- Email: [email protected]
NH3 system
Thin Solid Films 501 (2006) 31 – 34 www.elsevier.com/locate/tsf
H2 dilution effect in the Cat-CVD processes of the SiH4/NH3 system S.G. Ansari a, Hironobu Umemoto a,*, Takashi Morimoto a, Koji Yoneyama a, Akira Izumi b, Atsushi Masuda a,1, Hideki Matsumura a b
a School of Materials Science, Japan Advanced Institute of Science and Technology, Asahidai, Nomi, Ishikawa 923-1292, Japan Department of Electrical, Electronic and Computer Engineering, Kyushu Institute of Technology, Sensui, Tobata, Kitakyushu, Fukuoka 804-8550, Japan
Available online 19 August 2005
Abstract Gas-phase diagnostics in the catalytic chemical vapor deposition processes of the SiH4/NH3/H2 system were carried out to examine the effect of H2 dilution. The decomposition efficiency of NH3 showed a sharp decrease with the introduction of a small amount of SiH4, but this decrease was recovered by the addition of H2 when the NH3 pressure was low. On the other hand, at higher NH3 pressures, the decomposition efficiency showed a minor dependence on the H2 partial pressure. The addition of SiH4 to the NH3 system decreases the H-atom density by one order of magnitude, but this decrease is also recovered by H2 addition. H atoms produced from H2 must re-activate the catalyzer surfaces poisoned by SiH4 when the NH3 pressure is low. D 2005 Elsevier B.V. All rights reserved. Keywords: Chemical vapor deposition; Silicon nitride; Ammonia; Silane
1. Introduction Catalytic chemical vapor deposition (Cat-CVD), often called hot-wire CVD, is one of the most promising techniques for preparing thin amorphous silicon nitride (SiNX ) films at low substrate temperatures using SiH4 and NH3 as material gases [1,2]. SiNX films thus prepared can be used as gas- and water-resistant coatings for organic and inorganic devices and as interlayer insulating films for microelectronic devices. One of the problems in this technique has been the low decomposition efficiency of NH3 in the presence of SiH4. In the absence of SiH4, NH3 can be decomposed to NH2 and H with a decomposition efficiency of more than 50% [3]. However, the decomposition efficiency decreases sharply upon the introduction of a small amount of SiH4 [4]. This decrease has been attributed to the poisoning of the catalyzer surfaces by SiH4. Separating the catalyzers, one to decompose NH3 and another to decompose SiH4, is not easy because the diffusional rate of SiH4 is large under low * Corresponding author. Tel.: +81 761 51 1651; fax: +81 761 51 1149. 1 Present address: Strategic Industrialization Team, Research Center for Photovoltaics, National Institute of Advanced Industrial Science and Technology, Umezono, Tsukuba, Ibaraki 305-8568, Japan. E-mail address: [email protected] (H. Umemoto). 0040-6090/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2005.07.098
pressure conditions, such as those employed in conventional low-pressure CVD processes, and the prevention of catalyzer poisoning is difficult. Recently, it has been found that the addition of H2 improves not only the decomposition efficiency of NH3 in the presence of SiH4 but also the SiNX film quality [5– 7]. For example, Mahan et al. have shown that the content of N atoms in the films increases significantly with H2 dilution for a given NH3/SiH4 gas flow ratio [5]. H2 dilution also causes a reduction in the amount of N – H bonding in SiNX films [5]. Wang et al. have demonstrated that near perfect conformal surface coverage can be obtained on a 100-nm-scale object [6]. In the present work, a systematic study was carried out to determine the catalytic decomposition efficiency of NH3 in the SiH4/NH3/H2 system. Such information is essential for optimizing of the deposition conditions to prepare SiNX conformal films. The absolute H-atom densities were also measured under several conditions.
2. Experimental details The CVD chamber and other experimental apparatus were similar to those described elsewhere [3,4,8 – 12]. A
32
S.G. Ansari et al. / Thin Solid Films 501 (2006) 31 – 34
Decomposition efficiency of NH3
1.0
boxcar average-gated integrator system as a function of the wave number of the laser.
0.8 0.6
3. Results
0.4
3.1. Mass-spectrometric measurements
0.2
Fig. 1 shows the dependence of the decomposition efficiency of NH3 on the H2 flow rate. The NH3 flow rate was kept at 10 sccm, the H2 flow rate was varied from 0 to 90 sccm, whereas the SiH4 flow rate was 0 and 5 sccm. The decomposition efficiency of NH3 in its pure system is 85%, which decreases to 52% with the addition of 5 sccm SiH4. This efficiency recovers up to 80% with the addition of 90 sccm of H2. On the other hand, in the absence of SiH4, the decomposition efficiency shows a slight decrease with the increase in the H2 flow rate. This decrease may be attributed to the change in the retention time, which must depend on the total pressure. As listed in Table 1, in the absence of SiH4, the retention time decreases from 7.3 to 4.6 s, with an increase in the H2 flow rate from 0 to 90 sccm. The decomposition efficiency of SiH4 was more than 93% under all conditions shown in Fig. 1. The effect of H2 dilution was less efficient when the NH3 flow rate was high. Fig. 2 shows the variation in the amount of NH3 decomposed in unit time as a function of the NH3 flow rate, in the presence and absence of H2 (90 sccm) and SiH4 (5 sccm). The decomposition efficiency in the pure NH3 system was 85%, independent of the NH3 flow rate, below 35 sccm. This efficiency decreased with the addition of SiH4. The amount of decomposed NH3 shows saturation against the NH3 flow rate in the presence of SiH4. This decrease is recovered, although not completely, by the addition of H2 when the NH3 flow rate is below 30 sccm. At higher NH3 flow rates, however, the effect of H2 dilution disappears. The decomposition efficiencies as well as the amounts of material gases decomposed under typical conditions are summarized in Table 1.
0
20
40 60 H2 flow rate / sccm
80
100
Fig. 1. Dependence of the decomposition efficiency of NH3 on H2 flow rate in the presence (?) and in the absence (‚) of SiH4. The flow rates of NH3 and SiH4 were 10 and 5 sccm, respectively.
commercial quadrupole mass-spectrometer (Anelva, MQA200TS) was used to measure the absolute steady-state densities of the stable molecules, such as NH3 and SiH4. The decomposition efficiency of the material gas molecules can be obtained by comparing the signal intensities under catalyzer-heated and unheated conditions. The mass spectrometer was attached to the chamber through a sampling hole (0.4 mm in diameter), and was differentially pumped. The length of the catalyzer (0.5-mm-diameter W wire, Nilaco 99.95%) was 120 cm. The catalyzer temperature was fixed at 2270 K. The absolute densities of H atoms in the gas phase were determined by a vacuum-ultraviolet laser absorption technique [3,11]. The output of a dye laser at 729.6 nm was doubled in frequency by a h-BaB2O4 crystal and then tripled by a mixture of Kr and Ar to produce Lyman-a light at 121.6 nm. The typical pulse energy of the laser at the doubled stage (364.8 nm) was 10 mJ, which corresponds to 2.5 MW at the peak power. After passing through the CVD chamber, the Lyman-a laser beam entered a detection vessel filled with NO. The NO+ ion current was measured with a
Table 1 Decomposition efficiencies and the absolute amounts of decomposition in unit time under various gas-flow conditions Condition
a
1 2a 3 4 5 6 7 8 9 10 a
Flow rate (sccm)
Total pressure (Pa)
NH3
SiH4
H2
500 500 10 10 10 10 30 30 30 30
0 10 0 5 5 0 0 5 5 0
0 0 0 0 90 90 0 0 90 90
Umemoto et al. [4].
20 20 2.7 5.6 19 17 11 14 26 23
Decomposition efficiency (%)
Amount of decomposition in unit time (10 5 mol s 1)
NH3
NH3
50 5 85 52 80 80 87 41 40 80
SiH4 60 95 93
96 98
17 1.7 0.6 0.4 0.6 0.6 1.8 0.9 0.8 1.7
Retention time (s)
SiH4 0.4 0.3 0.3
0.3 0.3
1.1 1.1 7.3 10.2 4.9 4.6 10.0 10.9 5.7 5.2
2.0 1.5 1.0 0.5
0
10
20 30 NH3 flow rate / sccm
Fig. 2. The amount of NH3 decomposed in unit time as a function of NH3 flow rate in pure NH3 (>), NH3/SiH4 (g) and NH3/SiH4/H2 (?) systems. The flow rates of H2 and SiH4 were 90 and 5 sccm, respectively.
3.2. Vacuum-ultraviolet laser absorption measurements of H atoms The absorption spectra of H atoms are illustrated in Fig. 3. Unfortunately, the dynamic range of the absorption technique is not wide. It was impossible to measure the Hatom density under the same conditions as those for mass spectrometric measurements because the optical density was too thick. In addition, SiH4 strongly absorbs vacuumultraviolet light around 122 nm [13,14]. As a result, the catalyzer length was reduced to 24 cm while the SiH4 flow rate was kept at 1 sccm. The catalyzer temperature was the same as that employed in the mass spectrometric study, 2270 K. The distance between the catalyzer and the detection zone was 10 cm. From the minimum transmittances of the spectra, the absolute densities of H atoms are evaluated to be 2.4 1010, 1.7 1011 and 2.5 1011 cm 3, respectively, under the three conditions shown in Fig. 3: (a) NH3 (10 sccm)/SiH4 (1 sccm), (b) NH3 (10 sccm) and (c) NH3 (10 sccm)/SiH4 (1 sccm)/H2 (90 sccm). An addition of 1 sccm of SiH4 decreases the H-atom density by one order. This decrease is recovered by adding 90 sccm of H2. The effect of SiH4 addition on the decrease in H-atom density is much larger than that on the decomposition efficiency of NH3. This difference suggests the importance of the H-atom loss processes in the gas phase, such as H + SiH4 Y H2 + SiH3, as well as those on the substrate or chamber wall surfaces.
4. Discussion Mahan et al. as well as Liu et al. have recently suggested that H atoms produced by the catalytic decomposition of H2 contribute to efficient NH3 dissociation [5,7]. In the absence of SiH4, however, the decomposition efficiency of NH3 showed only a minor dependence on the H2 pressure, which strongly suggests that the reaction between H atoms and NH3 molecules, H + NH3 Y NH2 + H2, is of minor importance. This is reasonable since this reaction is 17 kJ mol 1
33
endothermic [15] and its activation energy is as large as 58 kJ mol 1 [16]. As shown in Fig. 2, when the NH3 flow rate is less than 30 sccm, the amount of NH3 decomposed in the presence of SiH4 increases with the introduction of H2, although the retention time decreases. SiH4 poisons the tungsten surfaces to decompose NH3, but the H atoms produced by the catalytic decomposition of H2 act as cleaning agents for Sicontaminated tungsten surfaces to restore the NH3 decomposition efficiency. Since H atoms etch silicon compounds from metal chamber walls to eject SiH4 [8], a similar etching may take place on heated tungsten surfaces. The H-atom density decreases when SiH4 is added to H2 [10]. In this case, we have attributed this decrease to the Hatom loss processes on chamber walls coated with silicon compounds. This is because the time responses in both the decrease and the increase were slow, taking several minutes for the change. On the other hand, in the present NH3/SiH4 system, the response was much faster. Therefore, the decrease in the decomposition efficiency should rather be attributed to the poisoning of the catalyzer surfaces by SiH4. Three problems remain to be addressed. The first is the asymptotic behavior of the NH3 decomposition efficiency against the SiH4 flow rate, which has been discussed in our previous publication [4]. When the NH3 flow rate is 500 sccm, the NH3 decomposition efficiency levels off at 5% when the SiH4 flow rate is more than 5 sccm. We have explained this independence by the presence of more than two kinds of active sites on the catalyzer surfaces, but will present another possible explanation here, taking into account the NH3 flow rate dependence. The second is the NH3 flow rate (pressure) dependence of the NH3 decomposition efficiency in the presence of SiH4. Table 1 shows that the decomposition efficiency is 52% when the flow rates of NH3 and SiH4 are 10 and 5 sccm, respectively, but decreases to 41% when the NH3 flow rate is 30 sccm, although the retention time increases. The third problem is why the decomposition efficiency is not improved by the addition of H2 when the NH3 flow rate is more than 30 sccm.
1.0 transmittance
NH3 decomposed / 10-5 mol s-1
S.G. Ansari et al. / Thin Solid Films 501 (2006) 31 – 34
0.8
(a)
0.6 0.4 0.2 0.0 82257
(b)
(c)
82258 82259 82260 wavenumber / cm-1
82261
Fig. 3. Vacuum-ultraviolet laser absorption spectra of H-atoms. (a) NH3 (10 sccm)/SiH4 (1 sccm), 3.5 Pa. (b) NH3 (10 sccm), 3.2 Pa. (c) NH3 (10 sccm)/ SiH4 (1 sccm)/H2 (90 sccm), 4.7 Pa.
34
S.G. Ansari et al. / Thin Solid Films 501 (2006) 31 – 34
One of the possible explanations for the first problem is that the decomposition of NH3 takes place not only on heated tungsten surfaces but also on tungsten silicide, although the efficiency is less. When the catalyzer is completely covered with silicide, no more decrease in the decomposition efficiency should be observed with an increase in the SiH4 flow rate. A similar explanation may be possible for the second problem. The NH3 decomposition efficiency on surfaces with silicide may be less than on bare tungsten surfaces but higher than that for silicide covered with NH3. In other words, the decomposition on the second layer should be less than that on the first layer. Then, the total decomposition efficiency may decrease with the increase in the NH3 pressure (flow rate). The residence time of NH3 on tungsten surfaces is much shorter than that of SiH4 and the saturation in NH3 decomposition cannot be observed in the absence of SiH4 when the NH3 flow rate is less than 30 sccm. It is more difficult to answer the final problem. NH3 adsorbed on silicide formed on tungsten surfaces may prevent the attack of H atoms to break Si –W bonds. It has been demonstrated that the addition of H2 not only enhances the decomposition efficiency of NH3 but also improves both SiNX film quality and step coverage [5– 7]. The decrease in the H-atom content in the SiNX films can be attributed to the presence of an excess amount of H atoms in the gas phase, which may abstract H atoms from the SiNX growing surfaces. There could be several explanations for the improvement in step coverage in the H2 added system. One is the competition between deposition and etching. Under conventional conditions, such as Condition 2 in Table 1, the decomposition efficiency of NH3 is much smaller than that in the presence of H2 (Condition 5 or 9), but the amount of NH3 decomposed in unit time is larger. Accordingly, the deposition rate is larger under conventional conditions. On the other hand, etching is slower because the H-atom density is smaller. Step coverage may be controlled by a delicate balance between deposition and etching. Another possible explanation is based on local heating. The substrate surfaces may be heated locally by the H-atom recombination processes when the H-atom density in the gas phase is high [17]. Then, the precursor radicals, such as SiH3, may migrate more smoothly on surfaces and the coverage will become more conformal.
5. Conclusion The catalytic decomposition efficiency of NH3 in the presence of SiH4 increases with the addition of H2 when the
NH3 pressure is low. This can be explained by the reactivation of the catalyzer surfaces by H atoms produced by the catalytic decomposition of H2. The H2 dilution effect is less clear when the NH3 pressure is high. NH3 molecules adsorbed on catalyzer surfaces may prevent the re-activation of the catalyzer. H atoms produced by the catalytic decomposition of H2 may also result in the improvement of the film quality.
Acknowledgements This work is supported in part by the Investigation for Innovative PV Technology from the New Energy and Industrial Technology Development Organization. S.G. Ansari acknowledges a postdoctoral fellowship from the Japan Society for the Promotion of Science.
References [1] H. Matsumura, Thin Solid Films 395 (2001) 1. [2] S. Osono, Y. Uchiyama, M. Kitazoe, K. Saito, M. Hayama, A. Masuda, A. Izumi, H. Matsumura, Thin Solid Films 430 (2003) 165. [3] H. Umemoto, K. Ohara, D. Morita, T. Morimoto, M. Yamawaki, A. Masuda, H. Matsumura, Jpn. J. Appl. Phys. 42 (2003) 5315. [4] H. Umemoto, T. Morimoto, M. Yamawaki, Y. Masuda, A. Masuda, H. Matsumura, Thin Solid Films 430 (2003) 24. [5] A.H. Mahan, A.C. Dillon, L.M. Gedvilas, D.L. Williamson, J.D. Perkins, J. Appl. Phys. 94 (2003) 2360. [6] Q. Wang, S. Ward, L. Gedvilas, B. Keyes, E. Sanchez, S. Wang, Appl. Phys. Lett. 84 (2004) 338. [7] F. Liu, S. Ward, L. Gedvilas, B. Keyes, B. To, Q. Wang, E. Sanchez, S. Wang, J. Appl. Phys. 96 (2004) 2973. [8] Y. Nozaki, K. Kongo, T. Miyazaki, M. Kitazoe, K. Horii, H. Umemoto, A. Masuda, H. Matsumura, J. Appl. Phys. 88 (2000) 5437. [9] Y. Nozaki, M. Kitazoe, K. Horii, H. Umemoto, A. Masuda, H. Matsumura, Thin Solid Films 395 (2001) 47. [10] H. Umemoto, Y. Nozaki, M. Kitazoe, K. Horii, K. Ohara, D. Morita, K. Uchida, Y. Ishibashi, M. Komoda, K. Kamesaki, A. Izumi, A. Masuda, H. Matsumura, J. Non-Cryst. Solids 299 – 302 (2002) 9. [11] H. Umemoto, K. Ohara, D. Morita, Y. Nozaki, A. Masuda, H. Matsumura, J. Appl. Phys. 91 (2002) 1650. [12] H. Umemoto, T. Morimoto, M. Yamawaki, Y. Masuda, A. Masuda, H. Matsumura, J. Non-Cryst. Solids 338 – 340 (2004) 65. [13] M. Suto, L.C. Lee, J. Chem. Phys. 84 (1986) 1160. [14] U. Itoh, Y. Toyoshima, H. Onuki, N. Washida, T. Ibuki, J. Chem. Phys. 85 (1986) 4867. [15] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, 80th edR, Chemical Rubber Corp, Boca Raton, FL, 1999 – 2000. [16] T. Ko, P. Marshall, A. Fontijn, J. Phys. Chem. 94 (1990) 1401. [17] A. Matsuda, Thin Solid Films 337 (1999) 1.
H2 dilution effect in the Cat-CVD processes of the SiH4/NH3 system S.G. Ansari a, Hironobu Umemoto a,*, Takashi Morimoto a, Koji Yoneyama a, Akira Izumi b, Atsushi Masuda a,1, Hideki Matsumura a b
a School of Materials Science, Japan Advanced Institute of Science and Technology, Asahidai, Nomi, Ishikawa 923-1292, Japan Department of Electrical, Electronic and Computer Engineering, Kyushu Institute of Technology, Sensui, Tobata, Kitakyushu, Fukuoka 804-8550, Japan
Available online 19 August 2005
Abstract Gas-phase diagnostics in the catalytic chemical vapor deposition processes of the SiH4/NH3/H2 system were carried out to examine the effect of H2 dilution. The decomposition efficiency of NH3 showed a sharp decrease with the introduction of a small amount of SiH4, but this decrease was recovered by the addition of H2 when the NH3 pressure was low. On the other hand, at higher NH3 pressures, the decomposition efficiency showed a minor dependence on the H2 partial pressure. The addition of SiH4 to the NH3 system decreases the H-atom density by one order of magnitude, but this decrease is also recovered by H2 addition. H atoms produced from H2 must re-activate the catalyzer surfaces poisoned by SiH4 when the NH3 pressure is low. D 2005 Elsevier B.V. All rights reserved. Keywords: Chemical vapor deposition; Silicon nitride; Ammonia; Silane
1. Introduction Catalytic chemical vapor deposition (Cat-CVD), often called hot-wire CVD, is one of the most promising techniques for preparing thin amorphous silicon nitride (SiNX ) films at low substrate temperatures using SiH4 and NH3 as material gases [1,2]. SiNX films thus prepared can be used as gas- and water-resistant coatings for organic and inorganic devices and as interlayer insulating films for microelectronic devices. One of the problems in this technique has been the low decomposition efficiency of NH3 in the presence of SiH4. In the absence of SiH4, NH3 can be decomposed to NH2 and H with a decomposition efficiency of more than 50% [3]. However, the decomposition efficiency decreases sharply upon the introduction of a small amount of SiH4 [4]. This decrease has been attributed to the poisoning of the catalyzer surfaces by SiH4. Separating the catalyzers, one to decompose NH3 and another to decompose SiH4, is not easy because the diffusional rate of SiH4 is large under low * Corresponding author. Tel.: +81 761 51 1651; fax: +81 761 51 1149. 1 Present address: Strategic Industrialization Team, Research Center for Photovoltaics, National Institute of Advanced Industrial Science and Technology, Umezono, Tsukuba, Ibaraki 305-8568, Japan. E-mail address: [email protected] (H. Umemoto). 0040-6090/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2005.07.098
pressure conditions, such as those employed in conventional low-pressure CVD processes, and the prevention of catalyzer poisoning is difficult. Recently, it has been found that the addition of H2 improves not only the decomposition efficiency of NH3 in the presence of SiH4 but also the SiNX film quality [5– 7]. For example, Mahan et al. have shown that the content of N atoms in the films increases significantly with H2 dilution for a given NH3/SiH4 gas flow ratio [5]. H2 dilution also causes a reduction in the amount of N – H bonding in SiNX films [5]. Wang et al. have demonstrated that near perfect conformal surface coverage can be obtained on a 100-nm-scale object [6]. In the present work, a systematic study was carried out to determine the catalytic decomposition efficiency of NH3 in the SiH4/NH3/H2 system. Such information is essential for optimizing of the deposition conditions to prepare SiNX conformal films. The absolute H-atom densities were also measured under several conditions.
2. Experimental details The CVD chamber and other experimental apparatus were similar to those described elsewhere [3,4,8 – 12]. A
32
S.G. Ansari et al. / Thin Solid Films 501 (2006) 31 – 34
Decomposition efficiency of NH3
1.0
boxcar average-gated integrator system as a function of the wave number of the laser.
0.8 0.6
3. Results
0.4
3.1. Mass-spectrometric measurements
0.2
Fig. 1 shows the dependence of the decomposition efficiency of NH3 on the H2 flow rate. The NH3 flow rate was kept at 10 sccm, the H2 flow rate was varied from 0 to 90 sccm, whereas the SiH4 flow rate was 0 and 5 sccm. The decomposition efficiency of NH3 in its pure system is 85%, which decreases to 52% with the addition of 5 sccm SiH4. This efficiency recovers up to 80% with the addition of 90 sccm of H2. On the other hand, in the absence of SiH4, the decomposition efficiency shows a slight decrease with the increase in the H2 flow rate. This decrease may be attributed to the change in the retention time, which must depend on the total pressure. As listed in Table 1, in the absence of SiH4, the retention time decreases from 7.3 to 4.6 s, with an increase in the H2 flow rate from 0 to 90 sccm. The decomposition efficiency of SiH4 was more than 93% under all conditions shown in Fig. 1. The effect of H2 dilution was less efficient when the NH3 flow rate was high. Fig. 2 shows the variation in the amount of NH3 decomposed in unit time as a function of the NH3 flow rate, in the presence and absence of H2 (90 sccm) and SiH4 (5 sccm). The decomposition efficiency in the pure NH3 system was 85%, independent of the NH3 flow rate, below 35 sccm. This efficiency decreased with the addition of SiH4. The amount of decomposed NH3 shows saturation against the NH3 flow rate in the presence of SiH4. This decrease is recovered, although not completely, by the addition of H2 when the NH3 flow rate is below 30 sccm. At higher NH3 flow rates, however, the effect of H2 dilution disappears. The decomposition efficiencies as well as the amounts of material gases decomposed under typical conditions are summarized in Table 1.
0
20
40 60 H2 flow rate / sccm
80
100
Fig. 1. Dependence of the decomposition efficiency of NH3 on H2 flow rate in the presence (?) and in the absence (‚) of SiH4. The flow rates of NH3 and SiH4 were 10 and 5 sccm, respectively.
commercial quadrupole mass-spectrometer (Anelva, MQA200TS) was used to measure the absolute steady-state densities of the stable molecules, such as NH3 and SiH4. The decomposition efficiency of the material gas molecules can be obtained by comparing the signal intensities under catalyzer-heated and unheated conditions. The mass spectrometer was attached to the chamber through a sampling hole (0.4 mm in diameter), and was differentially pumped. The length of the catalyzer (0.5-mm-diameter W wire, Nilaco 99.95%) was 120 cm. The catalyzer temperature was fixed at 2270 K. The absolute densities of H atoms in the gas phase were determined by a vacuum-ultraviolet laser absorption technique [3,11]. The output of a dye laser at 729.6 nm was doubled in frequency by a h-BaB2O4 crystal and then tripled by a mixture of Kr and Ar to produce Lyman-a light at 121.6 nm. The typical pulse energy of the laser at the doubled stage (364.8 nm) was 10 mJ, which corresponds to 2.5 MW at the peak power. After passing through the CVD chamber, the Lyman-a laser beam entered a detection vessel filled with NO. The NO+ ion current was measured with a
Table 1 Decomposition efficiencies and the absolute amounts of decomposition in unit time under various gas-flow conditions Condition
a
1 2a 3 4 5 6 7 8 9 10 a
Flow rate (sccm)
Total pressure (Pa)
NH3
SiH4
H2
500 500 10 10 10 10 30 30 30 30
0 10 0 5 5 0 0 5 5 0
0 0 0 0 90 90 0 0 90 90
Umemoto et al. [4].
20 20 2.7 5.6 19 17 11 14 26 23
Decomposition efficiency (%)
Amount of decomposition in unit time (10 5 mol s 1)
NH3
NH3
50 5 85 52 80 80 87 41 40 80
SiH4 60 95 93
96 98
17 1.7 0.6 0.4 0.6 0.6 1.8 0.9 0.8 1.7
Retention time (s)
SiH4 0.4 0.3 0.3
0.3 0.3
1.1 1.1 7.3 10.2 4.9 4.6 10.0 10.9 5.7 5.2
2.0 1.5 1.0 0.5
0
10
20 30 NH3 flow rate / sccm
Fig. 2. The amount of NH3 decomposed in unit time as a function of NH3 flow rate in pure NH3 (>), NH3/SiH4 (g) and NH3/SiH4/H2 (?) systems. The flow rates of H2 and SiH4 were 90 and 5 sccm, respectively.
3.2. Vacuum-ultraviolet laser absorption measurements of H atoms The absorption spectra of H atoms are illustrated in Fig. 3. Unfortunately, the dynamic range of the absorption technique is not wide. It was impossible to measure the Hatom density under the same conditions as those for mass spectrometric measurements because the optical density was too thick. In addition, SiH4 strongly absorbs vacuumultraviolet light around 122 nm [13,14]. As a result, the catalyzer length was reduced to 24 cm while the SiH4 flow rate was kept at 1 sccm. The catalyzer temperature was the same as that employed in the mass spectrometric study, 2270 K. The distance between the catalyzer and the detection zone was 10 cm. From the minimum transmittances of the spectra, the absolute densities of H atoms are evaluated to be 2.4 1010, 1.7 1011 and 2.5 1011 cm 3, respectively, under the three conditions shown in Fig. 3: (a) NH3 (10 sccm)/SiH4 (1 sccm), (b) NH3 (10 sccm) and (c) NH3 (10 sccm)/SiH4 (1 sccm)/H2 (90 sccm). An addition of 1 sccm of SiH4 decreases the H-atom density by one order. This decrease is recovered by adding 90 sccm of H2. The effect of SiH4 addition on the decrease in H-atom density is much larger than that on the decomposition efficiency of NH3. This difference suggests the importance of the H-atom loss processes in the gas phase, such as H + SiH4 Y H2 + SiH3, as well as those on the substrate or chamber wall surfaces.
4. Discussion Mahan et al. as well as Liu et al. have recently suggested that H atoms produced by the catalytic decomposition of H2 contribute to efficient NH3 dissociation [5,7]. In the absence of SiH4, however, the decomposition efficiency of NH3 showed only a minor dependence on the H2 pressure, which strongly suggests that the reaction between H atoms and NH3 molecules, H + NH3 Y NH2 + H2, is of minor importance. This is reasonable since this reaction is 17 kJ mol 1
33
endothermic [15] and its activation energy is as large as 58 kJ mol 1 [16]. As shown in Fig. 2, when the NH3 flow rate is less than 30 sccm, the amount of NH3 decomposed in the presence of SiH4 increases with the introduction of H2, although the retention time decreases. SiH4 poisons the tungsten surfaces to decompose NH3, but the H atoms produced by the catalytic decomposition of H2 act as cleaning agents for Sicontaminated tungsten surfaces to restore the NH3 decomposition efficiency. Since H atoms etch silicon compounds from metal chamber walls to eject SiH4 [8], a similar etching may take place on heated tungsten surfaces. The H-atom density decreases when SiH4 is added to H2 [10]. In this case, we have attributed this decrease to the Hatom loss processes on chamber walls coated with silicon compounds. This is because the time responses in both the decrease and the increase were slow, taking several minutes for the change. On the other hand, in the present NH3/SiH4 system, the response was much faster. Therefore, the decrease in the decomposition efficiency should rather be attributed to the poisoning of the catalyzer surfaces by SiH4. Three problems remain to be addressed. The first is the asymptotic behavior of the NH3 decomposition efficiency against the SiH4 flow rate, which has been discussed in our previous publication [4]. When the NH3 flow rate is 500 sccm, the NH3 decomposition efficiency levels off at 5% when the SiH4 flow rate is more than 5 sccm. We have explained this independence by the presence of more than two kinds of active sites on the catalyzer surfaces, but will present another possible explanation here, taking into account the NH3 flow rate dependence. The second is the NH3 flow rate (pressure) dependence of the NH3 decomposition efficiency in the presence of SiH4. Table 1 shows that the decomposition efficiency is 52% when the flow rates of NH3 and SiH4 are 10 and 5 sccm, respectively, but decreases to 41% when the NH3 flow rate is 30 sccm, although the retention time increases. The third problem is why the decomposition efficiency is not improved by the addition of H2 when the NH3 flow rate is more than 30 sccm.
1.0 transmittance
NH3 decomposed / 10-5 mol s-1
S.G. Ansari et al. / Thin Solid Films 501 (2006) 31 – 34
0.8
(a)
0.6 0.4 0.2 0.0 82257
(b)
(c)
82258 82259 82260 wavenumber / cm-1
82261
Fig. 3. Vacuum-ultraviolet laser absorption spectra of H-atoms. (a) NH3 (10 sccm)/SiH4 (1 sccm), 3.5 Pa. (b) NH3 (10 sccm), 3.2 Pa. (c) NH3 (10 sccm)/ SiH4 (1 sccm)/H2 (90 sccm), 4.7 Pa.
34
S.G. Ansari et al. / Thin Solid Films 501 (2006) 31 – 34
One of the possible explanations for the first problem is that the decomposition of NH3 takes place not only on heated tungsten surfaces but also on tungsten silicide, although the efficiency is less. When the catalyzer is completely covered with silicide, no more decrease in the decomposition efficiency should be observed with an increase in the SiH4 flow rate. A similar explanation may be possible for the second problem. The NH3 decomposition efficiency on surfaces with silicide may be less than on bare tungsten surfaces but higher than that for silicide covered with NH3. In other words, the decomposition on the second layer should be less than that on the first layer. Then, the total decomposition efficiency may decrease with the increase in the NH3 pressure (flow rate). The residence time of NH3 on tungsten surfaces is much shorter than that of SiH4 and the saturation in NH3 decomposition cannot be observed in the absence of SiH4 when the NH3 flow rate is less than 30 sccm. It is more difficult to answer the final problem. NH3 adsorbed on silicide formed on tungsten surfaces may prevent the attack of H atoms to break Si –W bonds. It has been demonstrated that the addition of H2 not only enhances the decomposition efficiency of NH3 but also improves both SiNX film quality and step coverage [5– 7]. The decrease in the H-atom content in the SiNX films can be attributed to the presence of an excess amount of H atoms in the gas phase, which may abstract H atoms from the SiNX growing surfaces. There could be several explanations for the improvement in step coverage in the H2 added system. One is the competition between deposition and etching. Under conventional conditions, such as Condition 2 in Table 1, the decomposition efficiency of NH3 is much smaller than that in the presence of H2 (Condition 5 or 9), but the amount of NH3 decomposed in unit time is larger. Accordingly, the deposition rate is larger under conventional conditions. On the other hand, etching is slower because the H-atom density is smaller. Step coverage may be controlled by a delicate balance between deposition and etching. Another possible explanation is based on local heating. The substrate surfaces may be heated locally by the H-atom recombination processes when the H-atom density in the gas phase is high [17]. Then, the precursor radicals, such as SiH3, may migrate more smoothly on surfaces and the coverage will become more conformal.
5. Conclusion The catalytic decomposition efficiency of NH3 in the presence of SiH4 increases with the addition of H2 when the
NH3 pressure is low. This can be explained by the reactivation of the catalyzer surfaces by H atoms produced by the catalytic decomposition of H2. The H2 dilution effect is less clear when the NH3 pressure is high. NH3 molecules adsorbed on catalyzer surfaces may prevent the re-activation of the catalyzer. H atoms produced by the catalytic decomposition of H2 may also result in the improvement of the film quality.
Acknowledgements This work is supported in part by the Investigation for Innovative PV Technology from the New Energy and Industrial Technology Development Organization. S.G. Ansari acknowledges a postdoctoral fellowship from the Japan Society for the Promotion of Science.
References [1] H. Matsumura, Thin Solid Films 395 (2001) 1. [2] S. Osono, Y. Uchiyama, M. Kitazoe, K. Saito, M. Hayama, A. Masuda, A. Izumi, H. Matsumura, Thin Solid Films 430 (2003) 165. [3] H. Umemoto, K. Ohara, D. Morita, T. Morimoto, M. Yamawaki, A. Masuda, H. Matsumura, Jpn. J. Appl. Phys. 42 (2003) 5315. [4] H. Umemoto, T. Morimoto, M. Yamawaki, Y. Masuda, A. Masuda, H. Matsumura, Thin Solid Films 430 (2003) 24. [5] A.H. Mahan, A.C. Dillon, L.M. Gedvilas, D.L. Williamson, J.D. Perkins, J. Appl. Phys. 94 (2003) 2360. [6] Q. Wang, S. Ward, L. Gedvilas, B. Keyes, E. Sanchez, S. Wang, Appl. Phys. Lett. 84 (2004) 338. [7] F. Liu, S. Ward, L. Gedvilas, B. Keyes, B. To, Q. Wang, E. Sanchez, S. Wang, J. Appl. Phys. 96 (2004) 2973. [8] Y. Nozaki, K. Kongo, T. Miyazaki, M. Kitazoe, K. Horii, H. Umemoto, A. Masuda, H. Matsumura, J. Appl. Phys. 88 (2000) 5437. [9] Y. Nozaki, M. Kitazoe, K. Horii, H. Umemoto, A. Masuda, H. Matsumura, Thin Solid Films 395 (2001) 47. [10] H. Umemoto, Y. Nozaki, M. Kitazoe, K. Horii, K. Ohara, D. Morita, K. Uchida, Y. Ishibashi, M. Komoda, K. Kamesaki, A. Izumi, A. Masuda, H. Matsumura, J. Non-Cryst. Solids 299 – 302 (2002) 9. [11] H. Umemoto, K. Ohara, D. Morita, Y. Nozaki, A. Masuda, H. Matsumura, J. Appl. Phys. 91 (2002) 1650. [12] H. Umemoto, T. Morimoto, M. Yamawaki, Y. Masuda, A. Masuda, H. Matsumura, J. Non-Cryst. Solids 338 – 340 (2004) 65. [13] M. Suto, L.C. Lee, J. Chem. Phys. 84 (1986) 1160. [14] U. Itoh, Y. Toyoshima, H. Onuki, N. Washida, T. Ibuki, J. Chem. Phys. 85 (1986) 4867. [15] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, 80th edR, Chemical Rubber Corp, Boca Raton, FL, 1999 – 2000. [16] T. Ko, P. Marshall, A. Fontijn, J. Phys. Chem. 94 (1990) 1401. [17] A. Matsuda, Thin Solid Films 337 (1999) 1.