Journal of Electron Spectroscopy and Related Phenomena 72 (1995) 59-63
Surface Density of States of Sb/GaAs(ll0) and H:GaAs(110) by Metastable Deexcitation Spectroscopy L. Pasquali, A. Plesanovas, A. Ruocco, A.C. Tarabini and S. Nannarone. a I. Abbati t .b M. Canepa, L. Mattera and S. Terreni. c ~Dipartimento di Fisica, Universita' di Modena, Via Campi 213/a, 41100 Modena, Italy. bDipartimento di Fisica, Politecnico di Milano. CDipartimento di Fisica, Universita' di Genova. Metastable deexcitation spectroscopy (MDS) was applied to the study of the electronic structure of Sb/GaAs(110) and H:GaAs(110) systems. For both systems valuable insights into the surface electronic structure (effective surface density of states and surface charge density) were obtained. A surface to volume ratio extremely higher than photoemission was obtained for MDS. For both systems new features, never observed by more conventional surface spectroscopies, were detected.
1. I n t r o d u c t i o n The determination of surface and interface electronic structure (energy eigenvalues, their _k dispersion and charge density) is one of the main issues of spectroscopy of solid surfaces and interfaces. In present work MDS (metastable deexeitation spectroscopy) has been applied to study the systems S b / G a A s ( l l 0 ) and H:GaAs(ll0). The technique is based on an interatomic Auger type deexcitation involving the ls core hole of helium metastable atom (He*) impinging on the surface and electrons belonging to surface orbitals of the solid. Two different deexcitation processes could in principle dominate, depending on the relative values of the He" effective ionization energy, E~, in front of the surface and the solid ionization energy, ~b. Namely: i) if 4) > E,'., deexcitation occurs via resonant ionization plus Auger neutralization, RI+AN, (HelsVV); ii) if ~b < E,'-, deexcitation occurs via Auger deexcitation, AD, (HelsVHe2s). Similar deexcitation processes reduced by ions (INS, ion neutralization spectroscopy) were used in the past, . INS, however, could be affected by kinetic energy broadening, , and some sputtering. The use of neutral He* atoms takes these problems to minimum .
Both processes are driven by wave function tails superposition and consequently they give, in principle, information both on surface density of states (SDOS) and on surface charge density (SCD). It is important to stress that, in spite of its crucial role played in determining surface properties, like reconstruction or chemisorption, SCD of individual surface orbitals is up to date obtained only through theoretical calculations [4-6]. The extreme surface sensitivity of MDS and its capability in giving information on SDOS was widely demonstrated in the case of clean and adsorbate covered surfaces by a number of groups . More recently MDS was applied successfully by our group to clean GaAs(110) to get insights, comparing MD spectra with theoretical results, into SDOS and SCD . The results presented in this conference are to be intended as a natural further step in the MDS investigation of G a A s ( l l 0 ) of ref. : both S b / G a A s ( l l 0 ) and H:GaAs(110)form, at one monolayer (ML) coverage, ordered overlayers [9, 10] and reliable calculated SCD and SDOS are available [4-6]. Moreover these systems are interesting both from technological and physical point of view.
0368-2048/95 $09.50 © 1995 Elsevier Science B.V. All rights reserved SSDI 0368-2048(94)02335-2
60 2. E x p e r i m e n t a l The experiments were run in the ultra high vacuum (UHV) experimental system described in ref.  at a base pressure of-~ 5 x 10 -11 Tort. Clean surfaces were prepared by cleavage. 1 ML Sb films were obtained by evaporation from a resistively heated tungsten basket. H chemisorption was achieved by exposure of clean surfaces to atomic hydrogen, obtained by thermal dissociation of H2 molecules by a W filament heated at a temperature of --,2000 ° K. Saturation coverage was determined by monitoring the loss rate of Ga 3d-empty surface state transition by electron energy loss spectroscopy (EELS)
. The details of metastable atoms source (based on principles developped in ref. ) used in this experiment are described in ref. . A supersonic He atoms beam is obtained by gas expansion from 7 atm to 10 -3 Tort through a nozzle of 20 ~um of diameter. Collimation is achieved by a 0.5 m m (in diameter) skimmer and a 3 m m aperture 10 cm apart from the sample surface. Excitation is obtained by an electron gun coaxial with the beam operated at an energy of 400 eV. In this way 90% of the excited helium atoms are aestimated to be excited on singlet state. The beam intensity is , ~ 1 0 7 atoms per second at the sample surface. The metastable atoms beam impinged at 45 ° with respect to the sample normal along the [1,]',0] azimuth. The electrons were detected and energy analyzed by a hemispherical analyzer (Vacuum Generators ADES 400) positioned at 45 ° with respect to the sample normal along the [1,]-,0] azimuth. The analyzer was driven in retarding mode with a constant resolution of ,-- 0.5 eV. Surface cleanliness was checked routinely by AR-UPS (angle resolved ultraviolet photoemission spectroscopy). Spectra were taken at the beginning and at the end of every MDS run. As concluded from the persistence of surface features, the surface remained clean for hours under metastable beam exposure. However MDS itself revealed as the most sensitive tool for surface cleanliness check.
o : ' s°
0 0 ]>,
%s'" %',, 0
~4Lnetic [ n e r g y
Figure 1. MD spectra of S b / G a A s ( l l 0 ) (filled dots) and H : G a A s ( l l 0 ) ( s q u a r e s ) .
3. E x p e r i m e n t a l r e s u l t s a n d d i s c u s s i o n Experimental MDS data obtained for the S b / G a A s ( l l 0 ) surface and the H : G a A S ( l l 0 ) surface are presented in Fig. 1. The spectra onsets at the high kinetic energy side were determined by a linear extrapolation after an horizontal linear background subtraction, obtaining the values of 12.04-0.2 eV and 11.3-t-0.2 eV for Sb/GaAs(110) and H:GaAs(110), respectively. These values and the spectral shapes are the indication that He* deexcitation occurs via RI+AN. In this case the maximum of kinetic energy is given by Ek,max -E ~ - 2¢; moreover, under suitable hypothesis, the shape of the experimental spectrum reflects the autoconvolution of the SDOS calculated at the distance z where the process occurs, weighted by the transition matrix element [1, 3]. In this frame an effective SDOS is obtained through a deconvolution procedure, while insights into SCD come mainly from a comparison with theoretical results. In our case, after spline filtering, the
61 .- ..1 c, O0
(A) <, X, h,.
'.£)(} bl i q)
2b0 ,I . . . .
I .... f)
1 1 1",'
[[irl(.~rg), ' ( e V )
,% q 1() / ,\ [~r (?rL})' k ( X / )
"u~ C ©
-2o 9 '~s-7 '6-'~"4 - -5-2-r' ........
' . . . . . . . . . . .
~° "° ~'c
-9-8-7-6-5-4-5-2-1 0 Binding Energy (eV) Figure 2. S b / G a A s ( l l 0 ) : a) MD spectrum (symbols) and spline (continuous line) versus kinetic energy; b) first derivative of MD spectrum versus binding energy; c) deconvolution of MD spectrum versus binding energy. For the meaning of symbols see text.
Figure 3. H:GaAs(ll0): a) MD spectrum (symbols) and spline (continuous line) versus kinetic energy; b) first derivative of MD spectrum versus binding energy; c) deconvolution of MD spectrum versus binding energy. For the meaning of symbols see text.
effective SDOS as a function of binding energy was derived by using the deconvolution algorithm introduced by Boiziau et al.  and extensively discussed by Sesselmann et al. . The results are shown in Figs. 2 and 3. Deconvolution resuits were tested against mathematical artifacts by comparison with first derivative of experimental spectra . The comparison is made in Figs. 2 and 3, showing that any artifact was introduced by our deconvolution code. As evident from features of first derivative, two main groups of structures of effective SDOS originate from the experimental spectrum. The first
ones are related to the fine structure of the first positive slope in the kinetic energy range between 12 and 7 eV; similarly, the second ones are related to the positive slope in the interval between 6 and 2 eV. The deconvolution figure obtained for the S b / G a A s ( l l 0 ) surface (Fig. 2) presents, in good agreement with the first derivative, three main features at 1.8+0.3, 6.74-0.3 and 8.54-0.3 eV of binding energies and two shoulders at 3.54-0.4 eV and 4.74-0.4 eV, labelled P1-P5 in the order, and a minimum at 5.64-0.3 eV. P1 is the most prominent spectral feature. It
62 occurs in the region where AR-UPS [16, 17] and theoretical calculations [4, 5] locate the Ss and $5 Sb-induced states. Because of its energy position, P1 is tentatively assigned to $5, while $6 is thought to contribute to the energy tail at lower binding energy. Their prevalence is in agreement with the spatial distribution of SCD associated to these orbitals, according with the results of Bertoni et al. . Theory predicts also a higher SCD associated to $5 with respect to $6 which is consistent with present assignment. In the same framework the spectral weight in the region of P2 and P3 has to be assigned to emission from $4 and $3. Their lower weight with respect to P1 is consistent with SCD associated to these states, which charge is predicted by theory to be localized mostly in the space region between first Sb plane and substrate. Switching to Ph, both energy position and theoretical SCD conspire to assigning this structure to $2 and $1, both s-character orbitals localized on the Sb plane. It is important to to stress that these states have never been observed in UPS, probably depressed by cross section effects and masked by bulk emission. Eventually P4 could be assigned to the theoretically predicted C1 state on the base of its energy position. However C1 is, according to theory, a state which SCD is localized on the cation of the second substrate plane. This makes questionable the assignment of P1 to C1. Moreover, comparing deconvolution with first derivative, the deconvolution procedure seems to exagerate the relevance of P4. Concluding this part on Sb/GaAs(ll0), it is worth noting that the minimum found at 5.64-0.3 eV nicely fits with the energy of the middle of the stomach gap of projected bulk band structure of GaAs(ll0). Addressing the H:GaAs(110) spectrum, it is worthwhile to remark, at first, the stiking difference with respect to Sb/GaAs(ll0), as shown in Fig. 1. In particular it is to be noted that the intensity reduction in the high kinetic energy region and the It-induced contributions at -,,5.5 eV and ,-,7.5 eV of kinetic energy. Moreover the absence of a portion of spectrum with negative slope can be noticed. This fact is related to the appearence
of H-induced states in the stomach gap, as shown below. The effective SDOS obtained by deconvolution is shown in Fig. 3 together with the first derivative. Four peaks and a shoulder superimposed to the raising up background are present, in agreement with the first derivative, at 1.2+0.3, 2.5+0.3, 3.94-0.3, 5.44-0.3 and 7.14-0.3 eV of binding energy, labelled F1 through Fh. Their assignment can be done making reference to AR-UPS data of ref.  and theoretical calculations . F1 can be ascribed to surface state A6, in excellent agreement with theoretical results . This state is much more pronounced in MDS than in AR-UPS , where it gives only a weak shoulder. F2 and F3 occur at energies very close to those predicted by theory and observed by AR-UPS for the states Hs and H4, respectively . These are H-induced states which charge density is mainly concentrated towards vacuum, , in agreement with the well defined structures observed in UPS. The F4 energy position is close to those of H3 and H2 H-induced states . However, though predicted by theory, they are not visible in ARUPS . Also for these states, theory shows a SCD extending outward, justifing their clear observation in MDS. Finally, Fs does not correspond to any theoretically predicted electronic state of the H-covered GaAs(ll0) surface. In agreement with the hypothesis made in ref. , this shouder can be associated with some degree of surface disruption due to H, which is known to introduce defect states in the heteropolar gap. 4. C o n c l u s i o n s Metastable deexcitation spectroscopy was applied to the study of Sb/GaAs(ll0) at one monolayer coverage and H:GaAs(ll0) at saturation coverage. In both RI+AN was the dominant deexcitation process. The MD spectra of the two systems showed marked differences which can be related to the electronic structure (SDOS and SCD) of the two surfaces, with a surface to volume ratio much higher than UPS.
53 As far as Sb overlayer is concerned, relevant insights into bonding at the interface through a picture of effective SDOS and information on SCD were gained. The dangling bond character of $6 and $5 states was clearly confirmed, the back bond and bonding character of $4 and $3 states was put into evidence and, last but not least, for the first time the s-character $2 and $1 states were observed. Switching to H, a clear picture of the electronic structure (SDOS and SCD) of bonding of H with GaAs(110) was obtained. The agreement with theory is excellent and for the first time the states H3 and H2 were observed. Because of space constraints, only MDS d a t a followed by a brief discussion were presented. More extended presentation will be done in a forthcoming paper. Deconvolution played a significative role in getting insights into the electronic surface structure and in particular into the effective SDOS. However, being demonstrated the powerful of MDS in surface and interface problems, efforts in the direction of directly simulating the deexcitation process would result in considerable help in extracting information from experiments.
8. 9. 10.
14. 15. 16.
5. A c k n o w l e d g m e n t s One of us (S.N.) is indebited with G. Chiarello for drawing his attention to MDS spectroscopy and with H. Conrad for helpful suggestion and encouragement at the beginning of this work. F. Manghi is acknowledged for helpful discussion.
tL Abbati died in Modena on the 9th of September 1993. REFERENCES 1. 2. 3.
H.D. Hagstrum, Phys. Rev., 96 (1954) 336. H.D. Hagstrum, Y. Takeishi and D.D. Pretzer, Phys. Rev., 139 (1965) A526. W. Sesselmann, B. Woratschek, J. Kfippers, G. Ertl and H. Haberland, Phys. Rev. B, 35 (1987) 1547. C.M. Bertoni, C. Calandra, F. Manghi and E. Molinari, Phys. Rev. B, 27 (1983) 1251. C. Mailhiot, C.B. Duke and D.J. Chadi, Phys. Rev. B, 31 (1985) 2213.
F. Manghi, C.M. Bertoni, C. Calandra and E. Molinari, J. Vac. Sci. Technol., 21 (1982) 731. see for example: a) C. Huang, G.H. Rocker, H.J. Janish, A. Ludviksson, C.L. Cobb, H. Tochimara, H. Metiu and R.M. Martin, Surf. Sci., 241 (1991) 197. b) B. Woratschek, W. Sesselmann, J. Kfippers and G. Ertl, Phys. Rev. Lett., 55 (1985) 1231. to be published. C.B. Duke, A. Paton, W.K. Ford, A. Kahn and J. Carelfi, Phys. Rev. B, 26 (1982) 803. A. Ruocco, S. Nannarone, M. Sauvage-Simkin, N. Jedrecy, R. Pinchaux and A. Waldhauer, Surf. Sci., 307-309 (1994) 662. A. Plesanovas, A. Castellani Tarabini, I. Abbati, S. Kaciulis, G. Paolicelli, L. Pasquali, A. Ruocco and S. Nannarone, Surf. Sci., 307-309 (1994) 890. M. Canepa, C. Guarnaschelli, L. Mattera, M. Polese, S. Terreni and D. Truffelli, Rev. Sci. Instr., 62 (1991) 1431. G. Paolicelli, G. Panaccione, R. Cosso, S. Kaciulis, S. Nannarone, I. Abbati, M. Canepa, S. Terreni and L. Mattera, Vuoto, 23 (1994) 106. C. Boiziau, C. Garot, R. Nuvolone and J. Roussel, Surf. Sci., 91 (1980) 313. H.D. Hagstrum and G.E. Becket, Phys. Rev. B, 4 (1971) 4187. A. Tulke, M. Mattern-Klosson and H. Lfith, Solid State Commun., 59 (1986) 303. P. M£rtensson, G.V. Hansson, M. L£hdeniemi K.O. Magnusson, S. Wildund and J.M. Nicholls Phys. Rev. B, 33 (1986) 2259. C. AstaJdi, L. Sorba, C. Rinaldi, R. Mercuri S. Nannarone and C. Calandra, Surf. Sei., 162 (1985) 39.