electronegativities, respectively, for the atom under studyand its neighbor. The table below shows the values for the electronegativity andcovalent radius for the elements involved.Element Electronegativity, x Covalent Radius, r (A)Zn 1.65Ga 1.81 1.26As 2.18Se 2.55 1.16Equations (19) and (6) can now be used to determine iXq and AE, respectively. Inequation (6), if Aq is given the units of electron charge, R and r are in Angstromsand e2 is 14.4, then the chemical shifts, E are given in electron volts. These results145are summarized in the tables below. The chemical shifts, AEGa O.ILd lESe are shiftsrelative to the isolated atom whereas IEGaAS and AEZnSe are the chemical shiftsrelative to GaAs and ZuSe, respectively.Ga 3d Chemical ShiftCompound R (A) AEGa (eV) AEGaAs (eV)GaAs 2.448 0.034 0.19 -GaSe 2.3 0.128 0.66 0.47Ga2Se3 2.35 0.192 1.02 0.83Se 3d Chemical ShiftCompound R, (A) Ag ‘Se (eV) AEZuSe (eV)ZnSe 2.448 -0.183 -1.20 -GaSe 2.3 -0.128 -0.79 0.41Ga2Se3 2.35 -0.085 -0.57 0.63