Effects of low oxygen exposures on the electronic surface properties of GaAs (110)

Effects of low oxygen exposures on the electronic surface properties of GaAs (110)

Surface Science 80 (1979) 273-277 0 North-Holland Publishing Company EFFECTS OF LOW OXYGEN EXPOSURES ON THE ELECTRONIC SURFACE PROPERTIES OF GaAs (11...

324KB Sizes 9 Downloads 40 Views

Surface Science 80 (1979) 273-277 0 North-Holland Publishing Company

EFFECTS OF LOW OXYGEN EXPOSURES ON THE ELECTRONIC SURFACE PROPERTIES OF GaAs (110) * C.D. THUAULT, G.M. GUICHAR and C.A. SEBENNE Laboratoire de Physique des Solides, associkau Centre National de la Recherche Scientifique, UniversitkPierre et Marie Curie, 4 Place Jussieu, F-75005 Paris,France

The first stages of oxygen adsorption, for exposures from 0.1 to lo4 L on cleaved surfaces of n and p-type GaAs samples, are investigated by Photoemission Yield Spectroscopy. For any doping, the evolution of the density of states in the gap with oxygen exposures depends on the quality of the cleave. It is shown that, for good cleaves of the most highly doped n-type samples, an oxygen-induced band appears in the gap at exposures as low as 1 L. This effect is discussed, assuming changes in the initial structure of the clean surface.

1. Introduction

The clean and oxygen covered GaAs (110) surfaces have been the object of considerable interest during the past few years. Numerous techniques have been used to study the process of oxygen adsorption up to the monolayer range. How. ever, many questions still remain unanswered. The earlier investigations, using Low Energy Electron Diffraction and Auger Electron Spectroscopy, have shown that very low exposures of oxygen appear to have no effect on the surface structure: Dom et al. [l] reported that oxygen could be detected on the surface only after exposure of approximately lo2 L for n-type and 5 X 10’ L for p-type samples. In their measurements, the exposure necessary to obtain half a monolayer of oxygen (one oxygen atom per surface molecule of GaAs) was about 10’ L. In their early work, Gregory and Spicer [2] found no modification in their UPS measurements for oxygen exposures up to about 10’ L, but they reported Fermi level movements at about lo4 L for p-type surfaces. More recently, using Photoemission Yield Spectroscopy, Guichar et al. [3] reported that if only few defects were induced by cleavage, the Fermi level position was sensitive to oxygen exposures as low as 1 L. In fact, for the clean surface, the situation is the following [3]: The perfect surface has practically no surface states within the gap and there is no band bending at the surface for any doping. The real surface has defects which induces a band of surface states in the lower half of the band gap. On n-type samples with 3.0 X 1016 electrons/cm3, a defect * Work supported in part by DGRST, under Contract No. 76-0784. 273

214

CD. Thuault et al. /Effects of low oxygen exposures

induced band is generally observed; when it has a small amplitude (type I: “good” cleaves), the Fermi level remains close to the bottom of the conduction band and a small band of occupied surface states is observed which is attributed to the tail of the normally empty surface states overlapping the conduction band. When it has a larger amplitude (type II: “bad” cleaves), the Fermi level is found at about midgap and the intrinsic surface states are no longer observed. After 1 L of oxygen, type I cleaves gave similar results as clean type II cleaves, which remain unchanged at such an exposure. In agreement with these results, Liith et al. [4] detected, by ellipsometry and surface photovoltage measurements, band bending changes, for both type, at oxygen exposures less than 1 L. Moreover, on good cleaves, they observed new transitions involving empty or filled oxygen induced surface states in the band gap, for coverage close to half a monolayer. Similarly, recent UPS measurements [5] show changes of the Fermi level position and of the valence band structure after 1 L of oxygen.

2. Results In this work, samples of different dopings, (n = 1.O X 1018 cmm3,n = 4.9 X 1018 p=l.OXlO” cm -3 [6] were studied using Photoemission Yield Spectroscopy, which has been described previously [7]. Contact potential measurements with a vibrating probe are used to confirm the work function values obtained from the absolute photoemission threshold. Good cleaves of p-type samples show no band bending and an accurate determination of the ionization energy at 5.40 + 0.05 eV is obtained. However, on some cleavages, a depletion layer with a band bending of about 0.1 eV may be induced by extrinsic states. Oxygen exposures up to lo3 L do not affect the ionization energy and the density of states remains unchanged at such a dose. The results on n-type samples with n = 3.0 X 1016 cmm3 have been previously reported [3]. In fig. 1, the effective density of occupied states in the gap and the upper part of the valence band is drawn, showing typical results after different cleaves of two n-type samples with 1.0 X 1018 cme3 and 4.9 X 1018 free electrons per cm3, respectively. Curves (a) are examples of “bad” cleaves, characterized, as before, by a measurable band bending. The potential barrier height varies with the cleave and the doping level. Oxygen exposures up to about 1O2L have no effect on the density of states in the gap, which is dominated by the defect-induced band. The latter begins to decrease for oxygen exposures more than lo2 L. Curves (b) correspond to “good” cleaves, characterized, as before, by the absence of band bending. It is the most frequent case observed for these two highly doped samples, because, at such a doping level, a higher density of cleavageinduced defects are needed to produce band bending. After about 1 L of oxygen. Cm-3,

C.D. Thuault et al. /Effects of low oxygen exposures

215

ENERGY (eV)

Fig. 1. Effective density of states for clean cleaved and oxygen covered GaAs (110) surfaces, for two n-type samples with different doping levels. Curves (a) are for “bad” cleaves; curves (b) are for “good” cleaves; curve (c) is for “very good” cleaves (see text).

the small band of occupied surface states close to the conduction band, at 4.1 eV, vanishes for any doping level. Still in the 1 L exposure range, the most highly doped of our n-type samples show an increase of the gap state density. It is always observed on good cleaves of 4.9 X 10’s cmm3 samples. This effect is small and rarely seen on 1O’a cmm3 samples. It has not been observed on 3.0 X 1016 cme3 samples. Then, for any doping again, the defect-induced band and the oxygen-induced states, when observed, are gradually reduced for oxygen exposure beyond IO3 L. Case (c) shows a particular behaviour in the 1 L exposure range: an oxygen induced surface state band appears at about 5.1 eV, superimposed over the defect-

276

C.D. Thuault et al. /Effects of low oxygen exposures

induced band, which is exceptionally small in that case. After exposures above lo2 L, these two bands begin to decrease and become very weak beyond lo3 L.

3. Discussion Regarding the clean surface, the results presented here bring a confirmation of the model of electronic surface structure previously suggested for lightly doped n-type GaAs samples [3]. For any doping level, the existence of a band, the bottom of which is in the upper part of the gap and which is filled up to the Fermi level, is shown, The attribution of this band to the fiiled bottom of Ga-derived intrinsic states is in good agreement with theoretical calculations 181, which take into account atomic displacements within the surface unit cell, as revealed by Low Energy Electron Diffraction analysis [9]. Moreover, the Fermi level surface position on n-type samples depends strongly, first,, on the density of surface defects able to induce a charge transfer and second, on the bulk doping, in a way which appears qualitatively satisfactory: at a given doping, a band bending appears when the density of surface defects is large enough. The main effects of low exposures of oxygen already reported from Photoemission Yield Spectroscopy measurements [3] are again observed, but it appears that, as defects do, the doping level of the studied samples influences the initial process of oxygen adsorption. For bad cleaves, in all cases, lo4 L of oxygen remove completely the defectinduced band. No doping effect on the oxygen adsorption is observed. For good cleaves, whatever the Fermi level bulk position, a dose as small as 1 L is sufficient to remove the electrons of the partly filled surface states band in the upper part of the gap. For n = 3.0 X 1016 cmL3 and n = 1.0 X 1018 cmb3 samples, the band bending produced by such a dose implies changes in the surface charge involving 5 X 10” to 10l 2 electrons/cm2 ; the appearance of acceptor surface levels in the gap, due to oxygen, can explain this excess of negative charges on the surface. For the most highly doped samples, the decrease of the intrinsic band close to the conduction band is always accompanied by the growth of a new surface state band in the midgap region. Moreover, the magnitude of this band, after 1 L of oxygen, is clearly seen to be cleavage-dependent. The density of surface states appearing in the gap, can be approximately estimated by comparing with the bulk density of states: assuming the same matrix elements for photoemission from bulk and surface states and an escape depth of electrons of 15 A [l 11, about 6 X 1013 states/cm2 are found in the gap for the best cleave (fig. 1, case (c)). From Auger spectroscopy measurements [4,10], beyond 1 L exposure, the oxygen coverage is certainly less than 0.01 monolayer. So, the density of oxygen-induced surface states is larger than the density of oxygen atoms by more than an order of magnitude and cannot be attributed to localized states associated with oxygen on a one state-one atom basis.

CD. Thuault et al. /Effects of low oxygen exposures

277

The second aspect to be discussed is the fact that only highly doped n-type samples show oxygen-induced surface states. The common effect of low oxygen doses is to provoque, on good cleaves of n-type samples a change in the surface position of the Fermi level: from a negligible band bending, values from 0.2 to 0.5 eV, depending on the doping, are usually obtained. It is clear that the higher is the doping, the higher is the corresponding surface electrostatic field. Since the effect of surface reconstruction, as proposed by Lubinsky et al. [9] is to induce a charge transfer corresponding to a surface dipole, a change in surface structure in the presence of an electric field can be emphasized. A possible explanation of the appearance of oxygen-induced surface states on a very good cleave of a highly doped n-type GaAs (110) surface could, therefore, be the following: when the band bending developed by enough oxygen adsorption corresponds to a surface electric field of sufficient value, it induces a change in the surface atomic structure. It relaxes at least partially the reconstructured surface with a consequent change of its electronic structure. Surface defects can partly or. totally impair the process. In order to check this model, further experiments are needed and, in particular, LEED studies on differently doped samples would be the most interesting.

References [l] [2] [3] [4] [5] [6] [7] [8] [9] [lo] [ 1 l]

R. Dorn, H. Ltith and G. Russel, Phys. Rev. BlO (1974) 5049. P.E. Gregory and W.E. Spicer, Surface Sci. 54 (1976) 229. G.M. Guichar, C.A. Sebenne and G.A. Garry, Phys. Rev. Letters 37 (1976) 1158. H. Lttth, M. Btichel, R. Dorn, M. Liehr and R. Matz,Phys. Rev. B15 (1977) 865. P. Pianetta, I. Lindau, P.E. Gregory, C.M. Garner and W.E. Spicer, Surface Sci. 72 (1978) 298. GaAs samples: n = 3 X 1016 crnm3 andp = 1 X 1018 cmm3 from MCP Electronics (England); n = 1 X 10’ 8 cm” and n = 4.9 X lo1 a cmm3 from RTC (Caen, France). CA. Sebenne, D. Bolmont, G.M. Guichar and M. Balkanski, Phys. Rev. B12 (1975) 3280. J.R. Chelikowsky, S.G. Louie and M.L. Cohen, Phys. Rev. B14 (1976) 4724. A.R. Lubinsky, C.B. Duke, B.W. Lee and P. Mark, Phys. Rev. Letters 36 (1976) 1058. F. Proix and F. Houzay, to be published. This value is obtained from the absolute value of the bulk photoemission yield, taking into account the absorption coefficient and the density of states in the valence band (see ref. [71).