XPS and SIMS analysis of gold silicide grown on a bromine passivated Si(111) substrate

Applied Surface Science 137 Ž1999. 103–112

XPS and SIMS analysis of gold silicide grown on a bromine passivated Si ž111 / substrate B. Sundaravel a,1, K. Sekar a,2 , G. Kuri a,3, P.V. Satyam a,4 , B.N. Dev a,) , Santanu Bera b, S.V. Narasimhan b, P. Chakraborty c , F. Caccavale d b

a Institute of Physics, SachiÕalaya Marg, Bhubaneswar 751 005, India Water and Steam Chemistry Laboratory, Applied Chemistry DiÕision, BARC, IGCAR Campus, Kalpakkam 603 102, India c Saha Institute of Nuclear Physics, Sector-1, Block-AF Bidhan nagar, Calcutta 700 064, India d UniÕersita di PadoÕa, Dipartimento di Fisica, Õia Marzolo 8, 35131 PadoÕa, Italy

Received 8 May 1998; accepted 6 August 1998

Abstract When a thin film of Au Ž; 100 nm. deposited under high vacuum conditions on a chemically prepared Br-passivated SiŽ111. substrate was annealed around 3638C, epitaxial layer-plus-island mode growth of gold silicide was observed along with some unreacted gold in stringy patterns. This unreacted gold was removed by etching the sample in aqua regia. X-ray photoelectron spectroscopy ŽXPS. and secondary ion mass spectrometry ŽSIMS. measurements were carried out on these samples. SIMS results reveal that the height of the islands is about 1.2 mm and the siliciderSi interface is abrupt. XPS measurements were made after sputtering the sample surface at constant intervals of time. Si 2 p, Au 4 f, C 1 s and O 1 s photoelectrons were detected. XPS spectra of Si 2 p are resolved into three peaks corresponding to bulk Si, Si in silicide and Si in oxide. The Au 4 f 7r2 peak in the silicide is shifted by 1–1.2 eV towards higher binding energy compared to metallic Au. The shift of Si 2 p towards the higher binding energy in the silicide is understood from the higher electronegativity of Au, while the shift of Au 4 f 7r2 peak towards higher binding energy is known to be due to d-electron depletion to form an sd hybrid. The XPS peak intensity profile with sputtering time indicates that the thin uniform layer Ž; 5.5 nm. of gold silicide is sandwiched between a thin Ž; 2.8 nm. SiO 2 layer and the SiŽ111. substrate. q 1999 Elsevier Science B.V. All rights reserved. Keywords: Gold silicide; XPS; SIMS; Interface

)

Corresponding author. E-mail: [email protected]; Fax: q91-674-581142. Present address: Department of Electrical Engineering and Materials Technology Research Centre, The Chinese University of Hong Kong, Shatin, Hong Kong, China. 2 Present address: FB Physik, Universitaet Osnabrueck, D-49069 Osnabrueck, Germany. 3 Present address: Surface Analysis Group, Korea Research Institute of Standards and Science, Yusung P.O. Box 102, Taejon, 305606, South Korea. 4 Present address: B3155, Building 401, Argonne National Laboratory, Argonne, IL 60434, USA. 1

0169-4332r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 9 - 4 3 3 2 Ž 9 8 . 0 0 3 7 8 - X

104

B. SundaraÕel et al.r Applied Surface Science 137 (1999) 103–112

1. Introduction Metal-semiconductor interfaces have been studied extensively owing to their variety of interesting properties which are both scientifically and technologically important. AurSi interface has been of interest owing to its possible application as a metal-Si contact in integrated circuit devices. As Au atoms diffuse very fast and serve as deep-trap centers in Si, AurSi contacts have been found unsuitable for this purpose. But the interface is of interest for fundamental reasons. Though Au is usually regarded as a nonreactive stable noble metal, it has been reported to be very reactive on Si surfaces even at room temperature w1,2x. At elevated temperatures this reactivity can lead to epitaxial growth of gold silicide, which shows a fundamental phenomenon like shape transition in heteroepitaxial islands w3x. Growth mode, orientation and structure of noble metal ŽAu, Ag. films on semiconductors ŽSiŽ111.. have been studied by many groups w4x. For Au, at low temperature, surface compounds have been observed due to the high diffusion rate of Si through gold layers w5,6x. By studying the Si outdiffusion from crystalline substrates through very thin Au films Ža few monolayers., several authors have reported the formation of Au x Si Žwith x f 4. alloys before oxidation of Si at the top surface starts w7,8x. Chang and Ottaviani w9x observed silicon oxide formation on a deposited Au layer on silicon. They observed a 15 times stronger Si outdiffusion through an Au film from SiŽ100. substrates compared to that from SiŽ111. substrates. The corresponding ratio of oxide layer thickness Ž dŽ100.rdŽ111. . was ; 15 and they attributed this to different binding energies of the Si atoms at the substrate surface. Hiraki et al. w2x observed an oxide layer thickness ratio of dŽ110.rdŽ111. f 5 for Ž110. and Ž111. substrate orientations. Amorphous substrates show a much stronger Si outdiffusion than the crystalline substrates w10x. The presence of H at the interface further enhances this outdiffusion for both amorphous and crystalline substrates by approximately a factor of two. The metastable phase Au 4 Si has been detected on both interfaces of the Au film Žsilicon-oxiderAu as well as AurSi interfaces.. Upon condensation at high temperature Žbeyond ; 4008C. the growth of the Au film follows the

Stransi–Krastanov mode, i.e., layer plus islanding w11x. Green and Bauer obtained crystalline gold silicide with different structures upon annealing Au films on SiŽ111. substrates deposited at temperatures between 2008 and 4008C w12x. Oura and Hanawa have shown that crystalline gold silicide thin layers are formed in the AuŽ111.rSiŽ100. system after heat treatment at 100–4008C w13x. They reported that different silicides are formed on the Ž111. planes of an Au film according to whether they are grown on the Ž111. face of Si or on the Ž100. face. They have argued that the difference in imperfection andror strain in the Au film affects the structure of the silicide. Their results suggest that silicide layers are formed both on the outer surface of the Au film and at the AurSi interface. Most of the studies discussed above were performed under ultrahigh vacuum ŽUHV. environment and at low coverages of Au. Presence of a native oxide layer at the interface strongly influences the interdiffusion behavior across a metal–semiconductor interface. There have been many studies on this aspect. To have a clean interface one usually works under UHV conditions so that oxide formation can be avoided. Oxide growth can also be inhibited under non-UHV conditions by the passivation of surface dangling bonds. It has been shown that bromine passivates dangling bonds of the Ž111. surface of silicon w14–17x, and this passivated structure is stable for several days in dry air. This type of chemical passivation of the SiŽ111. surface has been found to hinder oxide growth on these surfaces w18–20x. Moreover, bromine–silicon, and in general halogen–silicon, is an important system in semiconductor technology. We have undertaken a study of growth of noble metal layers on bromine–passivated SiŽ111. surfaces. It is noteworthy that the interdiffusion behavior for Cu deposited on bromine passivated SiŽ111. Žhereafter denoted as Br–SiŽ111.. surfaces is similar to that for Cu deposited on atomically clean SiŽ111. surfaces under UHV conditions w18,19,21x. Detailed characterization of Br–SiŽ111. surfaces have been published elsewhere w20x. When Au was deposited on Br–SiŽ111. surfaces, and then annealed in high vacuum around 3638C ŽAu–Si eutectic temperature. for 30 min, several interesting features were observed. We observed epitaxial growth of gold silicide in Stranski–Krastanov

B. SundaraÕel et al.r Applied Surface Science 137 (1999) 103–112

mode with islands in triangular and trapezoidal shapes w3x along with stringy patterns of unreacted gold constituting a fractal structure w22x on a thin uniform background of gold silicide. These features are seen in the optical micrograph shown in Fig. 1. The growth of equilateral triangular islands reflect the three-fold symmetry of the underlying SiŽ111. substrate. The island edges are aligned along the three equivalent  1104 directions of the SiŽ111. substrate. The elongated islands, instead of growing along three equivalent  1104 directions, grow only along one preferential direction which has been attributed to the vicinality of the substrate. Detailed observations have been published elsewhere w23x. The elongated trapezoidal islands show the phenomenon of shape transition w3x which has recently been identified as a major mechanism for strain relief in heteroepitaxial systems w24x. Our gold siliciderSi system shows a shape transition with triangular islands going over to trapezoidal ones w3x. The composition of the triangular and trapezoidal islands were determined to be of Au 4 Si by energy dispersive X-ray analysis ŽEDX. w23x. From the color optical micrographs Žnot shown. it is evident that the color of the islands and the uniform thin layer are identical. They are not golden indicating that they are not unreacted gold. However, the stringy structure that is also present has golden color indicating that it is composed of unreacted gold. From EDX analysis we

Fig. 1. Optical micrograph of an AurBr–SiŽ111. sample annealed around 3638C ŽAu–Si eutectic temperature. in high vacuum for 30 min. Gold silicide island formation together with a stringy pattern of gold are seen. Field of view of the micrograph is 400 mm.

105

could not estimate the composition of the thin uniform silicide layer because this layer is too thin compared to the penetration depth of the incident electron beam in the EDX analysis. Thus, there is an excessive contribution of Si X-rays from the underlying silicon substrate. In our earlier studies, several questions remained unanswered: Ži. What is the chemical state of the gold silicide? Žii. What is the thickness of the silicide islands? Žiii. What is the thickness of the thin uniform silicide layer? Živ. What happened to Br atoms initially present at the interface? In the present work, we have attempted to answer these questions. Gold silicide is metastable and several metastable metallic phases, both amorphous and crystalline are obtained by quenching from the liquid or vapor states or by ion-beam irradiation w25x. The concept of stabilization of metastable phases by epitaxy has been discussed by Farrow w26x. We have demonstrated the crystallinity and epitaxial nature of the thin gold silicide layer and the triangular and trapezoidal islands by Rutherford backscattering spectrometry and ion channeling measurements w27x. When silicide is formed by annealing an Au film on Si, it is likely that different metastable silicides will be formed at different temperatures. The composition of Au x Si y alloy is found to vary with temperature w28,29x. There are many ultraviolet photoemission and X-ray photoelectron spectroscopy ŽXPS. studies on gold silicide formed by room temperature deposition of a few monolayers ŽML. of Au onto SiŽ111. w30–33x. Brillson et al. have reported photoemission studies on AuŽ2 nm.rSiŽ111. annealed at 2008C w34x. Kolmakov et al. have done synchrotron radiation photoemission studies on AuŽ6 ML.rSiŽ111.Ž7 = 7. annealed at 6278C w35x. There are no XPS studies on gold silicide formed by annealing AurSiŽ111. around the Au–Si eutectic temperature Ž3638C. to our knowledge. In this work we present XPS and secondary ion mass spectrometry ŽSIMS. studies on the AurBr–SiŽ111. system annealed in high vacuum around 3638C. From XPS studies, our aim is to determine the chemical state of the gold– silicon compound and the thickness of the thin uniform silicide layer formed during the Stranski– Krastanov growth. From SIMS we aim to determine the height of the gold silicide islands and the redistribution of interface Br.

106

B. SundaraÕel et al.r Applied Surface Science 137 (1999) 103–112

2. Experimental SiŽ111. wafers were cleaned in distilled water, methanol, trichloroethylene, again in methanol and finally etched in 48% hydrofluoric acid ŽHF.. A bromine–methanol solution Ž0.05% bromine by volume. was squirted onto the sample, while the sample was taken out of HF, taking care that the sample was not exposed to air prior to being wet by the bromine–methanol solution. By this procedure w14,15x or by cleaving a Si crystal in the bromine– methanol solution w16,17x it was shown that bromine saturates dangling bonds of the SiŽ111. surface. This chemisorbed bromine is stable on SiŽ111. surfaces in dry air w14–17x and this chemisorption hinders oxide growth w18–20x. Bromine–chemisorbed SiŽ111. surfaces have been characterized by XPS w20x. Au was deposited onto Br–SiŽ111. substrates by simple resistive heating evaporation from a W basket. The deposition was carried out under a base pressure of 4 = 10y6 Torr. The thickness of the Au layer as measured from Rutherford backscattering spectrometry ŽRBS. is ; 100 nm. Then the AurBr–SiŽ111. sample was annealed at 363 " 108C Žnear the gold– silicon eutectic temperature. under a pressure of 2 = 10y5 Torr for 30 min. The samples were taken out in air. When viewed under an optical microscope, the samples showed some island structures on top of a uniform thin layer of gold silicide along with some stringy unreacted gold ŽFig. 1.. The unreacted gold was removed by etching the sample in an aquaregia solution for about 5 min. This etched sample will henceforth be called as ‘gold siliciderSiŽ111.’ sample. XPS and SIMS studies were carried out on this gold siliciderSiŽ111. sample. XPS measurements were carried out a few days after the sample had been prepared using Al-K a X-rays Ž1486.6 eV, width 0.85 eV. from a VG-Scientific ESCALAB Mk200X machine and a hemispherical analyzer of 0.4 eV resolution. The pass energy of the electron energy analyzer was 20 eV. The energy scale of the spectrometer has been calibrated with pure Ag and Cu samples. The full-width at half maximum ŽFWHM. of the Ag M5Ž3d 5r2 . Ž367 eV. photoelectron line was measured to be 1.7 eV. The pressure in the XPS analysis chamber was ; 8 = 10y1 0 Torr. For sputtering, 5 keV Arq ions

were used with a beam current of 1.2 mA on the sample. The sputtering rate of the sample was 0.22 nmrmin. XPS measurements were carried out after sputtering the sample surface in intervals of 5 min. Si 2 p, Au 4 f, C 1 s and O 1 s photoelectrons were detected. For a comparison with elemental gold, we have made XPS measurements on an as-deposited thin Ž; 100 nm. gold film on a native oxide covered SiŽ111. substrate where there should not be any significant interdiffusion w18,19,23x. SIMS measurements were carried out using a CAMECA-ims 4f ion microscope. With a 14.5 keV Csq primary beam, secondary negative ions of Si, Au, Br and O were detected. The Csq beam current of 10 nA, was rastered over an area of 100 mm = 100 mm. The beam dimension was from less than 1 mm to 20 mm. The secondary ions emitted from a central region Ž10 mm diameter. were collected by the spectrometer by using a field aperture. The erosion speed was evaluated by measuring the depth of the erosion crater at the end of each analysis by a Tencor Alpha-step profilometer. The erosion speed was about 20 nm per minute. All the measurements were carried out under a pressure of 2 = 10y9 Torr inside the chamber. 3. Results and discussions 3.1. A. XPS results 3.1.1. Binding energy shift of Si 2p XPS measurements were made on gold siliciderSiŽ111. samples after sputtering the sample surface at intervals of 5 min. Altogether ten spectra were taken. Si 2 p, Au 4 f, C 1 s and O 1 s photoelectrons were detected. Fig. 2 shows some of the Si 2 p spectra taken at different sputtering times. One can see that the peak at higher binding energy ŽB.E.. is prominent in the spectrum taken before sputtering. The intensity of this peak decreases with a simultaneous increase in intensity of the peak at lower B.E. with sputtering time. The peak at the higher B.E. is broader and contains more than one component. All the spectra were fitted with the assumption of three Gaussian peaks in the spectrum, one due to bulk silicon and the other two chemically shifted components from gold silicide and silicon oxide. One such fit with three components for a Si 2 p spectrum,

B. SundaraÕel et al.r Applied Surface Science 137 (1999) 103–112

Fig. 2. Si 2 p XPS spectra of the gold siliciderSiŽ111. sample after sputtering at constant intervals of time.

taken after sputtering the sample for 20 min, is shown in Fig. 3. The average positions of the three components are 99.3, 100.6 and 103.0 eV. The peak appearing at 99.3 eV with a FWHM of 1.46 eV corresponds to elemental or bulk Si. The component appearing at 103.0 eV with a FWHM of 2.58 eV is due to Si in SiO 2 . The energy shift of 3.7 eV for SiO 2 is in agreement with earlier results on thin SiO 2 layers w20,36x. It is also evident from the fact that the reduction of this peak intensity with sputtering time is correlated with the reduction of the O 1 s peak intensity ŽFig. 5. which will be discussed later. Larger Ž2.58 eV. FWHM value of the SiO 2 peak compared to the FWHM of the elemental peak Ž1.46 eV. is perhaps due to the presence of a small mixture of suboxides SiO 2y x with different values of x. Contributions from different oxidation states of Si can give peak shifts of different amounts compared to bulk Si ŽSi1q: q1.0 eV; Si 2q: q2.0 eV; Si 3q: q2.9 eV; Si 4q: q3.7 eV. w36x. Our observed shift of 3.7 eV for the oxide peak implies that the component of the highest oxidation state, Si 4q or, SiO 2 is predominant although small contributions from lower and higher B.E. components may be responsible for the broadening. Although from the above discussions it appears that a B.E shift larger than 3.7 eV is unlikely, a B.E. shift of q4.4 eV has been observed in thermal SiO 2 , which was explained in terms of a ‘structure induced chemical shift’ w37x. There may be some ambiguity in the assignment of the peak at 100.6 eV. A peak approximately at

107

this energy would be expected due to the suboxide with Si oxidation state mainly as Si1q. However, this assignment cannot be accepted. If this were the case, the intensity of this peak would follow that of the oxide peak Ž103.0 eV. and the O 1 s peak as a function of sputtering time. From Fig. 5, we see that the trend is opposite. We assign the peak at 100.6 eV with a FWHM of 2.55 eV to Si 2 p states in the silicide, i.e., to bonding between Au and Si in gold silicide. This assignment is also validated by the similarity of the intensity variation of this component and Au 4 f 7r2 with sputtering time ŽFig. 5.. In our case, the AurBr–SiŽ111. sample was annealed close to the Au–Si eutectic temperature Ž3638C. and the gold silicide composition is Au 4 Si ŽAu–Si eutectic composition. as observed from the EDX results on gold silicide islands w23x, although the precise chemical phase is not known. The shift of Si 2 p towards the higher binding energy in gold silicide is consistent with a charge transfer from Si to Au due to a higher electronegativity of Au Ž2.4. than that of Si Ž1.8. w30,34x. Depending on the sample depth we observed the B.E. of 2 p silicide peak to vary between 100.6 and 101.0 eV—the higher B.E. corresponding to the surface Žno sputtering.. 3.1.2. Binding energy shift of Au 4f Au 4 f XPS spectra of elemental Au from an as-deposited Au film on native oxide covered SiŽ111.

Fig. 3. Si 2 p XPS spectrum of the gold siliciderSiŽ111. sample sputtered for 20 min. The spectrum is fitted with three Gaussian components corresponding to bulk Si, gold silicide and silicon dioxide.

108

B. SundaraÕel et al.r Applied Surface Science 137 (1999) 103–112

and reacted Au from the gold siliciderSiŽ111. sample sputtered for 20 min are shown in Fig. 4. Both the spectra have two distinct peaks from Au 4 f 7r2 and Au 4 f5r2 and they are fitted with Gaussian peaks. There is no Au 4 f component of elemental Au in the spectrum of gold siliciderSiŽ111.. An Au 4 f 7r2 peak appears at 83.8 eV in elemental gold with a FWHM of 1.18 eV and at 85.0 eV in the silicide with a FWHM of 1.56 eV. There is a shift of 1.2 eV towards higher B.E. which cannot be explained from the electronegativity difference between Si and Au. Many of the recent studies have shown similar Au 4 f shift towards higher B. E. in gold silicide w30–32,35x. Lu et al. w30x observed a shift of 0.77 eV towards higher B.E. in Au 4 f 7r2 in as-deposited AurSiŽ100. under UHV conditions and they have proposed a charge compensation mechanism to account for this. According to the charge compensation mechanism, Au gains s charge from Si but loses its d charge to form the sd hybrid so that the net charge flow into the Au site in the Au–Si alloy increases slightly w30x. From band structure calculations w37x and a XANES study of Au–implanted Si w38x it is known that Au loses approximately 0.14 d electron count upon Au–Si interaction in Si. There have been several studies of chemical shifts in the AurSi system with the Au film thickness in the monolayer ŽML. regime. Molodtsov et al.

Fig. 4. Au 4 f XPS spectra from the gold siliciderSiŽ111. sample sputtered for 20 min and from an elemental Au film deposited on a SiŽ111. substrate with native oxide.

w31x have observed a shift of 0.66 eV in Au 4 f binding energy in the room-temperature-deposited AurSiŽ111. system under UHV conditions. They have identified the silicide phase as Au 3 Si from the theoretical estimates of core-level photoelectron B.E. on the basis of the Miedema, Boom and de Boer scheme w31x. Kolmakov et al. w35x observed a shift of 0.8 eV towards higher B.E. in Au 4 f spectra upon annealing AuŽ6 ML.rSiŽ111.Ž7 = 7. at 6278C. Yeh et al. w32x in their photoemission spectroscopic studies reported a shift of around 0.6 eV for the Au 4 f 7r2 line in room-temperature-deposited AurSiŽ111.Ž7 = 7.. They observed the AuŽ6 ML.rSiŽ111. interface to be abrupt with an Au layer sandwiched between the SiŽ111. substrate and a surface silicide layer, although the uniform layer of Au was not expitaxial. This surface silicide film may not be Au 3 Si because the local atomic configuration can be different from the bulk phase as it is only 1–1.5 ML thick and resides on the surface w32x. Also the silicide layer in their case is not likely to be epitaxial as the sandwiched gold layer is not epitaxial. In our case the thin uniform gold silicide layer is ; 5 nm thick. This thickness as well as that for the silicide islands have been obtained from micro-RBS measurements using an ion beam of 2 mm diameter w39x. The thin silicide layer as well as the gold silicide islands are epitaxial as observed from the reduced backscattering yield when the incident Heq beam is aligned along the w111x and the w110x directions of the underlying Si substrate in RBSrchanneling experiments w27x. Our observed shift of 1.2 eV in gold silicide is larger than previously reported results. This larger shift may be attributed to a structure induced chemical shift w40,41x as the bonding in epitaxial and nonepitaxial systems is expected to be somewhat different. The concept of ‘structure induced chemical shift’ in SiO 2 was introduced by Grunthaner et al. w40x and invoked in explaining the larger chemical shift of 4.4 eV observed for thermal SiO 2 w41x while a shift of 3.7 eV is expected for the highest oxidation state ŽSi 4q .. It should be mentioned here that the B.E. of the Au 4 f 7r2 peak has shown some variation. The B.E. shift observed for the top layer, i.e., for the unsputtered sample, is 1.0 eV, which gradually rises to 1.2 eV with sputtering time. For the SiO 2 system a correlation between the magnitude of the chemical shift and the charge transfer due to

B. SundaraÕel et al.r Applied Surface Science 137 (1999) 103–112

change of Si–O–Si bond angle has been given w41x. A quantitative analysis of the chemical shift in our gold siliciderSiŽ111. system from the concept of ‘structure induced chemical shift’ is not yet possible, because the atomic positions in the unit cell of the silicide are not known, although various unit cell structures have been proposed w12,13,42x.

3.1.3. C 1s peak The C 1 s spectrum of the unsputtered sample was fitted with three Gaussian peaks at 284.2 eV, 285.0 eV and 286.4 eV. The peak at 286.4 eV indicates C–O bonding w20,41x. The peaks at 284.2 eV and 285.0 eV correspond to ‘aliphatic’ C peaks w41x, the peak at 285.0 being the prominent one. Upon sputtering for 5 min the peak due to C–O bonding vanished. After sputtering for 10 min there was a shift of 0.8 eV and 0.3 eV towards lower binding energy in the ‘aliphatic’ C peaks at binding energies 284.2 eV and 285.0 eV, respectively. This is due to the direct bonding of aliphatic ŽCH x . groups with Si w43x. The aliphatic C was probably incorporated during etching of the sample. In Fig. 5, the sputter depth profile of C 1 s component at 284.7 eV is shown. Within the silicide layer, C is at the background level.

Fig. 5. Intensities of the Si 2 p components in bulk Si, in gold silicide and in oxide and Au 4 f 7r 2 , O 1 s and C 1 s peaks as a function of sputtering time Ždepth.. The vertical dashed lines demark the top oxide layer, the silicide layer underneath and the substrate Si.

109

3.1.4. XPS depth profile The variation of XPS peak intensities of the three components of Si 2 p, Au 4 f 7r2 , C 1 s and O 1 s with sputtering time is shown in Fig. 5. The sputter depth profiles of the oxide component of Si 2 p and O 1 s show the same decreasing trend and they are prominent up to a sputtering time of 12.5 min Žthe inflexion point for the oxide intensity.. Hence there is an SiO 2 layer of thickness ; 2.8 nm at the top surface of the gold siliciderSiŽ111. sample. The Au 4 Si phase is of metallic character permitting high mobility of Si atoms w8x. The Si atoms from the substrate reach the top surface by diffusing through the gold silicide layer and get oxidized. The formation of oxide by diffusion of Si atoms through gold silicide is also consistent with other earlier results w2,7–10x. After sputtering for 12.5 min, the intensities of the Si 2 p component in the oxide and O 1 s decreases further and the silicide component of Si 2 p keeps on rising and is prominently seen up to a sputtering time of 37 min. The depth profile of Au 4 f 7r2 intensity from gold silicide in the interval of 12.5–37 min shows higher intensity ŽFig. 5.. This indicates that there is a uniform silicide layer of thickness ; 5.5 nm which has grown before the gold silicide islands Žrefer to Fig. 1. started developing on it during the Stranski–Krastanov growth. The thickness of ; 5.5 nm for the thin uniform layer is consistent with our micro-RBS results w39x. In order to reach this conclusion we ignored the effect of the thick islands because their surface coverage is quite low and thus the X-ray beam mainly probes the uniform layer. If the surface coverage of islands were higher, Au 4 f 7r2 intensity would have continued undiminished for a much longer sputtering time. After ; 25 min of sputtering, the intensities of both the silicide component of Si 2p and Au 4 f 7r2 signals begin to decrease. This decrease is not as sharp as the rise in the first part. This is expected because of the thicker gold silicide islands which would contribute for a longer sputtering time. The Si 2 p signal from the bulk Si increases almost monotonically with sputtering time. The substrate appears to be exposed, except in the region of islands, beyond ; 37 min of sputtering. From these discussions we obtain the following structure of the gold siliciderSiŽ111. sample: Ž2.8 nm. SiO 2rŽ5.5 nm. gold siliciderbulk-silicon with a small coverage of thicker

110

B. SundaraÕel et al.r Applied Surface Science 137 (1999) 103–112

gold silicide islands on the 5.5 nm thin uniform gold silicide layer. We can see a considerable amount of Au 4 f 7r2 intensity in Fig. 5 in the sputtering interval 0–12.5 min corresponding to the oxide layer. There may be some gold silicide within the oxide layer. The presence of gold oxide is ruled out as the free energy of formation of gold oxide is practically zero and gold does not bond to oxygen w44x. The B.E. shift of Si 2 p and Au 4 f 7r2 from the silicide in this region is slightly different compared to that from the underlying silicide. There is a maximum deviation at zero sputtering time with the Si 2 p B.E. of 101.0 eV and Au 4 f 7r2 B.E. of 84.8 eV. This is due to a different chemical environment of gold silicide at the top layer. If we compare the ratio of intensities of the silicide component of Si 2 p and Au 4 f 7r2 in the intervals 0–12.5 min and 12.5–37 min, we can see that the intensity ratio in the oxide layer is different from that in the silicide layer. The transition probability of the photoelectron emission from Si 2 p and Au 4 f 7r2 core levels in this oxide environment may be different from that in the silicide layer. This may be responsible for the difference in the intensity ratio between the silicide component of Si 2 p and the Au 4 f 7r2 in the oxide layer compared to the intensity ratio in the underlying silicide layer seen in Fig. 5. 3.2. B. SIMS results SIMS depth profiling was performed on the triangular and trapezoidal islands and the uniform thin layer of gold silicide regions. The measurements were made about a month after the samples had been prepared. Fig. 6 shows one such SIMS depth profile taken from a bigger triangular island Ž40 mm one side.. The secondary ions were collected from the central area of this triangular island. From Fig. 6, the Si and the Au signals are found to be almost steady upto 60 min of sputtering. Beyond this the Si signal remains steady whereas the Au signal has been found to abruptly drop by a few orders of magnitude Žnot shown in the figure.. The measurement was stopped here. The height of the island, as measured to be the depth of the crater Žfor 60 min of sputtering., has been found to be 1.2 mm. Measurements made on several islands show a comparable island thickness. When the secondary signals

Fig. 6. SIMS spectra from an island on the gold siliciderSiŽ111. sample. The depth profile of Si, Au, O and Br are shown. After 60 min sputtering the Au signal drops by a few orders of magnitude.

were collected from an area on the flat region Žthin uniform layer of gold silicide., it was observed that the gold signal vanished abruptly. This shows that the flat uniform gold silicide layer is very thin. This is what was observed in the XPS studies as well Ždescribed in Section 3.1. where the sample area contained predominantly the uniform layer and in the micro-RBS studies w39x. Abrupt diminishing of gold signal beyond 60 min of sputtering shows that the silicide–Si interface is reasonably sharp. SIMS results show the presence of oxygen throughout the gold silicide island thickness with higher O counts coming from a top layer. This enhanced O counts seem to be correlated with the Br counts. The bromine–passivated SiŽ111. surface on which Au was deposited contains a very small amount of Br, less than 1r3 ML. The question remains: what happens to Br upon Au deposition and annealing? A conjecture is that Br might behave as a surfactant like As w45x —both of which passivate the SiŽ111. substrate. Although the Br counts are close to the noise level, from the SIMS depth profile we notice a definite trend of Br diffusing away from the interface towards the outer part of the gold silicide layer—a general trend for surfactants in surfactant mediated growth. From SIMS we notice a redistribution of Br towards the surface. However, in XPS such a small concentration cannot be detected. As it is, the XPS signal of Br from samples where the whole amount

B. SundaraÕel et al.r Applied Surface Science 137 (1999) 103–112

of Br is confined within a fraction of a nm is very weak w20x. From the oxygen rich top layer in the SIMS depth profile, it may be safe to assume that this is predominantly an oxide layer. However, it also contains Au, perhaps in the form of silicide like the top layer of the sample used in the XPS studies, except that the oxide layer is much thicker here. ŽAfter etching both the samples were kept in air—the XPS sample for a few days and the SIMS sample for over a month.. Clearly, the gold silicide has a different environment in the top layer consistent with the observed difference in the chemical shifts for both Si 2 p silicide and Au 4 f 7r2 components in the XPS studies discussed in Section 3.1.

4. Conclusion XPS and SIMS studies were carried out on Au Ž; 100 nm. films deposited on a chemically prepared Br-passivated SiŽ111. substrate and annealed around the Au–Si eutectic temperature. Chemical shifts in the XPS peaks of Si 2 p and Au 4 f signals in the annealed sample provide evidence for Au–Si bonding. Gold silicide with an average composition of Au 4 Si grows epitaxially on Br–SiŽ111. in Stranski–Krastanov mode. XPS depth profile with Arq sputtering gave the thickness of the thin uniform layer of Au 4 Si to be ; 5.5 nm with an oxide layer of ; 2.8 nm thickness on the top of the silicide layer. SIMS depth profile analysis on the triangular and trapezoidal islands showed that the heights of the islands are nearly constant with a value of about 1.2 mm. A constant height of the islands is assumed in the theory of shape transition. In our previous analysis of shape transition, we presented a rough estimate of the thickness Ž; 1 mm. of the gold silicide islands. In the present study we determined the island height and demonstrated its constancy. The sudden drop of the Au signal at the end of the gold silicide region in the SIMS study indicates that the siliciderSi interface is abrupt. For the uniform thin silicide layer, the immediate drop of the Au signal in the SIMS profile indicates that this layer is very thin, consistent with the XPS results and previous micro-RBS results. Br, initially present at the AurBr–SiŽ111. interface, has been found to diffuse

111

towards the surface Žouter part. of the silicide islands upon annealing—a behavior similar to surfactant mediated growth.

Acknowledgements This work was partly supported by ONR Grant No. N00014-95-10130 ŽProject No. USIF 9405..

References w1x A. Hiraki, M.A. Nicolet, J.W. Mayer, Appl. Phys. Lett. 18 Ž1971. 178. w2x A. Hiraki, E. Lugujjo, J.M. Mayer, J. Appl. Phys. 43 Ž1972. 3643. w3x K. Sekar, G. Kuri, P.V. Satyam, B. Sundaravel, D.P. Mahapatra, B.N. Dev, Phys. Rev. B 51 Ž1995. 14330. w4x G. Lelay, Surf. Sci. 132 Ž1983. 169. w5x A. Hiraki, E. Lugujju, M.A. Nicolet, J.W. Mayer, Phys. Status. Solidi ŽA. 7 Ž1971. 401. w6x T. Narusawa, S. Komiya, A. Hiraki, Appl. Phys. Lett. 22 Ž1973. 389. w7x L. Hultman, A. Robertson, H.T.G. Hentzell, E. Engstroem, P.A. Psaras, J. Appl. Phys. 62 Ž1987. 3647. w8x A. Hiraki, Surf. Sci. Rep. 3 Ž1984. 357. w9x C.A. Chang, G. Ottaviani, Appl. Phys. Lett. 44 Ž1984. 901. w10x A.A. Pasa, H. Paes Jr., W. Losch, J. Vac. Sci. Technol. A 10 Ž1992. 374. w11x E. Bauer, H. Poppa, Thin Solid Films 12 Ž1972. 167. w12x A.K. Green, E. Bauer, J. Appl. Phys. 47 Ž1976. 1284. w13x K. Oura, T. Hanawa, Surf. Sci. 82 Ž1979. 202. w14x J.A. Golovchenko, J.R. Patel, D.R. Kaplan, P.L. Cowan, M.J. Bedzyk, Phys. Rev. Lett. 49 Ž1982. 560. w15x G. Materlik, A. Frahm, M.J. Bedzyk, Phys. Rev. Lett. 52 Ž1984. 441. w16x B.N. Dev, V. Aristov, N. Hertel, T. Thundat, W.M. Gibson, Surf. Sci. 163 Ž1985. 457. w17x B.N. Dev, V. Aristov, N. Hertel, T. Thundat, W.M. Gibson, J. Vac. Sci. Technol. A 3 Ž1985. 975. w18x K. Sekar, P.V. Satyam, G. Kuri, D.P. Mahapatra, B.N. Dev, Nucl. Instr. and Meth. B 71 Ž1992. 308. w19x K. Sekar, P.V. Satyam, G. Kuri, D.P. Mahapatra, B.N. Dev, Nucl. Instr. and Meth. B 73 Ž1993. 63. w20x K. Sekar, G. Kuri, D.P. Mahapatra, B.N. Dev, V.S. Ramana, S. Kumar, V.S. Raju, Surf. Sci. 302 Ž1994. 25. w21x T. Nakahara, S. Okhura, F. Shoji, T. Hanawa, K. Oura, Nucl. Instr. and Meth. 45 Ž1990. 467. w22x K. Sekar, G. Kuri, P.V. Satyam, B. Sundaravel, D.P. Mahapatra, B.N. Dev, Solid State Commun. 96 Ž1995. 871. w23x K. Sekar, G. Kuri, P.V. Satyam, B. Sundaravel, D.P. Mahapatra, B.N. Dev, Surf. Sci. 339 Ž1995. 96.

112

B. SundaraÕel et al.r Applied Surface Science 137 (1999) 103–112

w24x J. Tersoff, R.M. Tromp, Phys. Rev. Lett. 70 Ž1993. 2782. w25x N.M. Jisrawi, W.L. McLean, N.G. Stoffel, M.S. Hegde, C.C. Chang, D.L. Hart, D.M. Hwang, T.S. Ravi, B.J. Wilkens, J.Z. Sun, T.H. Geballe, Phys. Rev. B 43 Ž1991. 7749, and references therein. w26x R.F.C. Farrow, J. Vac. Sci. Technol. B 1 Ž1983. 222, and references therein. w27x B. Sundaravel, K. Sekar, P.V. Satyam, G. Kuri, B. Rout, S.K. Ghose, D.P. Mahapatra, B.N. Dev, Ind. J. Phys. 70A Ž1996. 687. w28x T.R. Anantharaman, H.L. Luo, W. Klement, Nature 210 Ž1966. 1040. w29x G.A. Anderson et al., Mater. Sci. Eng. 7 Ž1971. 83. w30x Z.H. Lu, T.K. Sham, P.R. Norton, Solid State Commun. 85 Ž1993. 957. w31x S.L. Molodtsov, C. Laubschat, G. Kaindl, A.M. Shikin, V.K. Adamchuk, Phys. Rev. B 44 Ž1991. 8850. w32x J.-J. Yeh, J. Hwang, K. Bertness, D.J. Friedman, R. Cao, I. Lindau, Phys. Rev. Lett. 70 Ž1993. 3768. w33x L. Braicovich, C.M. Garner, P.R. Skeath, C.Y. Su, P.W. Chye, I. Lindau, W.E. Spicer, Phys. Rev. B 20 Ž1979. 5131.

w34x L.J. Brillson, A.D. Katnani, M. Kelly, G. Margaritondo, J. Vac. Sci. Technol. A 2 Ž1984. 551. w35x A. Kolmakov, J. Kovac, S. Gunther, L. Casalis, L. Grego¨ ratti, M. Marsi, M. Kiskinova, Phys. Rev. B 55 Ž1997. 4101. w36x G. Hollinger, F.J. Himpsel, Appl. Phys. Lett. 44 Ž1984. 93. w37x L.F. Mattheiss, R.E. Dietz, Phys. Rev. B 22 Ž1990. 1663. w38x Z.H. Lu, T.K. Sham, M. Vos, A. Bzowski, I.V. Mitchell, P.R. Norton, Phys. Rev. Rap. Commun. B 45 Ž1992. 8811. w39x B. Rout, B. Sundaravel, S.K. Ghose, G. Kuri, K. Sekar, D.P. Mahapatra, B.N. Dev, H. Bakhru, A.W. Haberl, Solid State Physics ŽIndia. 39C Ž1996. 39. w40x F.H. Grunthaner, P.J. Grunthaner, R.P. Vasquez, B.F. Lewis, J. Maserjian, J. Vac. Sci. Technol. 16 Ž1979. 1443. w41x M. Grundner, H. Jacob, Appl. Phys. A 39 Ž1986. 73. w42x F.H. Baumann, W. Schroeter, Philos. Mag. Lett. 57 Ž1988. 75. w43x R.C. Gray, J.C. Carver, D.M. Hercules, J. Electr. Spectrosc. Rel. Phen. 8 Ž1976. 343. w44x H. Dallaporta, A. Cros, Appl. Phys. Lett. 48 Ž1986. 1357. w45x M. Copel, M.C. Reuter, E. Kaxiras, R.M. Tromp, Phys. Rev. Lett. 63 Ž1989. 632.