J. MATER. CHEM., 1994, 4(12), 1821-1826 Studies of the Effects of NF, on the Growth of Polysilicon Films by Low-pressure CVD Part 1.-Effect on Growth Rate Michael L. Hitchman,* Junfu Zhaot and Sarkis H. Shamlian Department of Pure and Applied Chemistry, University of Strathclyde, 295 Cathedral Street, Glasgo w, UK G7 1x1 The control of crystalline quality of polysilicon prepared by low-pressure chemical vapour deposition (LPCVD) is important for device applications. A number of reaction parameters affect this quality and, in general, it has been found previously that layers grown in the amorphous state show good structural perfection and low strain on annealing. We have investigated a new strategy for producing good crystalline polysilicon by attempting to etch layers during the deposition process in order to reduce the size of crystallites in the layers.In this paper we report on the effect of nitrogen trifluoride on the growth rate of polysilicon. We interpret the decreased growth rate in the presence of NF, in terms of an etching effect by molecular fluorine and as a result of growth inhibition by strong surface adsorption of the NF3. The mechanism of this inhibition is discussed. One of most widely used materials for the fabrication of modern VLSI circuits is polycrystalline silicon, commonly referred to as polysilicon. It is used for the gate electrode in MOS devices, for the fabrication of high-value resistors, for diffusion sources to form shallow junctions, for conduction lines, and for ensuring ohmic contact between crystalline silicon substrates and overlying metallisation structures.The standard method of preparation of polysilicon layers in modern integrated circuit technology is the technique of low-pressure chemical vapour deposition (LPCVD).' Typically, SiH, is pyrolysed with a temperature in the range 580-650 "C and at a pressure in the range 25-130 Pa. However, the crystalline quality of the polysilicon deposited can, depending on deposition parameters, be very variable. For most, if not all, of the applications mentioned above there is the need to control carefully the degree of crystallinity of the layers since poorly crystallised material can lead to high internal stress (e.g. dislocations and stacking faults) which, in turn, can produce unwanted defects in an active device area.In recent years considerable efforts have been made to establish the deposition conditions required for obtaining high-quality crystalline layers. This topic has recently been reviewed.' Factors affecting the crystalline structure of polysilicon have been shown to include deposition temperature,2 partial press- ure of silane,, length of time for the dep~sition,~ and type of dopant used.5 All of these factors have an influence on the relative rates of deposition or layer growth (j,) and crystalline growth (jCg).In general, whenjd>jcg then the deposited film will be largely amorphous or consist of very small crystallites, and when j, <jcgthen the film will show extensive crystallis- ation.Actually, the situation is a little more complicated than this simple picture would suggest because crystallisation depends upon a nucleation step and associated with this there is usually an induction time, which means in some instances j, <jcgbut, nevertheless, films can remain arnorph~us.',~ Not-withstanding this complication, the important feature from a practical device point of view is that it has been found2 for polysilicon layers grown in the amorphous state that on post- deposition annealing the highly desirable properties of good structural perfection and low strain are obtained. This is in contrast to material which is either partially crystalline or fully crystalline on deposition and which on annealing shows ~~ t Present address: Department of Applied Chemistry, Taiyuan University of Technology, Taiyuan, Shanxi 030024, P.R.China. considerable lattice disturbances with some poor and some good crystallisation, leading to device defects. The need to control carefully the degree of crystallinity of polysilicon layers grown by LPCVD, which effectively means conti olling the relative rates of layer growth, is therefore apparent. The influence of deposition parameters such as deposition temperature, pressure and time as well as the presencc of in situ dopants on degree of crystallinity has been investigated previously, as mentioned above. We have considercd an alternative strategy for trying to produce good crystalline polysilicon by introducing into the CVD reactor an add] tional reactant which could partially etch the polysilicon as it is deposited.The rationale behind this strategy is that gas etching of the layer during the deposition process will ieduce the size of crystallites in the layer leading to an eflective decrease in crystalline growth rate. In particular, high-mergy surface sites such as would be associated with cryslalline defects and protuberances could be preferentially etched and this would lead to a better quality crystalline structure. To investigate this concept we have chosen NF, as a possible etchant gas for a number of reasons. It has been shown, for example, that NF, can be used to etch polys~licon~ and single-crystal silicon7 with etch rates as high as 1 pm min-'. This, though, has been in plasma reactors which will give rise to charged as well as neutral fluorine-conttiining molecules for reaction with silicon to form volatile SiF, species.However, it has also been shown that silicon can be etched by both molecular8~9 and atomic fluorine" and, fL rther-more, it has been reported'' that at temperatures in excess of 300°C that NF, dissociates homogeneously to forin F,; thermodynamic calculations (see discussion below) lent 1 sup-port to this. Experimental studies', have also shown that NF, undergoes spontaneous dissociative chemisorption on :,ingle- crystal silicon. Although at low temperatures up to 200°C it has been shown that13 NF3 does not spontaneously etch Si, we found in preliminary experiments that for typical polysil- icon deposition temperatures etching by NF, of a polyca'ystal- line silicon layer on a sapphire wafer did, in fact, OCCUI.Also a mass-spectroscopic investigation of the effect of temperature on NF, decomposition has shown14 that it does undergo pyrolysis at 600 "C. Therefore, NF, seemed a good candidate to investigate the idea of in situ etching during deposition in order to try and improve polysilicon layer quality. Other fluorine-containing compounds such as HF, F,, CIF, and XeF, might also be suitable, but NF, is less drfficult J. MATER. CHEM., 1994, VOL. 4 and dangerous to handle than most of these alternative materia1s.l5.l6 In this paper we examine the effect of NF, on polysilicon growth rate and discuss reasons for the observed effects.In a second paper we shall report on the effect of NF, on the degree of crystallinity of as-deposited polysilicon layers and on the crystalline quality of such layers after thermal ar~nealing.'~A third paper will discuss the effect of NF, on the chemical composition of polysilicon deposited by LPCVD.~~ Experimental The LPCVD reactor, the gas-handling system and typical deposition procedures and conditions have been described in detail previ~usly.'~~~' Briefly, the reactor consisted of a 95 mm (id) vitreous silica furnace tube with a three-zone heater. A flat temperature profile of various temperatures in the range 560-670 "C was maintained to within f1"C in the reactor zone.Pure SiH, and NF, (Air Products plc) were used as separate source gases with He being used as a diluent gas for some experiments. Depositions were made with 100% SiH, and with mixtures of SiH,-He, SiH,-NF, and SiH,-He-NF,. In all cases the total gas flow was kept constant at 100 sccmt Silane flows were in the range 5-100 sccm, NF, flows in the range 1-5 sccm and He flows adjusted accordingly to give a total flow of 100 sccm. The total pressure in the reactor was kept constant at 65 Pa by means of a control valve in the exhaust line to the rotary pump. Films were deposited on either 2 in sapphire ( 1102) or silicon (111) substrates. The wafers were placed parallel to each other and perpendicular to the flow with a spacing of 5mm in a fused silica boat capable of holding 50 wafers.Slots not used by test wafers were filled with dummy silicon wafers. Measurements of film thickness, crystallinity and composition were made, but in this paper we only report on thickness measurements. These were made by etching a concentric ring pattern for layers deposited on sapphire substrates2' and then measuring a thickness profile with a Sloan Dektak IIA profilometer. Results on layer crystallinity and composition are reported on in subsequent publication^.^^*^^ Results and Discussion Fig. 1 shows the variation of growth rate (j)with mole fraction of silane (x,)at constant temperature (580 "C), total pressure (65 Pa) and total gas flow (100 sccm). In the absence of NF, the dependence of j on x, is comparable to that reported earlier2' and is characteristic of a reaction system following Langmuir-Hinshelwood kinetics.'l The growth rate varies from 1.9 to 6.2 nm min-' for x, in the range 0.05-0.6.On adding a constant amount of NF, (1 sccm) to the SiH,-He mixture there is a dramatic fall-off in growth rate. For example, at a silane mole fraction of 0.15 the growth rate is 0.01 nm min-' with NF, compared with ca. 3 nm min-' at the same value of x, but in the absence of NF,. At x,=O.6 the values of growth rate with and without NF, are 3.8 and 6.2nm min -respectively. For x,<0.15 any polysilicon film deposited, even over relatively long run times (e.g. several hours), was too thin to measure reliably. All of these results indicate that, as expected, there could be some etching of polysilicon because of NF, addition.This effect is shown in a different way in Fig. 2 where growth rate is plotted as a function of mole ratio (y) of NF, : SiH, for a constant mole fraction of silane. However, it can be seen that if etching is occurring then the rate is not t Standard cm3 min-' "i 0 0 0 7 Ic 0.-E 0 0 0 0 0 0 Fig. 1 Dependence of polysilicon growth rate (j)on mole fraction of silane (x,) from a mixture of SiH, and He: 0,without NF,; 0,with NF, (1 sccm); & = 580 "C; P, =65 Pa; total gas flow, 100 sccm 30r I 0.00 0.04 0.08 0.12 Y Fig. 2 Dependence of polysilicon growth rate (j)on mole ratio (y) of NF, and SiH,: Td=580 "C; PT=65 Pa; total gas flow, 100 sccm; .x$=0.5 simply related to the gas-phase concentration of NF, since the fall in growth rate is not a linear function of 7.It can be also seen from Fig. 2 that the growth rate does not fall to zero for large y. In Fig. 1 the largest value of y, corresponding to x,=O.15, is 0.067 for which j~0.In Fig. 2 for the same value of y, j z7.0 nm min -'. The main experimental difference between the results in the two figures is the deposition temperature, 580°C for the results in Fig. 1 and 650°C for those in Fig. 2. The effect of temperature on growth rate is shown in more detail in Fig. 3. The total overall growth rate (j,) under any specific set of conditions can be taken as the difference between the deposition rate (j,) and the etch rate (j,) (i.e.jT=jd-je), each of which can be expressed in the form of general rate equations, and where kd and k, are the appropriate rate constants andf'(C,) and g(cN) are functions of the concentrations of SiH, (C,) and NF, (CN),respectively. The temperature dependence of J. MATER. CHEM.. 1994, VOL. 4 40 r 30 t7 t 0 10 -0 00 0 0 560 600 640 680 deposition temperature/"C Fig. 3 Dependence of polysilicon growth rate (j) on deposition tem- perature from a mixture SiH, and He: ?, without NF,; @, with NF, ( 1 sccm); P, = 65 Pa: Y, =0.5;total gas flow = 100 sccm the two rates can then be written as and (4) From the data in Fig. 3 for deposition without NF,, an Arrhenius-type plot can be made and the activation energy for deposition (Ea,d) can therefore be obtained.From the difference between growth without and with NF, the acti- vation energy for etching (Ea,e)can similarly be obtained. Fig. 4 shows the plots of eqn. (3) and (4) and from the slopes Ea,d is calculated to be 134.1 k0.3 kJ mol-I and Ea,e to be 130.4k0.8kJ mol-'. Clearly, the differences in growth rate between 580 and 650'C in the presence of NF, cannot be due to differences in the energies of activation of the deposition and etching processes. The second term on the right-hand side of eqn. (3) and (4) would therefore appear to be more deposition ternperature/OC 680 640 600 560 61 II I I 31.,., 1.04 1.08 1.12 1.16 1.20 10%~ Fig.4 Arrhenius-type plots for polysilicon total overall growth with NF, (jT),for growth without NF, (j,) and for 'etching'(j,): A, without NF, (j,): @, with NF, (1 sccm) (jT);W,je=jd-jT dominant in this context, and, since pre-exponential factors (A)are usually only weakly temperature dependent, the depen- dence of concentrations of active species on temperaturc may have a significant influence on growth and etching. We therefore now briefly examine the thermodynamics of formation of possible etchant species. Gibbs energies of forma- tion of such species can be calculated using data from JANAF thermochemical tables.22 For NF, gas-phase dissociation we assume that there are three possible reactions which can occur NF, $ NF,+F (1) NF, $ NF+F, (11) 2NF, S-N,+3F2 (1111 The Gibbs energy changes associated with each of these reactions as a function of temperature are shown in Fig.5 from which it can be seen that only reaction (111) is energeti- cally feasible. In fact, under typical deposition conditiclns of 560-67O'C and a total pressure of 65 Pa the extent cf this reaction is effectively loo%, therefore F, but not F' could be an etchant species. If F' were able to be formed then it would probably be more effective at etching polysilicon than F, since the activation energies for the production of fluorinated silicon species (e.g. SiF, and SiF,) from etching with F' lie in the range 8.7-14.5 kJ molp','' while for the formation of the predominant species SiF, from etching with F, the actiLation energy is in the range 38.2-66.5 kJ mol-';*,9 this is not unexpected since the reaction of F, requires breaking of an F-F bond.Of particular interest for the results presented here is that an activation energy in the range of 38.2-66.5 kJ mol-' for F, etching is significantly lower than the activation energy of 130 kJ mol-' determined from Fig. 4 for apparent etching. Therefore, if F, does etch the polysilicon as it is deposited then this process probably does not incorpor- ate the rate-determining step. Another possible etchant species to be considered is HF which could be generated by reaction between NF, and H,. In our system the H, would come from the decomposition of the SiH, when silicon is deposited.Possible reactions are 2NF,+H2 + 2HF+4F+N2 (IV) 2NF3+3H, $ 6HF +N2 (V) 2NF, +H, + 2HF +2NF2 (VI) 2NF, +H, +-2HF +2F2+N2 (VII) NF,+H2 $2HF+NF (VTTT) 0 -2001 ' I 800 850 900 950 1000 TIK Fig.5 Calculated dependence of Gibbs energies of reaction (hG) on thermodynamic temperature for reactions (I)-( 111): 2,reacti,m (1); @, reaction (11); M, reaction (111) The Gibbs energy changes for all of these reactions are large and negative under typical deposition conditions. For example, at ca. 600°Cand a total pressure of 65 Pa the AG values range from ca. -418 kJ mol-' for reaction (VI) to ca. -1750kJ mol-' for reaction (V). Therefore, energetically, HF could be readily formed in the gas phase and made available as an etchant for Si.Possible etching reactions by HF of silicon can be represented by the general equation Si + nHF +SiF, + n/2H2 (IX) where 1<n<4. If we assume that NF, in the reaction is completely converted to HF by reaction with H2 then the maximum partial pressure of HF, corresponding to 5 cm3 min-' of NF,, will be 10 Pa and the Gibbs energy changes for all cases of reaction (IX), except when n= 1, are negative and so etching by HF might occur. However, to our knowl- edge, no thermal etching of Si by HF at the temperature we have used has been reported. Indeed, anhydrous gaseous HF can be used to etch SiO, on Si without attack on the underlying material.23 There is therefore probably a kinetic hindrance to silicon etching.So if etching is occurring, leading to the effective fall in growth rate, then it would seem to have to involve molecular fluorine. The situation could be considerably complicated, though, by the fact that thermal dissociation of NF, may occur not only homogeneously but also heterogeneously with incorporation of nitrogen and fluorine at the surface.', This could then lead to growth rate inhibition analogous to that suggested in the case of polysilicon growth in the presence of PH, for in situ n-type d~ping,'.'~,~~ and for SIPOS growth in the presence of N20.25,26 In the case of PH, because of its lone pair of electrons one might expect an interaction with dangling bonds at a silicon surface, and evidence for the strong dissociative adsorption of PH, has been obtained by However,Meyerson and co-w~rkers.~~~~~ the strength of the Si-P bond (364kJ mol-')29 and its electronegativity difference (0.3)29are both significantly lower than the corre- sponding values for an Si-N bond (439kJ mol-' and 1.1, respectively) and, especially, for an Si-F bond (553kJ mol-' and 2.1).Also, all of the values are greater than those for an Si-Si bond (226kJ mol-') and an Si-H bond (299 kJ mol-' and 0.2).Therefore, NF, would be expected to be readily adsorbed on a silicon surface and to prevent the adsorption of the growth species SiH,. This adsorption could be dissoci- ative', and could involve, as indicated above, bonding with either N or F. The overall reaction scheme could then be represented by the following sequence of reactions: g SiH,(a) eSiH2(a) Si,H,(a) 3-x NF,(a)+T F,(g) (XV) SiH,(a)-+Si(s)+ 2H,(a) (XVII) Si2H,(a)+2Si(s) + 3H,(a) (XVII) SiH,(a)+Si(s) + H2(a) (XTX J.MATER. CHEM., 1994, VOL. 4 The equilibrium of reaction (X) lies well o~er to the left.30331 The reaction between silane and silylene h,ts been shown to be extremely and therefore the equilibrium lies well to the right. Tn the absence of any surface inhibition process, though, the primary route for polysilicon growth will be uia SiH, adsorption and diss~ciation,~~ although there is probably some growth uia SiH, and Si2H, as well. Silane itself is not expected to be strongly adsorbed and there is good evidence that this is the case.27,35,36 Silylene, on the other hand, as a biradical would be expected to interact re,tdily with silicon dangling bonds and to be strongly adsorbed.37 Disilane is certainly not as strongly adsorbed as silylene, but is consider-ably more strongly adsorbed than ~ilane.~"~, The argument for strong NF, adsorption has been given above.Hydrogen is known to inhibit polysilicon Since in our system no hydrogen gas is introduced into the reactor as a carrier gas and the only hydrogen present will be that from SiH, dissociation [reactions (X), (XVI1)-(XIX)], we assume this is readily desorbed and the equilibrium of reaction (XVI) will lie well to the left-hand side. The total growth rate (j,) of polysilicon can be represented as the sum of the growth rates from all silicon species adsorbed and undergoing heterogeneous decomposition where ki' is the heterogeneous rate constant for decomposition of adsorbed species SiHi, n is the total number of surface sites and OSiH, is the fractional coverage of any species SiHi.Following the same algebraic arguments given previously24 it can be shown that the inverse of the observed growth rate ( 1/',) is related to the mole ratio (1)) of NF, to SiH, by where A and B contain expressions involving kinetic and thermodynamic constants for the set of reactions (X)-(XIX) and n, is the gas-phase concentration of SiH4. Thus a plot of l/', us. y should be linear with an intercept corresponding to the deposition of polysilicon in the absence of NF,.Fig. 6 shows such a plot with data taken from Fig. 2 and the predicted relationship is found to hold. Note that in the case of PH, adsorption a similar plot was only linear up to y !z 3 x lop3but for greater values of y than this there was a deviation below linearity. This was attributed" to the frac- tional coverage of PH, reaching a limiting value of less than unity. Hence the inhibition effect would fall away and the 0.00 0.00 0.04 0.08 0.12 Y Fig.6 Test of eqn. (6), where 7 is the mole ratio of NF, to SiH,. Data points taken from Fig. 2 J. MATER CHEM., 1994. VOL. 4 growth rate would tend to level off. The results with NF, do not show any such deviation and this would be consistent with the greater adsorption tendencies of NF, mentioned above. Another interesting and significant difference between the results obtained with PH, and NF, arises from the fact that in the former case there is a marked variation of growth rate with radial distance on a wafer, but with NF, the layers are uniform across a wafer.For example, for polysilicon growth in the presence of PH, with y=4 x lo-, the growth rate at the edge of a wafer was found to be about 50% higher than in the centre,,, but with all of the mole fractions of NF, used in this study the deposited layers were uniform to within a few per cent across a wafer. In the case of PH, the radial variation of growth rate has been attributed to the blocking of kinetically controlled growth through SiH, adsorption and decomposition [reactions (XII) and (XVII)] and the trans- port-limited growth through SiH, and Si,H, adsorption and decomposition [reactions (XIIT), (XIV), (XVIII) and (XIX)].This mechanism seems not to be operating in the case of NF, and this could be because of the stronger affinity between Si and F. Any very reactive silylene or disilane formed in the gas phase by reactions (X) and (XI) could rapidly be removed by NF, to form Si-F containing species which would be very stable and would not be potential growth precursors. In this case growth of polysilicon would continue to be from 1825 are adsorbed more strongly on a silicon surface, may show a lower overall activation energy for silicon growth. In any case, the effect on E, is relatively small but with NF, there is a significant change in E, from 149 to 97 kJ mol-’.At a low mole fraction of silane a higher activation energy than in the absence of NF, is not unexpected if the NF, is inhibiting growth and adsorption of silane becomes the rate-determining step. As the mole fractron of silane increases then again the equilibrium of reaction (X) will be pushed over to the right, but now there is compctition between a homogeneous reaction with NF, and a hetero-geneous reaction with Si for the SiH, and Si,H, formal. The reactions with NF, can be written as SiH, $ SiH,+H, (XI SiH, +NF, -+A SiH, +SiH, +Si,H, (XI) Si,H, +NF, +B I XXT) where A and B represent NF,-Si complexes. The r..tte of formation of these NF,-Si complexes can be readily shown to be given by d([A] + [B])/dt zKO.’ [SiH,] (ko[NF,] +k, [SiH,] 1 (7) where K is the equilibrium constant for reaction (X), A.and k,SiH, alone and this process is always under kinetic ~ontrol~~.~’ and no radial variation of growth rate would be expected. If this is the case, the NF, is acting as a scavenger, rather like the wafer cages used in the deposition of P-doped polysilicon and semi-insulating polysilicon (SIPOS) to reduce the radial variation of growth rate. This could be an interesting and useful alternative approach for obtaining uniform layers in these systems. Some support for the concept of homogeneous reaction between NF, and Si,H, or SiH, is given by results for the apparent activation energies determined in the absence and presence of NF, as a function of silane mole fraction (Fig.7). Without NF, the apparent activation energy varies from a maximum value of 134 kJ mo1-l to a minimum value of 124 kJ mol-’. This is understandable if one remembers that there are alternative routes for polysilicon deposition and that as the mole fraction of silane increases so equilibria (X) and (XI) will be pushed towards the formation of SiH, and Si,H,, respectively, and the decomposition of these products, which \\\ 0.2 0.4 0.6 0.8 1.0 1.2 XS Fig. 7 Apparent activation energies as a function of silane mole fraction: m, without NF,; a, with NF, (1 sccm); P,=65 Pa; total gas flow, 100 sccm are the rate constants for reactions (XX) and (XXI), respectively, and it has been assumed that Si,H, is a reactive intermediate.Thus, for a fixed mole fraction of NF,, increasing the mole fraction of SiH, will increase its rate of disappearance through the reaction of SiH, and Si,H, with NF,, mak,ing it less available for deposition. As the temperature increascs this rate of consumption in the reactor before the test wafer will be enhanced so that there will be proportionally less SiH, for deposition at the test wafer at higher temperatures than at lower temperatures. Now it should be remembered that plots of the form shown in Fig. 4 are not true Arrhenius plots since the ordinate is not In k but rather In j,; i.e. the logarithm of the reaction rate rather than of the rate constant.The reaction rate in addition to reflecting the variation of the rate con- stant with temperature also contains the concentration of reac-tant, and if this concentration is less at higher tem-peratures than at lower temperatures then this will lead to an apparent lower activation energy than would have been obtained if the reactant concentration had remained constant over the whole temperature range. Conclusions The rationale behind this study of the effect of NF, on polysilicon growth by LPCVD was to try and introduce an etchant into the reactant gas mixture to reduce the size of deposited crystallites and to obtain films which show good crystalline quality on annealing. The results do indeed show an effective etching effect because of a decreased growth rate on addition of NF,, and thermodynamic consideration:, have suggested that etching could occur by molecular fluorine. However, it has also been suggested that the fall in growth rate with NF, addition could be enhanced because of arlsorp- tion of this species on the silicon surface leading to a blocking of adsorption by the growth precursor, SiH,.There is also evidence that NF, consumes the highly reactive molecule Si2H6 and the biradical SiH, in the gas phase, which means that, unlike other inhibited silicon growth processes, such as phosphorus-doped polysilicon and SIPOS growth, then is no radial non-uniformity of layer thickness on a wafer. The fact, though, that NF, reduces the polysilicon growth rate by adsorption and blocking of surface sites instead of orily by 1826 J.MATER. CHEM., 1994, VOL. 4 etching, as originally expected, does not necessarily mean that it will not assist in producing better quality layers of polysil-icon when they are subsequently annealed. It has been shown,2 for example, that phosphorus addition, which reduces the polysilicon growth rate in the same manner that NF, does, can lead to the production of higher-quality films. In the next paper in this series we shall report on the effect of NF, on the crystalline quality of polysilicon. 13 14 15 16 17 18 D. E. Ibbotson, J. A. Mucha, D. L. Flamm and J. M. Cook, J. Appl. Phys., 1984,56,2939. M. L. Hitchman and H. C. Shi, to be published. T. R. Torkelson, F. Oyen, S. E.Sadek and V K. Rowe, Toxicol. Appl. Pharmacol., 1962,4, 770. E. H. Vernot, C. C. Haun, J. D. McEwen and G. F. Egan, Toxicol. Appl. Pharmucol., 1973,26, 1. M. L. Hitchman, J. F. Zhao and S. H. Shamlinn, J. Muter. Chem., 1994,4, 1827. M. L. Hitchman, J. F. Zhao and S. H. Shamlian, J. Muter. Chern., 1994, 4, 1835. We gratefully acknowledge the award of an SERC grant for equipment, and of financial support from Taiyuan University of Technology, Air Products plc and Epichem Ltd for J.F.Z. and from the University of Strathclyde for J.F.Z. and S.H.S. 19 20 K. F. Jensen, M. L. Hitchman and W. Ahmed, in Proc. 5th Eur. Con$ CVD, ed. J. 0. Carlsson and J. Lindstrom, University of Uppsala, Uppsala, 1985, p. 144. M. L. Hitchman, J. Kane and A. E. Widmer, Thin Solid Films, 1979,59,231. Air Products plc and Epichem Ltd.are also thanked for the provision of gases for this research. 21 22 A. Clark, The Theory of Adsorption and ('atu Press, New York, 1970. JANAF Thermochemical Tables, 3rd edri. J. Phys. Chem. Rej. Data, 1985, 14, suppl. 1. References 23 24 R. Cleavelin and G. Duranko, Semiconduct. In[., 1987,10( 12), 94. M. L. Hitchman, W. Ahmed, S. Shamlian and M. Trainor. 1 2 3 4 5 6 7 8 9 10 11 M. L. Hitchman and K. F. Jensen, in Chenzicul Vupor Deposition-Principles and Applicutions, ed. M. L. Hitchman and K. F. Jensen, Academic Press, London, 1993, p. 159. G. Harbeke, L. Krausbauer, E. F. Steigmeier, A. E. Widmer, H. F. Kappert and G. Neugebauer, RCA Rev., 1983,44,187. P. Joubert, B. Loisel, Y. Chouan and L.Haji, J. Electrochem. Soc., 1987,134,2541. E. Kinsbron, M. Sternheim and R. Knoell, Appl. Phys. Lett., 1983, 42, 835. M. L. Hitchman, C. W. Jones, J. F. Zhao and S. H. Shamlian, Adv. Muter. Opt. Electron., 1993,2, 123. A. J. Woytek, J. T. Lileck and J. A. Barkanii, Solid State Technol., 1984,27( 3). 172. K. M. Eisele, in Proc. Syrnp. Plasma Etching, Electrochemical Society, Pennington, NJ, 1981, p. 174. J. A. Mucha, V. M. Donnelly, D. L. Flamm and L. M. Webb, J. Phys. Chem., 1981,85, 3529. M. Chen, V. J. Minkiewicz and K. Lee, J. Electrochem. Soc., 1979, 126,1946. M. J. Vasile and F. A. Stevie, J. Appl. Phys., 1982,53, 3799. J. A. Barkanic, D. A. Bohling and D. M. Reynolds, A Review and Safety Considerations of Dry Etching Using Nitrogen TriJEuoride, Air Products and Chemicals Inc., Allentown, PA, 1986. 25 26 27 28 29 30 31 32 33 34 35 36 37 38 Chemtronics, 1987,2, 147. M. L. Hitchman and J. Kane, J. Crystal Grourh, 1981, 55,485. M. L. Hitchman, Vacuum, 1984,34,377. B. S. Meyerson and W. Olbricht, J. Electroclrem. Soc., 1984, 131, 2361. B. S. Meyerson and M. L. Yu, J. Electrochm. Soc., 1984, 131, 2366. D. R. Lide, Handbook of Chemistry and Ph! sics, 71st ed. CRC Press, Boca Raton, FL, 1990-1991, p9-86: J. H. Purnell and R. Walsh, Proc. R. Soc. London, A, 1966, 293, 543. C. G. Newman, H. E. O'Neal, M. A. Ring, F. Leska and N. Shipley, Znt. J. Chem. Kinet., 1979,21, 1167. P. John and J. H. Purnell, J. Chem. Soc., 1973,69, 1455. G. Inoue and M. Suzuki, Chem. Phys. Lett., 1985, 122,361. J. M. Jasinski and J. 0.Chu, J. Chem. Phys., 1988,88, 1678. S. M. Gates, D. B. Beach, R. Inbihl, B. A. Scott and J. E. Denmuth, J. Vac. Sci. Technol., 1987, 5, 628. R. J. Buss, P. Ho, W. G. Breiland and M. E. Coltrin, J. Appl. Phys., 1988,63, 2808. J. G. E. Gardiniers, L. J. Giling, F. de Jong and J. P. van der Eerden, J. Crystal Growth, 1990, 104, 727. K. F. Roenigk and K. F. Jensen, J. Electrocht-Bm. So(,., 1985, 132, 448. 12 J. A. Shorter, J. G. Langan and J. I. Steinfeld, Surf Sci., 1989, 219, L560. Paper 4/02718C; Receired 9th Muy, 1994