- Email: [email protected]
An analysis of the TiN plasmachemical vapor deposition process based on optical emission spectroscopy measurements
Thin Solid Films 398 – 399 (2001) 343–348
An analysis of the TiN plasmachemical vapor deposition process based on optical emission spectroscopy measurements ¨ b S. Petera,*, H. Giegengacka, F. Richtera, R. Taberskyb, U. Konig a
¨ Chemnitz, 09107 Chemnitz, Germany Technische Universitat b Widia-Valenite, 45021 Essen, Germany
Abstract Optical emission spectroscopy measurements and film deposition experiments were performed for better understanding and control of the plasmachemical vapor deposition (PCVD) process used in the coating production of WidiayValenite. The effects of the titanium chloride flow, the nitrogen flow, the mean discharge current density and the pressure were investigated. The coatings were characterized with respect to deposition rate, composition, crystallographic structure and Vickers microhardness. In particular, the dependence on process parameters of emission signals (linesybands) from 12 species (Ti, Tiq , Cl, Clq, H2, H, q q N2, Nq 2 , N, N , Ar and Ar ) involved into the TiN deposition process was analyzed. Additional spatially resolved OES measurements of the 12 species revealed in particular that the emission of Tiq arises only from near the cathode surface. Based on an empirical model it was shown that the emission of atomic titanium is related to the intrinsic titanium deposition rate whereas the emission of ionized titanium results from the etching of titanium not yet bounded to nitrogen from the surface. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: Optical emission spectroscopy; Plasma diagnostics; Plasmachemical vapor deposition; Titanium nitride
1. Introduction Plasmachemical vapor deposition (PCVD) in pulsed glow discharges is a promising method for the industrial low temperature deposition of hard coatings on threedimensional substrates. Introduced into production for the first time by WIDIA in 1989 w1x, PCVD TiN-coated cemented carbides are used in many metal cutting applications. The PCVD processes for wear resistant TiN coatings are predominantly based on a gas mixture TiCl4–H2–N2–Ar operated in the pressure range of a few hundred Pa. The relations between in-situ measured discharge parameters and the coating properties have been investigated by several authors w2–8x. Non-invasive optical emission spectroscopy (OES) was used in many studies to determine information on the active species and the dominating plasmachemical reactions in the
discharge. However, the results obtained often differed strongly. According to Rie et al. w3x the TiN growth rate showed a direct dependence on intensities of ionized titanium. Rodrigo et al. w6x concluded from their experiments that titanium atoms and N2 (A3Sq u ) metastables are the major species contributing to the supply of activated titanium and nitrogen to the substrate surface. This conclusion was supported by an agreement between the film growth rate and the multiplied emission intensities of the indicated species. Mogensen et al. w7x investigated the PCVD of TiN in a large-scale industrial coating plant. The deposition rate was found to be strongly correlated with the Nq 2 OES signal and not the Tiq signal, when varying the total pressure and the nitrogen flow. In this paper, we present the results of a parametric study, including reliable OES signals from a larger number of emitting species than reported before.
* Corresponding author. Tel.: q49-371-561-8253; fax: q49-371561-8258. E-mail address: [email protected] (S. Peter). 0040-6090/01/$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 0 - 6 0 9 0 Ž 0 1 . 0 1 3 7 5 - X
344
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
Table 1 Standard values and investigated ranges of process parameters Substrate temperature Pressure Total gas flow (H2qN2qAr) Argon flow Nitrogen flow Titanium tetrachloride flow Discharge current density Pulse duration time Pause duration time
5508C 250...300...450 Pa 400 sccm 50 sccm 16...60...80 sccm 2.4...3.6...4.8 sccm (additional) 0.2...0.5...0.7 mAycm2 50 ms 80 ms
2. Experimental The deposition of titanium nitride films and the insitu study of the growth process by optical emission spectroscopy were carried out in a parallel-plate reactor. Hard metal substrates were inserted into the cathode, which was temperature-controlled by an additional resistance heater. The unipolar pulsed power supply was used in the current regulation mode. The PCVD process was investigated using spatially resolved OES. Optical emission was sampled from the negative glow and from the cathode fall region. Spectra were recorded in the spectral region from 250 to 1000 nm with a resolution of approximately 0.1 nm using a LN2-cooled CCD-array. In an earlier study we selected reliable signals (linesybands) of 12 species (Ti, Tiq, q q Cl, Clq, H2, H, N2, Nq 2 , N, N , Ar and Ar ) involved in the TiN deposition process (see Table 2 in Peter et al. w9x). Two different collimators and measuring principles were applied. Both collimators were directed parallel to the cathode surface. The collimator with 15 mm inner diameter was used to find out correlation between process parameters and OES signals. Its large diameter ensures a low sensitivity of signals on changes in the spatial distribution of the optical emission caused by variations of process parameters. In contrast, the distribution of lineyband intensities in the region between the cathode and the negative glow was measured using a collimator with 0.6 mm inner diameter. The movement of the light sampling unit (collimator, quartz lens and quartz fiber) was realized with a directly coupled precision feed-through. More details on the deposition chamber, the deposition process and the OES system are given elsewhere w9 x . To find out the correlation between process parameters, TiN coating characteristics and in-situ measurements, a systematic variation of individual parameters with the others kept constant at standard conditions (see Table 1) was performed. The TiN coating thickness was measured using the ball-grinding method. X-Ray diffraction measurements were made with filtered CuKa radiation. The CuKa2 part was eliminated using the Rachinger-correction.
Fig. 1. Emission signals of (a) nitrogen and argon species and (b) titanium and chlorine species as a function of the precursor flow.
Energy dispersed electron microanalysis measurements were carried out to determine the chlorine and oxygen concentrations in the coatings. 3. Results 3.1. Dependence of the OES-signals on the process parameters Figs. 1–4 show the influence of titanium chloride flow, nitrogen flow, mean discharge current density and
Fig. 2. Emission signals of (a) nitrogen and argon species and (b) titanium and chlorine species as a function of the nitrogen flow.
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
345
3.2. Distribution of the OES-signals within the cathode region For a better understanding of the TiN PCVD process and a reliable interpretation of the relations presented above, the distribution of OES signals between the cathode and the negative glow of the discharge was investigated. Complete spectra were recorded in order to enable the correction for background. In Fig. 5 the distribution of the normalized (mean value equals unity) emission intensity of all 12 measured species is given. A cathode fall length of approximately 8 mm is estimated from this OES measurement. The emission of ionized titanium near the cathode surface is at least one order of magnitude higher than in the negative glow (see Fig. 5a). All other measured neutral atomic radicals (H, N and Cl) show somewhat higher emission intensities close to the cathode as compared with N2, Nq 2 , Nq, H2, Ti, Clq, Ar and Arq. Fig. 3. Emission signals of (a) nitrogen and argon species and (b) titanium and chlorine species as a function of the mean discharge current density.
pressure on the emission intensities of 10 species. The intensities have been normalized to unity for standard process conditions (see Table 1) and averaged for all lines used for a certain species (see Table 2 in Peter et al. w9x). Within the investigated range of parameters we observed an interesting behaviour of several lineyband intensities: ● The emission intensity of argon (constant argon flow in all experiments) depends linearly on the discharge current density (Fig. 3a). It does not change with the precursor flow (Fig. 1a) and changes little with the nitrogen flow (Fig. 2a) and the pressure (Fig. 4a). ● For Ar, N2, Cl and N, the intensity ratios Iwionx y Iwneutralx of ionized and respective neutral species are nearly constant when the process parameters (cf. Table 1) are varied. In contrast to that IwTiqx yIwTix changes strongly with process conditions. ● With increasing precursor flow the Tiq emission increases nearly linearly (Fig. 1b). It decreases with increasing nitrogen concentration (Fig. 2b) and to a lower extent also with increasing pressure (Fig. 4b). As a remarkable fact, the Tiq signal does not increase with the discharge current but passes through a flat maximum (Fig. 3b). ● The signal from atomic titanium rises with about the third power of the discharge current density (Fig. 3b). This indicates a complex formation process. ● Also the signals from chlorine species (Cl and Clq) increase with increasing precursor flow (Fig. 1b) and much stronger with increasing discharge current (Fig. 3b).
3.3. Dependence of the TiN coating characteristics on the process parameters An increasing deposition rate was found for increasing current density, nitrogen flow and pressure. In contrast, with increasing TiCl4 flow the deposition rate decreases. For all the coatings the X-ray diffraction experiments revealed the expected face-centered cubic (NaCl) structure. The lattice constant was determined from the dominating (200) TiN reflection and typically a value of approximately 0.424 nm was found. This is very close to the lattice constant for stoichiometric TiN: a0s 0.42417 nm w10x.
Fig. 4. Emission signals of (a) nitrogen and argon species and (b) titanium and chlorine species as a function of the pressure.
346
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
deposition rate Rat. x of the film forming species (xsTi, N, Cl): Rxat.s
Fig. 5. Distribution of the normalized emission intensities of 12 species within the cathode region (standard parameters).
The chlorine concentration in the TiN coatings varied between 2 and 5 at.%. The oxygen content was F1 at.% for all deposited layers. The microhardness HV0.05 of all coatings was within the range of 2000–2400. A more detailed report about the dependence of the TiN coating characteristics on the process parameters will be published elsewhere. 4. Discussion Due to the specific principles of the generation of optical emission in the discharge, the OES analysis of the PCVD for TiN can give only selected information, mainly about the density of excited species. Considering the behaviour of the argon, atomic nitrogen and atomic hydrogen intensities as well as the behaviour of the q intensity ratios (IwArq x yIwArx, IwNq 2 x yIwN2 x, IwCl x y q IwClx and IwN x yIwNx) one can conclude that within the investigated range of process parameters (see Table 1): ● the electron energy distribution is nearly constant; and ● the electron density is influenced by the discharge current density only. These are favorable conditions for analysis since the measured emission intensities from excited species give information about their ground state densities. To correlate the species density in the gas phase with the deposition characteristics we calculated the atomic
rfdfAcoatNA CA x tdep. C 8 xAMx
(1)
where : rf is the density of the film (assumed to be 5 gycm3 w5x); df is the thickness of the film; tdep. is the deposition time; Acoat is the coated area (cathode area, 240 cm2); NA is Avogadro’s number (6.02=1023 ymol); CA x is the atomic concentration of species x in the A coating (determined under the assumption CA TisCN suggested by the X-ray results); and Mx is the molar mass of x. At the deposition temperature of 5508C typically only a few percent of chlorine supplied with the TiCl4 are incorporated in the TiN films (see also w5,11,12x). Therefore, chlorine deposition is a small loss of a large amount of Cl in the gas phase and cannot be expected to be measured by OES. In contrast, a high portion of the supplied titanium (approx. 1.6=1018 atomsys at standard parameters) is deposited. So we correlated our OES results with the atomic titanium deposition rate. The correlation was done using an empirical model: Rmodel sC1µŽIwTix. ayC2ŽIwTiqx. b∂ŽIwNx. c Ti
(2)
where C1 and C2 are constants and a, b and c are exponents similar to ‘reaction orders’. The model is based on the following assumptions and considerations: ● the Ti emission intensity IwTix is a measure of the free titanium in the gas phase and is correlated with the intrinsic titanium deposition rate; ● the Tiq emission intensity IwTiq x) is related to the removal of titanium from the surface by etching of atoms not yet bonded to nitrogen; and ● nitrogen is the required bond partner of Ti for the TiN deposition. In the investigated parameter range the N2 flow was generally limiting the deposition rate. Although no specific role of atomic nitrogen could be deduced within the present study, its emission intensity, IwNx, was selected as a measure of active nitrogen — because of its high reactivity and, in comparison to Nq or Nq 2 , its high concentration. For another promising reactive species, the NH radical, unfortunately no reliable signal could be selected (for the 360 nm NH emission see Fig. 2 in Peter et al. w9x). In Fig. 6a–d the atomic titanium deposition rate is correlated with the empirical model. The fit parameters are C1s9=1017 atomsys, C2s0.2, as0.15, bs1.0 and cs0.5. It should be mentioned that the reaction order of nitrogen (and hydrogen) in thermal CVD of TiN was also found to be 0.5 w13x.
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
347
the pulsed DC) of their optical emission intensities. The emission of Tiq starts prior to others (see Fig. 5c in Peter et al. w9x). In contrast, the signal from atomic titanium (see Fig. 5b in Peter et al. w9x) was detected later and decreased not so suddenly as the other emissions but similar to the current (see Fig. 4 in Peter et al. w9x). 5. Conclusions
Fig. 6. Comparison of the atomic titanium deposition rate (solid line) with the model based on OES intensities (dotted line) in dependence on the TiCl4 flow (a), the N2 flow (b), the current density (c) and the pressure (d).
A parametric study, including OES signals from 12 emitting involved into the PCVD of TiN was performed. Except in the current series, the discharge was operated in constant current mode. That resulted in nearly constant excitation conditions and thereby the individual influence of the titanium chloride flow, the nitrogen flow, and the pressure could be analyzed. Using an empirical model it was shown, that the combination of the emission signals from atomic titanium, ionized titanium and atomic nitrogen could describe the deposition rate of titanium to a large extent. Atomic nitrogen was selected as the active nitrogen species necessary to deposit TiN, although its specific role could not be revealed within this study. The dependence of the Ti and Tiq signals on the process parameters investigated was specific and often contrary. The emission of atomic titanium was found to be a measure of the intrinsic Ti deposition rate. In contrast, the Tiq signal arises from the etching of titanium not yet bound to nitrogen from the surface. The measured spatial distribution of the Tiq emission is also in agreement with this hypothesis. Further investigations are required in order to understand the special excitation mechanism responsible for the Tiq emission. References
The simple empirical model based on OES signals fits well to the experimental atomic deposition rate. The remaining deviations probably result on the one hand mainly from inaccuracies in determining the atomic deposition rate of titanium (homogeneity of the deposition rate and the stoichiometry across the coated cathode area). On the other hand, the model is very simplified. It does not take into account for example the interaction of energetic ions and neutrals with the growing film. But, it confirms our hypothesis about the interaction of deposition and etching reactions in the PCVD of TiN from TiCl4–H2–N2–Ar mixture. It is worth mentioning that, although in some parameter regions the emission intensities of neutral and ionized titanium show similar trends, their origin and meaning are absolutely different. A further significant difference between Ti and Tiq was observed in the temporal behaviour (over the period of
w1x U. Konig, ¨ R. Tabersky, H. van den Berg, Surf. Coat. Technol. 50 (1991) 57–62. w2x Y. Ishii, T. Shibata, Y. Yoshino, H. Ichimura, K. Kobayashi, J. Ceram. Soc. Jpn. Int. Ed. 100 (1992) 1167–1173. w3x K.T. Rie, A. Gebauer, J. Wohle, ¨ Plasma Chem. Plasma Process. 13 (1) (1993) 93–101. w4x R. Hochreiter, J. Laimer, H. Stori, ¨ D. Heim, Surf. Coat. Technol. 74y75 (1995) 443–449. w5x J. Patscheider, L. Shizhi, S. Vepreck, Plasma Chem. Plasma Process. 16 (3) (1996) 341–362. w6x A.B. Rodrigo, C. Lasorsa, M. Shimozuma, F. Alvarez, P. Perillo, J. Phys. D: Appl. Phys. 30 (1997) 2397–2402. w7x K.S. Mogensen, S.S. Eskildsen, C. Mathiasen, J. Bottiger, Surf. Coat. Technol. 102 (1998) 41–49. w8x M. Eckel, P. Hardt, M. Schmidt, Surf. Coat. Technol. 116–119 (1999) 1037–1041. w9x S. Peter, F. Richter, R. Tabersky, U. Konig, ¨ Thin Solid Films 377y378 (2000) 430–435.
348
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
w10x JCPDS, International Center for Diffraction Data, Swarthmore, PA, 1992. w11x J. Laimer, H. Stori, ¨ P. Rodhammer, ¨ Thin Solid Films 191 (1990) 77–89. w12x K.S. Mogensen, N.B. Thomsen, S.S. Eskildsen, C. Mathiasen, J. Bottiger, Surf. Coat. Technol. 99 (1998) 140–146.
w13x N. Popovska, S. Poscher, P. Tichy, G. Emig, H. Ryssel, in: M.D. Allendorf, C. Bernard (Eds.), Chemical Vapor Deposition, Proceedings of The Fourteenth International Conference and EUROCVD-11, 97-25, Electrochem. Soc, Pennington, NJ, 1997, pp. 592–599.
An analysis of the TiN plasmachemical vapor deposition process based on optical emission spectroscopy measurements ¨ b S. Petera,*, H. Giegengacka, F. Richtera, R. Taberskyb, U. Konig a
¨ Chemnitz, 09107 Chemnitz, Germany Technische Universitat b Widia-Valenite, 45021 Essen, Germany
Abstract Optical emission spectroscopy measurements and film deposition experiments were performed for better understanding and control of the plasmachemical vapor deposition (PCVD) process used in the coating production of WidiayValenite. The effects of the titanium chloride flow, the nitrogen flow, the mean discharge current density and the pressure were investigated. The coatings were characterized with respect to deposition rate, composition, crystallographic structure and Vickers microhardness. In particular, the dependence on process parameters of emission signals (linesybands) from 12 species (Ti, Tiq , Cl, Clq, H2, H, q q N2, Nq 2 , N, N , Ar and Ar ) involved into the TiN deposition process was analyzed. Additional spatially resolved OES measurements of the 12 species revealed in particular that the emission of Tiq arises only from near the cathode surface. Based on an empirical model it was shown that the emission of atomic titanium is related to the intrinsic titanium deposition rate whereas the emission of ionized titanium results from the etching of titanium not yet bounded to nitrogen from the surface. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: Optical emission spectroscopy; Plasma diagnostics; Plasmachemical vapor deposition; Titanium nitride
1. Introduction Plasmachemical vapor deposition (PCVD) in pulsed glow discharges is a promising method for the industrial low temperature deposition of hard coatings on threedimensional substrates. Introduced into production for the first time by WIDIA in 1989 w1x, PCVD TiN-coated cemented carbides are used in many metal cutting applications. The PCVD processes for wear resistant TiN coatings are predominantly based on a gas mixture TiCl4–H2–N2–Ar operated in the pressure range of a few hundred Pa. The relations between in-situ measured discharge parameters and the coating properties have been investigated by several authors w2–8x. Non-invasive optical emission spectroscopy (OES) was used in many studies to determine information on the active species and the dominating plasmachemical reactions in the
discharge. However, the results obtained often differed strongly. According to Rie et al. w3x the TiN growth rate showed a direct dependence on intensities of ionized titanium. Rodrigo et al. w6x concluded from their experiments that titanium atoms and N2 (A3Sq u ) metastables are the major species contributing to the supply of activated titanium and nitrogen to the substrate surface. This conclusion was supported by an agreement between the film growth rate and the multiplied emission intensities of the indicated species. Mogensen et al. w7x investigated the PCVD of TiN in a large-scale industrial coating plant. The deposition rate was found to be strongly correlated with the Nq 2 OES signal and not the Tiq signal, when varying the total pressure and the nitrogen flow. In this paper, we present the results of a parametric study, including reliable OES signals from a larger number of emitting species than reported before.
* Corresponding author. Tel.: q49-371-561-8253; fax: q49-371561-8258. E-mail address: [email protected] (S. Peter). 0040-6090/01/$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 0 - 6 0 9 0 Ž 0 1 . 0 1 3 7 5 - X
344
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
Table 1 Standard values and investigated ranges of process parameters Substrate temperature Pressure Total gas flow (H2qN2qAr) Argon flow Nitrogen flow Titanium tetrachloride flow Discharge current density Pulse duration time Pause duration time
5508C 250...300...450 Pa 400 sccm 50 sccm 16...60...80 sccm 2.4...3.6...4.8 sccm (additional) 0.2...0.5...0.7 mAycm2 50 ms 80 ms
2. Experimental The deposition of titanium nitride films and the insitu study of the growth process by optical emission spectroscopy were carried out in a parallel-plate reactor. Hard metal substrates were inserted into the cathode, which was temperature-controlled by an additional resistance heater. The unipolar pulsed power supply was used in the current regulation mode. The PCVD process was investigated using spatially resolved OES. Optical emission was sampled from the negative glow and from the cathode fall region. Spectra were recorded in the spectral region from 250 to 1000 nm with a resolution of approximately 0.1 nm using a LN2-cooled CCD-array. In an earlier study we selected reliable signals (linesybands) of 12 species (Ti, Tiq, q q Cl, Clq, H2, H, N2, Nq 2 , N, N , Ar and Ar ) involved in the TiN deposition process (see Table 2 in Peter et al. w9x). Two different collimators and measuring principles were applied. Both collimators were directed parallel to the cathode surface. The collimator with 15 mm inner diameter was used to find out correlation between process parameters and OES signals. Its large diameter ensures a low sensitivity of signals on changes in the spatial distribution of the optical emission caused by variations of process parameters. In contrast, the distribution of lineyband intensities in the region between the cathode and the negative glow was measured using a collimator with 0.6 mm inner diameter. The movement of the light sampling unit (collimator, quartz lens and quartz fiber) was realized with a directly coupled precision feed-through. More details on the deposition chamber, the deposition process and the OES system are given elsewhere w9 x . To find out the correlation between process parameters, TiN coating characteristics and in-situ measurements, a systematic variation of individual parameters with the others kept constant at standard conditions (see Table 1) was performed. The TiN coating thickness was measured using the ball-grinding method. X-Ray diffraction measurements were made with filtered CuKa radiation. The CuKa2 part was eliminated using the Rachinger-correction.
Fig. 1. Emission signals of (a) nitrogen and argon species and (b) titanium and chlorine species as a function of the precursor flow.
Energy dispersed electron microanalysis measurements were carried out to determine the chlorine and oxygen concentrations in the coatings. 3. Results 3.1. Dependence of the OES-signals on the process parameters Figs. 1–4 show the influence of titanium chloride flow, nitrogen flow, mean discharge current density and
Fig. 2. Emission signals of (a) nitrogen and argon species and (b) titanium and chlorine species as a function of the nitrogen flow.
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
345
3.2. Distribution of the OES-signals within the cathode region For a better understanding of the TiN PCVD process and a reliable interpretation of the relations presented above, the distribution of OES signals between the cathode and the negative glow of the discharge was investigated. Complete spectra were recorded in order to enable the correction for background. In Fig. 5 the distribution of the normalized (mean value equals unity) emission intensity of all 12 measured species is given. A cathode fall length of approximately 8 mm is estimated from this OES measurement. The emission of ionized titanium near the cathode surface is at least one order of magnitude higher than in the negative glow (see Fig. 5a). All other measured neutral atomic radicals (H, N and Cl) show somewhat higher emission intensities close to the cathode as compared with N2, Nq 2 , Nq, H2, Ti, Clq, Ar and Arq. Fig. 3. Emission signals of (a) nitrogen and argon species and (b) titanium and chlorine species as a function of the mean discharge current density.
pressure on the emission intensities of 10 species. The intensities have been normalized to unity for standard process conditions (see Table 1) and averaged for all lines used for a certain species (see Table 2 in Peter et al. w9x). Within the investigated range of parameters we observed an interesting behaviour of several lineyband intensities: ● The emission intensity of argon (constant argon flow in all experiments) depends linearly on the discharge current density (Fig. 3a). It does not change with the precursor flow (Fig. 1a) and changes little with the nitrogen flow (Fig. 2a) and the pressure (Fig. 4a). ● For Ar, N2, Cl and N, the intensity ratios Iwionx y Iwneutralx of ionized and respective neutral species are nearly constant when the process parameters (cf. Table 1) are varied. In contrast to that IwTiqx yIwTix changes strongly with process conditions. ● With increasing precursor flow the Tiq emission increases nearly linearly (Fig. 1b). It decreases with increasing nitrogen concentration (Fig. 2b) and to a lower extent also with increasing pressure (Fig. 4b). As a remarkable fact, the Tiq signal does not increase with the discharge current but passes through a flat maximum (Fig. 3b). ● The signal from atomic titanium rises with about the third power of the discharge current density (Fig. 3b). This indicates a complex formation process. ● Also the signals from chlorine species (Cl and Clq) increase with increasing precursor flow (Fig. 1b) and much stronger with increasing discharge current (Fig. 3b).
3.3. Dependence of the TiN coating characteristics on the process parameters An increasing deposition rate was found for increasing current density, nitrogen flow and pressure. In contrast, with increasing TiCl4 flow the deposition rate decreases. For all the coatings the X-ray diffraction experiments revealed the expected face-centered cubic (NaCl) structure. The lattice constant was determined from the dominating (200) TiN reflection and typically a value of approximately 0.424 nm was found. This is very close to the lattice constant for stoichiometric TiN: a0s 0.42417 nm w10x.
Fig. 4. Emission signals of (a) nitrogen and argon species and (b) titanium and chlorine species as a function of the pressure.
346
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
deposition rate Rat. x of the film forming species (xsTi, N, Cl): Rxat.s
Fig. 5. Distribution of the normalized emission intensities of 12 species within the cathode region (standard parameters).
The chlorine concentration in the TiN coatings varied between 2 and 5 at.%. The oxygen content was F1 at.% for all deposited layers. The microhardness HV0.05 of all coatings was within the range of 2000–2400. A more detailed report about the dependence of the TiN coating characteristics on the process parameters will be published elsewhere. 4. Discussion Due to the specific principles of the generation of optical emission in the discharge, the OES analysis of the PCVD for TiN can give only selected information, mainly about the density of excited species. Considering the behaviour of the argon, atomic nitrogen and atomic hydrogen intensities as well as the behaviour of the q intensity ratios (IwArq x yIwArx, IwNq 2 x yIwN2 x, IwCl x y q IwClx and IwN x yIwNx) one can conclude that within the investigated range of process parameters (see Table 1): ● the electron energy distribution is nearly constant; and ● the electron density is influenced by the discharge current density only. These are favorable conditions for analysis since the measured emission intensities from excited species give information about their ground state densities. To correlate the species density in the gas phase with the deposition characteristics we calculated the atomic
rfdfAcoatNA CA x tdep. C 8 xAMx
(1)
where : rf is the density of the film (assumed to be 5 gycm3 w5x); df is the thickness of the film; tdep. is the deposition time; Acoat is the coated area (cathode area, 240 cm2); NA is Avogadro’s number (6.02=1023 ymol); CA x is the atomic concentration of species x in the A coating (determined under the assumption CA TisCN suggested by the X-ray results); and Mx is the molar mass of x. At the deposition temperature of 5508C typically only a few percent of chlorine supplied with the TiCl4 are incorporated in the TiN films (see also w5,11,12x). Therefore, chlorine deposition is a small loss of a large amount of Cl in the gas phase and cannot be expected to be measured by OES. In contrast, a high portion of the supplied titanium (approx. 1.6=1018 atomsys at standard parameters) is deposited. So we correlated our OES results with the atomic titanium deposition rate. The correlation was done using an empirical model: Rmodel sC1µŽIwTix. ayC2ŽIwTiqx. b∂ŽIwNx. c Ti
(2)
where C1 and C2 are constants and a, b and c are exponents similar to ‘reaction orders’. The model is based on the following assumptions and considerations: ● the Ti emission intensity IwTix is a measure of the free titanium in the gas phase and is correlated with the intrinsic titanium deposition rate; ● the Tiq emission intensity IwTiq x) is related to the removal of titanium from the surface by etching of atoms not yet bonded to nitrogen; and ● nitrogen is the required bond partner of Ti for the TiN deposition. In the investigated parameter range the N2 flow was generally limiting the deposition rate. Although no specific role of atomic nitrogen could be deduced within the present study, its emission intensity, IwNx, was selected as a measure of active nitrogen — because of its high reactivity and, in comparison to Nq or Nq 2 , its high concentration. For another promising reactive species, the NH radical, unfortunately no reliable signal could be selected (for the 360 nm NH emission see Fig. 2 in Peter et al. w9x). In Fig. 6a–d the atomic titanium deposition rate is correlated with the empirical model. The fit parameters are C1s9=1017 atomsys, C2s0.2, as0.15, bs1.0 and cs0.5. It should be mentioned that the reaction order of nitrogen (and hydrogen) in thermal CVD of TiN was also found to be 0.5 w13x.
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
347
the pulsed DC) of their optical emission intensities. The emission of Tiq starts prior to others (see Fig. 5c in Peter et al. w9x). In contrast, the signal from atomic titanium (see Fig. 5b in Peter et al. w9x) was detected later and decreased not so suddenly as the other emissions but similar to the current (see Fig. 4 in Peter et al. w9x). 5. Conclusions
Fig. 6. Comparison of the atomic titanium deposition rate (solid line) with the model based on OES intensities (dotted line) in dependence on the TiCl4 flow (a), the N2 flow (b), the current density (c) and the pressure (d).
A parametric study, including OES signals from 12 emitting involved into the PCVD of TiN was performed. Except in the current series, the discharge was operated in constant current mode. That resulted in nearly constant excitation conditions and thereby the individual influence of the titanium chloride flow, the nitrogen flow, and the pressure could be analyzed. Using an empirical model it was shown, that the combination of the emission signals from atomic titanium, ionized titanium and atomic nitrogen could describe the deposition rate of titanium to a large extent. Atomic nitrogen was selected as the active nitrogen species necessary to deposit TiN, although its specific role could not be revealed within this study. The dependence of the Ti and Tiq signals on the process parameters investigated was specific and often contrary. The emission of atomic titanium was found to be a measure of the intrinsic Ti deposition rate. In contrast, the Tiq signal arises from the etching of titanium not yet bound to nitrogen from the surface. The measured spatial distribution of the Tiq emission is also in agreement with this hypothesis. Further investigations are required in order to understand the special excitation mechanism responsible for the Tiq emission. References
The simple empirical model based on OES signals fits well to the experimental atomic deposition rate. The remaining deviations probably result on the one hand mainly from inaccuracies in determining the atomic deposition rate of titanium (homogeneity of the deposition rate and the stoichiometry across the coated cathode area). On the other hand, the model is very simplified. It does not take into account for example the interaction of energetic ions and neutrals with the growing film. But, it confirms our hypothesis about the interaction of deposition and etching reactions in the PCVD of TiN from TiCl4–H2–N2–Ar mixture. It is worth mentioning that, although in some parameter regions the emission intensities of neutral and ionized titanium show similar trends, their origin and meaning are absolutely different. A further significant difference between Ti and Tiq was observed in the temporal behaviour (over the period of
w1x U. Konig, ¨ R. Tabersky, H. van den Berg, Surf. Coat. Technol. 50 (1991) 57–62. w2x Y. Ishii, T. Shibata, Y. Yoshino, H. Ichimura, K. Kobayashi, J. Ceram. Soc. Jpn. Int. Ed. 100 (1992) 1167–1173. w3x K.T. Rie, A. Gebauer, J. Wohle, ¨ Plasma Chem. Plasma Process. 13 (1) (1993) 93–101. w4x R. Hochreiter, J. Laimer, H. Stori, ¨ D. Heim, Surf. Coat. Technol. 74y75 (1995) 443–449. w5x J. Patscheider, L. Shizhi, S. Vepreck, Plasma Chem. Plasma Process. 16 (3) (1996) 341–362. w6x A.B. Rodrigo, C. Lasorsa, M. Shimozuma, F. Alvarez, P. Perillo, J. Phys. D: Appl. Phys. 30 (1997) 2397–2402. w7x K.S. Mogensen, S.S. Eskildsen, C. Mathiasen, J. Bottiger, Surf. Coat. Technol. 102 (1998) 41–49. w8x M. Eckel, P. Hardt, M. Schmidt, Surf. Coat. Technol. 116–119 (1999) 1037–1041. w9x S. Peter, F. Richter, R. Tabersky, U. Konig, ¨ Thin Solid Films 377y378 (2000) 430–435.
348
S. Peter et al. / Thin Solid Films 398 – 399 (2001) 343–348
w10x JCPDS, International Center for Diffraction Data, Swarthmore, PA, 1992. w11x J. Laimer, H. Stori, ¨ P. Rodhammer, ¨ Thin Solid Films 191 (1990) 77–89. w12x K.S. Mogensen, N.B. Thomsen, S.S. Eskildsen, C. Mathiasen, J. Bottiger, Surf. Coat. Technol. 99 (1998) 140–146.
w13x N. Popovska, S. Poscher, P. Tichy, G. Emig, H. Ryssel, in: M.D. Allendorf, C. Bernard (Eds.), Chemical Vapor Deposition, Proceedings of The Fourteenth International Conference and EUROCVD-11, 97-25, Electrochem. Soc, Pennington, NJ, 1997, pp. 592–599.