J. MATER. CHEM., 1994, 4( 12), 1827-1834 Studies of the Effects of NF, on the Growth of Polysilicon Films by Low-pressure CVD Part 2.-Effect on Crystallinity Michael L. Hitchman,* Junfu Zhaot and Sarkis H. Shamlian Department of Pure and Applied Chemistry, University of Strathclyde, 295 Cathedral Street, Glasgow, UK GI IXL In this paper the effects of silane mole fraction, deposition temperature and the addition of NF, on the crystallinity of as-deposited polysilicon films and on the crystalline quality of annealed films have been investigated with the aid of Raman spectroscopy and scanning electron microscopy (SEM). It has been found that without the addition of NF, it is necessary not only to have a low deposition temperature, as previously suggested in the literature, but also to have a high silane mole fraction in order to obtain amorphous films from which high-quality material can be obtained upon annealing.With the addition of NF, it is has been found that amorphous layers may be grown, and hence good quality annealed films obtained, at much higher deposition temperatures than is normally possible. As has been discussed in the Introduction to the previous paper,' polycrystalline silicon is widely used in VLSI fabri- cation. In all of the applications it is necessary to control the crystallinity of the layer carefully, since poorly crystallised material can give rise to high internal stress which can produce undesirable defects in active device areas. Factors affecting the crystalline structure of polysilicon include2 deposition temperature,, silane partial pressure,, deposition time' and dopant type.6 From the point of view of layer and device quality it has been found3 that partially or fully crystalline as-deposited polysilicon layers when annealed show significant lattice disturbances whereas post-deposition annealing of as- deposited amorphous layers gives material with the desirable features of good structural perfection and low strain.The most successful strategy3 for achieving these desirable proper- ties has been to grow at a low temperature where the deposition rate (j,) is greater than the crystalline growth rate ( and largely amorphous material is obtained. High .jcg)235 silane partial pressure4 and short deposition times5 can also favour the deposition of amorphous layers, but less attention has been paid to these factors.We have investigated an alternative approach of trying to etch polysilicon partially as it deposits with the aim of reducing crystallite size, leading to an effective decrease in crystalline growth rate, and of removing preferentially high- energy surfaces sites which might be associated with crystalline defects. As reported in the previous paper,' the gas we chose to explore this concept, NF,, did lead to a net decrease in growth rate and, hence, an apparent etching effect. However, while a consideration of the thermodynamics of the decompo- sition of nitrogen trifluoride did suggest that the observed fall in growth rate with the addition of this reactant could be due to in situ etching of the deposited layer by molecular fluorine, the results also suggested that adsorption of the NF, and the blocking of adsorption by the growth precursor SiH4 could occur. However, the fact that NF, reduces the polysilicon growth rate by inhibiting growth instead of just by etching, as originally expected, does not necessarily mean that it will not help in the production of high-quality crystalline layers on annealing.The n-type dopant PH, also reduces the polysil- icon growth rate through an adsorption rnechani~m,~ but at the same time it produces better quality crystalline material on annealing than undoped polysilicon grown under identical -tPresent address: Department of Applied Chemistry, Taiyuan University of Technology, Taiyuan, Shanxi 030024, P.R.China. conditiom2 Therefore, we have investigated the effect of NF, on the crystalline quality of polysilicon and report on the results in this paper. Experimental The LPCVD reactor, the gas-handling system and typical deposition procedures and conditions have been described in detail previo~sly.~,~ The particular conditions for this study have been given in the previous paper.' Briefly, depositions were made with 100% SiH, and with mixtures of SiH4-He, SiH,-NF, and SiH,-He-NF,. The total gas flow was always 100sccmS with silane flows in the range 5-100sccm, NF, flows in the range 1-5 sccm and He flows adjusted to give the balance of 100 sccm.Films were deposited on silicorx(111) substrates with the wafers held in a fused silica boat p.iralle1 to each other and perpendicular to the gas flows. Annealing of samples was carried out in a double-v alled, fused silica tube. The annealing gas, nitrogen, was Ibassed between the two walls before being allowed to the inner tube in order to preheat it. Standard loading and unloading procedures were used with the time taken for these steps being ca. 30min. Annealing was carried out at 950'C and at atmospheric pressure for 2-3 h; there were no observable changes in the crystallinity of the layers after annealing ft )r 2 h. The degrees of crystallinity of as-deposited layers arid the crystalline quality of annealed layers were examined by R aman ~cattering.~ The experimental arrangement for R aman measurements of the films has been described in detail The technique employed an argon laser with a wavelength of 514.5 nm and 500 mW power.The slit width used was either 5 or lOcm-', depending on the sample, and the spectrum was scanned from 400 to 600cm-'. The sample was held in a holder at the optimum angle of 35" for maximum Raman intensity," and the laser light was focused onto a 1 mm diameter spot on the sample. The Raman-scattered light was collected at the standard angle of 90" to the incident light. The spectra were either recorded directly as a trace on a plotter or from a computer printout. The spectra of anilealed samples were compared with that obtained from single-crystal silicon.The surface morphology of as-deposited and annealed layers was also examined by SEM. The instrument used was a JEOL $: Standard cm3 min-'. JSM-840A with a magnification of x30000 and a beam energy of 15 or 25 keV. Results and Discussion Effect ,of Mane Mole Fraction on Crystallinity Fig. 1 shows the effect of silane mole fraction (x,) on the Raman spectra of as-deposited layers of silicon grown from silane-helium mixtures at a temperature of 580°C. The fre- quency and lineshape of the Raman spectrum of silicon films reflect the degree of crystallinity., In crystalline silicon the line due to lattice vibrations occurs at 522cm-', but for amorphous silicon the lack of lattice periodicity allows scat- tering from all vibrational modes and this produces a broad, less intense line centred at 483 cm-'.For films grown under conditions where there is a transition between a low degree of crsytallinity (i.e. mainly amorphous) and full crystallinity (ie. fully polycrystalline) a superposition of the two types of spectra will be obtained. In Fig. 1 this range of spectra can be seen. At a high mole fraction of silane (x,=O.5) a broad, very low intensity peak at ca. 481 cm-' is observed, corre- sponding to amorphous material. Harbeke et al., reported that for Td <580 "C amorphous silicon is grown from pure SiH,. However, they did not examine the effect of the partial pressure of SiH,. In their deposition system they had a total pressure (P,) of 65 Pa (0.5 Torr) and no other gas was added.So their silane partial pressure was also ca. 65 Pa (cf. our value of ca. 33 Pa). From the results of Joubert et a/., one would expect for both these partial pressures and for &= 580'C the as-deposited films to be largely amorphous, as indeed is found. At a partial pressure of ca. 3 Pa, corresponding approximately to x,=0.05, Joubert's results would indicate a fully crystalline layer. This is seen to be the case here since the spectrum in Fig. 1 for x,=O.O5 is a sharp peak at ca. 522 cm-', although there is a slight shoulder on the low- wavenumber side of the peak which probably indicates that the film is not fully crystalline. The SEM pictures also clearly showed the transition from crystalline to amorphous layers as the mole fraction of silane was increased. Fig.2 gives I.,.,.,.,. T13000 0.3 0.4 0.5 300 400 500 600 waven urnbedcm-' Fig. 1 Effect of silane mole fraction (x,) on the Raman spectra of as- deposited silicon films: gas mixture, SX-FIe; & =580 "C; P, =65 Pa J. MATER. CHEM., 1994, VOL. 4 -1pm Fig. 2 SEMs for as-deposited films: ((7) x,=O.O5, (h) s,=O.5; Td= 580 "C; P, =65 Pa representative SEMs for layers grown from low (x, =0.05) and high (x,=0.5) silane mole fractions, and the changeover from a surface with many small crystallites to one which, apart from impurity particles, is smooth is apparent. The work of Joubert et suggests a sharp transition between amorphous and crystalline material, but our results (Fig.1) show that there is more likely to be a gradual transition between the two states. One factor which compli- cates comparisons of degrees of crystallinity of polysilicon layers grown by different groups of workers is the effect of deposition time. This has been discussed elsewhere,2." but essentially if long growth runs are used then unintentional annealing with partial crystalline growth, during deposition. can occur leading to apparent differences in crystallinity. These differences can be exacerbated by leaving deposited films in the reactor for varying lengths of time after deposition. Unfortunately, while it is possible to minimise the effect of unintentional post-deposition annealing by removing the samples from the reactor immediately after the end of a growth run, the actual deposition time often needs to be varied to obtain layers of a desired thickness, as growth rates change with deposition parameters.In our studies deposition times varied between 60 and 240min without NF, and between 60 and 360min with NF,, and no samples were allowed to remain in the reactor after the end of growth runs. The slow transition from amorphous through mixed crystal- line to almost fully crystalline material is what one might expect for a solid-state process, and is comparable to the gradual transition found by Harbeke et aL3 and ourselves' as the deposition temperature has been varied. The reason why the changeover with silane partial pressure occurs is, of course, that at high values of x, the rate of arrival of SiH4 at the surface, and its subsequent adsorption and decomposition by approximately first-order kinetics' will be high and there will be little time for surface diffusion and ordering, especially at low temperatures.At low x,values, on the other hand, surface species have more time to find crystalline growth sites before the arrival of further adsorbed species. The presence of J. MATER. CHEM.. 1994, VOL. 4 1829 adsorbed hydrogen from the decomposition of SiH, will also hinder surface migration and this effect will be greater the higher the initial partial pressure of silane. Harbeke et aL3 examined with Raman scattering the effect of deposition temperature on the crystalline perfection of silicon films after annealing.Joubert et d4did not report on crystalline quality for films grown with different silane partial pressures. Fig. 3 shows typical examples of Raman spectra obtained after annealing a mainly amorphous (x, =0.4), as-deposited film (Fig. 3A), and after annealing a partially crystal- line (x, =0.3), as-deposited film (Fig. 3B). The complete con- version in both cases to crystalline material after annealing is apparent. Fig. 4 summarizes values of peak intensity as a function of x, for as-deposited and annealed films. For low values of x, (0.05 to 0.2) where films grow as practically fully crystalline layers, there is, not unexpectedly, a large change in Raman intensity, and hence crystallinity, on annealing.For a mixed crystalline film (x,=O.3) there is a slight increase in line intensity on annealing [cf. Fig. 3B] corresponding to a greater degree of crystallinity. For the amorphous, as- deposited films (x,=O.4 to 0.6) the line intensity and the extent of crystallinity have increased significantly; SEM pic-tures essentially confirmed these observations. The maxima in Fig. 4 at M, ~0.3may correspond to changes in the texture of the crystalline grains with silane mole A measure of the crystalline perfection of a film which has been annealed is the Raman line~hape.~ Annealed materials of low distortion and internal strain will give a lineshape resembling that of bulk single-crystal silicon, while strong and extended tails of the Raman line indicate highly distorted or strained material which is undesirable for critical device applications. Fig.5 shows the Raman full linewidth at 1/10 height as a function of silane mole fraction as well as the 8000 A 6000 4000 yv) 2000 u)c) cr3-8 Y.-v) 30300 400 500 600 g 10000 .-C a5 8000U 6000 4000 2000 300 400 500 600 wavenum ber/cm-' Fig. 3 Comparison of Raman spectra for (a)as-deposited and (6)anne-aled films: A, x, =0.4; B, x, =0.3; Td= 580 "C;P, =65 Pa; T,=950 "C 10000I -8000 7 I v) v)YC0' 6000-<x c.-v)C +C.z 4000-a I5CT \ 2000 -Ll1 OJ 0.0 0.2 0.4 0.6 XS Fig. 4 Raman peak intensity as a function of silane mole fracti In for as-deposited (A)and annealed (0)films.Deposition and annealing conditions as for Fig. 3 ............................................................................. singlecrystal silicon 0.0 0.2 0.4 0.6 XS Fig.5 Raman linewidth as a function of silane mole fracticin for annealed films. Deposition and annealing conditions as for Fig 3. corresponding linewidth for single-crystal silicon. The differ- ence between the high silane mole fraction group, with I alues close to the bulk value, and the low mole fraction group is notable. This indicates that whilst there is little disturhance of the crystal lattice for the annealed high x, group, fcllr the low x, group the material after annealing remains in a highly disturbed state with some poor and some good crystal1is;ition. Clearly for good device quality polysilicon a high partial pressure of silane should be used for deposition.Effect of Deposition Temperature on Crystallinity .~As has already been mentioned, Harbeke et ~1 have shown that deposition temperature is also an important parameter which influences film growth and crystalline structure. Fig 6(a) and (b) show Raman spectra for films grown froni gas compositions with x,=0.5 and 1, respectively, at diflerent deposition temperatures. Because of the high values of .Y,, at Td < 580 "C amorphous material is obtained in both cases (c$ Fig. 1). As Td increases, more crystalline material is grown; J. MATER. CHEM., 1994, VOL. 4 1’l’”l’” (a1 the representative SEM pictures in Fig.7 show this effect as well. This is consistent with results obtained by Harbeke et ~1.~ 3000 counts s-’ After annealing, all of the films were shown to be fully crystalline from the Raman spectra. Fig. 8 \bows the Raman Td 1°C full linewidth at 1/10 height, and again amorphous. as-deposited material shows in both cases little strain with linewidths being close to that of single-crystal silicon. For higher-temperature polycrystalline as-deposited materials there are considerable distortions and strains in the annealed layers, particularly for those grown from [he diluted silane. The SEM pictures support this (Fig. 9)-where the very poor crystalline quality for an as-deposited film grown at 670 C is 620 in strong contrast to the rather better quality material grown at, for example, 600°C.The results with pure silane are very similar to those obtained by Harbeke et 01,~ who also used undiluted silane. The better results for the annealed material C 580,-grown at 600°C with pure silane than those for the corre- sponding material produced using diluted silane may be Q,c.E 300 400 500 600 C a associated with the higher amount of adyorbed hydrogen, resulting from the higher partial pressure of silane, which inhibits the surface migration of silicon kpecies leading to fewer opportunities for crystallisation. This could also be the reason why at higher temperatures (T,>620 C) the pure silane still gives less strain and distortion than diluted silane.The reason why a deposition temperature in the region of 500-600 ’C appears to be critical in terms of quality has been discussed el~ewhere.~.~ The conclusion thus far is that in order to obtain good quality crystalline polysilicon for device applications it is necessary to grow amorphous material for subsequent annealing. Harbeke et d3have previously claimed that a deposition temperature of T,<580 C is required to achieve this. The results presented in this section confirm this, but in addition we have shown that it is necessary to have a high partial pressure of silane as well. We now examine the effect on crystallinity of adding NF, to the reactant gases. [r (b: 5000 counts s-’ T,I’C- , .600 560 300 400 500 600 wavenumber/cm-’ Fig. 6 Effect of deposition temperature (Td)on the Raman spectra of as-deposited silicon films: (a)x, =0.5 (SiH,-He), (h)x,= 1 (pure SiH,) -1pm Fig.7 SEMs for as-deposited films showing the effect of deposition temperature on crystalline growth: (a) 600 C, (h) 620 C. (c) 650 C, (d) 670 C; s,=0.5; P, =65 Pa J. MATER. CHEM., 1994, VOL. 4 for growth at 670°C. These observations are consistent with .-I .-0) A= 6ol zt singlecrystal silicon 20I I I I I 1 540 580 620 660 T,j 1°C Fig. 8 Ranian linewidth as a function of deposition temperature (T,) for annealed films: A,x, =0.5 (SiH,-He); ,x, = 1 (pure SiH,) Effect of NF, on Crystallinity Fig. 10(a) and (b)show Raman spectra for films grown from gas compositions with x,=O.5 and 1, respectively, at different deposition temperatures but in each case with NF, (1 sccm) added to the reactant gas mixture.From these spectra it can be seen that the NF, has a marked effect on the crystallinity of the as-deposited films. Whereas in the absence of NF, amorphous films could be grown only at temperatures less than ca. 580 "C with x, 30.5 (cf. Fig. 6), here with a mole fraction of NF, (xN)of only 0.01, amorphous films can be obtained at temperatures as high as 640 "C.The SEM pictures in Fig. 11 show the changeover from a microcrystalline surface obtained at 620 "C to a distinctively polycrystalline surface the model of NF, etching and adsorption presented in the previous paper' since a gas which both etches and is strongly adsorbed would reduce the possibilities for crystalli.;ation, although at high temperatures the adsorption effect would be offset by weakening of the bond between NF, and the surface.The significant effect of NF, is shown in a different manner in Fig. 12, where increasing the mole fraction of NF, allows amorphous films to be obtained at 650'C. Although the signals are very noisy, there is an indication that they are becoming broader and shallower as xN increases, suggesting that the extent of amorphous growth increases with addition of NF,. All samples grown in the presence of NF, were annealed under exactly the same conditions as those grown uithout NF,. Fig. 13 shows Raman spectra obtained after annealing the samples, whose spectra for the as-deposited state are shown in Fig.12. In all cases the material has undcrgone crystallisation, but there are still strong indications of mixed crystallinity in practically all cases, with the shoulders on the low-wavenumber side of the peaks suggesting strain and distortion in the layers. This is confirmed by Fig. 14 and 15 which show Raman full linewidths at 1/10 height as a function of % for x,=O.S and 0.1 (Fig. 14) and as a function Ctf NF, mole fraction (Fig. 15). With an NF, mole fraction of xN= 0.01 the best results are obtained with diluted silane at high deposition temperatures (7'' 3 650 "C). This is also shown in Fig. 15 for results obtained at %=650'C where the best quality layer is found for xN=0.01.From both plots, though, the deviation from the single-crystal linewidth is ca. 15 cm-' which is higher than the ca. 10 cm-' determined from Fig. 8 and the value of ca. 3 cm-' from Fig. 5, both cases uithout NF, addition. Thus although the addition of NF, to the reactant gas mixture can give rise to amorphous filn~s and subsequent annealing of these films can, in accord with the corresponding effects of silane partial pressure and depc sition Fig. 9 SEMs for annealed films showing the effect of deposition temperature on crystalline growth: (a) 600 "C,(b)620 "C, (c) 650 "C, (d) 670 "C; x, =0.5;PT=65 Pa J. MATER. CHI,M., 1994, VOL. 4 2000 counts s-’ 620 I 600 ;c”””””””””” I 400 500 600 1 Td 1°C 640 620 600 580 560 300 400 500 600 wavenumber/cm-’ Fig.10 Effect of deposition temperature (G)on the Raman spectra of as-deposited silicon films grown in the presence of NF, (1 sccm): (a)x,=O.5 (SiH,-He), (b)x,= 1 (pure SiH,) temperature, lead to good quality crystalline layers, there is slightly more strain in these layers than in those grown at low temperature and high silane mole fraction. This difference could arise from incorporation of small amounts of N and F into the layers as a result of the strong adsorption of NF,. In the next paper we show that this is, in fact, the case and that although high-temperature annealing causes most of the flu- orine to diffuse out from the layer, some nitrogen is retained as an impurity. Recry stallisation will not occur readily in impurity regions and consequently poorer quality layers will be obtained on annealing.This effect is expected to be more marked the higher the NF, mole fraction and the lower the deposition temperature, since both these parameters will give higher NF, adsorption. This is indeed observed (see Fig. 14 and 15). Conclusions It has been demonstrated that an additive gas can be used to affect the degree of crystallinity of LPCVD polysilicon films, although not only for the reasons originally expected. The -1pm Fig. 11 SEMs for annealed films showing the etfect of deposition temperature on crystalline growth: (a) 620 C, (h) 650 C. (c) 670 C: x,=0.5; P, =65 Pa use of NF, can produce amorphous films from which low- strain, annealed good quality crystalline polysilicon can be obtained upon annealing.However, amorphous growth results not only because of in situ etching but also because of strong adsorption of NF,, which inhibits the surface diffusion required for crystalline growth on the growth surface. The best quality annealed films produced with NF, are obtained at low NF, mole fraction (x, GO.01) and high deposition temperature (G3650 “C),which is of interest if it is necessary to deposit polysilicon at temperatures greater than 650 C in order to be compatible with other stages of device production. In the absence of NF,, the work has shown that it is necessary not only to use low deposition temperatures to achieve amorphous films from which good quality crystalline material can be obtained upon annealing, as recommended by Harbeke et but also to have a high silane mole fraction.Fig. 16 summarises the results obtained for the degree of crystallinity of as-deposited films as a function of x, and T, both in the absence and presence of NF, (xN=0.01). We gratefully acknowledge the award of an SERC grant for equipment, and 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. Air Products plc and Epichem Ltd. are also thanked for the J. MATER. CHEM., 1994, VOL. 4 T II 1000 counts s-1 I E 0.047 0.041 0.034 0.028 0.022 0.016 300 400 500 600 wavenumber/cm-' Fig.12 Effect of NF, mole fraction (xN)on the Raman spectra of as- deposited silicon films: x,=0.5; & =650 T;P, =65 Pa 2000 counts s-' I 300 400 500 600 waven um berlcm-' Fig. 13 Effect of NF, mole fraction (xN)on the Raman spectra of annealed silicon films. Deposition conditions as for Fig. 12; T,= 950 "C. 80 -60-40 -Fig. 14 Raman linewidth as a function of deposition temperature (Td) for annealed films: A,x, =0.5 (SiH,-He); , x, = 1 (pure Si H4) Fig. 15 Raman linewidth as a function of NF, mole fracllion (xN). Deposition and annealing conditions as for Fig. 13. 1.o 0.8 0.6 0.4 0.2 m rn 0.0 I I I 7I Fig. 16 Dependence of degree of crystallinity of as-deposited films on silane mole fraction (x,) and deposition temperature (T,).Without NF,: ,films totally crystalline; m,films totally amorphous; m, films partially crystalline.With NF, (xN=0.01): 0,films totally amorph- ous; 8,films partially crystalline. provision of gases for this research. We also acknowledge helpful discussions with Prof. W. E. Smith and Dr. B. Rospendowski. References M. L. Hitchman, J. F. Zhao and S. H. Shamlian, J. Muter. Chem., 1994, 4, 1821. M. L. Hitchman and K. F. Jensen, in Chemical Vapor Deposition- Principles and Applications, 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. J. MATER. CHl M., 1994, VOL. 4 E. Kinsbron, M. Sternheim and R. Knoell, Appi. Phys. Lett., 1983, 42, 835. M. L. Hitchman, C. W. Jones, J. F. Zhao and S. H. Shamlian, Adz;. Muter. Opt. Electron., 1993,2, 123. M. L. Hitchman, W. Ahmed, S. Shamlian and M. Trainor, Chemtronics, 1987, 2, 147. K. F. Jensen, M. L. Hitchman and W. Ahmed. in Proc. 5th Eur. Conf. on 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. W. Lang, in Proc. 12th. Int. Conf on Raman Spectroscopy, ed. J. R. During and J. F. Sulkian, Wiley, Chichester, 1991. p. 816. Paper 4/02725F; Received 9th Muy, 1994