Carbonyl sulfide (OCS) as a sulfur-containing precursor in MOCVD: a study of mixtures of Me,Cd and OCS in the gas and solid phases and their use in MOCVD Matthew J. Almond," Brian Cockayne,b Sharon A. Cooke: David A. Rice," Philip C. Smithb and Peter J. Wrightb "Department of Chemistry, University of Reading, Whiteknights, Reading, Berks, UK RG6 6AD bDRA (Malvern), St. Andrews Road, Malvern, Worm., UK WR14 3PS Mixtures of dimethylcadmium (Me,Cd) and carbonyl sulfide (OCS) have been examined in the gas and solid phases over a wide range of temperatures. No interaction is observed between Me2Cd and OCS in a 1 :1molar ratio at room temperature in the gas phase, nor is any interaction detected in the solid phase at liquid-nitrogen temperature. On heating the 1:1mixtures of Me2Cd and OCS to 250 "C in a sealed vessel, gaseous products are formed.These consist of methane, carbon monoxide and ethane in an approximately 12 :2 :1 molar ratio, although a large excess of unreacted OCS is also present showing that this compound does not react fully with Me2Cd at 250 "C. In a flow system at 300 "C only methane and carbon monoxide are formed, in the molar ratio 6 :1, although the amount of reaction of the OCS is much less (as evidenced by a higher proportion of unreacted OCS). When the flow reaction is repeated at 450 "C more of the OCS reacts and the proportion of carbon monoxide in the gaseous reaction products is much higher. Using a commercial MOCVD apparatus, high-quality layers of cadmium sulfide are obtained from Me,Cd-OCS mixtures.Temperatures in the range 350-450 "C lead to somewhat slow growth rates which only reach 1 pm h-' when a 200-fold molar excess of OCS over Me2Cd is used. A small amount of prereaction is observed, but only when H, is used as the carrier gas. This is attributed to the formation of very small concentrations of H,S by reaction of OCS with H,. The resulting epitaxial layers have good thickness and electrical uniformity. These experiments confirm that OCS may be used as a precursor for the growth of thin layers of CdS by MOCVD. However, the large excess of OCS required here suggests that the compound might be more useful for doping than for the growth of pure layers of CdS. In order to exploit fully the technique of metal-organic chemi- cal vapour deposition (MOCVD) for the growth of thin layers of 11-VI semiconductor materials it is necessary to optimise the precursor compounds used.Such work is of considerable importance given the use of 11-VI compounds in blue-green light-emitting diodes (LEDs) and lasers.'-4 For many of these applications, controlled doping of the semiconductor layer is required. In this respect molecular beam epitaxy (MBE) has, to date, demonstrated better control than has MOCVD. It is only by careful selection of the most suitable precursor com- pounds that MOCVD will yield epitaxial layers of the relevant purity and uniformity for successful exploitation in devices. The subsequent problem of doping control can then be addressed. It is probably true that most of the possibilities for experi- ment within the rather limited range of group I1 precursors have been tried.5" Such experiments have included the use of adducts of Me2Cd with Lewis base donor molecules, e.g.TMEDA (Me,NCH2CH2NMe2), as the metal-containing pre- cursor, but these precursors are not successful in reducing prereaction when the metal is cadmium.5b In recent times attention has turned to the sulfur source where an extensive choice of reagents exists. The use of the hydride, hydrogen sulfide, provides good-quality layers in terms of optical proper- ties and purity with growth possible at low temperatures (300-400 "C). The problem with this precursor is that there is also an unwanted prereaction which gives rise to non-uniform- ities in the layer properties.The use of a number of alkyl group sources has been studied. These sources have included both monoalkyl compounds, e.g. methylthiol,6 and dialkyl compounds, e.g. diethyl~ulfide.~In addition, five- and six- membered heterocyclic ring compounds have been considered.* In the majority of cases the use of these types of compounds leads to a higher growth temperature (>4OO"C) and an associated degradation in the optical and electrical properties of the layers. However, the uniformity of the layer thickness and the elimination of unwanted prereaction are benefits for these compounds. Thus far, the best compromise in terms of optical and electrical properties and uniformity is the series of monosubs tit uted sulfur compounds (e.g.methyl t hiol ). Recent work has demonstrated the success of ButSH as the sulfur- containing precursor for the growth of thin layers of 11-VI compounds at temperatures around 325 OC9 In recent publi- cations we reported the use of the thiirane propylene sulfide as a precursor for both ZnS" and CdS" by MOCVD. In the current work we have turned our attention to the compound carbonyl sulfide, OCS. To our knowledge this compound has not been studied in MOCVD systems, although the related compound CS2 has been investigated in a preliminary series of experiments as a precursor, with Me,Zn, for ZnS forma- tion.12 There were three main reasons for our selection of OCS as a precursor. First is the well known behaviour of OCS to act as a source of sulfur atoms in a range of chemical reaction^.'^-'^ Second is that OCS is likely to generate sulfur in a very clean chemical reaction in which the only byproduct is carbon monoxide gas, so avoiding the problems associated with many organosulfur compounds whereby carbon may be incorporated as an impurity in the semiconductor layer.It is thought that the carbon impurity arises via the intermediacy of organic radicals which are formed upon decomposition of the organosulfur precursor. Third, an impetus was given to our research by our observations that OCS is often found as an impurity in commercially supplied samples of H2S. Thus IR spectra of gaseous samples of H2S obtained direct from the cylinder in which it was supplied invariably show the character- istic bands of OCS, of which the most intense is the feature arising from the v(C=O) vibration centred at around 2060cm-l.Since H,S is the most usual sulfur precursor for growth of thin layers of sulfide semiconductor materials it is of importance to ascertain the reactivity of OCS with Me,Cd under MOCVD conditions. Our experimental strategy has followed closely that employed when propylene sulfide was assessed as a precursor." First, we have looked for any evidence of molecular interactions between OCS and Me2Cd J. Muter. Chem., 1996, 6(lo), 1639-1642 1639 in the gas phase at room temperature and in the condensed solid phase at 77 K Secondly, we have studied the thermolysis of mixtures of OCS and Me,Cd in both static and flow systems over a range of temperatures Lastly, we have used OCS as the sulfur source in a commercial MOCVD reactor in order to grow layers of device-quality CdS Experimental An all-glass vacuum line was used for all the experiments at Reading except for the Grignard preparation of dimethyl- cadmium l6 l7 Gas chromatographic (GC) analyses were per- formed using a Perkin Elmer 8420 capillary gas chromatograph fitted with a 25 m Chrompack capillary column Detection was by flame ionisation detection (FID) Carbonyl sulfide (Aldrich, stated purity 96+ %) was used direct from the cylinder in which it was supplied Gas-phase IR measurements were made by containing samples of the vapours produced from the various reactions in an evacuated gas cell (path length ca 10cm) fitted with KBr windows A Perkin Elmer 1720-X Fourier Transform interferometer was used to record the IR spectra with a resolution of f2 cm-' 'Cold cell' experiments were executed as described in detail elsewhere Experiments were carried out in which mixtures of Me,Cd and OCS were heated under vacuum in static systems or under nitrogen in flow systems For static experiments an approxi- mately 1 1 mixture of Me,Cd and OCS (ca 1 5 mmol of each reagent) was sealed into a Pyrex vessel of volume ca 150 cm3 This vessel was placed in an oven and was held at a temperature of 250°C for 24 h The gaseous products were analysed by GC Similar sealed Pyrex vessels were also prepared containing only pure dimethyl cadmium or pure carbonyl sulfide Again these were heated and the evolved gases analysed by GC measurements In a separate experiment, Me,Cd and OCS were allowed to react under flow conditions A mixture of dimethylcadmium and a large excess of carbonyl sulfide was allowed to pass through a heated Pyrex tube, of length ca 40 cm and diameter ca 5 cm, under a flow of dry nitrogen The temperature, ca 300 "C, was controlled by a Eurotherm temperature controller Unreacted material, after passing through the reactor, was collected in a trap held at -78 "C The gases evolved during the course of the reaction, and which passed through the -78 "C trap, were collected and analysed by GC Because the degree of reaction of OCS in this experiment was small the run was repeated but the temperature of the reactor was increased to 450°C Growth of CdS by MOCVD The growth of the epitaxial layers was carried out at the Defence Research Agency (DRA), Malvern under atmospheric pressure in a conventional horizontal MOCVD reactor described previously for similar experiments l9 Palladium-diffused hydrogen was used as the carrier gas The sample of dimethyl cadmium was kept in a temperature-controlled bath and its flow rate into the reactor was varied by adjusting the temperature of the bath Carbonyl sulfide was admitted to the reactor direct from the cylinder in which it was supplied, its flow rate was adjusted by means of a mass-flow controller Growth was carried out in a water-cooled reactor cell with the substrates mounted on an RF heated graphite susceptor The substrates used were GaAs single-crystal wafers orien- tated on the (100) plane Prior to growth the wafers were polished and etched as described elsewhere l9 The prepared wafers were loaded into the reactor which was subsequently flushed with high-purity hydrogen Immediately before growth the wafers were heated to 500 "C for 10 min, after which the 1640 J Muter Chem , 1996, 6(lo), 1639-1642 temperature was reduced to the growth temperature and layer growth was initiated The layers were grown for 1 h After growth the reactor was flushed with hydrogen before the substrates and layers were allowed to cool to room temperature and removed Cleave-and-stain techniques in conjunction with an optical microscope were used to determine the layer thickness The surface morphologies of the layers were examined using both optical microscopy and a scanning electron microscope (SEM) In addition, one layer grown under optimum conditions was selected and its electrical properties were assessed using Hall effect measurements Secondary ion mass spectrometry (SIMS) was performed to determine the nature of any impurities present in the layer of CdS Results and Discussion A gaseous mixture of dimethylcadmium and carbonyl sulfide in a 1 1 molar ratio was examined by FTIR spectroscopy The spectrum of the mixture corresponded exactly to a super- position of the spectra of the two individual components Hence it is concluded that these reagents do not form any gaseous adduct at room temperature Moreover, as described in more detail elsewhere," there is no detectable interaction between these reagents in the condensed solid at 77 K This latter finding is in contrast to the results obtained from solid mixtures of dimethylcadmium and propylene sulfide at 77 K,1° where a weak interaction was observed Thus there is no evidence for the formation of any adduct between Me,Cd and OCS under the various conditions of our experiments The thermal decomposition of 1 1 molar ratio mixtures of Me,Cd and OCS was studied in static systems at 250°C As controls, the thermolyses of both pure compounds under identical conditions were determined individually On heating a sample of pure Me,Cd (2 5 mmol Me,Cd in a volume of ca 125 cm3) to 250 "C for 24 h, a range of gaseous alkanes and alkenes were formed (see Table 1) The data obtained are in accord with those obtained previously 2o Upon heating a sample of OCS (2 0 mmol in a volume of 217 cm3) under the same conditions, no reaction or decomposition occurred Having obtained data on the pure compounds, we proceeded to study the reactions of approximately 1 1 (ca 15 mmol of each reagent) molar mixtures of Me,Cd and OCS at 250°C After heating for 24 h, then allowing to cool, the gases produced were methane, carbon monoxide and ethane in an approxi- mately 12 2 1 molar ratio An excess of unreacted OCS was also present The appearance of the reactor had changed from colourless to orange with a layer of a solid material being deposited on the inside of the reactor The important finding from these experiments is that OCS itself does not decompose at a temperature as low as 250"C, but reacts in the presence of Me,Cd In an attempt to mimic more closely an MOCVD reactor we thus proceeded to study the 'flow' reaction between Me,Cd and OCS entrained in a flow of nitrogen gas A furnace was placed around the reactor and the temperature was raised to 300+ 5 "C The gas stream emerging from the reactor was passed through a trap held at -78 "C and the gases not trapped were collected and analysed by GC The flow system was held at 300+ 5 "C for ca 1 5 h during which time a very slight yellow colouration could be observed on the inside of the reactor in the heated zone only Only small quantities of gaseous products were evolved at this point of the reaction, such products consisted of CO and CH, However, a large quantity of unreacted OCS was collected from the reactor in the -78 "C trap The temperature was increased to 400 "C and another gas sample was taken for GC measurements This gave a similar result to that obtained at 300"C, but the presence of a small amount of C2H6 was noted as was an Table 1 Gas chromatographic analysis of the products obtained after heating samples of Me,Cd and carbonyl sulfide assignment" Me,Cd (static)b, 250 "c' mixture (static)b, 250 "Cd mixture (fl~w)~, 300 "c" mixture (fl~w)~, 450"Cf CH4 CH3 -CH3 51.7 28.7 80.9 6.4 85.7 17.2 20.7 CH, =CH, 12.3 CH,CH,CH3 0.5 CH, =CHCH, 6.8 co 12.8 14.3 62.1 "All values quoted are molar percentages of gases with an accuracy of +0.1%.bThe static systems were heated in an oven at the stated temperature for 24 h; in the flow experiments heating was continued for 1.5 h. 'The vessel was cooled to -78 "C before the measurements were taken. d21.8% of the OCS had reacted in this experiment. "1.6% of the OCS had reacted in this experiment. f3.3% of the OCS had reacted in this experiment: note that the degree of reaction of the OCS is, as expected, much less in the flow than in the static system but that it increases substantially with increasing temperature. increase in the relative amount of CO.Increasing the tempera- ture to 450°C produced a much higher concentration of CO (as evidenced by the GC measurement, see Table 1) and so it can be concluded that more OCS has reacted at the higher temperatures. A deeper yellow-orange solid was observed on the side of the reactor at this temperature. A sample of this yellow-orange coating was analysed by X-ray powder diffrac- tion. The d-spacings and the relative intensities obtained from the X-ray data were in accord with those of cubic CdS.,' It is noteworthy that the production of CO gas appears to be linked to the deposition of solid CdS, since greatly increased amounts of both of these products are observed when the temperature is raised to 450 "C. From these preliminary studies, it is apparent that the system merited further detailed investi- gation.Accordingly growth experiments were performed reacting Me2Cd and OCS in a conventional MOCVD reactor at DRA, Malvern. Growth experiments by MOCVD All growth experiments were performed using the (100)surfaces of GaAs crystals as substrates. The initial series of growth runs were carried out keeping the reagent concentrations fixed, and the growth time constant at 60min, but varying the growth temperature. Throughout these growth experiments the molar ratio, OCS :Me2Cd (VI :11), was maintained at 25 :1. In Fig. 1 the results obtained are illustrated graphically. The growth rate of cadmium sulfide is seen to increase with growth temperature from 300 to 400°C.In the temperature range 400-450 "C the growth rate remains approximately constant. In the second series of growth runs the growth temperature was fixed and the concentration of OCS in the reactor varied (see Fig. 2). The vapour-phase ratio of the reagents was increased from 25 : 1 to 200 :1. The growth rate is seen to rise 0.0 1 J 300 400 500 growth temperature/"C Fig. 1 Variation of CdS growth rate as a function of growth temperature. Dimethyl cadmium flow rate 25 cm3 min-'; the Me,Cd to OCS molar ratio is 1 :25. 0 50 100 flow rate of OCS/C~~min-' Fig. 2 Variation of CdS growth rate as a function of carbonyl sulfide flow. Growth temperature 350 "C; dimethyl cadmium flow rate 25 cm3 min-'. The Me2Cd to OCS molar ratio is varied from 1 :25 to 1:200.as the OCS to Me2Cd ratio is increased over this range. The quality of the layers was generally good. The best quality layers were obtained at growth temperatures of 375-400 "C and with an OCS flow rate of 80 cm3 min-' (i.e. OCS :Me2Cd ratios of ca. 200: 1). A small amount of prereaction was observed in each growth experiment when Me,Cd and OCS were used as precursors and H, as the carrier gas, although the extent of this was only very slight when compared with the amount of prereaction observed when using Me2Cd and H2S as precursors. The prereaction observed here is thought to arise from the reaction between OCS and H2, which forms a small amount of H,S. This is a similar reaction to that observed by Takata et al.when using CS2 and Me,Zn as precursors to grow layers of ZnS by M0CVD.l' The prereac- tion was not seen in the flow experiments carried out at Reading where N2 was used as the carrier gas. The findings suggest that He might be preferable to H, as the carrier gas when using a commercial MOCVD apparatus to grow layers of CdS using these precursors. Hall measurements show that the layers have a bulk carrier concentration of 3 x 1017 atoms ~m-~. This level of impurities, while acceptable, is somewhat high for a successful semicond- ucting device ( 1015-1017 atoms cm-3 is normally considered to be the desirable range for impurity levels). SIMS measure- ments provide information as to what these impurities are. First, there is evidence that gallium atoms (which here will act as an n-type dopant) have diffused into the layer from the GaAs substrate.Secondly, there is evidence for the presence of halogens (which will also act as n-type dopants) in the sample. These probably arise from the preparation of the OCS sample and it should be noted that no steps were taken in these experiments to purify the OCS. Doubtless these impurity levels could be reduced by more rigorous purification. J. Mater. Chem., 1996, 6(lo), 1639-1642 1641 Conclusions Growth of CdS at 350 "C using OCS and Me,Cd as precursors produces high-quality layers of CdS suitable for semiconductor device applications A flow rate of OCS of 80 cm3 min-' (I e a high OCS to Me2Cd molar ratio of ca 200 1) gives the optimum growth rate of CdS layers of ca 100 pm h-' The slow growth rate of CdS layers under these conditions means that OCS may have limited application in the growth of layers of pure CdS However, it could well find use in doping experiments where a precursor yielding low levels of sulfide would be desirable It is noteworthy that there is no sign of carbon incorporation in any of the layers of CdS grown using OCS as a precursor Lastly, it is apparent that OCS is much less reactive with Me,Cd than is H,S under MOCVD con- ditions This finding implies that the impurity of OCS com- monly encountered in commercial samples of H2S is unlikely to perturb significantly the MOCVD growth of layers of CdS when using H2S as the group 16 precursor It is of interest to compare briefly the results of growth experiments on CdS using propylene sulfide" or carbonyl sulfide as the sulfur source The former precursor gives much faster growth rates as it has a much higher reactivity with Me,Cd, as evidenced by the 'static' and 'flow' thermolysis experiments carried out in this laboratory Moreover, it is noteworthy that while neither Me,Cd and OCS nor Me,Cd and propylene sulfide show any interaction at room tempera- ture, the latter pair of reagents, unlike the former, does show a weak Lewis acid-base interaction in the solid phase at 77 K In summary, it is apparent that propylene sulfide is a more suitable sulfur source than is carbonyl sulfide in MOCVD, but that carbonyl sulfide might find specific appli- cations in doping experiments The authors wish to thank J Newey and D Osborne (DRA, Malvern) respectively for the SIMS and thickness and electrical measurements S A C gratefully acknowledges the Research Board of the University of Reading for providing a studentship OBritish Crown Copyright 1995/DRA Published with the permission of the Controller of Her Bntannic Majesty's Stationery Office References 1 M A Haase, J Qui, J M DuPuydt and H Cheng, Appl Phys Lett, 1991,59,1272 2 J M Gaines, R R Drenten, K W Haberern, T Marshall, P Mensz and J Petrozzello, Appl Phys Lett, 1993,62,2462 3 N Nakayama, S Itoh, H Okuyama, M Ozawa, T Ohata, K Nakano, M Ikeda, A Ishibashi and K Akimoto, Electron Lett, 1993,29,2194 4 J M DuPuydt,M A Haase,S Guha, J Qui,H Cheng,B J Wu, G E Hofler, G Meis-Haugen, M S Hagedorn and P F Baude, J Cryst Growth, 1994, 138,667 5 (a) P J Wright, B Cockayne, P J Parbrook, P E Oliver and A C Jones, J Cryst Growth, 1991, 108, 525, (b) M J Almond, M P Beer, K Hagen, D A Rice and P J Wnght, J Muter Chem, 1991,1,1065 6 S Fujlta, M Isemura, T Sakmoto and N Yoshimura, J Cryst Growth, 1988,87,581 7 P J Parbrook, A Kamata and T Uemoto, Jpn J Appl Phys, 1993,32,669 8 P J Wnght, R J M Gnffiths and B Cockayne, J Cryst Growth, 1984'66,525 9 D F Foster, I L Patterson, L D James, D J Cole-Hamilton, D N Armitage, H M Yates, A C Wright and J 0 Williams, Adv Muter Opt Electron, 1994,3, 163 10 M J Almond,M P Beer,S A Cooke,D A RiceandH M Yates, J Muter Chem, 1995,5,853 11 M J Almond, B Cockayne, S A Cooke, D A Rice, P C Smith and P J Wright, J Muter Chem, 1995,5, 1351 12 S Takata, T Minami, T Miyata and H Nanto, J Cryst Growth, 1988,86,257 13 T-L Tso and E K C Lee, J Phys Chem, 1984,88,2781 14 G Cook and 0 D Krogh, J Chem Phys, 1981, 74, 841 and refs therein 15 (a)M Hawkins and A J Downs, J Phys Chem, 1984, 88, 1527, 3042, (b)M Hawkins, M J Almond and A J Downs, J Phys Chem, 1985,89,3326 16 E Krause, Ber, 1917,50, 1813 17 P R Jacobs, E D Orrell, J B Mullin and D J Cole-Hamilton, Chemtronics, 1986, 1, 15 18 M J Almond,S A Cooke,D A RiceandL A Sheridan, J Phys Chem ,1995,99,14641 19 P J Wright and B Cockayne, J Cryst Growth, 1982,59,148 20 D A Rice, PhD Thesis, University of Exeter, UK, 1967 21 JCPDS, International Centre for Diffraction Data 1985, file number 10-454 Paper 6/03483G, Received 20th May, 1996 1642 J Muter Chem, 1996,6(10), 1639-1642