J. MATER. CHEM., 1994, 4( l), 29-34 Chemomechanical Polishing of Gallium Arsenide and Cadmium Telluride to Subnanometre Surface Finish Evaluation of the Action and Effectiveness of Hydrogen Peroxide, Sodium Hypochlorite and Dibromine as Reagents Laurence McGhee; Scott G. McMeekin,bt Irene Nicol," Max I. Robertsonb and John M. Winfield"* a Department of Chemistry, University of Glasgow, Glasgow, UK G12 8QQ Logitech Ltd., Old Kilpatrick, Dunbartonshire, UK G60 5EU Aqueous hydrogen peroxide and sodium hypochlorite in the pH range 7-9 are more effective chemomechanical polishing reagents for gallium arsenide than is dibromine in methanol. Sodium hypochlorite is an acceptable alternative to hydrogen peroxide for gallium arsenide: it also produces good-quality surface finishes on cadmium telluride over the same pH range.The results of dip-etch and polishing reactions, studied using [825r]-labelled dibromine, atomic absorption spectroscopy (AAS) and pH or concentration variation, are used to propose a model for chemomechanical polishing of these materials. Solution etching and chemomechanical polishing of group 14, 13-15 and 12-16 semiconductors are technologically import- ant processes. They are related, since in both cases topochem- ical reactions are potentially involved. In a chemomechanical polishing process an etching reaction, in which material is removed from the semiconductor surface by dissolution, is moderated by mechanical means. Etching and polishing reac- tions have both received considerable attenti~nl-~ and the general principles involved are well known. They include the formulation of etchant reagents as mixtures of oxidants and complexing agents.However, the investigations have been made largely on an empirical basis and many of the events postulated to occur on a semiconductor surface rest on unproven assumptions or have been inadequately described. Increasingly, this lack of detailed knowledge is a barrier to the production of the high-quality surface finishes required for device manufacture. The current requirement of a chemo- mechanical polishing process is that it should be capable of achieving rapidly a subnanometre finish. This means that the variation between high and low points over a short distance on the surface, as measured by a stylus instrument, should be 1 nm or less.Many semiconductor materials are not robust, hence the mechanical forces experienced by the sample should be kept to a minimum. There is, therefore, a requirement for new and more selective reagents. As a basis for developing new reagents we have made a detailed investigation of the behaviour of dibromine (Br,), hydrogen peroxide( H,O,) and sodium hypochlorite(NaOC1) on gallium arsenide(GaAs) and cadmium telluride(CdTe). The use of Br,, usually in MeOH, for polishing GaAs'-' and CdTe'*6 has been widespread for many years and H202 has received some attenti~n.'-~'~-'~ However, these reagents have never been compared under similar conditions. The hypo- chlorite anion is the central member of the series of reagents represented by -0-0-, -0-X and X-X, where X represents a halogen, but it has been used far less Our approach has been to combine chemomechanical-polishing experiments, carried out with careful control of reagent concentration and pH where appropriate, with physicochemical studies of dip-etch reactions.A preliminary account of part of this work has appeared'* and is part of a t Present address: Department of Electronic and Electrical Engin-eering, University of Glasgow, Glasgow, UK G12 SQQ. general investigation of polishing and etching of electronic, optoelectronic and optical rnaterial~.'~ Experimental General Methods The acidities of the solutions used in the dip-etch and polishing reactions were determined using a standard pH electrode (Russell pH Ltd) with an EIL 7050 meter. A Perkin-Elmer Lambda 9 spectrometer was used to monitor the decomposition of hypochlorite solutions as the pH was varied.The removal of substrate material to polishing or etching solutions was monitored using a Perkin-Elmer 1100B atomic absorption spectrophotometer in conjunction with an MHS-10 hydride generator to detect soluble As, and an HGA 700 graphite furnace to detect soluble Te. Soluble Cd was determined using the flame method. No attempt was made to determine soluble Ga using AAS since its sensitivity to detection is considerably lower than those of the other three elements. Spectrosol standard solutions were employed throughout and standards were matrix matched to the ana- lytes.Sample solutions from etches or polishes required no pretreatment other than dilution. Replicate analyses were performed in all cases. Data presented are specimen results from series of experiments having reproducible outcomes. Reflectance IR spectra of etched wafer surfaces were obtained using a Nicolet IRIS dedicated FTIR microscope. In selected cases, wafer surfaces were mapped by taking 10 random sampling points over a 5 mm x 4 mm area. In dip-etch experi- ments, wafers were weighed before and after a reaction to correlate mass changes with soluble material produced. Polishing experiments were performed using Logitech CP4000 (reagents Br,-MeOH and H202-NH3) or Logitech PM2A polishing equipment, in the latter case with the addition of a PP5 vacuum jig.CS2Br]-labelled dibromine, used to determine the Br, interaction with a wafer surface, was produced from [*'Br]-labelled ammonium bromide by reaction with H2S04 in ~acuo.'~Neutron irradiations, "Br(n, y)82Br, were carried out in the Scottish Universities Reactor Research Centre, East Kilbride, with a neutron flux of 3.6 x 10'' neutron cm- s-' for 3 h. The activity of the CS2Br]-labelled ammonium bromide was ca. 100 mCi. Labelled solids were transferred to Glasgow ca. 24 h after irradiation to ensure complete decay of short-lived components. The specific activity of the [82BrJ-bromine labelled Br, was determined by counting a measured aliquot of Br, in AnalaR chloroform (1cm3).CS2Br] count rates were determined using a sodium iodide scintillation counter with a scaler ratemeter (Ecko and NE). The usual background corrections were made and radiochemical purity was deter- mined by half-life measurement (tl,, =35.3 h). Preparation of Reagents The etchant solutions were prepared from analytical grade materials and were freshly prepared prior to each experiment, since all decomposed to some extent at room temperature. In order to make meaningful comparisons in polishing and dip-etching experiments, it was necessary to determine pre- cisely the extent of decomposition of a reagent. The peroxide reagent was prepared by dropwise addition of aqueous NH, (Pronalys, M & B, 33% wt./wt.) to aliquots (60.0 cm3) of H,O, (AnalaR, BDH, 100 vol.33% wt./wt.) until the pH reached 6.54, 7.4 or 8.0. The solutions were stored in plastic bottles and aliquots taken for analysis every few hours. Each aliquot was diluted x 100, acidified with 0.1 mol dm-3 H2S04 (20 cm3) and titrated against standard 0.020 mol dm-, KMnO, solution. The higher the pH the faster the decompo- sition. At pH 8.0, the oxidising ability was 25% of its original value after 40 h, at pH 7.4 it was 75% after 140 h and at pH 6.54 it was 78% after 160 h. All solutions were sealed during use since minute quantities of impurity catalysed the decomposition at irreproducible rates. Bromine-methanol mixtures (100cm3) were prepared from Br, and MeOH (AnalaR BDH) and the concentration determined by iodi- metric titration using 0.100 rnol dm-3 [S,O,]’-(Volucon M & B).Aqueous sodium hypochlorite (12% wt./wt., Spectrosol, BDH) was brought to the desired pH by dropwise addition of glacial acetic acid (Pronalys, M & B). Aliquots (5 cm3) of Br,-MeOH were pipetted into aqueous KI (1 g in 100 cm3) and the liberated iodine titrated against 0.100 mol dm-3 [S2O3I2-. Approximately 33% of the oxidising power of Br,-MeOH was lost over 24 h. After 6 days, the oxidising power decreased by ca. 50%. This behaviour was observed for Br,-MeOH solutions in the concentration range 0.1-0.4 mol drn-,. Available chlorine in NaOCl solutions (1:100 vol./vol. in water) was determined by iodimetry with 0.100mol dm-, [S2O3I2-over the pH range 5.4-8.1 and over 21 h.The conversion of NaOCl to HOCl was monitored using electronic spectroscopy. For [OCl] -A,,, =292 nm and for HOCl A,,, = 236 nm.17 The rates at which [OCll- was converted to HOCl and underwent decomposition were pH dependent. Solutions whose pH initially was 67 had a pH of 4.5 after 20 h. Their electronic spectra showed no evidence for [OCl] -. Solutions prepared at higher pH were stable for longer periods but all trace of [OCll-in the spectra had disappeared after 24 h. Provided the solutions were prepared immediately prior to etching, no significant change in pH occurred during the 0.5 h duration of an experiment as shown by control experiments. Dichlorine was evolved during decomposition of the solutions and iodimetric titration revealed a decrease in oxidising power after 0.5 h.Dip-Etch Experiments Wafer sections (2.5 cm x 2.0 cm) were etched in solutions (60.0 cm3) which were stirred continuously. The wafer sections were supported in a Pyrex glass holder inside a Pyrex reaction vessel. Etch times were 0.5 h or 10 min for GaAs, and 10 min for CdTe. Material lost to the solution was determined by mass change and by determination of soluble As, Cd or Te by atomic absorption spectroscopy. In separate experiments, the effects of concentration variation in Br,-MeOH, pH J. MATER. CHEM., 1994, VOL. 4 change in H,02 and both effects on etching with OCI--HOCl were determined. Reactions under anhydrous conditions were also undertaken using dry Br, (P,O, in vacuo) in dry dichloro- methane (molecular sieve type 3A).These experiments used a double-limbed Pyrex vacuum vessel incorporating a stop-cock to isolate one limb from the other (40cm3) which was attached to a Pyrex vacuum line ( lop4Torr). Concentrations ranged from 0.662 to 10.85 mol dm-3. Polishing Experiments The polishing experiments employed single-crystal GaAs ( 100) and polycrystalline CdTe wafers (2.5 cm x 2.0 cm). Typical conditions for H20,-NH, and Br,-MeOH polishes were a plate speed of 70rpm, a load of 400 gem-' and a reagent feed rate of 700 cm3 h-l. Polishing with H,O,-NH, was investigated over the pH range 6-8.5 since dip-etch experi- ments showed this to be the optimum. Polishing with Br,-MeOH was carried out over the concentration range 0.1-0.4 mol drn-,.The conditions used for [OCl] --HOCI polishing were a plate speed of 30 rpm and flow rate of 200 cm3 h-’ under minimal load. In all cases a velvet-napped polishing pad was used. Solutions were in the pH range 6.4-12.3 and polishing reactions were performed usually over 0.5 h. In each case the surface was examined by a Zeiss Normaski microscope, and surface roughness of the higher quality finishes was measured over a 5mm range using a Rank Taylor Hobson Talystep stylus instrument (2 pm radius stylus). Data were compared using the parameter R,, the arithmetic mean of the departures of the roughness profile from the mean line. The polishing of polycrystalline CdTe revealed excessive grain-boundary phases, and, consequently poorer finishes resulted.Hence, R, values for CdTe were significantly higher than for GaAs. In all cases the R, value was the result of several traces obtained over a 5 mm trace from different regions of a wafer. Radiotracer Experiments CS2Br]-Labelled dibromine was used to determine the adsorp- tion of Br, on a wafer surface and to follow the outcome of the adsorption process. Etching using radioactive bromine was carried out in the double-limbed vessel under vacuum as described for the anhydrous dip-etches (vide supra);each limb could be fitted into the well scintillation counter enabling rs2Br] count rates from solid and vapour to be determined separately at regular intervals. The count rate from vapour above the wafer was negligible compared with those from the wafer itself and could be ignored without serious error.Results Dip-Etching of GaAs and CdTe Single crystal (100) wafers of gallium arsenide were etched by both aqueous hydrogen peroxide and aqueous sodium hypo- chlorite solutions at room temperature over 0.5 h periods but in both cases the extent of reaction was highly pH dependent. Mass decreases using H,O, varied uniformly from 1.4% at pH 8.1 to 33.2% at pH 8.85 and using NaOCl they varied from 1.0% at pH 7.5 to 16.4% at pH 13.0. In both cases, mass decreases were <1% at lower pH values and reactions were very rapid at high pH; with NaOCl at pH 13.0, a black corroded surface resulted. For comparison, GaAs etching with Br,-MeOH (0.44 mol dm-3) produced mass decreases in the range 3.8-4.8% after 10 min.In contrast, polycrystalline cad- mium telluride wafers did not appear to be etched by H,O, or NaOCl under these conditions but etching with Br, in MeOH occurred rapidly. Mass decreases determined after J. MATER. CHEM., 1994, VOL. 4 10 min etches ranged from 1.8 to 24.9% for Br, concentrations in the range 0.12-0.83 mol dm-3. The effects of reagent concentration and/or pH were deter- mined by analyses of As or Cd and Te in the solutions after a predetermined time and are shown in Fig. 1-3. The data are presented as soluble As, Cd or Te, these being determined from the product of the concentration of the element and the volume of solution.For a given concentration of reagent, the quantity of As removed from GaAs by Br,-MeOH after 0.5 h was significantly greater than that removed by NaOCl (Fig. 1) and etching by the former reagent was detected readily after 10 min even below 0.1 mol dmP3. In all cases the relationship 160 I401 120 20/00.0 0.5 1.o 1.5 2.0 concentration/mol dm-3 Fig. 1 Dependence of soluble As on reactant concentration in dip-etch reactions of GaAs using Br,-MeOH for 0.5 h (A); Br,-MeOH for 10min (A); NaOCl for 0.5 h (0) 30I I 25 tI 15 0=tu, 10 y 02 4 6 8 10 12 14 PH Fig.2 Dependence of soluble As on pH in dip-etch reactions of GaAs for 0.5 h using H,O,-NH, (0)or NaOCl (A) 0.6Oe71 between soluble As and reagent concentration appeared to be linear to a good approximation.Dilution of NaOCl solutions inevitably led to a pH decrease (from 13.0 to 11.3 for the data in Fig. 1).Changes in dip-etch characteristics for GaAs arising from protonation of [OCl] -and deprotonation of H,O, were investigated by determi- nation of soluble As, using solutions prepared as described in the experimental section, with the results shown in Fig. 2. For both etchants these data provided a more sensitive test of the extent of reaction than was possible from simple weighing. In particular, the data indicated that little or no reaction occurred below pH 7.5; above this value the reaction was substantial particularly for the reagent H20,-NH, (Fig. 2). For H202-NH, the relationship between soluble As and pH was exponential; the scatter in the experimental data arises from the combination of data from two sets of reactions.The corresponding relationship for NaOCl etching was lcss well defined, owing to experimental difficulties in obtaining stable pH values in the pH region 7.5-9.5 with relatively small volumes of solution. The lack of a well defined relationship possibly reflects the complexity of the reagent, since [OCll-, HOCl and their decomposition products are all present.17 The most notable feature was the indication of a plateau between pH 9.5 and 11.0, suggesting a buffering effect; a similar plateau was found in mass decrease measurements. Combined, these observations suggest that the plateau was a real effect rather than an experimental artifact.The variation of Cd and Te removal from polycryuitalline CdTe wafers with Br, in MeOH is shown in Fig. 3. At all concentrations both elements were detected in solution after 10min. The data were not sufficient to define precise relation- ships but except for the lowest concentration of Br2 used, more Cd than Te was lost from a wafer. Although this could have been a reflection of wafer quality, it suggests that the Br,-MeOH reaction is not straightforward. Examination of GaAs wafers after etching with H,O,-NH, or Br,-MeOH by FTIR microscopy indicated a distinct difference in the distribution of hydroxy group density between the two surfaces. Bands assignable to surface terminal hydroxy groups were present in both cases but were better resolved after treatment with H,O,-NH,.Mapping this surface using the relatively sharp 1670 cm-I band, indicated a contoured surface in which peaks and valleys were well defined. The surface of GaAs after etching with Br,-MeOH was dmost featureless implying a lack of specificity in the etching reaction. In previous work it has been assumed that the products from etching of GaAs or CdTe by Br, in MeOH are the corresponding bromides, which for Gal", As"' and Te" will be hydrolytically unstable. It was therefore of interest to examine the behaviour of Br, towards these materials under rigidly anhydrous conditions. Etching of GaAs wafers by Br2-CH,C1, solutions in V~CUOled to small mass increases rather than the decreases observed with MeOH solution when trace water was not excluded.However, there was no obvious pattern discernible and labelling of Br, with the radioactive 0.3 q, led to the immediate detection of ["Br] activity from the solid. Thereafter there was a slow, continuous increase in the 0.2 C8,Br] solid count rate with a concomitant decrease in C8,Br] count rate for the vapour. This is illustrated in Fig. 4(a)for a 0.1 reaction between 82BrBr (2.3 mmol) and GaAs (1 rnmol). 0.0 0.0 0.2 0.4 0.6 concentration/mol dm-3 0.8 1.o Saturation of the GaAs surface by Br, was never achieved and the reaction could be accelerated in its later stages by addition of more ',BrBr. The uptake of bromine by GaAs determined for the specific count rate of 82BrBr and the solid tracer C8,Br] was adopted as a more sensitive probe.2 0.4 Exposure of a GaAs wafer to ',BrBr at room temperature Fig. 3 Dependence of soluble Cd (0)and Te (A)on Br, concentration count rate after 14 h was 3.2 mg atom Br(mmo1 GaAs)-'. for 0.5 h dip-etching of CdTe by Br,-MeOH Uptakes determined in a second experiment using 1.61 rnmol J. MATER. CHEM., 1994, VOL. 4 Polishing of GaAs and CdTe /< // ,' I 1 I I5 01 I I I5 0 50 100 150 200 250 300 350 0 0.5 0 I@) 0.211 -0.1 * A & L I I I I I Fig. 4 Reactions between 82BrBr and GaAs or CdTe; (a)increase in solid Cs2Br] count rate (0)and decrease in count rate of 82BrBr(0) with time; (b)comparison between Cs2Br] count rates from GaAs and CdTe under identical conditions.(0)GaAs; (A)CdTe; (*) indicates that volatile and weakly adsorbed material were removed at this time Br, were 0.75 and 2.1 mg atom Br (mmol GaAs)-' after 5.5 and 22 h respectively. After extensive exposure to Br, vapour, a viscous yellow oil was formed on the surface of GaAs and colourless needles on the vessel wall. Their physical appear- ance suggested that they may have been AsBr, and GaBr, respectively. Although the exposure of CdTe to ',BrBr resulted in an immediate uptake of radioactivity by the solid, the solid count rate showed no increase thereafter. Addition of further 82BrBr or the use of a larger quantity initially, had no effect. The behaviour of CdTe and GaAs wafer fragments (0.2 mmol) towards "BrBr vapour (0.05 mmol) at room temperature is compared in Fig.4(b). After 2 h, the count rate from GaAs was approximately seven times greater than that from CdTe. All volatile and weakly adsorbed materials were removed by pumping at this point. The CS2Br] count rate from CdTe was unaffected while that from GaAs decreased to 58% of its former value. There is a strong implication therefore that one of the products (possibly AsBr,) from the reaction between GaAs and Br, is volatile while the other (possibly GaBr,) is more persistent. The effect of an organic solvent on the reaction between 82BrBr and GaAs was examined by allowing the vapour to diffuse into a counting cell containing a GaAs wafer immersed in 1,2-dihydroxyethane. Methanol was unsuitable for this purpose owing to its volatility and C,H,(OH), has been used as an alternative solvent to MeOH in GaAs polishing. The behaviour observed was similar to that in the absence of a solvent but the uptake of CS2Br] activity appeared to be slower.Thus the solvent did not enhance the extent to which reaction occurred but rather exerted a moderating effect in limiting the concentration of Br, that diffused to the GaAs surface. Single-crystal GaAs( 100) wafers (20 x 25 mm) were polished very satisfactorily using aqueous H202 in the pH range 6.0-8.5. In all cases subnanometre surface finishes were obtained, R, being in the range 0.2-0.9 nm. Surfaces viewed through a Normaski optical microscope were featureless. The quantity of material removed during 5 min periods was con- stant at a given pH value over a 1.5 h polishing experiment but increased exponentially with increasing pH over the range 6.0-8.5 (see Fig.1of ref. 14). The relationship determined was log S=O.8 pH -5.26 (1) where pH refers to the H202 solution and S is the stock (material) removed (pm) by a 0.5 h polish, sample loading 400 g cmP2 and H202 solution flow rate 700 cm3 h-l. It is noteworthy that exponential relations with pH have been found for both polishing and dip-etch processes [Fig. 2(b)], the latter being displaced from the former by ca. 1 pH unit. Even after 5 min polishing at pH >8.5 the surface finish was less satisfactory, owing to excessive decomposition of the etchant and the formation of rough areas.Polishing for periods>0.5 h in the pH range 8.0-8.5 led to a similar situation. It is apparent therefore that careful pH control for this reagent is crucial. Dibromine in MeOH was a less satisfactory reagent since even with a low concentration (0.1% vol./vol.), 'orange peel' roughness7 was apparent; at higher concentrations this was visible to the naked eye (cf. Plate 1 in ref. 14). Stock removed during a 0.5 h polish increased linearly from 5 to 183 pm as the Br2 concentration was increased from 0.5-2.0% vol./vol. Surface roughness increased also from R, =2-9 nm. Determinations of total As collected over 0.5 h periods during polishing with Br,-MeOH and aqueous H202 gave compar- able results.For Br, concentrations of 0.1, 0.2, 0.5, 1.0 and 2.0% vol./vol. Values for As were 14, 93, 114, 247 and 322 mg respectively, and for H20, at pH 7.0, 7.5 and 8.0 the As values were 13, 45 and 257mg. Together with the stock removal measurements made, these data indicate that the overall rates of polishing GaAs with 1.0% vol./vol. Br, in MeOH or with 30% wt./vol. H202at pH 8.0 are comparable. The quality of the surface finish is therefore not related directly to the overall polishing rate. Although it was possible to polish CdTe using aqueous H202, at pH values somewhat lower than those used for GaAs, the process was slow and therefore aqueous NaOCl was investigated as an alternative. This reagent polished GaAs satisfactorily within 0.5 h over the pH range 8.0-8.5, R, ca.1 nm, the optimum pH being 8.3. Polycrystalline CdTe was polished to R, ca. 4nm over a similar pH range, optimally at pH 8.8. In both cases stock removal was small, as were the quantities of As and Cd detected in the polishing fluids (Table 1). The trend in the As data was for removal of As to Table 1 Total soluble arsenic and cadmium detected in polishing fluids collected after 0.5 h polishing of GaAs and CdTe with aqueous NaOCl-CH,C02H PH arsenic/mg cadmium/mg 6.4 5.0 7.2 7.2 1.5 7.6 4.3 0.4 8.0 3.9 2.0 8.3 11.2 1.5 8.8 1.2 9.2 12.5 0.8 10.6 0.4 12.3 18.1 0.3 J. MATER. CHEM., 1994, VOL. 4 be greater at high pH; the value of 18mg determined at pH 12.3 corresponded to 3 pm stock removal.In comparison, stock removal from GaAs by H,02 at pH=8.5 under other- wise identical polishing conditions, was 150 pm. The quantities of Cd detected in solution were very small at all pH values (Table 1)and Te was undetectable indicating that these values were less than 0.5 mg. Discussion and Conclusions In this work it has been demonstrated that aqueous H,O, in the pH range 6.0-8.5 is superior to the more widely used Br, in MeOH for chemomechanical polishing of GaAs wafers. Aqueous NaOCl in the pH range 8.0-8.5 is an acceptable alternative and this reagent will also polish polycrystalline CdTe satisfactorily; the quality of the surface finishes on CdTe obtained in this work is a reflection of the polycrystalline nature of the samples used.Chemomechanical polishing of semiconducting materials is often carried out with the addition of abrasives such as a-alumina, presumably to achieve shorter polishing times. This is unnecessary for GaAs and CdTe and thus the problem of sub-surface damage associated with undue mechanical action is minimised.'' Some insight has also been gained into the reasons for the differences in polishing behaviour by reference to the etching reactions of the reagents. Such studies have been made in the past usually by contact gauge or optical microscopy measure- ments or by mass changes. The advantage of our approach which involves investigation of both the wafer and the etchant solution is that a greater understanding of the physicochemical processes occurring is obtained.Our rationalisation of the observations made on the reactions of Br, and H,O, and NaOCl with GaAs and CdTe follows and is summarised in Scheme 1. However, before proceeding, some general com- ments are warranted. First, in accordance with previous we assume that in the case of GaAs oxidation occurs to give Ga"' and As"' compounds as the final products. Second, the outcome of the reaction involving GaAs is depen- dent on kinetic factors which relate to the removal of products from the surface rather than the thermodynamics of the redox reaction. The reaction between aqueous H202 and GaAs almost certainly leads to the formation of an oxohydroxy layer on the surface, as judged by FTIR analysis, which must be removed to enable further reaction to occur.The importance of pH in affecting dissolution of the oxidized surface has been recognised previ~usly.'~'~ The key observation made here is the exponential dependence, on pH of both dip-etching and chemomechanical polishing processes which may be a result of the higher solubility of the species that comprise the layer as the pH increases. In the latter case the layer is removed largely by mechanical wiping, since in the lower part of the effective pH range for polishing, 6.0-8.5, As removal under 1 redox Br, t GaAs H202+ GaAs I 1 adsorbed Br layer oxohydroxy layer I 2complexation GaBr, AsBr, Ga and As oxohydroxides 3 removal of products soluble Ga and As oxohydroxides Scheme 1 Reactions of GaAs with Br,-MeOH and H,0,-NH3 dip-etching conditions is ineffective (Fig.2). If during the mechanical wiping process, the sparingly soluble material forms a passivating layer, an element of selectivity becomes possible. Support for this suggestion is provided by the observation of varying hydroxy group density on a GaAs surface that has been etched using H202-NH,. The passivating layer concept was first described in the context of polishing of silica'' and we have recently identified chemically the constituents of the layer formed when silica is polished by an hydrogendifluoride reagent.I5 Chemomechanical polishing of GaAs by aqueous NaOCl has many similarities with its polishing by H,O,; the process is pH dependent and good surface quality results.Whilst surface passivation of polycrystalline CdTe occurs with aque- ous H202 or NaOC1, polishing with NaOCl is feasible despite the more tenacious character of the surface layer (see Table 1). The passivating layers are most likely to be oxidic in nature with the added possibility of chloro species when using NaOCl. Dichlorine, formed by the reaction of [OCll- or HOCl with C1- anion, inevitably present in commercial hypochlorite solutions,20 could also be one of the redox-active species involved. Chlorine-36 radiotracer experiments to test this possibility are currently in progress. Cg2Br]-radiotracer experiments indicate that bromination of a CdTe surface passivates it towards further reaction with Br, but the dissolution process in MeOH does not confer sufficient selectivity for Br,-MeOH to be a good polishing reagent towards CdTe.One of the so far unexplained features of the etching of CdTe by Br,-MeOH is the higher Cd level detected in solution for the reaction. Recent studies"*22 of CdTe etching by aqueous Br, in the presence of HBr indicate that the reaction is first order with respect to Br,, it has a low activation energy,' indicative of a diffusion controlled reaction' and that the redeposition of Te observed during the reaction occurs uia the production of an undefined soluble Te species.22 Clearly the reactions occurring in the CdTe-Br, system are complex and require further investigation.Adsorption of bromine on GaAs is followed by a redox reaction which continues as long as Br, is present. There appears to be no mechanism for controlling the reaction in this case and the surface finish obtained from chemomechan- ical polishing with this reagent suffers as a result. The authors thank DTI, SERC and Logitech Ltd for financial support of this work, partly through the DTI Nanotechnology Initiative. The staff of the Scottish Universities Research Reactor Centre, East Kilbride and Dr. D. James, Nicolet Instruments Ltd, Warwick are thanked for assistance with neutron irradiations and FTIR microscopy respectively. References 1 B. Tuck, J. Mater. Sci., 1975, 10,321. 2 D. J. Stirland and B. W. Straughan, Thin Solid Films, 1976, 31, 139.3 W. Kern, RCA Rev., 1978,39,278. 4 C. S. Fuller and H. W. Allison J. Electrochem. SOC., 1962,109,880. 5 M. V. Sullivan and G. A. Kolb, J. Electrochem. SOC., 1963, 110, 585. 6 S. L. Reidinger, D. W. Snyder, E. I. KO and P. J. Sides, Muter. Sci. Eng., 1992, B15, L9. 7 J. C. Dyment and G. A. Rozgonyi, J. Electrochem. Soc., 1971, 118,1346. 8 Y. Mori and N. Watanabe, J. Electrochem. SOC., 1978,125, 1510. 9 E. Haroutiounian, J. P. Sandino, P. ClCchet, D. Lamouche and J. R. Martin, J. Electrochem. SOC., 1984,131,27. 10 J. J. Kelly and A. C. Reynders, Appl. Surface Sci., 1987,29, 149. 11 I. Barycka and I. Zubel, J. Muter. Sci., 1987,22, 1299. 12 S. H. Jones and D. K. Walker, J. Electrochem. Soc., 1990, 137, 1653. 34 J. MATER. CHEM., 1994, VOL. 4 13 V. L. Rideout, J. Electrochem. SOC.,1972,119,1778;H. Hartnagel 18 D. F. Weirauch, J. Electrochem. SOC.,1985, 132, 250. and B. L. Weiss, J. Muter. Sci., 1973,8, 1061. 19 L. M. Cook, J. Non-Cryst. Solids, 1990, 120, 152; N. J. Brown, 14 S. G. McMeekin, M. Robertson, L. McGhee and J. M. Winfield, Precision Engineering, 1987,9, 129. J. Muter. Chem., 1992,2, 367. 20 J. W. Mellor, Comprehensive Treatise on Inorganic and Theoretical 15 D. S. Boyle, J. A. Chudek, G. Hunter, D. James, M. I. Littlewood, Chemistry, Longmans, Green and Co, London, 1946, vol. 2, L. McGhee, M. 1. Robertson and J. M. Winfield, J. Muter. Chem., p. 245; G. Peintler, I. Nagyphl and I. R. Epstein, J. Phys. Chem., 1993, 3, 903; M. Beveridge, L. McGhee, S. G. McMeekin, 1990,94,2954. M. I. Robertson, A. Ross and J. M. Winfield, J. Muter. Chem., 21 A. S. Kovalenko, B. L. Druz, E. I. Gusakova and L. P. Tikhonova, 1994,4, 119. J. Appl. Chem., USSR (Engl. Transl.), 1992,65.466. 16 L. McGhee, P. R. Stevenson and J. M. Winfield, unpublished 22 P. F. Vengel, V. N. Tomashik and A. V. Fornin, J. Appl. Chem., work, 1992. USSR (Engl. Transl.), 1992,65,751. 17 L. C. Adam, 1. Fabian, K. Suzuki and G. Gordon, Inorg. Chem., 1992,31,3534. Paper 3/04289H; Received 2 1st July, 1993