Ti overlayer growth on oxidized GaAs(110) versus Ti oxidation on physisorbed O2 on GaAs(110) at 25 K

201

Surface Science 254 (1991) 201-208 North-Holland

Ti overlayer growth on oxidized GaAsf 110) versus Ti oxidation on physisorbed 0, on GaAs( 110) at 25 K Y.Z.

Li, D.J.W.

Department Received

Aastuen,

of Materials

26 October

J.M.

Seo, U.S.

Ayyala

Science and Chemical Engineering

1990; accepted

for publication

and

J.H.

Weaver

University of Minnesola,

19 February

Minneapaiis,

MN 55455, USA

1991

Interfacial reactions for Ti deposition onto oxidized GaAs(ll0) have been studied for oxide thicknesses up to - 10 A. The oxidized GaAs surfaces were produced by low-temperature X-ray-enhanced techniques using condensed N,O at 25 K. Titanium deposition at 300 K reduced the surface Ga and As oxides and a Ti oxide layer was formed. Schottky barrier studies showed that the Fermi level pinning position for these interfaces was closer to the valence band maximum than for Ti growth on clean, oxide-free GaAs(ll0). We also investigated Ti interactions when deposited directly onto solid Oz condensed on GaAs(ll0) at 25 K. In this case, the Ti adatoms were completely oxidized to Ti 4.t-like bonding when there was sufficient 0, for reaction. This process follows a thermodynamically favorable pathway and results in a GaAs surface with little semiconductor disruption.

1. Introduction

The control’ted growth of thin insulating films on GaAs has attracted considerable interest because such heterostructures are needed in microelectronic device applications. There are essentially two different approaches that have been used to synthesize these heterostructures. One involves the growth of the native oxides of GaAs (e.g, refs. [l-lo]) while the other involves the formation of a nonnative insulating film on the GaAs surface (e.g., refs. [11,12]). Here, we investigate the properties of both native and non-native insulating films on GaAs(ll0) formed at 25 K under conditions that alter the growth parameters and the resulting structures. At 300 K, the sticking coefficient of 0, on GaAs(ll0) is very small [l] and thick oxide layers cannot be grown. However, recent photoemission investigations [2,3] of GaAs(ll0) onto which multilayers of molecular 0, had been condensed at 20 K showed that Ga and As oxides could be grown in a controtled fashion by changing the X-ray exposure. Those studies revealed the coexistence of As,O,-like and As,O,-like species. While these ~39-6028/91/$03.50

0 1991 - Elsevier Science Publishers

novel oxide films may be useful in quantum electronic device applications, the electrical properties of metal-oxide-semiconductor heterostructures derived from them have not been investigated and, indeed, the stability of the native oxide during metal overlayer formation is uncertain. The first part of this paper emphasizes X-ray photoemission (XPS) investigations of Ti deposition on oxidized GaAs(ll0) at 300 K following native oxide formation at 25 K. Titanium was chosen because of its high reactivity with oxygen and because of the large band gap of TiO* [13]. TiO, is also attractive because it has been well studied in the context of catalysis and photoelectrochemistry 114-171. We will show that the native oxides are unstable when exposed to Ti because the Ga- and As-oxide bonding configurations are replaced by TiO,-like structures. The Schottky barrier formed on these oxidized GaAs surfaces has a final E, pinning position that is closer to the valence band maximum (VBM) than that found for Ti growth directly onto cleaved GaAs(ll0). The second part of the paper examines R-oxide formation on GaAs(l10). This is done by depositing Ti onto condensed 0, on GaAs(ll0) at 25 K.

B.V. (North-Holland)

202

Y.Z. Li ef al. / I n?e~~aciaI reacfiomfor Ti de~osifiononto oxidized Ga&l IO)

Growth at low temperature allows for kinetic control over interface chemistry and structure. The resulting overlayer reflects the competition between 0, reaction with Ti adatoms and with the GaAs support.

2. Experimental The X-ray photoemission experiments were conducted in a ultrahigh vacuum chamber having a base pressure of 8 x lo-” Torr. Unpinned mirror-like GaAs(l10) surfaces were obtained by in situ cleaving. Low-temperature studies were made possible by attaching the sample holders with a Cu braid to the cold stage of a closed cycle He refrigerator. The base temperature of the sample was 25 K, as measured with a Au-Fe/chrome1 thermocouple attached to the sample holder. For studies of Ti growth on oxidized GaAs(llO), we used samples that were p-type, Zn-doped at 1 x 1017 and 2 x 10” cmp3, and n-type, Si-doped at 1 x 1O*7cme3. N,O was condensed onto clean cleaved surfaces at 25 K, and subsequent exposure to X-rays led to the production of the GaAs surface oxide, following the procedures discussed in detail in ref. [lo]. N,O was used as an oxidant rather than 0, because its condensation temperaTorr (compared to ture is - 64 K at 1 X lo-” - 25 K for 02) and because molecular dissociation of N,O provides 0 and N, (dissociation energy 1.7 eV) [l]. Surface N,O exposures were done at typical pressures of 5 X 10--’ Torr at 25 K. Thereafter, the pressure was reduced to less Torr within a few minutes by than 4 x lo-” combined ion and cryo pumping. The typical operating pressure during data acquisition was 1 X IO-“’ Torr. Oxidation was induced by illumination with AlKo (1486.6 eV) and MgKa (1253.6 eV) radiation. Completion of reaction was judged by the disappearance of emission from the N,O species and the stability of the Ga and As core level emission. After reaction was complete, the sample was warmed to 300 K and Ti adatoms were deposited. The typical Ti deposition rate was - 2 A/min for low coverages and - 4 A/min for high coverages, as measured with a quartz crystal thickness monitor. AlKa X-rays were used for

and 0 1s core levels while studies of the AsZp,,, Mg Ka X-rays were used for Ga2p,,, and Ti 2p investigations. For studies of Ti interactions with 0, physisorbed on GaAs(llO), the samples were n-type, Si-doped at 1 x 10” and 2 x 10” cme3. Typical 0, dosing conditions were 5 X lo-* Torr for 100 s for 5 L exposures and 1 X lo-’ for 100 s for 10 L exposures. Ti was deposited onto the 0, multilayers at 25 K, and the amount of Ti was again monitored with a thickness monitor.

3. Ti overlayer growth on oxidized GaAs(l10) Fig. 1 shows the evolution of Ti2p and AsZp,,, core level energy distribution curves (EDCs) when Ti was deposited onto GaAs-oxide prepared with 4 L of N,O exposure (5 X lo-’ Torr, 80 s). Prior to Ti deposition, the As2p,,, emission showed a GaAs peak and an oxide peak (lower right EDC). The chemical shift of 3 eV indicates that the As-oxide was As,O,-like. Since these measurement were done at 300 K, the amounts of other As-

Relative Binding Energy (eV) Fig. 1. Ti 2p and As 2p,,, evolution during Ti deposition on an oxidized GaAs(ll0) surface using 4 L N,O condensation. The measurements were performed with the (p-type) sample surface at 300 K. With 2 A deposition, the oxide component of As2p has nearly disappeared. This, plus the apparent formation of TiOz (Ti4” ), demonstrates the transfer of oxygen from As oxide to 7-i oxide bonding configurations. Further Ti deposition is accompanied by the appearance of Ti3+ and TiZ+, and eventually Ti metal.

Y.Z. Li et al. / Interfacialreacfions for Ti deposition onto oxidized GaAs(ll0)

oxides were negli~ble, although such bonding configurations existed at 25 I( [lo]. The thickness of the oxide can be estimated by studying the photoemission intensities of the As 2p,,, substrate component and the oxide component. If the mixture of Ga-oxide and As-oxide is assumed to be uniform and the photoelectron mean free path is 7 A (ref. [18]), then the oxide thickness can be estimated to be - 2.5 A. This corresponds to the average disruption of - 1.3 ML of the GaAs(ll0) surface. (The formation of thicker oxide layers will be discussed below.) The results of fig. 1 show that emission from As,O, is significantly reduced when 2 A of Ti are deposited onto this disrupted oxide layer, indicating the loss of As,OJ-like bonding. Equivalent results for Ga show the same pattern as the oxide is reduced. Such reactions of Ti to As-oxide and Ga-oxide have also been observed previously on a native GaAs-oxide prepared using a different technique [9]. Examination of the Ti 2p spectrum for 2 A shows that the Ti atoms are in nominal 4 + configurations, based on the chemical shift of 4.65 eV relative to Ti metal. If we assume that - 1.5 A of Ti completely reduces the As,O, and Ga 203 configurations to form TiO,-like bonding, we would again conclude that - 1.3 ML of GaAs was oxidized by the low temperature photo-induced oxidation with 4 L N,O exposur? As can be seen in fig. 1, Ti deposition beyond 2 A is accompanied by the evolution of the Ti oxide from TiO,-like to lower oxidation states and, ultimately, Ti metal nucleates and grows. (The relative binding energies of Ti3+ and Ti2+ are from ref. 1141, where the reduction of TiO, by a Ti overiayer was studied and Ti3’, Ti*+, and Ti’ configurations were obtained.) To investigate the growth mode of Ti on this novel GaAs-oxide layer, we determined the attenuation of the As~P,,~, Ga2p,,,, and 01s core emission features as a function of deposition, as shown in fig. 2. The kinetic energies of the relevant core levels are as given so that the photoelectron mean free path for electrons emitted from 0 1s states should be considerably longer than for As and Ga while the mean free path for As should be close to that for Ga. It is clear from fig. 2 that photoelectrons from the three elements have quite different attenuation behaviors. The rapid Ga 2p,,,

203

K.E.-955eV

-5 -6 -

0

IO

20

30

40

Ti coverage &) Fig. 2. Attenuation curves of AsZp,,a, Ca2p,,, and 01s core levels during Ti deposition on an oxidized p-GaAs(ll0) surface using 4 L N,O condensation. I($) is the total photoemission intensity (including substrate and oxide components) measured at a Ti coverage. The attenuation of Ga2p indicates layer-bylayer growth for Ti, and the slower As2p atten~tion indicates the mixing of As in the Ti overlayer after liberation from As-O bonding. The slow 0 attenuation indicates oxygen segregation at the surface.

attenuation (l/e length = 6 A) shows that the Ti overlayer grows approximately in a layer-by-layer fashion, with no evidence for large clusters. The As2p,,, emission, however, does not attenuate as quickly as Ga, and this indicates that there is intermixing of As in the growing Ti overlayer. We note that analogous mixing of As has been discussed for direct Ti deposition onto GaAs(ll0) [19]. The 0 Is attenuation is very slow (l/e length > 40 A), and this indicates that oxygen atoms out-diffuse to the near-surface region. It is worth noting that this oxygen out-diffusion behavior in Ti overlayers is consistent with what has been reported for Ti/CuO and Ti/YBa,CuO,_, interfaces 1201. To rule out the possibility that the slow 01s attenuation was caused by adsorption of 0 molecules from the vacuum system, a control experiment was conducted for Ti deposition directly onto GaAs; the amount of 0 adsorption was so small that it can be neglected. Thus, the 0 emission reflects a changing distribution of 0 atoms during overlayer evolution. To further study the metal-GaAs-oxide interface reaction, we grew a thicker oxide at 25 K on GaAs(llO). First, we exposed the cleaved GaAs (110) surface to 20 L of N,O and induced oxida-

Y.Z. Li et al. / Interfacial reactions for Ti deposition onto oxidized GaAs(ll0)

As 2p

As,O,

As,O,

6

6

Relative

4

2

Binding

Ga As

0

-2

Energy

-4

(eV)

Fig. 3. Evolution of AsZp,,, emission for Ti deposition on an n-GaAs surface that had been oxidized by 60 L N,O condensation. As has three distinct bonding configurations in GaAs, As,O,, and As,O,. Ti deposition causes the reduction of the surface oxides. The EDCs have the same vertical scale in counts per second for direct comparison.

tion with X-ray illumination. When oxidation was complete, we repeated the procedure twice to get a total exposure of 60 L N,O. Based on the evolution of As oxide features, we conclude that this is approximately the ultimate thickness for the GaAs oxide prepared by this method. We estimate the oxide thickness to be - 9 A using the As2p,,, photoemission intensities of the substrate and the oxide. While the mechanism that determines the ultimate thickness is not fully understood, it should be dictated by the balance between the rate of photo-induced desorption of N,O and the rate of oxygen in-diffusion and reaction with GaAs. The results of fig. 3 illustrate the evolution of the As 2p,,, emission as a function of Ti deposition on this 60 L GaAs-oxide thin film at 300 K. The coexistence of As,O,-like and As,O,-like bonding is evident, as previously shown, when there is an abundant supply of oxygen during low temperature photo-induced oxidation [2,3,10].

Once Ti is deposited, however, the As,O, intensity decreases drastically and the As,O, intensity increases. This early behavior can be explained in a straightforward fashion by assuming that oxygen in the top layer of As,O, is scavenged by Ti to form TiO, and As,O, converts to As,O,. With increasing Ti deposition, As,O, is also reduced. This is particularly evident after most of the As,O, has been consumed. We cannot specify the bonding of this reduced As, but the attenuation behavior of As2p,,, shown in fig. 2 demonstrates that there are As atoms intermixed in the heterogeneous TiO, and Ti overlayer. In principle, very high resolution studies might be able to distinguish As in these configurations, but it is likely that there are a range of inequivalent configurations that contribute to give undifferentiated features. Parallel studies show no discernible change in the full width at half maximum of Ga2p,,, emission at low Ti coverage. This indicates that the Ga-oxide is not reduced by Ti in the same fashion as the As-oxide. Most likely, this indicates sequential reactions of Ti- to As-oxide and then to Ga-oxide due to reaction energetics. In this study, we did not observe discernible features related to Ti-Ga and Ti-As bonding configurations. Such species are certainly present and have been identified in synchrotron radiation photoemission studies but they have not been observed in XPS studies [17,19]. Note that the deposition of only - 4 A of Ti atoms completely consumes the oxygen that was previously bonded to As. This again indicates that - 5 ML (10 A) of As-oxide originally formed. An important aspect of metal-semiconductor interface formation is the Fermi level ( EF) position in the surface bandgap. In device-related research, different techniques have been developed to passivate clean GaAs surfaces using oxygen [21], H,S [22], Na,S [23], and (NH,),S [23,24] to reduce the interface state density. Schottky barriers have then been studied when metal overlayers were deposited on the treated GaAs surfaces. Different barrier heights have been reported for differently treated interfaces [21-241. In the present experiments, E, movement in the bandgap has been investigated using the photoelectron kinetic energy of the AsZp,,, substrate component, giving the results shown in fig. 4.

a m~im~m

around I.5 A. This E, movement is related to chemical reaction at the interface at this particular coverage. As discussed earlier, one monolayer of Ti consumes most of the oxygen atoms to release As from As-O. With further deposition, Ti nucleation is initiated and the overlayer starts to develop a metallic skin. As a thicker Ti overlayer is formed, the final E, pinning position converges to - 1.0 eV below the CBM. The final pinning position is determined by the solid curve of fig. 4 and is verified by studies of the Ga2p,/, emission. fn contrast, it has been established (251 that the final E, pinning position for Ti deposition on p-type and n-type GaAs(l.IO) at room temperature is - 730 meV from the CBM. The different final E, position is attributed to a different density of states profile inside the surface band gap, as caused by interfacial reactions. Note that the final E, position in fig. 4 is different from that of ref. [21] where the interfacial layer at the Ti/GaAs interface produced a position closer to the CBM than for the clean Ti/GaAs interface. This d~ffer~~~ indicates a novel interface formed with the technique presented here.

Fig. 4. Fermi level movement as a function af Ti deposition on the oxidized surface. The p and n-type surfaces were oxidized with 4 and 60 L N,O exposures, respectively. The movement of E, toward the CBM at 1-2 w deposition reMects chemical changes at the interface. The final E, pasition 1.0 eV from the CBM differs from that found for Ti/GaAs interface formation.

Prior to Ti deposition, E, was piuned by surface oxidation. For both n- and p-type GaAs substrates, Ti deposition first moved E, toward the GBM, but then drew it toward the VBM to reach a final pinning position. The peak thus formed has

ri O? I

/

86420-2

/

,

,

+4

+3

+2

metal

GoAs

8

’

Relative Binding Energy (eV) Fig. 5. c) Is, Ti Zp, and As~P,/~ EDCs for a variety of Oa condensations and Ti depositions. Curve (I) is for the clean cleaved GaAs(llO) surface at 25 K. Curves (2) were measured after 10 L Oa and 5 A Ti were condensed on the clean surface at 25 K. They show moIecuIar Oa and oxygen in As- and ‘Ii-derived oxide states. Curves (3) were acquired after reaction was complete for the conditions of curves (2). They show the disappearance of molecular Oz and the completion of Ti4+ formation. Curves (4) were obtained after the sample was warmed to 300 K. Tlzey show no change relative to the fully-oxidized state at 25 K. Curves (5) show that the deposition of an additional 5 A of Ti at 300 K completely reduced the As-oxide while producing severat different Ti confi$urations. Curves (6) show the result of an experiment in which 10 L 4 and then 10 A Ti were deposited onto clean GaAs(lfO) at 25 K. In this case, the supply of oxygen was sufficient to completely oxidize Ti into Ti4’*’ and to form smaIi amounts of As-oxide.

206

Y. Z. Li et al. / Interfacial reactions

4. T&oxidation on physisorbed 0, on GaAs( 110) at 25 K The previous section emphasized reaction of the native oxides of GaAs by Ti to form TiO structure. Here, we discuss chemical reactions between Ti and 0 on GaAs(ll0) at 25 IS where the 0 is in the form of multilayers of molecular species. Fig. 5 shows 0 Is, Ti 2p, and As 2p,,, core level EDCs for a variety of conditions with Ti and 0 present on cleaved GaAs(ll0). Curve 1 represents the As2p,,, emission for the clean cleave. The three spectra labeled 2 show that both Ti and As were oxidized when 10 L of OZ and then 5 A of Ti were condensed on the GaAs(lI0) surface and investigated with XPS. All of the Ti was effectively converted into Ti-0 bonding configurations. Although molecular 0, was still the dominant form of 0 following Ti deposition and oxidation, the molecular form diminished and ultimately disappeared under continuous illumination by Al Ka or Mg Kcr X-rays at 25 K (total flux of - 10’” photons/cm2 for AlKa and - lOI photons/cm’ for Mg KcY). This reflects the consumption of oxygen to form fully-coordinated Ti4+-like bonding and the balance between photon-stimulated substrate reaction and photon-stimulated 0, desorption. Indeed, the oxide component of the 0 Is spectra increased in relative intensity as the Ti atoms were fully oxidized (curves labeled 3). The reaction products formed in this way were then stable because the EDCs of all three core levels remained the same upon warming to 300 K, releasing some of the kinetic constraints imposed by formation at 25 K (compare curves labeled 3 and 4). When an additional 5 A of Ti was deposited at 300 K onto this TiO,/GaAs-oxide)/ GaAs structure, there was oxygen atom transfer from the As oxide to the Ti atoms to form more Ti oxide (curves 5). Indeed, the As-oxide emission almost vanished. This reflects the higher heat of formation for TiO, (-AH,? = 944 kJ/mol) than that for As,O, ( -AH/ = 655 kJ/mol). In this case, however, there was insufficient oxygen to completely oxidize the 5 A of added Ti, and Ti atoms existed in (nominal) Ti4+, Ti3”, Ti’+, and metallic Ti bonding configurations [14]. One can estimate the amount of 0 needed to

for Ti deposition onto oxidized

GaAs(ll0)

completely oxidize the Ti overlayer by simply counting. Assuming the the sticking coefficient of 0, on GaAs(llO) was unity at 25 K, then 10 L 0, exposure would correspond to an 0 atom surface density of 1.1 X 1016 cme2. The number of atoms represented by 1 A of Ti corresponds to 5.9 X lOI cme2. Therefore, 10 L 0, could react with 9.3 A Ti to completely oxidize Ti into TiO,. To demonstrate this correspondence between Ti and 0 exposures, the EDCs labeled 6 in fig. 5 were obtained in a separate experiment when 10 L 0, and then 10 A Ti were condensed on GaAs(ll0) at 25 K. In this case, the molecular 0, was consumed to produce mainly Ti4+. Small amounts of As-oxide persisted, consistent with a slight surplus of oxygen and the assumption that the sticking coefficient of 0, on GaAs(ll0) at 25 K was close to unity. Fig. 6 provides additional insight into the chemical reactions between Ti and 0, at 25 K on a GaAs(ll0) surface. When 10 L 0, and then 20 A Ti were condensed on cleaved GaAs(llO), the Ti metal configuration appeared to dominate, as shown in curve 1. The contribution from Ti-0 was not as evident because of the attenuation of the photoelectrons by the Ti overlayer. As shown above, 10 L P2 is only enough to completely oxidize - 9.3 A of Ti to form TiO, (curve 6 of fig. 5), and the Ti overlayer is dominantly metallic. When 5 L of additional 0, was first condensed on the surface at 25 K, the Ti4+ component appeared and molecular 0, was clearly present (fig. 6, curve 2). After X-ray illumination, the molecular 0, desorbed from the surface and Ti atoms were present in Ti4’ and metallic Ti configurations (curve 3). At first glance, the persistence of metallic Ti might be surprising because there was abundant 0, initially present at the surface for oxidation. It can be understood, however, by noting that 0, was desorbed during X-ray illumination by photon-stimulated desorption [26]. We therefore conclude that the reaction rate for atomic Ti on solid 0, at 25 K differs from that for solid 0, on Ti. To further emphasize the difference in apparent reaction rates caused by photo-induced 0, desorption, we condensed another 5 L 0, and measured the EDCs labeled 4 in fig. 6. Under X-ray ilthese spectra evolved into those lumination,

Y.Z. Li et al. / Interfacial reactions for Ti deposition onto oxidized GaAs(ll0)

01s

Ti 2p hu=1253.6eV

hv

TiO,l

q

T=25K

1486.6eV Ti 0, i

+loL

0,

58 Ti

1

I

8

4

0

Relative

the surface. Even at 25 K, the solid-vacuum interface provides an oxygen sink. Thus, diffusion of oxygen atoms will always cause oxidation of Ti when Ti is in a lower oxidation state than Ti4+. When condensed oxygen is trapped between the Ti overlayer and the GaAs surface and there is more oxygen than Ti atoms for the formation of Ti oxide in the overlayer, then oxygen atoms will completely oxidize the overlayer Ti and GaAs at the interface. Eventually, the diffusion of oxygen through the TiO, overlayer becomes dominant, and the oxygen atoms are either consumed by oxidation or desorbed from the surface.

5. Conclusion

metal 12

207

,/

1

”

6

Binding

/

6

4

2

0

-2

-4

Energy (eV)

Fig. 6. Curves (1) show Ti2p and 01s EDCs after clean GaAs(ll0) was exposed to 10 L 0, and then 20 A Ti. The oxygen atoms appear to be in oxide configuration while metallic Ti dominates the Ti emission because there was not enough oxygen to completely oxide the Ti atoms. Curves (2) show that the condensation of 5 L of additional oxygen yields greater amounts of TiO, as well as molecular 0,. After X-ray irradiation, the molecular 0, desorbed, leaving both metallic and oxidized Ti. Another 5 L 0, condensation [curves (4)] produced no discernible Ti oxidation and the oxygen molecules were desorbed upon extended irradiation [curves (5)]. Capping this system with 10 L 0 and 5 A Ti produced curves (6) but X-ray irradiation failed to eliminate metallic Ti emission in the probed region.

labeled 5 as molecular oxygen was almost completely desorbed from the surface. However, the Ti2p EDCs showed the persistence of metallic Ti. Considering the scenario that the desorption of 0, might be faster than dissociation and transport of oxygen through the surface oxide of the Ti layer, we added an additional 10 L 0, and then 5 A Ti to form a cap of TiO, in an attempt to change the balance. Curves 6 and 7 in fig. 6 show the resulting Ti2p and 01s EDCs immediately after cap formation and after X-ray illumination. Evidently, this capping procedure did not yield the oxidation of the remaining elemental Ti. Thus we conclude that 0 diffusion at 25 K is not sufficient to penetrate the TiO, layer to cause further subcutaneous oxidation once a layer of TiO, is formed at

We have investigated the formation of Ti overlayers on GaAs oxides grown by a low-temperature photo-induced oxidation technique. Titanium is found to react readily with two forms of Asoxide, namely As,O, and As,O,. With only a few monolayers of metal deposition, the As-oxide is dissociated, giving oxygen atoms to the metal overlayer to form a metal-oxide. Further growth of Ti on this interface follows a layer-by-layer mode, as demonstrated by the attenuation behavior of substrate GaAs photoelectrons. Both As and 0 are found to mix in the increasingly metallic Ti overlayer. Because of the different profile of electronic density of states inside the band gap, the final E, position at the interface formed is different from that formed at pristine interfaces. The oxidation processes for Ti deposited onto solid oxygen condensed on GaAs(ll0) at 25 K yield TiO,-like bonding. This thin oxide film is stable at 25 and 300 K. The interface between the Ti-oxide and the GaAs can be controlled so that there is no significant GaAs-oxide formation. This technique of oxide growth on GaAs is promising since the atomic and chemical properties of the interface can be well-controlled.

Acknowledgements

This work was supported by the Army Research Office under DAAL03-88-K-0093. Helpful

208

Y.Z. Li

et al. / Interfacial

reactions

discussions with G.D. Waddill and T.R. Ohno are gratefully appreciated.

References [l] K.A. Bertness,

T.T. Chiang, C.E. McCants, P.H. Mahowald, A.K. Wahi, T. Kendelewicz, I. Lindau and W.E. Spicer, Surf. Sci. 185 (1987) 544. PI S.G. Anderson, J.M. Seo, T. Komeda, C. Capasso and J.H. Weaver, Appl. Phys. Lett. 56 (1990) 2510. T. Komeda, C. Capasso and [31 J.M. Seo, S.G. Anderson, J.H. Weaver, Phys. Rev. B 41 (1990) 5455. I. Lindau, CM. Garner and W.E. Spicer, 141 P. Pianetta, Phys. Rev. B 18 (1978) 2792. I51 K. Stiles, D. Mao and A. Kahn, J. Vat. Sci. Technol. B 6 (1988) 1170. [61 C.F. Yu, M.T. Schmidt, D.V. Podlesnik, E.S. Yang and R.M. Osgood, Jr., J. Vat. Sci. Technol. A 6 (1988) 754. R. Ludeke, Y. Jugnet, J.F. Morar and F.J. 171 G. Landgren, Himpsel, J. Vat. Sci. Technol. B 2 (1984) 351. B 4 [81 G. Hughes and R. Ludeke, J. Vat. Sci. Technol. (1986) 1109. J.R. Waldrop and R.W. Grant, J. Vat. [91 S.P. Kowalczyk, Sci. Technol. 19 (1981) 611. D.J.W. Aastuen, U.S. [lOI J.M. Seo, Y.Z. Li, S.G. Anderson, Ayyala, G.H. Kroll and J.H. Weaver, Phys. Rev. B 43 (1990) 9080. [ill H. Hasegawa, M. Akazawa, H. Ishii and K. Matsuzaki, J. Vat. Sci. Technol. B 7 (1989) 870. [121 Q.D. Qian, J. Qiu, M. Kobayashi, R.L. Gunshor, M.R. Melloch and J.A. Cooper, Jr., J. Vat. Sci. Technol. B 7 (1989) 793.

for Ti deposihn

onto oxidized GaAs(l10)

F.A. Grant, Rev. Mod. Phys. 31 (1959) 646. G. Rocker and W. Giipel, Surf. Sci. 181 (1987) 530. V.E. Henrich, Prog. Surf. Sci. 14 (1983) 175. A.J. Bard, J. Photochem. 10 (1979) 59. F. Xu, D.M. Hill, P.J. Benning, and J.H. Weaver, J. Vat. Sci. Technol. A 7 (1989) 2593. G.D. Waddill and J.H. [181 M.C. Schabel, I.M. Vitomirov, Weaver, J. Electron Spectrosc. Relat. Phenom., in press. M. de1 Giudice, J.J. Joyce and J.H. [191 M.W. Ruckman, Weaver, Phys. Rev. B 33 (1986) 2191. 1201 H.M. Meyer III, J.H. Weaver and K.C. Goretta, J. Appl. Phys. 67 (1990) 1995. [211 M.T. Schmidt, D.V. Podlesnik, H.L. Evans, CF. Yu, E.S. Yang and R.M. Osgood, Jr., J. Vat. Sci. Technol. A 6 (1988) 1446; M.T. Schmidt, Q.Y. Ma, D.V. Podlesnik, R.M. Osgood and E.S. Yang, J. Vat. Sci. Technol. B 7 (1989) 980. [221 T. Tiedje, K.M. Colbow, D. Rogers, Z. Fu and W. Eberhardt, J. Vat. Sci. Technol. B 7 (1989) 837. MS. Hedge, L.A. Farrow, C.C. Chang and ~231 C.J. Sandroff, J.P. Harbison, Appl. Phys. Lett. 54 (1989) 362; C.J. Sandroff, M.S. Hedge and CC. Chang, J. Vat. Sci. Technol. B 7 (1989) 841. M.R. Melloch and T.E. Dungan, Appl. v41 M.S. Carpenter, Phys. Lett. 53 (1988) 66; M.S. Carpenter, M.R. Melloch, B.A. Cowans, Z. Dardas and W.N. Delgass, J. Vat. Sci. Technol. B 7 (1989) 845. I.M. Vitomirov, C.M. Aldao, S.G. Ander1251G.D. Waddill, son, C. Capasso, J.H. Weaver and Z. Liliental-Weber, Phys. Rev. B 41 (1990) 5293. T. Komeda, J.M. Seo, C. Capasso, G.D. (261 S.G. Anderson, Waddill, P.J. Benning and J.H. Weaver, Phys. Rev. B 42 (1990) 5082. [131 U41 [151 (161 [171