J. MATER. CHEM., 1994, 4( 12), 1835-1842 Studies of the Effects of NF, on the Growth of Polysilicon Films by Low-pressure CVD Part 3.-Effect on Composition Michael L. Hitchman,*" Junfu Zhao,+" Sarkis H. Shamlian," Stanley Affrossman," Mark Hartshorne," Ewa A. Maydellb and Hamid Kheyrandish*" " Department of Pure and Applied Chemistry, University of Strathclyde, 295 Cathedral Street, Glasgow, UK G11XL b Department of Physics and Applied Physics, University of Strathclyde, Glasgow, UK G4 ONG Centre for Thin Film and Surface Research, University of Saiford, Salford, UK M5 4WT A range of analytical techniques (FTIR, SIMS, AES, SNMS, XPS) has been used to study the effect of adding NF, to an LPCVD gas mixture of SiH,-He on the composition of the resulting as-deposited and annealed polysilicon.Comparisons have been made with CVD silicon nitride and with polysilicon deposited in the absence of NF,. It has been shown by all the techniques used that for LPCVD in the presence of NF,, nitrogen is incorporated into the layer, but quantitative analysis with SNMS and XPS shows that such films are very similar to polysilicon and contain less than ca. 5 atom% of nitrogen. FTIR and SIMS also have revealed the presence of fluorine in as-deposited layers from LPCVD with NF,, but results from SIMS suggest that its concentration is very low and decreases by a factor of about two on annealing. It is concluded that while polysilicon deposited in the presence of NF, does contain small quantities of N and F, nevertheless the material may have interesting properties, with the use of NF, improving the surface quality.In the first paper in this series' we outlined the uses of polycrystalline silicon in VLSI fabrication and we pointed out that annealing of as-deposited amorphous films gives the highly desirable features of good structural perfection and low strain necessary for the good quality crystalline material required for device applications. The strategies for achieving this have been discussed',2 and we have reported on an alternative approach involving trying to etch the polysilicon partially as it deposits with the aim of reducing the crystallite size, leading to an effective decrease in the crystalline growth rate, and thus obtaining a less crystalline and more amorphous material.The gas we used as a potential etchant was NF,. However, although this did lead to a net decrease in growth rate,' as would be expected if etching were occurring, we have shown that, in addition to in situ etching of the deposited layer, adsorption of this reactant and the blocking of adsorp- tion of the growth precursor SiH4 could also occur. Nevertheless, we have also demonstrated2 that the use of NF, can produce amorphous films from which low strain and undistorted good quality crystalline polysilicon can be obtained upon annealing. The best quality annealed films produced with NF, are obtained with a low mole fraction of NF, (xNGO.01) and high deposition temperature (&3 650 "C), in contrast to results without NF, where typically it is necessary to use Td,<580"C.Also, in comparison with the growth of amorphous layers without NF,, there is slightly more strain after annealing the NF, treated layers. This we have suggested could be due to the incorporation of small amounts of N and F into the layers as a result of the strong adsorption of NF,. In this paper we report the results of qualitative analysis with FTIR, secondary-ion mass spectrometry (SIMS) and Auger electron spectroscopy (AES), and of quantitative analy- sis with secondary neutral-atom mass spectrometry (SNMS) and X-ray photoelectron spectroscopy (XPS) of films grown with and without NF,. t Present address: Department of Applied Chemistry, Taiyuan University of Technology, Taiyuan, Shanxi 030024, P.R.China. j: Present address: MATS UK, Wavertree Boulevard South, Wavertree Technology Park, Liverpool, UK L7 1PG. Experimental The LPCVD reactor, the gas-handling system and rypical deposition procedures and conditions have been described in detail previo~sly~,~ with the particular conditions used for the study being given in the earlier papers in this series 1,2 To allow a ready comparison between the results for the karious techniques all the results reported in this paper were obtained using the same set of samples. The film deposited mithout NF, (labelled sample 11)was grown from an SiH,-He mixture onto a silicon( 111) substrate with the mole fraction (x,) of silane being 0.1. The deposition temperature was 600'C and the total pressure (PT)in the reactor was 65 Pa.The thickness of this film as determined by surface profilometry' was 465 nm. The layer deposited with NF, (sample 111) had x,=O 5 and PT=65 Pa, Td=670 "C and xN=0.0093. This film was 456 nm thick and had a silvery, metallic appearance, exactly the same as that of the polysilicon sample. Both samples I1 and I11 were annealed at 950 "C under 1 atm of nitrogen. In order to have a nitrogen-containing reference material, a Sam pie of silicon nitride (sample I) was also analysed by the karious techniques. This sample was prepared by a standard thermal CVD method and was supplied by Philips Components Ltd., Southampton. The thickness of this layer as determined by ellipsometry was 120nm.The film showed a uniform blue interference colour. FTIR analysis of the layers in the range 390-2000 cm-' was carried out using a Nicolet spectrometer (Model 51: DX), although in all cases significant spectral features were only observed at wavenumbers less than ca. 1400 cm-' and so just the range from 390 to 1380cm-' is reported on here. Static SIMS data were obtained using a VG Quadrupole SIMS LAB3 instrument. A Ga' ion beam was used for monitoring nitrogen and fluorine as well as oxygen in the layers. With a 3.5 keV Ga' beam of current intensity 1 nA, rastered over an area of ca. 110 pm x 110 pm, sputter etch rates were low. However, in order to remove any surface contamination effects the sample was sputter-cleaned for ca.10 min arid it is estimated that material to a depth of 6-12 nm was removed during this time. Both positive and negative SIMS analysis was then carried out. Quantification in SIMS is generally ambiguous, except for low concentrations of impurities in homogeneous matrices. This is because the process of ionisation on the surface is critically dependent on the chemistry of the surface. For this reason the samples were also analysed using electron impact J. MATER. CHE:M., 1994, VOL. 4 annealing occurred, possibly because of leakage of oxygen into the atmospheric-pressure annealing system. Fig. l(a) and (b) show spectra for sample 111, polysilicon deposited in the presence of NF,, for as-deposited and annealed material, respectively.The spectra of both layers SNMS. This is a relatively new surface analysis te~hnique~.~ show similar peaks to those found for sample I1 at and is a valuable, complementary method to SIMS since it is 609-611 cm-' and at 471-473 cm-'. The ax-deposited layer not significantly influenced by matrix effect^.^ In e-beam spectrum has a band at 1094cm-' correhponding to the SNMS the positive and negative SIMS signals are suppressed Si-0 bending mode, but this is not apparent for the annealed and sputtered neutrals are ionised at some distance away from the sample surface using electron impact post-ionisation. Thus the signal is largely independent of any matrix effects and depends only on the electron impact ionisation cross-sections, the sputter yields and the collection efficiencies of the neutrals. The technique can therefore be calibrated by using known standards and so serves as a useful quantitative complement to SIMS.By using a switching facility between a STMS operational mode to a SNMS mode the same instru- ment can be used for both techniques and a wide range of sensitivities from the sub-ppb level upwards can be obtained. A detailed description of the procedures for SNMS has been given previ~usly.~ For the SNMS studies in this work, the system used was a VG Scientific SIMSLAB in the Centre for Thin Film and Surface Research at the University of Salford. The primary analysing beam used was either 10 keV Ar+ or 30 keV Ga' . With the former beam the crater dimension was 370 pm x 490 pm while with the latter it was 100 pm x 80 pm. The etch rates were determined using surface profilometry; typical etch rates were in the range 0.4-0.8 nm s-'.The instrument was also used in the dynamic negative SIMS mode for monitoring of fluorine, oxygen and other impurities. All analyses were carried out with a system pressure of ca. lop8Pa. AES was used for surface compositional analysis. This was performed in the same system as used for the static SIMS. The Ga+ ion beam was again used to remove surface contami- nation prior to electron beam irradiation for AES analysis. XPS data were obtained with a Vacuum Science Workshop X-ray anode, using AI-Ka radiation, and a 100mm hemi- spherical analyser. The photoelectron intensities were meas- ured at a take-off angle normal to the surface. Results FTIR Spectroscopy The main feature of the spectrum for sample I, the silicon nitride layer, was a strong absorption peak at 845 cm-'.This is characteristic of the Si-N asymmetric stretching mode.8 A weak band at 460 cm-' probably arose from an Si breathing mode,9 while a very weak band at 1085cm-' could be associated with an Si-0 bending mode,g arising from the incorporation of small amounts of oxygen impurity in the layer. The absence of any band at ca. 610 cm-' showed there was no Si-Si bonding, while the absence of any band in the region of 1200 cm-' indicated there was no, or very little, hydrogen incorporated into the layer, as would be expected for thermal CVD material; this is in contrast with plasma CVD material where significant amounts of hydrogen incor- poration can occur.1o In the spectra for as-deposited and annealed polysilicon, sample 11, three peaks of interest could be identified in each case.A peak at ca 610 cm-' could be assigned to the Si-Si bending mode," while that at 475 cm-' was again due to the Si breathing mode. Neither of these two peaks showed any significant changes in position or shape on annealing. A third sample, probably because it is masked by the broad band centred around 897cm-'. This broad band could be associ- ated with some Si-N bond stretching in the material. However, the presence of the peak at 611 cm-' shows that if N is present then it does not form stoichiometric Si,N,.This conclusion is supported by the silver, metallic appearance of sample 111; if the sample had had a composition close to that of Si,N, then one would have expected to have seen an interference colour. The situation is complicated, though, by the possible pres- ence of F in the layers. Fig. l(a)for the as-deposited film has a peak at 847cm-' which could be attributed to an Si-N stretching mode, but in addition there are peaks at 964 and 791 cm-' which could be associated with SiF, and SiF, stretching modes, re~pectively.'~-'~ On annealing, the three individual peaks disappear and the region of the spectrum between 750 and 1100 cm-' is occupied by the broad band mentioned above and centred at 897cm-'.The spectrum is 94 88 82 76 h8 v 70 m Y '"1 70 I 1310 1080 850 620 390 wave nu mbe rkm-' peak at 1093 cm-' in the as-deposited layer and 1086 cm-' Fig. 1 FTIR spectra for polysilicon films with NF, additionin the annealed layer was again associated with Si-0 bending, (sample 111). (a) As-deposited: & =670 'C, xS=0.5, xN=0.0093. and a broadening and slight intensification of this peak after (b)Annealed: T,=950 "C J. MATER. CHEM., 1994, VOL. 4 very similar to that reported by Fujita et a1." who have prepared fluorinated silicon nitride films by plasma CVD of a mixture of SiF,, N, and H,. They observed a strong, but broad peak in the region 750-1100 cm-' and centred on 900 cm-', which they also assigned to Si-N stretching and which they pointed out would veil absorptions due to Si-F, and N-F, bonding. An absorption at cu.1032 cm-' would be expected for a N-F,, band (m= 1-3) and this may be related to the small shoulder at 1026 cm-I in Fig. I@). However, on the basis of the relative bond energies of 271 kJ mol-' for N-F and 535 kJ mo1-I for Si-F, it would be expected that the majority of the fluorine would be bonded to Si rather than N, and the spectrum in Fig. 1(a) is indicative of this. Compared with Fig. 1(a),the apparent disappearance of the absorption peaks associated with Si-F, bonds [Fig. l(b)] could be simply due to the diffusion of F out of the layers during the high-temperature anneal. The loss of fluorine from a fluorinated silicon nitride by such a mechanism has been suggested by Fujita and Sasaki." However, they suggested this would occur only in the presence of hydrogen in the layer when the slightly higher H-F bond energy of 570.3 kJ mol-I compared with 552.7 kJ mol-' for the Si-F bond would favour HF formation and desorption to the gaseous phase.This is not likely to be an efficient mechanism in our case because there is little evidence for much hydrogen incorpor- ation in the layer. Peaks in the region 2000-2300 cm-' would be expected'' for Si-H symmetric stretching and in the region 3340-3350 cm-' for N-H symmetric stretching. No structure was observed in either of these regions for the as- deposited material. The small peak at ca. 1144 cm-' in Fig. l(a) could be associated with an N-H stretch,8 but if this is the case, then there is only a very small amount of hydrogen in the film, as would be expected for a film deposited at 670'C.Therefore, it is unlikely that on annealing all of the F will be lost from the layer and we conclude that the peaks due to residual Si-F bonds are not seen because they are masked by the broad band centred on 897 cm-l. The reason why the broad band replaces the individual peaks could be associated with changes in bonding on annealing. Fig. 1 (u) clearly indicates that during deposition there is some Si-N and Si-F bonding. Annealing will lead to molecular rearrangement and crystalline growth with prob- able strengthening of these bonds. The individual absorption bands will therefore become larger leading to a greater overlap and eventually an envelope encompassing the individual absorptions.The central frequency of this envelope at 897 cm-' is consistent with this picture. In pure Si,N, the Si-N absorption maximum is ca. 850 cm-'. An Si-N bond will be polarized (Si6+ -Nap) because of the higher electro- negativity of N (3.0) compared with Si (1.8).17 If fluorine, which has an even higher electronegativity (4.0), is present in the film and is also bonded to Si then the Si will be more positively ionised and the force constants will be enhanced. Hence any absorption due to Si-N bonding will be shifted to a higher frequency, as is found. The FTIR analysis therefore suggests that polysilicon films deposited in the presence of NF, will contain N and F as impurities, although the layers probably do not comprise stoichiometric Si,N,.The observations made in ref. 2 concern-ing the slightly greater strain observed in annealed polysilicon layers grown in the presence of NF, than in those grown in the absence of NF, (resulting from N and F impurities) are thus seen to be justified. Further evidence for these impurities is given by other analysis techniques. SIMS Fig. 2(u) and (h) show positive and negative SIMS data for the silicon nitride layer, sample I. The spectra show a large 1837 10~~ I 1o4 1o3 1o2 10'-'m loo 4.-C 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.C.100.0 m /z Fig. 2 SIMS data for a silicon nitride film (sample I): (a) positive SIMS, (b)negative SIMS number of peaks and only a few have been selected in the discussion which follows for illustrative purposes.No unex-a pectedly, the presence of Si is clearly seen (e.g. 28Si+, "Si,' ) as is the presence of N (e.g. 7oSi2N+, 42SiN-). The appvarance of 0 (e.g. l6O+,45SiOH+, 72Si20+, l6O-, I7OH-, 44SiO-, 60Si02-, 76Si03-) is also not unusual because, even though a surface layer of ca. 20nm had been removed before the analysis, CVD of Si,N, is not normally carried out in ii high-vacuum integrity reactor. Therefore, contaminatioli with oxygen can readily occur. The presence of 0 is coirsistent with the evidence given by FTIR for traces of Si-0 bonding in the layer. The apparent appearance of mass 19, which could correspond to F in the layer, is believed to be an electron-stimulated desorption signal and not a true SIMS signal for this sample.Fig. 3(a) and (b)show the positive and negative STh4S data for polysilicon without NF, addition (sample 11).Again there are a large number of peaks, but the important point to note is the absence of any peaks which could be associated with N; the peak at m/z= 14 can be attributed to 28Si2+. As in the case of Si,N, the presence of 0 is consistent with the type of LPCVD reactor used for film growth and also it agrees with the FTIR spectrum for sample TI. Sample 11 i:b, apart from oxygen contamination, a reasonably pure sample of poly silicon. Fig. 4(u) and (h)show the positive and negative SIRIS data for sample 111, the polysilicon layer with NF, addition.In the positive spectrum there is ample evidence of Si (e.g. 28Si2+, 28Si+, %i2+), but there is also the appearance of a peak at m/z=70 corresponding to Si,N+, which was also present in sample I, but not in sample 11. So the suggestion from FTIR of the presence of N (Fig. 1) is supported by SIMS. Further confirmation is given by negative SIMS which clear1-v shows 42SiNp, as was obtained for the Si3N4sample. The presence 1838 1 SiOH' i 0-lo4j 1.OH-1o2 10' 1.-no 0.0 10.0 20.0 30.0 40.0 50.060.070.0 80.0 90.0 100.0 m Jz Fig. 3 SIMS data for a polysilicon film without NF, addition (sample IT): (a) positive SIMS, (h)negative SIMS 1o7 1o6 1o5 Ga' 1o4 Si+ 1o3 1o2 10' 1o2 10' '*"OO 10.0 20.0 30.0 40.0 50.060.0 70.0 80.0 90.0 100.0 m Jz Fig.4 SIMS data for a polysilicon film with NF, addition (sample 111): (a)positive SIMS, (h)negative SIMS J. MATER. CHI:M., 1994, VOL. 4 of F in this sample is indicated by the large peak at m/z= 19 in the negative SIMS as well as by the peaks at m/z=47 in both spectra. A further indication of F in the layer is given by the small peaks at m/z=66 and 85 in the positive SIMS which could correspond to the appropriatc SiF, + species. Again the evidence given by FTIR is supported by the SIMS, and one can conclude that polysilicon grown in the presence of NF3 does incorporate some N and F into the layer. It is difficult to quantify the amount of N in the film though because of matrix effects and the presence.for example, of 28si2 + in the spectrum as well which swamps the weak N+ signal arising from the low-efficiency nitrogen secondary ion formation. Other techniques are needed to give more infor- mation about the extent and nature of the N and F presence in sample 111. AES Auger electron spectroscopy (AES) is a useful complementary method of surface analysis to static SIMS, sampling as it does the top few nm of a film.'' Fig. 5(a)shows the Auger spectrum for the Si3N, film (sample I). This spectrum was recorded after sputtering ca. 20nm of the layer. Before this was done there was an additional peak at 270 eV corresponding to the C KLL signal. In the sputtered sample the Si LMM peak at 97eV is clearly seen as is the N KLL peak at 380eV.The presence of 0 is again revealed (0KLL) at 510 eV. Sputtering of a further 30-50 nm showed that the presence of 0 persisted 60I t 3 1601 1 " 100 200 300 400 500 600 700 800 kinetic e ne rgy/eV Fig. 5 AES for samples I (a),I1 (h)and I11 (c) J. MATER. CHEM., 1994, VOL. 4 even quite deep into the layer. Very significantly, there is no evidence for F (F KLL) at 650 eV, although, as pointed out below, AES is a much less sensitive technique than SIMS.18 Fig. 5(b) shows the Auger spectrum for sample TI, pure polysilicon, after ca. 30 nm had been removed by sputtering; the only element found is Si at 98 eV (Si LMM). The low sensitivity of AES is probably the reason for not detecting 0 in the layer, although there is possibly a very small peak at 510 eV corresponding to 0 KLL.Fig. 5(c) is the Auger spectrum for sample I11 taken at a depth of ca. 34 nm, and the only element found is Si (Si LMM at 97.1 eV). There is no evidence for 0, N or F which is in contrast to the FTIR and SIMS results. Auger spectroscopy typically has a sensitivity18 of ca. 1 atom% which can make the technique of limited value, but in this case it does strongly suggest that whilst N and F are certainly present in polysilicon grown in the presence of NF,, as indicated by FTIR and SIMS, the layer is by no means 'pure' silicon nitride, which is again consistent with the FTIR results. We now examine this idea with more quantitative methods of analysis.SNMS Fig. 6 shows SNMS data for the silicon nitride sample. As with the static SIMS data [Fig. 2(a)] Si is readily seen (e.g.28si2+ 28si+ , 56Si2+) as is the presence of N (e.g. 42SiN+).7 Oxygen is again found (e.g. l60+, ,%OH+). Fig. 7 shows an SNMS depth profile for sample I. The SIN' profile clearly shows the sharp interface between the film and the substrate, with no evidence of any N in the silicon substrate. The profile for m/z=14 which can arise from both Si2+ and N+ also shows a sharp signal decrease as the interface is crossed and N is no longer present. The increase at the interface in the rn/z=28 profile for Si+ is consistent with the changeover from an Si,N, film to pure Si. Quantitative analysis based on the stoichiometry of Si,N, showed that the relative sensitivity factors for 14Nf and 42SiN+ were 0.06 and 0.02, respectively; both these values agree well with the values of 0.06 and 0.03 obtained by independent analysis by the equipment manufacturer.These relative sensitivity factors can be used to quantify any nitrogen found in the other samples analysed. Fig. 8 is an SNMS depth profile for the polysilicon layer deposited in the absence of NF, (sample 11). This SNMS profile, not unexpectedly, shows no change at the interface region, indicating, as the SIMS did, that the layer is reason- ably pure. Fig.9(a) shows the SNMS depth profile for as-deposited polysilicon in the presence of NF,. The presence of N is lo5k 1o4 Si+ 0.0 6.0 12.0 18.0 24.0 30.0 36.0 42.0 48.0 54.0 60.0 m lr Fig.6 Static SNMS data for a silicon nitride film (sample I) t SIN' \ \ loo-' ' ' ' " ' ' . .0.0 106.3 212.7 319.0 425.3 531.7 t/s Fig. 7 Dynamic SNMS depth profile for sample I t 1 0.0 237.4 474.8 712.2 949.6 1187.0 t /s Fig. 8 Dynamic SNMS depth profile for sample I1 apparent from the 42SiN+ profile, with a sharp step at the interface with the silicon substrate. However, in contrast to the Si,N4 sample (Fig. 7) there is only a small change at the interface for the profile for rn/z=14 corresponding to Si2+ and N+. Although an enlargement of this profile showed the small step more clearly, there was obviously less contribution of N to the m/z= 14 signal for sample 111 than for sample I, again supporting the evidence of FTIR and AES that sample 111 is not pure Si,N,.A calculation of the N content of the film, taking into account the contribution of doubly charged silicon to the signal for m/z =14, showed it to be certainly <5%, which is reasonably consistent with the AES result above. The SNMS depth profiles [Fig. 9(b)] for sample I11 after annealing are very similar to those for the as-deposited layer. A magnification of the scale for the profile for m/z= 14 again illustrated the contributions of both Si2+ and N+ to the signal. Most importantly, analysis of the spectrum showed that the nitrogen content of the film remained essentially J. MATER. CHEM., 1994, VOL. 4 lo4 Si2', N+ :2102 rn Y ,l@L .0.0 180.9 361.8 .542.6 723.5 904.4 tls Fig. 9 Dynamic SNMS depth profile for sample 111: (a)as-deposited, (h)after annealing unchanged after annealing. This strongly suggests that the peak at 897 cm-' in the FTIR spectrum [Fig. l(b)] does arise from the Si-N stretching mode. Any evidence for the role of F in the layer, though, is not given by the SNMS analysis. As has already been indicated, SNMS is not inherently as sensitive as SIMS and so this is why it was not possible to detect any F in sample 111. Therefore, negative SIMS depth profiling was carried out in the same apparatus on sample 111. Fig. lo@)shows the depth profile obtained with this technique; the 19F- is prominent. Unfortunately, it is not possible to quantify the amount of F present, although its normalized yield relative to 28Si-is 13: 1.The oxygen concentration in the film is very low, estimated to be certainly less than 1 part in lo4. SIMS being more sensitive than SNMS shows up the ubiquitous C in the film with, as is often the case, a higher level at the near-surface region. Fig. 10(b) gives the negative SIMS depth profile for sample 111 after annealing. The most significant features of this depth analysis are the very different profile for 19F-and a normalized yield relative to '*Si- of 6: 1. These changes indicate that not only has there been a redistribution of the fluorine but also more than half of it has diffused out during 0- 10' Ll loo2 0.0 478.2 956.5 1434.7 1913.0 2391.2 10' inoIV 0.0 340.8 681.6 1022.4 1363.8 1703.9 tIs Fig.10 Dynamic SIMS depth profile for sample 111: (a)as-deposited, (b)after annealing annealing. The FTIR spectrum [Fig. l(b)] is consistent with this observation since we concluded that it was unlikely that all of the F had been lost from the layer on annealing, with the shift of the Si-N peak to a higher frequency indeed suggesting that there was still some F present in the film. The 0 profile in Fig. lo@)is also consistent with the FTIR results [cj Fig. l(a) and l(b)] since the two to three times higher l60-SIMS signal at the surface than in the as-deposited layer shows that some oxidation has probably occurred during the nitrogen atmospheric-pressure anneal.In summary, the SNMS analysis has confirmed the findings of FTIR and SIMS that for polysilicon grown in the presence of NF, (sample 111) N is present, but only at a level of < 5 atom%; i.e. the film is certainly not stoichiometric Si,N,. The SNMS analysis also shows that after annealing there is essentially no change in the N content. The lack of sensitivity of SNMS does not allow any F to be detected in sample 111, but SIMS depth profiling does reveal it and, furthermore, shows that about 50% of the F is lost from the layer on annealing. The remaining F is, in view of the strong Si-F bond energy, undoubtedly bonded to Si and this probably J. MATER. CHEM., 1994, VOL. 4 E, (exp.)sample lev Ib 101.9IF 99.4 IIId 99.2 a Full width at half height.Si,N,. Table 1 XPS binding energies Si 2p N 1s 0 Is E, (lit.) FWHH" Eb (exp.) Eb (lit.) Eb (exp.) /e v lev lev IeV lev 101.9 2.4 397.3 397.7 532.2 99.4 2.1 --532.6 99.4 2.1 397.2 397.7' 532.3 'Poly Si. Poly Si+NF,. For Si,N,. FTIR spectrum, which can be attributed to Si-N stretching, to a higher frequency than one normally associates with that mode. and escape depths using Wagner's sensitivity factors." Fig. 11(a) and (b)show the peaks for Si 2p and N Is, respect- ively, for silicon nitride, sample I. Data for the binding energies (E,) and compositions are summarised in Tables 1 and 2, respectively. Peaks for 0 Is and C 1s were also found, but they are not shown here, although the data for oxygen are also summarised in the tables.The presence of oxygen is consistent with the results of the other methods of analysis Si 2p N Is (6) Fig. 11 XPS for the silicon nitride film (sample I): (a) Si 2p spectrum, (b)N Is spectrum Si 2p I[111111111~11111l111 100 110 120 130 EdeV .gc I 3 $ EdeV Fig. 13 XPS for the polysilicon film with NF, addition (sample TIT): (a) si 2P spectrum, (b)N 1s spectrum Table 2 XPS composition (atom '30) sample Si 2p" N 1s 0 Is I 0.48 0.36 0.16 I1 0.69 0.00 0.31 111 0.72 0.02 0.26 " Note that silicon compositions include all core contributions. reported above. The rather high oxygen content of 16 atom% probably results, as mentioned earlier, from contamination of the sample during deposition and from the fact that XPS monitors only a few monolayers at the surface.The Si 2p signal shows only one peak at the expected binding energy for silicon nitride and the nitrogen signal also correspmds to that of the nitride. The quantitative amounts of nitrogen and oxygen suggest that the surface silicon is in the form of nitride and oxide in the ratio 3: 1, and the fact that the Si ?p peak half width is greater than that shown by either of the two polysilicon samples (cJ Fig. 12 and 13) is consisterit with contributions from Si-0 bonds and with a separation of only 1.5 eV between the Si 2p peaks for Si,N, and for SiOz (Eb= 103.4 eV19). For the polysilicon film deposited without NF, addition the XP spectrum is shown in Fig.12 and the Si 2p signal consists of a major peak corresponding to elemental silicon with an E, of 99.4 eV and a contribution from oxidised silicon 3.6 eV higher and separated by a distinct valley; both Ebs are in good agreement with literature values." The signal at ca. 17 eV above the core signal is associated with the elemental silicon bulk plasmon. Comparison of the 0 1s and Si 2p oxide signals shows the ratio of the oxygen to oxidised silicon to be 1.8: 1, which suggests that the silicon had an overlayer of SiO,; this is contrary to the AES result, but in that case some 30 nm were sputtered off before the measurement was made. Fig. 13 shows the XPS result for the polysilicon film deposited with NF, addition (sample 111).The peak due to the silicon bulk plasmon is again in evidence, while thc Si 2p peak appears at 99.2eV which is very close to that for pure Fig. 12 XPS for the polysilicon film without NF, addition (sample 11) silicon,-indicating that the environments of the Si atclms in this sample and sample I1 are very similar; in other words, sample 111 is very much like a polysilicon layer. This is in contrast to sample I where clearly the layer is very different from polysilicon. However, N is definitely present in sample 111, and, although the N 1s peak occurs at a binding energy close to that expected for nitride, it can be seen from Table 2 that the layer contains only ca. 2 atom% of N, which is in agreement with the SNMS calculated value of 65atom%.The oxygen content again appears to be high and the Si 2p signal, as for sample 11,also shows a contribution from Si-0 bonds. For sample 111, though, the valley between the two Si 2p peaks is less pronounced than for sample I1 and this is postulated to arise from a small Si 2p nitride signal which reduces the valley depth. However, the contribution from the Si 2p nitride is more marked than expected from the nitrogen to oxygen composition ratio, but it should be remem- bered that these elemental amounts are derived assuming uniform distributions whereas the oxide is probably again an overlayer; the calculated 0 to N ratio is therefore exaggerated. No evidence was found in any of the spectra for F in the films.XPS is a technique which has a detection limit compar- able to that of AES (i.e. at the atom% level) and the earlier evidence for only small amounts of F in sample I11 is supported by its absence from the XPS results. Therefore, once again one sees that growing polysilicon in the presence of NF, does introduce a small amount of N, but the layer is still much more like pure polysilicon than stoichiometric Si,N,. Conclusions Polysilicon deposited by LPCVD in the absence and presence of NF, has been analysed by FTIR, SIMS, AES, SNMS and XPS. All of the techniques have shown that the incorporation of nitrogen occurs during polysilicon deposition in the pres- ence of NF,. However, a comparison with corresponding analyses of CVD silicon nitride has shown that the film is not stoichiometric Si,N,, and quantitative analysis by SNMS and XPS, in fact, shows the nitrogen content to be less than ca.5 atom%. Fluorine is also revealed by FTIR and SIMS to be present in the film deposited in the presence of NF,, and although it was not possible to obtain a quantitative measure of the amount of fluorine present, the fact that other less sensitive techniques, such as AES, SNMS and XPS, did not show any evidence of fluorine indicates that the amount was probably significantly less than 1atom%. SIMS analysis also suggested that annealing of the film caused about 50% of the fluorine to diffuse out of the layers. Earlier work2 has shown that growing polysilicon in the presence of NF, can yield good quality annealed layers even when the deposition temperature is much higher than is normally possible.The results in this paper show that this advantage can be offset by the incorporation of N and F into the layers. However, since the amount of this elemental contamination is small, especially for F, this may not be a significant disadvantage, especially if a material with proper- ties intermediate between those of a true semiconductor and J. MA'TER. CHEM.. 1994, VOL. 4 an insulator is required, (as in the case of using N,O with SiH, to produce SIPOS20). Studies are now needed on the electrical characteristics of polysilicon deposited with the aid of NF, and on its use in simple devices to investigate the potential of the material for microelectronic applications.An investigation of the use of other fluorine-containing species (e.g. F2, HF, ClF, and XeF,) for the control of the crystalline quality of polysilicon might also be interesting, and may also help to gain a better understanding of the chemistry of the deposition process. However, from a practical point of view, these gases would generally be more difficult and dangerous to handle than NF,. 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 provision of gases for this research.References 1 M. L. Hitchman, J. F. Zhao and S. H. Shamlian. J. Muter. Chem., 1994,4, 1821. 2 M. L. Hitchman, J. F. Zhao and S. H. Shamlian. J. Muter. Chem., 1994,4, 1827. 3 K. F. Jensen, M. L. Hitchman and W. Ahmed, in Proc. 5th Eur. Conj: on CVD, ed. J. 0.Carlsson and J. Lindstrom, University of Uppsala, Uppsala, 1985, p. 144. 4 M. L. Hitchman, J. Kane and A. E. Widmer, Thin Solid Films, 1979,59,231. 5 J. Tumpner, R. Wilsch and A. Benninghoven, J. Vac. Sci. Technol. A, 1987,5,1186. 6 R. Wilson, J. A. Van den Berg and J. C. Vickerman, Surf Interface Anal., 1989,14, 393. 7 J. S. Colligon, H. Kheyrandish, J. M. Walls and J. Wolstenholme, Thin Solid Films, 1991,200,293. 8 J. Kanicki, MRS Symp. Proc., 1988, 118,671. 9 G. Lucovsky and D. V. Tsu, J. Vuc.Sci. Technoi. A, 1987,5,2231. 10 S. E. Alexandrov, M. L. Hitchman and S. H. Shamlian, Adv. Muter. Opt. Electron., 1993,2, 301. 11 F. A. Johnson and W. Cochran, in Rep. Int. Conf. Phys. Semiconduct., ed. A. C. Stickland, Institute of Physics, London, 1962, p. 498. 12 K. Yamamoto, T. Nakanishi, H. Kasahara and K. Abe, J. Non-Crystal. Solids, 1983,59-60, 213. 13 C. R. Bailey, J. B. Hale and J. W. Thompson, J. ('hem. Phys., 1937, 5,274. 14 E. A. Jones, J. S. Kirby-Smith, P. J. H. Woltz and A. H. Neilsen, J.Chem. Phys., 1951,19,242. 15 S. Fujita, H. Toyoshima and A. Sasaki, J. Appl. Phys., 1988, 64, 3481. 16 S. Fujita and A. Sasaki, Proc. Symp. Silicon Nitride and Silicon Oxide Thin Insulating Films, 1987,87, 535. 17 N. N. Greenwood and A. Earnshaw, Chemistrv of the Elements, Pergamon, Oxford, 1984, pp. 473,431. 18 S. Wolf and R. N. Tauber, Silicon Processing for the VLSI Era, Vol.1-Process Technology, Lattice Press, Sunset Beach, CA, 1986, ch. 17. 19 D. Briggs and M. P. Seah, Practicul Surface Analysis, Wiley, Chichester, 1990. 20 M. L. Hitchman, Vacuum, 1984,34,377. Paper 4/02728K; Receiued 9th May, 1994