Faraday Discuss. Chem. SOC., 1990, 89, 51-63 X-Ray Absorption Spectroscopy under Conditions of Total External Reflection: Application to the Structural Characterisation of the Cu/GaAs( 100) Interface S. Pizzini and K. J. Roberts*+ Department of Pure and Applied Chemistry, Uniuersity of Strathclyde, Glasgow GI 1 X L G. N. Greaves SERC Daresbury Laboratory, Warrington WA4 4 A D N. T. Barrett LURE Bztiment 209D, Centre Uniuersitaire, 91 405 Orsay Cedex, France I. Dring and R. J. Oldman ICI Chemicals and Pol-ymers plc, The Heath, Runcorn WA7 4QE The theoretical principles and instrumentation requirements of X-ray absorp- tion spectroscopy under conditions of total external reflection are described with particular reference to their application in the characterisation of solid/solid interfaces.The advantages of combining XANES, EXAFS and reflectivity in a single series of measurements are highlighted through the structural characterisation of oxide layers on Cu and GaAs( 100). The data reveal the surface of Cu deposited on float-glass to comprise a macroscopic mix of metallic C u and structurally disordered oxide. The latter appears to have a local structure close to that observed in Cu10 but with a higher oxygen coordination af, the surface. The surface of commercial-grade GaAs( 100) has ca. 7-9 A of disordered oxide in which G a coordinates to oxygen both tetrahedrally and octahedrally whilst As is only found in tetrahedral sites. Strong correlated second-shell cation coordinations around both G a and As central atoms reveal a single non-stoichiometric surface oxide depleted of As at the surface.10 and 100 A thick Cu films deposited on GaAs(100) are found to be completely oxidised with the oxide being more disordered and having a higher oxygen coordination than those deposited on float-glass. For 10 A thick films, Cu appears coordinated to Ga and As through the oxygen which decorates the inner surfaces of micro-voids and fissures in the disordered oxide on GaAs. 1. Introduction Solid/solid interfaces play a key role in many physical and chemical processes and their non-destructive structural characterisation is of significant technological interest. The structural nature of the interface between metals and semiconductors is of fundamental importance to device technology. The formation of a passivated oxide layer at this interface either during or after fabrication can have a significant impact on electrical conductivity and hence device performance and reliability.Additionally, the structural control of such interfaces is important in MOS technology for the preparation of well defined Schottky barriers. ; Also at SEKC Daresbury Laboratory, Warrington WA4 4AD. 5152 The Cu/GaAs( 100) Interface Experimental studies have so far concentrated mostly on metallic depositions on clean GaAs surfaces, usually in UHV conditions.'-' The chemistry at such clean interfaces is now reasonably well understood' and predicts the formation of alloyed phases associated with topochemical reaction between the semiconductor and its overlay- ing metal. Photoemission studies'-' of Cu/GaAs show evidence for intermixing between Cu with the substrate, resulting in the formation of copper arsenides and the segregation of metallic gallium.However, commercial-grade GaAs has a native oxide coating,x and thus in practical electronic devices the presence of a thin insulator between the metal and the semiconductor9 is assumed. Although some XPS studies'" of the Cu/GaAs( 100) system have shown the oxide layer on GaAs wafers to be unaffected by the metal overlayer, the understanding of the whole interface region including the extent and form of the oxide layer on the substrate, the reaction of the metal overlayer with the substrate and the structure of the metal overlayer is far from complete. X-Ray absorption spectroscopy under conditions of total external reflection' ' - I z allows surface sensitive studies to be carried out under realistic operational conditions (i.e.in air, under dense gases or under liquid). In this paper we outline the theoretical principles and instrumental requirements for this technique and describe its application to the structural characterisation of the Cu/GaAs( 100) interface. The latter is presented in three stages: first, to examine the surface structure of Cu thin films on an inert glass substrate; secondly, to determine the local atomic structure of the oxide layer on a polished GaAs( 100) wafer;14 and finally to determine the Cu environment within Cu overlayers on GaAs( 100). 2. Glancing-angle X-Ray Absorption Spectroscopy 2.1. Background Theory X-Rays can be made surface-sensitive by using glancing-angle geometry.When a monochromatic X-ray beam is incident on a flat surface there is a well defined angle of incidence 4c below which the angle of refraction is zero and total external reflection takes place. Total external reflection is possible at the interface between air and a condensed medium because the refractive index of most materials is less than 1 in the X-ray part of the electromagnetic spectrum." The critical angle 4c is related to the refractive index of the condensed medium ( n 2 ) by Snell's law: (1) where 6 = (4nNe2/m)1/E2, p = pA/4n; N is the surface atomic density (atoms cm-'), E is the X-ray photon energy and p is the absorption coefficient of the medium. As p and S are of the order of 10~'-lO-h the numerical value of 4c is small (typically of the order of a few mrad for most materials).Far from an absorption edge p can be neglected cos 4c = n, = 1 - 6 - ip and cos 4c can be expanded to give: 4 , - ( 2 6 4,( mrad) = 20p I where p is the density of the sample (g cm - 3 ) . or in practical units: ?/ E (keV) ( 3 ) From electromagnetic theory, for glancing angles 4<< bc the refracted wave does not penetrate the medium but propagates parallel to the surface, vibrating in a plane perpendicular to it, and is coupled to the standing wave generated by the incident and refractive waves." The penetration depth z of X-rays in the medium is governed by the damping of the evanescent wave in the direction perpendicular to the surface. As 4 decreases z falls to a minimumih such that: (4) Zm,n = A / (47dh.I.S. Pizzini et al.53 zmin is typically ca. 20-50 A for most materials for hard X-ray photons. As 4 is increased past +c the penetration rapidly increases, approaching the limiting case for 4 >> & c , where z = (sin 4 ) p . From the Fresnel equations, the X-ray reflectivity coefficient for photons of energy E incident on a perfectly smooth surface with glancing angle 4 is given by:'6 h - ~ [ 2 ( h - 1)]"2 R( E, x) = h + x [ 2 ( h + l)]''? where x = 4/qhc, h = x'+(x'- 1)'/?+ Y' and Y = P / S . This equation thus shows that the reflectivity R ( E ) is related to both the real (8) and imaginary (P) parts of the refractive index. Thus the reflected X-rays scattered from a sample at energies greater than a component absorption edge will contain extended absorption fine structures (ReflEXAFS) in both P ( E ) and S( E ) , related by the Kramers-Kronig equations.R ( E ) can be written" as: R(P, 6 ) = R(P0, &J+AR(P -Po, -80) (6) where Po and 6,) refer to the non-oscillatory part of p and 6 and ( P - Po) and (6 - &,) are, respectively, the P - and S-EXAFS. The difference between these two components ( A R ) can be expressed as: (7) The relative weights of the p- and S-EXAFS are determined by the function (dP/dS). Martens and Rabe" have shown that the latter approaches zero for glancing angles 4 << 4c. In this condition, the 6-EXAFS can be neglected and the ReflEXAFS are essentially P-EXAFS. By inverting eqn ( 5 ) , the relation between P ( E ) [ e . g . p ( E ) ] and A R ( p -Po, S-Sd = (d~/dP).~CO,,\,[(P -Po)+(dp/dS).,c"",,(S -&)I- R ( E ) can Since the (1 - W/(I be extrapolated.In the case 4 << &c we obtain:" oscillations in S are small compared with the oscillations in the function + R ) , 6( E ) can be approximated to a monotonic function beyond an absorp- tion edge and the ReflEXAFS are therefore quite simply related to the EXAFS oscillations in the absorption coefficient. The same treatment cannot be applied to the case @<< 4c, where the oscillations in S dominate." In this case a Kramers-Kronig analysis must be applied to derive the 6-EXAFS from the experimental R ( E ) . " In this work we restrict ourselves to the case where 4<< 4', where we can use the approximation p == 1 - R, derived from eqn (8) by assuming A R << R.The resulting absorption spectra were normalised, background-subtracted and modelled using the curved-wave least-squares fitting routine EXC'URVXX. I' 2.2. Instrumentation Working at the small glancing angles required for total external reflection geometry imposes tight constraints on incident-beam collimation and sample alignment. We have developed a dedicated instrument"' for Refl EXAFS investigations on the Daresbury SRS (fig. 1 ) ; it has the following features: ( a ) beam definition is provided by precision tungsten carbide slits 50pm thick which are positioned before and after the sample; ( h ) the sample is set on a horizontal goniometer (step size 0.25 mdeg) on which is mounted a motorised translation stage (step size 1 p m ) which is used to align the sample precisely at the focus on the goniometer arc, ( c ) the incident ( I < ) ) and reflected ( I , ) intensities as function of photon energy are measured with ion chambers.A third ion54 I _ _ _ _ _ _ The Cu/GaAs( 100) Interface Fig. 1. Schematic diagram of the experimental set-up for the glancing-angle instrument. chamber ( I , ) is available to record the transmission spectrum of a model compound simultaneously and so provide for precise energy calibration and measurement of any chemical shift. Experimentally the sample is first accurately aligned on the goniometer and a reflectivity curve is collected at a fixed energy close to the absorption edge of interest. From this R ( 4 ) dependence suitable angular positions, subject to the restrictions imposed by the approximations discussed in section 2.1, are chosen.R( E ) is then measured at the desired glancing angles ( 4 ) . ReflEXAFS measurements are carried out on station 9.2"' at the Daresbury SRS, which can provide photons in the 5-35 keV energy range. An Si(220) harmonically rejecting double-crystal monochromator is used; when convoluted with the extended source height of 300pm and the height of the precision entrance slits, it provides an angular resolution of<0.01 mrad. This corresponds to an energy resolution of ca. 5 x lop5, which is more than adequate for these kind of measurements. 3. Application to Cu/GaAs( 100) 3.1. Materials LEC Czochralski grown polished GaAs( 100) wafeTs, cut 2" off-axis, were obtained from ICI Wafer Technology.Cu thin films, ca. l000A thick, on float-glass substrates and 100 and 10 A thick on GaAs( 100) were vacuum evaporated at base pressures typically ca. 10- ' Tort-.? The evaporation rate was monitored using a quartz thickness monitor. 3.2. Cu on Float-Glass The reflectivity curve of the Cu film on glass, recorded at an energy of 8.68 keV is shown in fig. 2( a ) and reveals a critical angle ( 4, = 0.33') and hence a film density less than that expected for pure Cu (4,=0.36"). ReflEXAFS spectra at three glancing angles (4, = 0.3+,, b2 = 0.64, and 4l = 0.754,), corresponding to penetration depths of ca. 18, 21 and 26 A, respectively, were recorded." Fig. 3 compares the near-edge absorption spectrum of the Cu film (angle &) with the transmission spectra of some model compounds and shows that the surface of the C u film is partially oxidised.The edge position and characteristic edge feature indicates that oxidised Cu is present at the surface predominantly as C u ' . Fig. 2 ( b ) and ( c ) show the Cu K-edge k'-weighted experimental ReflEXAFS spec- trum (angle 4?) and the corresponding Fourier transform. The fitted parameters to the .; 1 Torr 1 0 1 315/760 PaS. Plzzini et al. 55 1.0 0.8 L .- > L 0 . 4 .- - (d E 0.2 C 0. 0 0. 00 0. 15 0. 30 0. 45 0. 60 (PIa 4 6 8 10 k l k ' 1.2 0. 0 0 2 4 6 8 10 R I A Fig. 2. Experimental data for 1000 A thick Cu film deposited on float-glass. ( a ) Reflectivity curve recorded at an energy of 8.92 keV also showing angular positions used to record ReflEXAFS spectra.( h ) k'-weighted ReflEXAFS spectrum recorded for glancing angle 4:. ( c ) Fourier transform of ( h ).56 1.2 0. 8 * .- c a & 0.4 m e, v 2 0. 0 .fl B 2 -0.4 -0. 8 The Cu/GaAs( 100) Interface -. .. -. -. .- .- -20 0 20 40 60 80 100 energy/ eV Fig. 3. The near-edge spectra of model compounds Cu ( a ) , Cu,O ( b ) , CuO ( c ) tecorded in transmission compared with the ReflEXAFS spectra of Cu/glass at angle & ( d ) , 100 A Cu/GaAs ( e ) and 10 8, Cu/GaAs d f ) (cf: section 5). The energy values have been scaled to the approximate position of the absorption K-edge for metallic Cu at 8.779 keV. data together with the crystallographic parameters of the model compounds are given in table 1. These show that whilst the film surface is oxidised a substantial amount of residual mTtallic Cu is also present.The latter is evidenced by the strong coordination at ca. 2.5 A which is due to the first shell of f.c.c. metallic Cu. This shell is ca. 50% of the amplitude for bulk Cu and is complemented by an oxygen shell present at ca. 1.85 A for the angles 4 , and 42 and at ca. 1.90 A for &. The former is close to the first oxygen shell in Cu20. The Cu-Cu distances at 2.8-3.0 A are probably due to cation shells of the oxide phase and should be compared with the second-neighbour shell at 3.00 A in bulk Cu20. The splitting of this shell on the ReflEXAFS data suggests that the surface layer has a more distorted structure than bulk Cu20. The weightings of the coordination numbers of the first (oxygen)- and second (metallic copper)-neighbour shells are consistent with a mixed metal-oxide surface layer with each phase giving a contribution of ca.50%. However, the angular dependence of the coordination numbers of the 0 and Cu cation shells are indicative of a composition which varies slightly with depth, the oxide phase being more pronounced in the outmost surface layer. The longer Cu-0 distance typical of CuO, obtained for the lowest penetration depth (angle 4 , ) confirms this trend by being indicative of a higher associ- ation to oxygen in the outmost surface layer. 3.3. The Oxide Coating on GaAs(100) Fig. 4(a) shows the reflectivity curve from a polished 3 in? GaAs( 100) wafer at an energy of 11.78 keV, which is close to the ASK absorption edge. Associated ReflEXAFS spectraS.Pizzini et al. 57 Table 1. Shell distances R ( A ) , coordination numbers N and Debye-Waller factors 2 0 2 (Az) obtained from least-squares fits of the ReflEXAFS spectra of the Cu film deposited on glass atom N 0 1.4 c u 5.2 c u 3.9 c u 6.1 c u 3.7 0 1.2 c u 6.6 c u 2.6 c u 5.0 c u 1.3 0 1.1 c u 6.7 c u 2.3 c u 3.2 c u 3.0 R 2rr2 41 1.91 2.53 2.81 3.05 3.43 42 1.85 2.52 2.77 2.98 3.45 43 1.86 2.54 2.81 3.01 3.49 0.010 0.018 0.015 0.027 0.029 0.0 10 0.018 0.015 0.027 0.029 0.010 0.018 0.015 0.027 0.029 ~~ ~ Cu (metal) c u z o CUO atom N R atom N R atom N R c u 12 2.556 0 2 1.841 0 4 1.947 c u 6 3.615 c u 12 3.007 0 2 2.766 0 6 3.526 c u 4 2.884 c u 4 3.071 c u 2 3.159 c u 2 3.410 c u 2 3.727 The three glaqcing angles + I , 4z and 43 are as indicated in fig.2(a). The crystallographic values of N and R ( A ) for Cu, Cu20 and CuO are also reported. at two glancing angles ( 4 , = 0.604, and 42 = 0.704,), corresponding to penetration depths of ca. 25 and 30 A, respectively, were recorded above the ASK and GaK absorption edges. The ASK-edge k'-weighted ReflEXAFS spectrum (angle 4,) and the correspond- ing Fourier transform are shown in fig. 4( b ) and ( c ) . The structural parameters obtained from the least-squares fitting of these data are given in table 2 together with the crystallographic parameters of some model compounds. These results show the surface layer to be partially oxidised whilst the persistence of correlations characteristic of the GaAs substrate indicates that the oxide thickness is less than the X-ray penetration depth.The closeness of the ReflEXAFS fits to the GaAs bulk structure confirms that phase separation between the GaAs substrate and its oxygen-rich surface phase occurs. The approximate thickness of this layer, estimated from the penetration depth and the observed coordination numbers, is ca. 7-9 A." Ga and As atoms have different oxygen environments in the 2urface oxide. There is an oxygen neighbour shell for both Ga and As atoms at ca. 1.7 A, close to that found for the tetrahedral coordination in GaAsO, or three-fold coo!dination in As,03. However, Ga atoms have an additional oxygen shell at ca. 1.95 A which is typical of58 The Cu/GaAs( 100) Interface 1.0 2 0.8 .- > .- * 2 0. 6 2 B g 0. 2 vl 0. 4 .- - 2 c 0. 0 0. 0 0. 1 0. 2 0. 3 cpl" I 4 6 8 10 k1A-I 0 2 4 6 8 10 R I A Fig.4. Experimental data for polished GaAs. ( a ) Reflectivity curve recorded at an energy of 11.78 keV, showing the positions of the two glancing angles 4, and 4: at which ReflEXAFS spectra were recorded. ( h ) As K-edge k'-weighted ReflEXAFS spectrum recorded at glancing angle 4 , . ( c ) Fourier transform of ( h ) .Faraday Discuss. Chem. SOC., 1990, vol. 89 S. Pizzini et ai. (Facing p . 5 9 )S. Pizzini et al. 59 Table 2. Shell distances R ( A ) , coordination numbers N and Debye-Waller factors 2 a 2 (A’) obtained from least-squares fits of Ga and ASK-edge ReflEXAFS spectra of the GaAs( 100) wafer for two glancing angles 4 , and 42 < 4?) (M stands for As or Ga) Ga-0 Ga-0-M Ga-As Ga-Ga R N 2u’ R N 2u’ R N 2u2 R N 2a2 __ 41 1.72 0.6 0.001 2.88 3.1 0.006 2.46 3.0 0.008 3.92 10.1 0.026 1.95 2.2 0.002 3.15 3.4 0.005 42 1.66 1.3 0.008 2.81 3.0 0.005 2.49 2.4 0.022 3.97 4.5 0.018 1.96 2.7 0.003 3.10 5.4 0.008 As As-0 As-0-M As-Ga As- As R N 20’ R N 2a’ R N 2a’ R N 2u2 __ - ____________ 41 1.68 0.9 0.002 2.84 1.1 0.012 2.43 3.1 0.012 3.95 10.4 0.028 3.09 1.0 0.012 42 1.68 1.1 0.001 2.84 1.0 0.006 2.44 3.0 0.012 3.97 9.0 0.026 3.10 1.2 0.007 Ga-0 As-0 Ga- As Ga-Ga As- As R N R N R N R N R N GaAs 2.44 4 3.99 12 GaAsO, 1.77 4 3.16 4 1.72 4 3.16 4 Ga,O, 1.92 3 2.83 1 2.08 3 2.94 3 3.3 1 3 As20, ‘ I 1.79 3 3.22 3 The crystallographic values of R ( A ) and N for GaAs and some Ga and As oxides are also reported.I’ Arsenolite phase of As,O,. the octahedral coordination of Ga,O,. Thus while As in the surface oxide seems to be exclusively in a tetrahedral environment, Ga exists in both tetrahedral and octahedral coordinations.The Ga (As) coordination shell distances at ca. 3 A resemble the second-she11 distances of Ga and As oxides (table 2). Two cation shells are present, at the same distances (ca. 2.8 and ca. 3.1 A) for both Ga and As central atoms. These are probably associated with the two different oxygen coordinations for Ga. The shell at ca. 2.8 A possibly arises from a coordination between cations both in tetrahedral environment, whilst the shell at ca. 3.1 A shell is related to the coordination between octahedral Ga and a tetrahedral cation (As or Ga). The matching of these second-shell distances for both Ga and As central atoms show the oxide to be a single component which can thus be modelled as a microscopically random mixture of tetrahedral and octahedral sites,60 The Cu/GaAs( 100) Interface with Ga occupying both and As only the former four-fold site.Conversely, if the surface comprised a macroscopic mixture of the separate oxide phases, as suggested before," then only one second shell would be present to complement the single oxygen shell for As. Comparison between the Ga and As environments as a function of glancing angle shows no significant variation for As concentration while indicating a distinct increase in Ga coordination. Together with the angle dependence of the 0-shell amplitudes, this trend indicates a greater Ga association to 0 in the outmost oxide layer and a correspond- ing As depletion.Calculation of the oxygen-bridging cation-cation bond angles for the ca. 2.8 A correlation show this to be close to tetrahedral, which indicates that oxygen atoms may initially enter the GaAs structure in substitutional sites. The depletion of As atoms from the GaAs( 100) surface suggests that these are the sites where the oxygen substitution preferentially takes place. However, the weighting of the coordination numbers of the oxygen and the Ga (As) oxide shells suggest that this substituted structure is less well packed than the substrate. Also, a consistent number of octahedral sites is present in the oxide layer, especially in the near surface. Plate 1 shows a structural model based on this analysis. This was constructed as a continuous random network (CRN)".'" built on a GaAs(100) substrate.Oxygens were attached to the dangling bonds at the surface and the CRN was built up from a mixture of octahedral and tetrahedral Ga atoms, tetrahedral As atoms and bridging 0 atoms. Bond lengths were closely constrained to the measured oxygen distances, but the tetrahedral bond angle was allowed to vary. An open structure results. The presence of two-fold coordinated oxygens and the variable cation coordination combine to encourage the formation of micro-voids (typical size ca. 5 A ) and fissures whose internal surfaces are oxygen-rich. 4.3. Cu Overlayers on GaAs(100) Fig. 5 ( a ) compares the reflectivity curves for 100 and 10A thick Cu overlayers on GaAs( 100) at an energy of 8.68 keV. The measured critical angles (4cl =0.310° and Scz = 0.278") should be compared with those calculated from the electron density for a clean surface which for Cu, Cu,O and GaAs are 0.364, 0.310 and 0.284", respectively.The relationship between 4c and density [eqn (3)] shows the thicker Cu overlayer to have a density similar to that of Cu,O. As the thickness of the lOA Cu overlayer is less than the penetration depth of X-rays, the reflectivity process is expected to be determined by the GaAs substrate interface, with the difference in 4,.. with respect to the calculated value being probably due to the presence of a graded interface." Comparison of the near-edge spectra of the two Cu overlayers on GaAs( 100) [fig. 3( e ) and (f)] with the model-compound data show subtle differences with respect to the Cu/glass data discussed in section 3.2.Whil2t the position of the edge feature indicates the presence of Cu' cations in both the 10A and the 1008, overlayers, the additional presence of a shoulder at ca. 7 eV above the edge, close to that observed in CuO, suggests a non-stoichiometric oxide comprising both Cu' and Cu" cations. Additionally, the XANES spectra for Cu/GaAs are typical of a completely oxidised surface, in contrast to Cu/float-glass which show residual features typical of metallic Cu. Background-subtracted ReflEXAFS spectra for 100 A ( 4 , = 0.44,) and 10 8, (& = 0.5 thick Cu films on GaAs are overlayed and presented together with their corresponding Fourier transforms in fig. 5 ( 6) and ( c ) . The structural parameters obtained from least-squares fits of the data are resumed in table 3.In both the thinner and the thicker overlayer we find that Cu is completely oxidised. In the thicker layer no significant long-range order is observed. This is consistent with a disordered oxide overlayer being formed from the amorphous oxide covering GaAs( 100).S. Pizzini et al. 61 1.0 3 0.8 .- > .- c) 2 0. 6 -0 2 % 0.4 8 0.2 .- - c 0. 0 0. 6 15 10 5 M h -Y * s o -5 -10 2 4 6 8 10 k / k ’ 0 1 2 3 4 5 R / A Fig. 5. Experimental data for Cu on GaAs showing overlayed data from 100 A (-) and 10 8, thick (. - - * . * ) Cu overlayers on GaAs(100). ( a ) Reflectivity curves recorded at an energy of 8.68 keV. ( b ) Cu K-edge k3-weighted ReflEXAFS spectra. ( c ) Fourier transforms of ( b ) . The C u - 0 distances in the Cu oxide are longer than the more ordered phase observed on glass and are more typical of four-fold coordination in CuO.The more extensive Cu oxidation on GaAs compared to float-glass substrate is presumably due to the open structure of the oxide layer on GaAs. The cracks and micro-voids in this layer will encourage mass-transport and hence the reactivity of this interface, Cu initially bonding covalently via lone-pair electrons of the 0 atoms decorating the internal surfaces. In the thinner Cu layer, additional cation correlations are found at higher distances. Whilst the similar back-scattering amplitudes of Cu, Ga and As make an unambiguous assignment of these correlations very difficult, the amorphous nature of the oxide62 The Cu/GaAs( 100) Inferface Table 3.Shell distances R ( A ) , coordination numbers N and Debye- Waller factors 2a' (A') obtained from least-squares fits of Fourier-filtered CuK-edge ReflEXAFS spectra of Cu/ GaAs( 100) atom N R 2 2 100 A 0 3.2 1.95 0 2.3 2.60 0.01 1 0.008 10 A 0 4.3 2.00 0.009 0 2.4 2.8 I 0.009 M 3.0 (10.2) 3.00 (kO.003) 0.010 M 5.3 (k0.5) 3.80 (10.004) 0.010 M is Cu, As or Ga. The values reported in parentheses refer to the variations in the parameters obtained for M = Cu, As or Ga. discussed above seems to preclude the presence of Cu cations at these larger distances. Additionally, as the first Cu-M (M = Ga, As or Cu) correlation is calculated at a distance (ca. 3.00*0.03 A) close to the cation-cation distance in the oxide coating on GaAs (table 2); this suggests that Cu trapped in the voids in the surface oxide correlates with the surrounding cations.The open oxide structure on GaAs should promote Cu inter- diffusion into the oxide. Such cation diffusion is commonly observed in chalcogenide glasses." Alternative models suggesting the formation of arsenides and alloys can also be ruled out. Cu-As first-shell distances in Cu,As or Cu,As are typically ca. 2.45-2.5 A, whilst Cu-Ga distances in Cu-Ga alloy are expected at ca. 2.55 A. The absence of coordination shells at such short distances tends to support previously published data"' in excluding the formation of Cu arsenides or Cu-Ga alloy at the Cu/GaAs(100) interface. In the thicker Cu overlayer Cu-M correlations are not present and the oxygen coordination number is smaller.I t would seem that the reactivity of Cu in the structure of the thinner Cu overlayer catalyses extensive oxidation when more Cu is present. Where Cu passivates on the surface of float-glass, leaving approximately half of the Cu as metal, the opposite appears to be the case for GaAs( 100) where all the Cu becomes oxidised. As the Cu coverage builds up it would appear that the surface oxide generated is increasingly disordered, explaining the absence of Cu-M correlations and the reduced oxygen coordination numbers. 4. Conclusions X-Ray absorption spectroscopy under conditions of total external reflection provides a useful technique for the structural characterisation of 'real' surfaces and interfaces. Structural information from XANES, EXAFS and reflectivity can be combined in a single series of measurements to provide a cross-correlated structural probe.By varying the glancing angle within the total external reflection region, quite subtle depth-resolved structural variations can be characterised. This technique has been applied to a structural investigation of the Cu/GaAs( 100) interface. Examination of Cu thin films deposited on float-glass substrates reveals the surface to consist of a mixture of metallic and oxide (predominantly Cu') phases. The oxide is found to have fairly long-range structural order but with a local environmentS. Pizzini et a]. 63 distorted with respect to that expected for a bulk Cu' oxide. The surface oxide of GaAs( 100) appears to have an open amorphous structure consisting of microscopically random distributions of octahedrally coordinated Ga and tetrahedrally coordinated Ga and As.Cu deposited onto the GaAs surface is completely oxidised, producing an amorphous phase of four-fold Cu. At the interface correlations are seen, which suggest the Cu bonds to the oxygen-rich internal surfaces present in the voids and micro-fissures which characterise the natural oxidised surface of GaAs( 100). No trace of arsenides or Cu-Ga alloy formation is present at the interface. Further applications of these techniques can be envisaged to other 'real' interfaces such as those involved in catalysis, electrochemical processes, ion-implantation and low-dimensional solids. Developments associated with these techniques are also cur- rently in p r ~ g r e s s , ~ ~ * ~ ' notably the incorporation of powder X-ray diffraction within our general set-up, which is directed towards a complete characterisation of interfaces using synchrotron radiation.This work has been supported by an S.E.R.C. cooperative research project in collabor- ation with ICI Chemicals and Polymers plc. We gratefully acknowledge S.E.R.C. for provision of beam time on the Daresbury SRS, and K. Singer who prepared the Cu overlayers on GaAs. S.P. acknowledges ICI Chemicals for the financial support of a research assistantship. References 1 S. H. Pan, D. Mo, W. G. Petro, I . Lindau and W. E. Spicer, J. Vac. Sci. Technol. B, 1983, I , 593. 2 W. G. Petro, T. Kendelewicz, I . Lindau and W. E. Spicer, P h j ~ . Rev. B, 1986, 34, 7089. 3 W. G.Petro, I. A. Babalola, P. Skeath, C. Y. Su, I . Hino, I. Lindau and W. E. Spicer, J. Vac. Sci. Technol., 1982, 21, 585. 4 I . Lindau, T. Kendelewicz, N. Newman, R. S. List, M. D. Williams and W. E. Spicer, Surj: Sci., 1985, 162, 591. 5 R. E. Viturro, C. Mailhiot, J. L. Shaw, L. J. Brillson, D. La-Graffe, G. Margaritondo, G. D. Pettit and J. M. Woodall, J. Vac. Sci. Technol. A, 1989, 7, 855. 6 R. Cao, K. Miyano, T. Kendelewicz, I . Lindau and W. E. Spicer, Appl. Phj-s. Lett., 1988, 53, 210. 7 J. F. McGilp and A. B. McLean, J. Phjx C, 1988, 21, 807. X K. A. Bertness, J . J. Yeh, D. J. Friedman, P. H. Mahowald, A. K. Wahi, T. Kendelewicz, I . Lindau and W. E. Spicer, Phj~s. Rec. B, 1988, 38, 5406. 9 S. M. Sze, Phj~sics of.Semicotiductor Detliceq (Wiley, New York, 1981). 10 S. P. Kowalczyk, J. R. Waldrop and R. W. Grant, J. Vae. Sci. Technol., 1981, 19, 611. 1 1 G. Martens and P. Rabe, PI?j*s. Star. Sol. ( a ) , 1980, 58, 415. 17 S. J . Gurman and R. Fox, Philos. Mag. B, 1986, 54, L45. 1.1 S. M. Heald, H. Chen and J . M. Tranquada, Phj>s. Rev. B, 1988, 38, 1016. 14 N. T. Barrett, G . N. Greaves, S. Pizzini and K. J. Roberts, Surf Sci., 1990, 227, 337. I 5 R. W. James, The Crj.stalline State, Vol. / I , The Optical Principles of the Djfraction ot'x-raj-s ( G . Bell and Sons, 1969). I6 L. G. Parratt, Ph.rs. Rer., 1954, 95, 359. 17 L. Bosio, R. Cortes, A. Defrain and M. Froment, J. Electroanal. Chem., 1984, 180, 265. I X S. J. Gurman, J. fhj*s. C', 1988, 21, 3699. 19 S. Pizzini, K. J. Roberts, G. N. Greaves, N. Harris, P. Moore, E. Pantos and R. J. Oldman, Rec. Sci. /t?s/runi., 1989, 60, 2525. 2t) G . P. Diakun, G . N. Greaves, S. S. Hasnain and P. D. Quinn, Daresbury Laboratory Report, I)L/SC'I/TM38E (1984). 21 S. Pizzini, K. J . Roberts, I . Dring, R. J . Oldman and G. N. Greaves, P/ij.sica B, 1989, 158, 676. 22 W. M. Lau and R. P. Bult, Mater. Lett., 1987, 5, 88. 23 I). E. Polk and D. S. Roudreaux, Phj-s. Rep., 1973, 31, 92. 24 G. N . Grea\,es and E. A. Davis, Philos. MU,^., 1974, 29, 1201. 75 A. T. Steel, G. N . Greaves, A. P. Firth, A. E. Owen, J. Non-C'rj-st. Solids, 1989, 107, 1 5 5 . 26 N . Barlow. C. Hrennan, S. E. Doyle, G. N. Greaves, M. Miller, A. H. Nahle, K. J. Roberts, J. Robinson, J . N. Sherwood and F. C. Walsh, Reil. Sci. Instrum., 1989, 60, 2386. 27 <;. Derbyahire, B. Dobson, G . N. Greaves, N. Harris, P. Mackle, P. R. Moore, K. J . Roberts, N . Allinson, J . Nicoll, S . Doyle and R. J . Oldman, Rer. Sci. Inctrirm., 1989, 60, 1897. Paper 0/00293C; Receiued 18rh Jatiuar!., 1990