J. MATER. CHEM., 1991, 1(4),663-666 Use of Phenylarsine in the Atmospheric Pressure Metal Organic Chemical Vapour Deposition of GaAs on Si(lO0) Neil R. Dennington, Andrew C. Wright and John 0. Williams* Solid State Chemistry Group, Department of Chemistry, UMIST, PO Box 88, Manchester M60 lQD, UK The Ill-V compound semiconductor GaAs has been grown on silicon substrates by atmospheric pressure metal organic chemical vapour deposition (APMOCVD) using trimethylgallium (TMGa) and phenylarsine (PAS). The GaAs exhibited high crystalline quality (X-ray rocking curve FWHM 430 arc seconds) and a specular morphology. Electrical properties were poor with an inversion from p- to n-type within the layers. This is attributed to both dislocations present because of the lattice mismatch at the GaAs/Si interface and doping effects due to impurities introduced from the PAS source.Strong emission due to free-exciton transitions may be seen in the photoluminescence spectra of the samples. Keywords: Gallium arsenide; Metal organic chemical vapour deposition; Photoluminescence spectroscopy; Secondary-ion mass spectrometry; Transmission-electron microscopy The heteroepitaxial growth of GaAs on Si has attracted great interest in recent years owing to the potential use of this system in future electronic devices.' There are, however, several intrinsic difficulties with the GaAs/Si system which reduce the GaAs crystal quality. These include the 4.1% lattice mismatch,2 the formation of antiphase boundaries (APBs),, the large differences in thermal expan~ivities~ and problems of interface charge neutrality.' The severity of these problems has been greatly reduced by the use of two-step growth techniques6 combined with the use of heat cycling7 in situ or ex situ growth annealing,* iso-electronic dopingg and the use of strained-layer superlattices as buffer layers between the GaAs and silicon substrate.lo Conventionally, trimethylgallium (TMGa) and arsine (ASH,) have been used as Group I11 and Group V sources in the MOCVD-growth of GaAs." Arsine is highly toxic at low concentrations [uiz. 0.05 ppm (RL)] l2 and, accordingly, alternative safer As-con- taining precursors for MOCVD have been sought. l3 Several alternative Group V sources have been suggested including tert-butylarsine (TBA)', trimethylarsine (TMA)I4 diethylarsine (DEA)" ethylarsine (EA)16 and phenylar~ine'~ (PhAsH, or PAS).These liquid sources are safer than gaseous arsine since they can be stored and dispensed more easily. They yield GaAs epitaxial layers of varying crystalline quality and electrical properties (see e.g. ref. 18). Prior to the acceptance of any new source material it must be proved to be fully capable of any growth previously performed using ASH,. In this paper we demonstrate the utility of PAS in the growth of GaAs on Si(lO0) and indicate the purity limitations of the presently available precursor material. Experimenta1 Atmospheric pressure MOCVD was used to grow 3 pm thick GaAs layers on Sb-doped Si( 100) & 3"substrates.Trimethyl- gallium and phenylarsine (Epichem), were used as precursor materials. Similar sources have been employed previously in the preparation of homoepitaxial GaAs and the results pub- lished.l7 Details of source impurity concentrations are given in Table 3 of ref. 17. Immediately prior to growth the Si substrate was cleaned using a procedure described else-where,"" and baked out in situ at 1000 "C for 20 min under flowing HZ.In the two-step growth procedure used, a lOOA thick GaAs layer was deposited at 500 "C and a V: I11 ratio of 40: 1, with the remainder of the GaAs layer being grown at 650 "C and a V: I11 ratio of 15 :1. The conditions yielded a growth rate of ca. 3 pm h-' and based on extensive studies of growth parameter variation^'^" were optimum for GaAs growth using the present precursors. The GaAs epilayers were characterised using optical microscopy (Olympus-B4-2), scanning electron microscopy (Philips SEM 525), transmission electron microscopy (Philips, EM430), double-crystal X-ray diffraction (Bede QC-l), capaci- tance-voltage (C-V) profiling and dynamic secondary-ion mass spectrometry (SIMS).Photoluminescence (PL) spectra were recorded in the temperature range 10-300 K following excitation with laser radiation at 514 nm. Experimental details of these techniques may be found else~here.'~~.'~~ Results Fig. 1 compares the surface morphology of GaAs on Si substrate under identical conditions (see Experimental) using trimethylgallium and (a) arsine (b) PAS.To the naked eye the surfaces are mirror smooth but in the optical micrographs compared surface features are observed in the two cases. On close examination these surfaces are inferior to those of homoepitaxial GaAs.lgb A high degree of crystallinity is demonstrated by the narrow double-crystal X-ray rocking curve FWHM values, which are as low as 430 arc seconds. Using similar conditions with trimethylgallium and arsine, corresponding half widths of ca.450 arc seconds have been fo~nd.'~" A typical C-V profile, through a (100) GaAs layer grown using PAS, is shown in Fig. 2. Most striking is the apparent change from p- to n-type behaviour towards the heterointer- face. An increase in p-type carrier concentration is observed from the GaAs surface (depth =zero) towards the heterointer- face (depth=3 pm) until at a distance of 1-2 pm from the layer surface a change to n-type behaviour occurs. Carrier concentration then falls before increasing towards the hetero- interface. Fig.3 shows typical low temperature (10 K) PL spectrum for GaAs layers nucleated at different temperatures. Because of strong absorption at the excitation wavelength (514 nm) spectral information is obtained from the uppermost CQ. 1 pm of the GaAs layer. The spectra are dominated by a narrow, J. MATER. CHEM., 1991, VOL. 1 Fig. 1 Optical micrographs of GaAs surfaces grown on Si surfaces under identical experimental conditions using trimethylgallium and (a) arsine (6)PAS depth/pm Fig.2 A typical C-V profile through a GaAs epitaxial layer. The GaAs surface corresponds to depth zero resolved Si,, (e, A) emission at 1.462 eV. A weak but resolved free-exciton transition (FE) is observed at 1.489 eV, together with a weak shoulder at 1.470eV attributed to a carbon- related (e, A) transition.lgb Increasing levels of Si towards the GaAs layer surface is shown by SIMS which also reveals relatively sharp, discrete GaAs/Si interfaces (Fig. 4). Both C and Zn contamination is found in the layers. Fig. 5 shows cross-sectional transmission electron micrographs (TEM) taken of GaAs/Si interfaces grown from (a)TMGa and ASH, (b) TMGa and PAS under similar conditions. These micrographs show the similarity of the two structures in terms of defect content and also allow clear demarcation of specific imperfections such as dislocations (D) stacking faults (S) and microtwins (M) in the epitaxial GaAs.Discussion The ability to grow epitaxial GaAs of acceptable crystalline quality on Si(100) using PAS and trimethylgallium has been demonstrated by this work. However, only poor electrical behaviour is exhibited by this material. This is thought to I 1 I I I 1 1.54 1.51 1.48 1.45 1.42 1-39 energylev Fig. 3 10 K PL spectra of GaAs epitaxial layer as a function of V :I11 ratio during growth reflect more the insufficient purity of the PAS source and peculiarities of the GaAs/Si system rather than an intrinsic inability to grow GaAs using a combination of trimethylgal- lium and PAS.We have, indeed, shown on a previous occasion that homoepitaxial GaAs of excellent crystalline quality and with background electron concentration of 5 x lo1’ cm-3 and J. MATER. CHEM., 1991, VOL. 1 depth (arb. units) 5:l 0:l 5:1 0: 1 25:l lo3i /t 10 10 depth (arb. units) Fig. 4 Dynamic SIMS profiles for (a) silicon (b) C and (c) Zn of GaAs/Si interfaces as a function of V :I11 ratio during growth. The GaAs surface is at the origin and the profiles show relatively sharp interfaces and an increasing level of Si towards the GaAs surface electron mobility of 20 000 cm2 s-’ V-’ can be grown using this pair of sources.’ The p-type nature of the surface layers is accounted for by the Zn and C incorporation.Both of these species are known to act as acceptors in GaAs, and Zn is known to be present in the present batches of PAS.17 However, the change to n-type behaviour towards the heterointerface cannot be explained in terms of Zn and C contamination alone. It is known that dislocation density and defect concentration increases towards the heterointerface2’.22 and we have con- firmed this by cross-sectional TEM carried out on our layers (see Fig. 5). It is also known that dislocations act as donors in GaAs.22*23 Therefore, it is likely that it is the dislocations that determine the electrical behaviour of the material close to the heterointerface. It is apparent from the electron microg- raps that at ca.1.5-2 pm from the GaAs surface the layer dislocation density increases and this corresponds to the distance where n-type doping (N,) is in excess of p-type doping (NA).Consequently, the background carrier concen- tration (N, -NA)changes to n-type. It is, however, important to remember that this sharp transition from p- to n-type behaviour is highly unlikely to be seen within any actual layer and is in fact an artefact inherent in the C-V profiling technique. This is due to the existence of a depletion region24 Fig. 5 Cross-sectional TEM micrograph of GaAs epilayer grown on Si using (a)TMGa and ASH, (b) TMGa and PAS. Various defects can be identified such as dislocations (D), stacking faults (S) and microtwins (M) in the GaAs. In (b)position of the interface with the Si substrate is marked (I) within the material during profiling which extends from ca.0.2 to 3.0 pm. At each stage of the C-V technique, material is etched away removing the depletion region and allowing the majority carrier concentration to be expressed. In a ‘real’ sample the region over which the alteration takes place will be depleted of carriers. This region may be several micrometres thick. Thus, the real change of electrical behaviour, although representing the variation in doping levels within the layers, is probably not as abrupt as indicated in Fig. 2. The presence of the FE emission in the PL spectrum of the GaAs epilayers is evidence for the high quality of the material. The dominance of the SiGa (e, A) emission is due to the high levels of Si, confirmed by SIMS, in the surface of the layers.This Si is thought to occupy sites along dislocations where it remains electrically ne~tral.~’ The higher surface concen-tration of Si is thought to be due to a vacancy/dislocation- enhanced diffusion mechanism. Following the high-tempera- ture bake out, the surface layer of the substrate will contain many-vacancies.26 During the growth of the GaAs buffer layer, vacancies will be generated at the sample surface allowing Si to diffuse into them. This process will continue throughout growth with Si moving out of the substrate with the growth surface like a ‘wall’. Subsequent diffusion of Si from the substrate will be hindered by the GaAs layers, while the surface diffusion rate of Si may be considerably higher.26 Conclusions Our experiments have demonstrated an ability to grow high- quality GaAs on Si(100) without the complex growth pro- 666 J.MATER. CHEM., 1991, VOL. 1 cedures previously reported. More importantly, the viability of phenylarsine as an alternative As precursor in atmospheric pressure MOVPE has been demonstrated for GaAs/Si. Further purification of this precursor is required if better electrical properties are to be achieved, but characteristics of I1 12 13 G. B. Stringfellow, in Organometallic Vapour Phase Epitaxy: Theory and Practice, Academic Press, New York, 1989. C. Breckerridge, C. Collins, B. Hollomby and G. Lulham G., The Toxicologist, 1983, 3, Abs 93.R. W. Lum, J. K. Klingert and M. G. Lamont, Appl. Phys. Lett., 1987,50, 1151. the GaAs/Si system itself containing appreciable defects may be limiting factors dictating material quality. 14 15 C. Cooper, M. W. Ludowise, V. Aebi and R. L. Moon, Electron Lett., 1980, 16, 20. R. Bhat, M. A. Kozaond and B. J. Skromme, Appl. Phys. Lett., One of us (N.R.D.) would like to acknowledge the financial support of SERC throughout this study. Additionally, we would like to thank M. MacDonald and S. Hibbet for provid- ing PL and SIMS spectra, respectively. 16 17 1987,50, 1194. D. Schimtz, G. Strauch, V. Michno, A. Melas and H. Jurgensen, presented at EW-MO VPE 111, Montpellier, France, 1989. R. D. Hoare, 0.F. Z. Khan, J. 0. Williams, D. M. Frigo, D. C. Bradley, E. Chudzynska, P.Jacobs, A. C. Jones and S. A. Rush- worth, Chemtronics, 1989, 4, 78. 18 N. R. Dennignton and J. 0. Williams, in the press. References 19 (a) N. R. Dennington, PhD Thesis, University of Manchester, 1989; (b) N. Hunt and J. 0.Williams, Chemtronics, 1987, 2, 165; 1 R. P. Gale, J. C. C. Fan, B. Y. Tsaur, G. W. Turner and F. M. (c)N. D. Gerrard, D. J. Nicholas and J. 0.Williams, Chemtronics, Davis, I.E.E.E. Electron Device Lett., 1981, EDL-2, 169. 1988, 3, 17. 2 M. Akiyama, Y. Kawarada and K. Karinishi, J. Cryst. Growth, 20 K. Mohammed, J. L. Merz and D. Kasemset, Appl. Phys. Lett., 1984, 68, 21. 1983, 43, 103. 3 S. K. Ghandi and J. E. Ayers, Appl. Phys. Lett., 1988, 53, 21 H. L. Tsai and J. W. Lee, Appl. Phys. Lett., 1987, 51, 130.1204. 22 N. R. Dennington, A. C. Wright and J. 0.Williams, unpublished 4 5 6 W. A. Harrison, E. A. Grant, J. R. Waldrop and R. W. Grant, Phys. Rev. B., 1978, 18, 4402. H. Kroemer, J. Cryst. Growth, 1987,81, 193. M. Akiyama, Y. Kawarada and K. Kaminishi, J. Cryst. Growth, 1984,68, 21. 23 24 data. A. T. Macronder, S. N. G. Chu, K. E. Strege, A. F. Bloerike and W. D. Johnston, Appl. Phys. Lett., 1984, 44,615. R. W. Lum, J. K. Klingert, R. B. Bylsma, A.M. Glass, A.T. Macronder, T. D. Harris and M. G. Lamont, J. Appl. Phys., 7 M. Yamaguchi, A. Yamamoto, M. Tachikawa, Y.Hoh and 1988, 64, 6727. M. Sugo, Appl. Phys. Lett., 1988, 53, 2293. 25 M. E. Goodye, Semiconductor Device Technology, MacMillon, 8 9 10 R. W. Kaliski, C. R. Ito, D. G. McIntyre, M. Fenn, H. B. Kia, R. Bean, Z. Zanio and K. K. C. Ksieh, J. Appl. Phys., 1988, 64, 1196. W. Walukiewicz, Appl. Phys. Lett., 1989, 54, 2009. D. W. Nam, H. Holonyak, K. C. Hsieh, R. W. Kaliski, J. W. Lee, H. Scichijo, J. E. Elper, R. D. Burnham and T. L. Paoliu, Appl. 26 27 London, 1983, p. 61. A. Freundlick, J. C. Grenet, G. Neg, A. Leycuras and C. Verie, Appl. Phys. Lett., 1988, 52, 1976. J. Huang, M. Meyer and V. Pontikis, Phys. Rev. Lett., 1989, 63, 628. Phys. Lett., 1987, 51, 39. Paper 110I 125A; Received 1 1 th March, 1991