Chemical bonding and composition of silicon nitride films prepared by inductively coupled plasma chemical vapor deposition

Chemical bonding and composition of silicon nitride films prepared by inductively coupled plasma chemical vapor deposition

Surface & Coatings Technology 204 (2010) 2923–2927 Contents lists available at ScienceDirect Surface & Coatings Technology j o u r n a l h o m e p a...

560KB Sizes 0 Downloads 26 Views

Recommend Documents

No documents
Surface & Coatings Technology 204 (2010) 2923–2927

Contents lists available at ScienceDirect

Surface & Coatings Technology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s u r f c o a t

Chemical bonding and composition of silicon nitride films prepared by inductively coupled plasma chemical vapor deposition M. Matsuoka a,⁎, S. Isotani a, W. Sucasaire a, L.S. Zambom b, K. Ogata c a b c

Institute of Physics, University of São Paulo, Rua do Matão, Travessa R, 187, CEP 05508-090, São Paulo, SP, Brazil Faculdade de Tecnologia de São Paulo — CEETEPS, Praça Coronel Fernando Prestes, 30, CEP 01124-060, São Paulo, SP, Brazil Nissin Electric Company, Ltd., 47, Umezu-Takase-cho, Ukyo-ku, Kyoto 615-8686, Japan

a r t i c l e

i n f o

Available online 4 March 2010 Keywords: Silicon nitride coating Chemical vapor deposition X-ray photoelectron spectroscopy

a b s t r a c t Thin silicon nitride films were prepared at 350 °C by inductively coupled plasma chemical vapor deposition on Si(100) substrates under different NH3/SiH4 or N2/SiH4 gas mixture. The chemical composition and bonding structure of the deposited films were investigated as a function of the process parameters, such as the gas flow ratio NH3/SiH4 or N2/SiH4 and the RF power, using X-ray photoelectron spectroscopy (XPS). The gas flow ratio was 1.4, 4.3, 7.2 or 9.5 and the RF power, 50 or 100 W. Decomposition results of Si 2p XPS spectra indicated the presence of bulk Si, under-stoichiometric nitride, stoichiometric nitride Si3N4, oxynitride SiNxOy, and stoichiometric oxide SiO2, and the amounts of these compounds were strongly influenced by the two process parameters. These results were consistent with those obtained from N 1s XPS spectra. The chemical composition ratio N/Si in the film increased with increasing the gas flow ratio until the gas flow ratio reached 4.3, reflecting the high reactivity of nitrogen, and stayed almost constant for further increase in gas flow ratio, the excess nitrogen being rejected from the growing film. A considerable and unexpected incorporation of contaminant oxygen and carbon into the depositing film was observed and attributed to their high chemical reactivity. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Thin silicon nitride films are insulators and of great technological importance due to their potential applications in passivation layers, diffusion barriers, and dielectric materials in microelectronic and optoelectronic devices, being incorporated at the Si/SiO2 interface formed in Si-based transistors and memory devices [1,2]. This incorporation as gate dielectrics is known to improve the structural and electrical quality of the interface, resulting in the suppression of dopant diffusion across the interface as well as low leakage current. Furthermore, silicon nitride has been currently used in a number of industry applications, such as a high-temperature structural material and surface micromachining, because of high-temperature strength, good fracture toughness, excellent wear and oxidation resistance [3,4]. On the other hand, silicon oxynitride (SiNxOy) films, i.e., SiO2 films with nitrogen incorporated, are also used widely as a dielectric material, exhibiting remarkable advantages in gate dielectrics over the normal silicon dioxide [5–8]. These films have been prepared by some methods, such as nitridation and/or oxidation [7,9–11], reactive sputtering [2,5,12,13], plasma immersion ion implantation (PIII) [14] and chemical vapor

⁎ Corresponding author. Tel.: + 55 11 3091 6969; fax: + 55 11 3031 2742. E-mail address: [email protected] (M. Matsuoka). 0257-8972/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2010.02.071

deposition (CVD). Besides conventional low-pressure CVD [15], plasma enhanced CVD has been intensively used in the manufacturing processes through electron cyclotron resonance [6] and inductively coupled plasmas [16]. These films have been extensively investigated in recent years; however, some important aspects of these materials such as the bonding structure and composition of the nitride and oxynitride materials remain poorly understood to our knowledge, which would prevent the further improvement of the functional properties of these materials. The present study reports the deposition of silicon nitride on Si(100) substrates by inductively coupled plasmas CVD (ICP-CVD), using different gas mixture (NH3 and SiH4, or N2 and SiH4), as a function of two process parameters; the flow ratio of process gases, NH3/SiH4 or N2/SiH4, and radiofrequency (RF) power. The purpose of this study is to investigate the deposited films by means of X-ray photoelectron spectroscopy (XPS), to gain the chemical composition and bonding structure in the deposited films from the XPS spectra and to relate this chemical information to the two process parameters. 2. Experimental The silicon nitride films were synthesized in a CVD reactor at 350 °C. The reactor consists of a quartz chamber of 150 mm in diameter and 1.2 m in length, equipped with an electric furnace. The RF power at a frequency of 13.56 MHz is coupled to the plasma

2924

M. Matsuoka et al. / Surface & Coatings Technology 204 (2010) 2923–2927

through an inductor mounted around the periphery of the reactor. The detailed description is given elsewhere [16,17]. A Si(100) wafer of 7.5 cm in diameter was dipped in dilute HF solution to remove silicon oxide and immediately loaded to a wafer holder in the reactor. The process gases were admitted through suitable mass flow controllers into the reactor evacuated by a rotary pump and the working pressure in the reactor was kept constant at 8.0–9.3 Pa. The gas flow ratio used, NH3/SiH4 or N2/SiH4, was 1.4, 4.3, 7.2, or 9.5; the RF power, P, for NH3/ SiH4 was 50 or 100 W and that for N2/SiH4, 50 W. The films prepared were 80–130 nm in thickness measured with an ellipsometer using a wavelength of 632.8 nm. Analyses of elemental and chemical composition were carried out using XPS with incident Mg Kα radiation (hν = 1253.6 eV). The XPS spectra for the films were collected ex situ, after a sputter using 2 keV Ar+ ion beam for 2 min, from the film bulk eroded to a depth of several nm, even though compositional changes may be induced as a result of preferential sputtering and atomic mixing of the elements, and the binding energies were calibrated with respect to the C 1s peak (285.0 eV) due to adventitious carbon. Our study includes fitting of Si 2p and N 1s XPS spectra with the Doniach–Šunjić function [18] which has been widely used for XPS line shape fitting. 3. Results and discussion Four elemental species, Si, N, O, and C, were identified in the films by assignment of the corresponding signals observed in the XPS spectra. Fig. 1a and b shows two examples of the normalized Si 2p spectra for the films prepared at NH3/SiH4 = 1.4 and 9.5, respectively, and P = 100 W. Fig. 2a and b indicates the corresponding N 1s spectra and Fig. 3a and b exhibits the O 1s and C 1s spectra for the film prepared at NH3/SiH4 = 9.5 and P = 100 W. No substantial difference was observed in the O 1s and C 1s spectra, except for the variation in the integrated intensity of each spectrum. The dominant feature in the O 1s spectra is a main peak at 533.4 eV, which can be considered to originate wholly from an O–Si

Fig. 1. Si 2p XPS spectra for films prepared at P = 100 W and (a) NH3/SiH4 = 1.4; (b) NH3/SiH4 = 9.5.

Fig. 2. N 1s XPS spectra for films prepared at P = 100 W and (a) NH3/SiH4 = 1.4; (b) NH3/SiH4 = 9.5.

environment of SiO2 [10]; the principal peak in the C 1s spectra is observed at 285.0 eV and identified as amorphous carbon, graphite and/ or hydrocarbon.

Fig. 3. (a) O 1s and (b) C 1s XPS spectra for film prepared at NH3/SiH4 = 9.5 and P = 100 W.

M. Matsuoka et al. / Surface & Coatings Technology 204 (2010) 2923–2927

The composition ratio in the film can be estimated from A(N)/A (Si), where A(N) and A(Si) are the areas, after correcting with relative sensitivity factors due to the analyzer transmission and the photoionization cross section, for the whole N 1s and Si 2p spectra, respectively. Fig. 4a and b indicates the composition ratio and the oxygen content in the film calculated in a similar manner, respectively, as a function of NH3/SiH4 or N2/SiH4. In Fig. 4a, the composition ratio increases from around 0.5 to 0.8– 1.0, when NH3/SiH4 or N2/SiH4 changes from 1.4 to 4.3, and tends to become constant with further increase in gas flow ratio. The composition and structure of the film are known to be extremely sensitive to parameters such as energy and number of ions incident on the substrate [6]. The observed increase and constancy in composition ratio result, respectively, from incorporation of nitrogen into the film at lower gas flow ratios and etching of volatile nitrogen-containing species, such as NH and NH2, formed on the growing film surface by chemically enhanced desorption processes, which are commonly referred to as chemical sputtering [19], at higher gas flow ratios. On the other hand, contamination in the films by oxygen is surprisingly high (Fig. 4b), while the carbon impurity level in the films is relatively lower (3–8 at.%), except the film deposited with NH3/SiH4 = 7.2 and P = 50 W (13 at.%). It is known that the nitride

2925

deposition can readily incorporate impurities such as oxygen and carbon during deposition, due to their high chemical reactivity with the constituent elements of the film [20–22]. The oxygen impurity level in the film deposited with NH3/SiH4 increases with the increase in gas flow ratio until this level reaches 14 and 18 at.% at NH3/ SiH4 = 7.2 for P = 50 and 100 W, respectively, and becomes constant afterwards. In this case, the oxygen level for P = 100 W is always higher than that for P = 50 W at the same gas flow ratio. This fact can be understood as follows: the higher the RF power, the larger the number of high-energy ions, leading to elevated dissociation of residual water and oxygen gas in the reactor and higher sputtering yield of the inner-wall material of the quartz chamber. Further, the oxygen level in the film prepared with N2/SiH4 at P = 50 W tends to be lower than that in the films formed with NH3/SiH4 at the same gas flow ratio and RF power (Fig. 4b). This should be due to the fact that the electron-impact ionization cross section for NH3 is larger than that for N2 [23]. The Si 2p spectra are characterized by doublet terms Si 2p3/2 and Si 2p1/2 due to spin–orbit coupling. It is obvious from Fig. 1 that the line shape and the position of maximum intensity of the spectrum change with the gas flow rate, implying the existence of more than one doublet state. Our decomposition of Si 2p was tentatively done in terms of the following components: bulk Si (Si0), stoichiometric nitride Si3N4, oxynitride SiNxOy, and stoichiometric oxide SiO2, based on the literature values of the corresponding binding energies [6,8,9,14,24–27]. The contribution of SiO2 is due to the presence of the stoichiometric oxide in the O 1s spectra at a binding energy of 533.4 eV shown in Fig. 3a [10,24] and the formation of an intermediate phase between Si3N4 and SiO2, SiNxOy, has been reported in XPS studies [6,25]. The decomposition procedures for the XPS spectra are as follows: (1) the line shape function of the individual components is assumed to be the same for all the spectra and given by the Doniach–Šunjić function [18] convoluted with the instrumental resolution function represented by a Gaussian; (2) the spin–orbit splitting in the Si 2p spectra is assumed to be the same for all of the components and equal to 0.6 eV [8,28], and (3) the integrated intensity of the Si 2p1/2 peak relative to that of the Si 2p3/2 peak is equal to the spin–orbit multiplicity of 1/2. The decomposition results for the Si 2p spectra are also shown in Fig. 1 and the best fitting parameters obtained are given in Table 1. It is important to note that addition of a component with a binding energy intermediate between the Si0 and Si3N4 components, which is termed ‘Si1’, was necessary to fit successfully the Si 2p spectra. This component can be attributed to ‘under-stoichiometric’ silicon

Table 1 Best-fit parameters for Si 2p and N 1s XPS spectra. Spectrum Component

Binding energy (eV) This work

Si 2p3/2

Si0 Si1 (SiNx) Si3N4

N1s

Fig. 4. (a) Composition ratio A(N)/A(Si) and (b) oxygen content in the films, as a function of gas flow ratio. The solid lines are guides for the eyes.

SiNxOy SiO2 Spin–orbit splitting (eV) N1 (SiNx) Si3N4 N(–SiOx)3 O–N(–Si)2

Literature data

99.8 ± 0.1 98.6–99.2 [26], 99.2 [9], 99.4 [25], 99.6 [6], 99.8 [8,24] 100.8 ± 0.1 100.5 [27], 100.9 [26], 101.8 [9], 102.0 [20] 101.6 ± 0.1 101.3 [14], 101.4 [25], 101.7 [27], 101.8 [8], 102.0 [6,20], 103.2 [26] 102.5 ± 0.2 102.7 [25], 102.8 [27], 103.2 [6] 103.4 ± 0.1 103.0 [8,14], 103.6 [6,24,25], 103.7 [27] 0.60 0.60 [8], 0.61[26] 397.1 ± 0.1 397.4 [20,31] 397.8 ± 0.1 397.4[9], 397.5[7], 397.6 [25,30], 397.7 [6,20], 398.0 [24,31], 398.3 [29] 398.6 ± 0.1 398.2–398.4 [7] 399.4 ± 0.1 399.0–399.3 [7], 399.7 [25], 399.7–401.0 [8]

2926

M. Matsuoka et al. / Surface & Coatings Technology 204 (2010) 2923–2927

nitride [9,13] or SiC [29,30] by comparing its binding energy with published data; however, the possibility for SiC was discarded, because of no observation of SiC in the C 1s spectra and small amounts of C in the films. On the other hand, the N 1s spectra can be curve-fitted usually with three components at around 398, 399, and 400 eV in line with the literature. The first component is known to be due to Si3N4 [6,8,9,24,25] and the second and third components are attributed to the respective chemical bonds N(–SiOx)3 and O–N(–Si)2 [7], where N(–SiOx)3 indicates a chemical environment close to that of Si3N4 [N(–Si)3] with most of the second nearest-neighbor atoms of nitrogen replaced by oxygen and O–N(–Si)2 means a nitrogen atom bonded to one oxygen and two silicon atoms. To complete the decomposition of the N 1s spectra it was necessary to introduce a component at about 397 eV, which is named ‘N1’, into the decomposition procedure. The results are shown in Fig. 2 and the best fitting parameters obtained are given in Table 1. In connection with the N1 component, Kubler et al. [31] observed two components at 397. 4 and 398.0 eV in their N 1s spectra for deposited silicon nitride films and attributed them to nitrogen in subnitride and true nitride local environments, respectively. According to their argument, the lower binding energy of our N1 component compared with that of the Si3N4 component should be explained by the higher extra-atomic relaxation energy in subnitride environments and hence the N1 component can be assigned unequivocally to nitrogen in under-stoichiometric silicon nitride. In order to verify the consistency in our decomposition procedure and to obtain reliable chemical information, we calculated the corrected area of the components appearing in both the Si 2p and the N 1s spectra. For example, the Si3N4 component appears in the Si 2p and N 1s spectra: the corrected area of this component in the Si 2p spectra is indicated by A(Si3N4; Si) and that in the N 1s spectra, by A (Si3N4; N). Fig. 5a shows a relation between A(Si3N4; N) and A(Si3N4; Si), exhibiting a good correlation. From the inclination of the line adjusted in the figure, the compositional formula of this component can be calculated and is found to be equal to Si3N4.05. Other good correlations are obtained between A(N(–SiOx)3; N) and A(SiNxOy; Si), which is not shown here, and between A(N1; N) and A(Si1; Si), shown in Fig. 5b. The correlation results obtained between A(N(–SiOx)3; N) and A(SiNxOy; Si) and between A(N1; N) and A(Si1; Si) can be considered to be reasonable. The compositional formula reduced from Fig. 5b is found to be equal to SiN0.23, that is, under-stochiometric silicon nitride. In Ref. [20] a component observed at 397.4 eV, which corresponds to the N1 component, is referred to as SiNx. The respective terms ‘Si1’ and ‘N1’ are indicated in Table 1 by ‘Si1 (SiNx)’ and ‘N1 (SiNx)’, where x = 0.23, according to Refs. [20] and [31], and the correlation results above mentioned between the N1 and Si1 components. The relative area of each component identified in the Si 2p spectra was calculated to observe the behavior of the component with the variation in gas flow ratio, where the relative area of the Si3N4 component, for example, is given by A(Si3N4; Si)/A(Si). Fig. 6 displays the relative areas of the components for the films deposited with NH3/SiH4 at P = 50 W as a function of gas flow ratio. Common features of the relative areas observed in all the cases are as follows: (a) The Si0 and Si1 components decrease with increasing the gas flow ratio and tend to disappear for higher gas flow rations. The presence of these components in lower gas flow ratios is owing to the deficiency in nitrogen and oxygen in the film (Fig. 4). (b) The SiNxOy and SiO2 components act contrary. The SiNxOy component increases with increasing the gas flow ratio at the expense of Si3N4 in part and the SiO2 component starts appearing and increases with the increase in gas flow ratio. These increases in area correspond to the increasing incorporation of oxygen and/or nitrogen into the film. (c) The Si3N4 phase shows a slight decrease with the increase in gas flow ratio by its conversion into SiNxOy.

Fig. 5. (a) A(Si3N4; N) vs. A(Si3N4; Si) and (b) A(N1; N) vs. A(Si1; Si).

4. Conclusion The Si 2p spectra were decomposed in terms of bulk Si, understoichiometric nitride, stoichiometric nitride Si3N4, oxynitride SiNxOy, and stoichiometric oxide SiO2, and the N 1s spectra, in terms of understoichiometric silicon nitride, Si3N4, N(–SiOx)3, and O–N(–Si)2. These decomposition results could be explained self-consistently. The effect

Fig. 6. Relative area of each phase in the Si 2p XPS spectra for films prepared at P = 50 W with NH3/SiH4 as a function of gas flow ratio. The solid lines are guides for the eyes.

M. Matsuoka et al. / Surface & Coatings Technology 204 (2010) 2923–2927

of the gas flow ratio on the composition ratio can be explained as follows. At low gas flow ratios, the incorporation of nitrogen increases because of the high reactivity of nitrogen; the deficiency of nitrogen permits the presence of bulk Si and under-stoichiometric nitride in the film. Next follows the constancy in composition ratio with further increases of the gas flow ratio, in which excess nitrogen on the growing film surface is rejected by chemical sputtering. This results in the reduction and disappearance of bulk Si and under-stoichiometric nitride and in the appearance and enhancement of SiO2 and SiNxOy, by the incorporation of incident nitrogen and background oxygen. The contaminant oxygen level increases with the increase in RF power which leads to elevated dissociation of residual water and oxygen gas in the reactor. To our knowledge the contamination by oxygen has not been fully taken into account in silicon nitride film deposition. Acknowledgements Two of the authors (M.M. and W.S.) are grateful to Japan International Cooperation Agency (JICA), Japan, and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Brazil, respectively, for financial support to work in Japan and scholarship. References [1] [2] [3] [4]

M.L. Green, E.P. Gusev, R. Degraeve, E.L. Garfunkel, J. Appl. Phys. 90 (2001) 2057. R. Kärcher, L. Ley, R.L. Johnson, Phys. Rev. B 30 (1984) 1896. T.T.T. Hien, Ch. Ishizaki, K. Ishizaki, J. Ceram. Soc. Japan 113 (2005) 647. P.J. French, P.M. Sarro, R. Mallée, E.J.M. Fakkeldij, R.F. Wolffenbuttel, Sens. Actuators, A 58 (1997) 149. [5] F. Rebib, E. Tomasella, M. Dubois, J. Cellier, T. Sauvage, M. Jacquet, Thin Solid Films 515 (2007) 3480.

2927

[6] P. Cova, S. Poulin, O. Grenier, R.A. Masut, J. Appl. Phys. 97 (2005) 073518. [7] J.W. Kim, H.W. Yeom, Y.D. Chung, K. Jeong, C.N. Whang, Phys. Rev. B 66 (2002) 035312. [8] J.R. Shallenberger, D.A. Cole, S.W. Novak, J. Vac. Sci. Technol., A 17 (1999) 1086. [9] A.L. Gerrard, J.-J. Chen, J.F. Weaver, J. Phys. Chem. B 109 (2005) 8017. [10] H.-f. Li, S. Dimitrijev, D. Sweatman, H.B. Harrison, P. Tanner, J. Appl. Phys. 86 (1999) 4316. [11] C.H.F. Peden, J.W. Rogers Jr., N.D. Shinn, K.B. Kidd, K.L. Tsang, Phys. Rev. B 47 (1993) 15622. [12] M. Vila, E. Román, C. Prieto, J. Appl. Phys. 97 (2005) 113710. [13] A. Batan, A. Franquet, J. Vereecken, F. Reniers, Surf. Interface Anal. 40 (2008) 754. [14] G.J. Wan, P. Yang, R.K.Y. Fu, Zh.Q. Yao, N. Huang, P.K. Chu, J. Vac. Sci. Technol. A 23 (2005) 1346. [15] Y. Toivola, J. Thurn, R.F. Cook, G. Cibuzar, K. Roberts, J. Appl. Phys. 94 (2003) 6915. [16] L.S. Zambom, R.D. Mansano, R. Furlan, Vacuum 65 (2002) 213. [17] L.S. Zambom, R.D. Mansano, R. Furlan, P. Verdonck, Thin Solid Films 343–344 (1999) 299. [18] S. Doniach, M. Šunjić, J. Phys. C 3 (1970) 285. [19] W. Sucasaire, M. Matsuoka, K.C. Lopes, J.C.R. Mittani, L.H. Avanci, J.F.D. Chubaci, N. Added, V. Trava, E.J. Corat, J. Braz. Chem. Soc. 17 (2006) 1163. [20] I. Bertóti, Surf. Coat. Technol. 151–152 (2002) 194. [21] M. Matsuoka, S. Isotani, J.C.R. Mittani, J.F.D. Chubaci, K. Ogata, N. Kuratani, J. Vac. Sci. Technol., A 23 (2005) 137. [22] M. Matsuoka, S. Isotani, W. Sucasaire, N. Kuratani, K. Ogata, Surf. Coat. Technol. 202 (2008) 3129. [23] W. Hwang, Y.-K. Kim, M.E. Rudd, J Chem. Phys. 104 (1996) 2956. [24] P.D. Kirsch, J.G. Ekerdt, J. Vac. Sci. Technol., A 19 (2001) 2222. [25] B.S. Sahu, P. Srivastava, O.P. Agnihotri, S.M. Shivaprasad, J. Non-Cryst. Solids 351 (2005) 771. [26] J.M. Lackner, W. Waldhauser, R. Berghauser, R. Ebner, M. Beutl, G. Jakopic, G. Leising, H. Hutter, M. Rosner, T. Schöberl, Surf. Coat. Technol. 192 (2005) 225. [27] J. Szépvölgyi, I. Mohai, J. Gubicza, J. Mater. Chem. 11 (2001) 859. [28] J.P. Chang, M.L. Green, V.M. Donnelly, R.L. Opila, J. Eng Jr., J. Sapjeta, P.J. Silverman, B. Weir, H.C. Lu, T. Gustafsson, E. Garfunkel, J. Appl. Phys. 87 (2000) 4449. [29] K. Shimoda, J.-S. Park, T. Hinoki, A. Kohyama, Appl. Surf. Sci. 253 (2007) 9450. [30] S. Gallis, V. Nikas, E. Eisenbraun, M. Huang, A.E. Kaloyeros, J. Mater. Res. 24 (2009) 2561. [31] L. Kubler, J.L. Bischoff, D. Bolmont, Phys. Rev. B 38 (1988) 13113.