Structural properties of SiO2 films prepared by plasma-enhanced chemical vapor deposition

Structural properties of SiO2 films prepared by plasma-enhanced chemical vapor deposition

Materials Science in Semiconductor Processing 4 (2001) 43–46 Structural properties of SiO2 films prepared by plasmaenhanced chemical vapor deposition ...

110KB Sizes 1 Downloads 30 Views

Materials Science in Semiconductor Processing 4 (2001) 43–46

Structural properties of SiO2 films prepared by plasmaenhanced chemical vapor deposition Fabio Iacona*, Giulio Ceriola, Francesco La Via CNR-IMETEM, Stradale Primosole 50, 95121 Catania, Italy

Abstract SiO2 thin films have been prepared by plasma-enhanced chemical vapor deposition from SiH4 and N2O precursors by using different values of the N2O/SiH4 flow ratio (g). Rutherford backscattering spectrometry has been employed to obtain the O/Si atomic ratio of the films. Infrared spectroscopy has demonstrated that oxides having the same O/Si atomic ratio are characterized by a different structure. Indeed, from the analysis of the Si–O–Si stretching peaks, we have found that the peak frequency and full-width at half-maximum (FWHM) are dependent on g. Peak position and FWHM have been used to calculate the bond angle distribution of the films. The results have demonstrated the occurrence of a Si–O–Si bond angle relaxation phenomenon in films deposited by using a larger excess of N2O. # 2001 Elsevier Science Ltd. All rights reserved.

1. Introduction Silicon dioxide thin films are widely used in semiconductor device technology as gate insulators, for passivation and as intermetal dielectric layers. These films are either thermally grown through the oxidation of silicon or deposited by thermal or plasma-enhanced chemical vapor deposition (PECVD). The advantage of PECVD is that low deposition temperatures can be used, avoiding defect formation in the underlying silicon substrate, dopant diffusion and degradation of the metal layers. However, PECVD-prepared silicon dioxide films are usually affected by a high impurity concentration (hydrogen and nitrogen, H2O, Si–OH groups), and are characterized by a high porosity compared with those deposited at higher temperature [1]. To improve the quality of the deposited layers, PECVD processes can be optimized by varying some deposition (i.e. temperature, total and partial pressures, plasma power and frequency, etc.) and post-deposition (temperature and environment of the annealing process) parameters [2–4].

*Corresponding author. Tel.: +39-095-591243; fax: +39095-7139154. E-mail address: [email protected] (F. Iacona).

In this work, we present some results on the chemical and structural properties of SiO2 thin films prepared by PECVD from SiH4 and N2O precursors. We have mainly investigated the influence of the N2O/SiH4 flow ratio by using Rutherford backscattering spectrometry (RBS), ellipsometry and infrared (IR) spectroscopy. We will show that a variation of the N2O/SiH4 flow ratio can influence not only, as expected, the composition of the deposited layers, but also their structure.

2. Experimental SiO2 thin films, about 2000 A˚ thick, have been prepared by using a parallel plate plasma-enhanced chemical vapor deposition (PECVD) system, consisting of an ultra-high-vacuum chamber (base pressure 1  109 Torr) and a RF generator (13.56 MHz), connected through a matching network to the top electrode of the reactor; the bottom electrode is grounded and acts also as a sample holder. All deposition processes have been performed by using 50 W of input power. The substrates, consisting of 500 (1 0 0) Czochralski silicon wafers, have been heated at 3008C during the deposition. The source gases used are high-purity (99.99% or higher) SiH4 and N2O; the N2O/SiH4 flow ratio g has

1369-8001/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved. PII: S 1 3 6 9 - 8 0 0 1 ( 0 0 ) 0 0 1 3 0 - X

44

F. Iacona et al. / Materials Science in Semiconductor Processing 4 (2001) 43–46

been varied between 5 and 50, while keeping constant, at a value of about 150 sccm, the total gas flow rate. The total pressure during deposition processes has been maintained constant at a value of about 6  102 Torr. The structural properties of both as-deposited and annealed SiO2 films have been studied by Rutherford backscattering spectrometry (RBS), ellipsometry and infrared (IR) spectroscopy. RBS measurements were carried out by using a 2.0 MeV He+ beam in random configuration in order to measure film stoichiometry and thickness. Ellipsometry measurements have been performed with a AutoEL III ellipsometer (from Rudolph Research), operating at a fixed wavelength of 632.8 nm. Infrared transmission measurements were carried out by using a Perkin Elmer Fourier Transform Infrared spectrometer. The spectra were measured in the range 400–4000 cm1 with a resolution of 4 cm1.

3. Results and discussion The chemical composition of SiO2 films deposited by using different values of the N2O/SiH4 flow ratio (g) has been determined by using the RBS technique. The results are shown in Fig. 1, reporting the silicon concentration as a function of the g value. The figure shows that a very large excess of N2O (g values ranging from 15 to 50) is needed to obtain a stoichiometric SiO2 film (corresponding to a Si concentration of about 33 at%), while, for g lower than 15, films are substoichiometric, with a very steep increase of the Si concentration up to about 53.5 at% for g=5. Substoichiometric (SiOx, with x52) films are also characterized by a small (less than 10 at%) but detectable nitrogen concentration, due

Fig. 1. Si concentration, derived from RBS measurements, vs. the N2O/SiH4 flow ratio (g) for silicon oxide films prepared by PECVD.

Fig. 2. Refractive index (measured at l=632.8 nm) vs. the N2O/SiH4 flow ratio (g) for silicon oxide films prepared by PECVD.

to the use of N2O as gaseous precursor for film deposition; the nitrogen concentration does not remarkably depend on g, and it becomes negligible when the deposited layers approach the stoichiometric condition. The refractive index (n) of the oxide films, determined by ellipsometry, is shown in Fig. 2. Refractive index exhibits a marked dependence on the g value, clearly increasing for films with low g values. Note also that n constitutes a more sensible marker than the RBS technique of the characteristics of the SiO2 films, showing that the refractive index of samples with g=15 and 20 (stoichiometric according to the RBS data) is higher than films deposited by using higher g values. The high n values of the films deposited with lower g values are a direct consequence of their Si content. Indeed, the peculiar substoichiometric oxide structure can be described by using the random mixture model (RMM) [5], suggesting that SiOx films can be treated as a simple Si/SiO2 mixture, neglecting the existence of the silicon intermediate oxidation states. According to the RMM, neglecting also the small nitrogen content of the films, and following the procedure reported in Ref. [6] and based on the effective medium approximation (EMA), the refractive index n of a SiOx film can be evaluated from eSi  e eSiO2  e 0 ¼ fSi þ fSiO2 ; ð1Þ eSi þ 2e eSiO2 þ 2e where the dielectric function e is related to the refractive index by the relationship e ¼ n2 ; fSi and fSiO2 ; eSi and eSiO2 are, respectively, the fraction and the dielectric function of the two components of the mixture. The calculated values, however, are higher than the

F. Iacona et al. / Materials Science in Semiconductor Processing 4 (2001) 43–46

experimental ones; for instance, for a Si concentration of 53.5 at% (corresponding to g=5) by applying Eq. (1) we obtain n=2.1, to be compared with the experimental value of 1.79. The discrepancy could be partially explained by the presence of a void fraction, but it is mainly due to the inadequacy of the RM model. Indeed, it has been already demonstrated that SiOx films structure can be better described by the more complex random bonding model RBM [7], accounting also for the presence of substoichiometric tetrahedral units [8]. However, the application of the EMA analysis to such a model is not straightforward, due to the difficulty in defining the physical properties of the five different tetrahedral units (Si–Si4xOx, with x=0, 1, 2, 3, 4) predicted by the RBM. Fig. 3 shows the 600–1400 cm1 region of the infrared spectra relative to SiO2 films deposited by using different g values. This region contains both the main absorption peaks of a SiO2 film, i.e. those corresponding to the stretching (around 1050 cm1) and bending (around 810 cm1) vibrations of the Si–O–Si bonds. The figure clearly shows that the stretching vibration frequency depends on the g value; the peak can be found at about 1026 cm1 for g=10 and the frequency monotonically increases with g, up to a value of 1052 cm1 for g=50. This behavior is consistent with literature data reporting an almost linear dependence of the Si–O–Si stretching frequency on the composition of SiOx films [4] and it is due to the induction effect resulting from the replacing of highly electronegative O atoms with Si atoms in the network of the substoichiometric film. All the observed frequencies are remarkably lower than that observed in thermally grown SiO2 layers (about 1080 cm1) [9]. Also the full-width at half-maximum (FWHM) of the stretching peaks depends on the g value; in particular

Fig. 3. 600–1400 cm1 region of the infrared spectra of silicon oxide films deposited by using different values of the N2O/SiH4 flow ratio (g). The spectra are normalized to the intensity of the Si–O–Si stretching peak.

45

FWHM values are very high for low g values (up to about 200 cm1), while IR peaks become very narrower (less than 100 cm1) for high g values. Note also that the large differences around the 850 cm1 region of the IR spectra reported in Fig. 3 are mainly due to the presence of nitrogen in the oxide films; indeed, the absorption peak at about 875 cm1 observed for the sample with g=10 is due to the stretching vibration of the Si–N bonds; in agreement with the RBS data, the spectra clearly demonstrates that the N content becomes almost negligible already for g=15, and, accordingly, the peak corresponding to the bending vibration is again clearly visible at about 810 cm1 for g=20 and higher. Therefore, infrared spectroscopy is a very sensitive tool for the analysis of PECVD-prepared SiO2 films, being able to detect also little differences among the spectral characteristics of the films with the highest g values and characterized by almost similar values of composition and refractive index. The trends of both position and FWHM of the stretching peaks as a function of the g value are summarized in Fig. 4. From these values, by applying the central force model [10], the Si–O–Si bond angle for the various oxides can be derived from the equation n2s ¼



   1 2 a 4a ð1  cos yÞ þ ; 200pc m 3M

ð2Þ

where c is the speed of light (m/s), a the force constant (in N/m) of Si–O bonds, m and M are the mass (in kg) of O and Si atoms, respectively, y is the Si–O–Si bond angle and ns is the stretching frequency. Results are shown in Fig. 5, and demonstrate that by increasing the g value the bond angle increases from about 1278 (for g=15) to 1338 (for g=50). Note that oxides with g=5, 7.5 and 10 have been excluded from this analysis because, due to their relatively high nitrogen content, the Si–N bonds absorption is strong and prevents a careful analysis of the Si–O–Si stretching peak.

Fig. 4. Si–O–Si stretching peak frequency and FWHM vs. the N2O/SiH4 flow ratio (g) for silicon oxide films.

46

F. Iacona et al. / Materials Science in Semiconductor Processing 4 (2001) 43–46

Fig. 5. Bond angle and width of the bond angle distribution (Dy) vs. the N2O/SiH4 flow ratio (g) for silicon oxide films.

Therefore, IR data suggest the occurrence of a bond relaxation phenomena when in the gaseous mixture used for the deposition of SiO2 film a larger excess of N2O is added. However, this phenomenon seems to not really involve a worsening of the oxide film quality; indeed, by applying the central force model, we can also calculate the width of the distribution of the bond angles Dy, depending on both the stretching peaks frequency ns and FWHM (Dns), by using the following expression:   mn Dn 360 s s ð200pcÞ2 Dy ¼ : ð3Þ p a sin y Dy values relative to our SiO2 films are reported in Fig. 5; the data demonstrate that the dispersion of the bond angle values becomes remarkably narrower for the films characterized by the larger g values (and also larger bond angles), while for a true relaxation phenomenon we expect also an increase of the Dy value.

4. Conclusions In this chemical prepared We have

work, we have presented some results on the and structural properties of SiO2 thin films by PECVD from SiH4 and N2O precursors. found that, by maintaining constant all the

other main deposition parameters, the variation of the N2O/SiH4 flow ratio g represents a good method to control both the composition and the structure of the deposited films. Indeed, we have individuated two distinct region of g values; the first one, approximately defined by g values lower than 15, in which substoichiometric films are obtained; for such films both ellipsometry and IR data indicate the presence of Si–Si bonds, strongly influencing the refractive index and the absorption properties. The second region, for g>15, includes stoichiometric films; IR data have demonstrated that, inside this region, the variation of the N2O/SiH4 flow ratio allows to tailor the microstructural properties of the films; in particular we have been able to obtain larger Si–O–Si bond angles, while simultaneously decreasing their dispersion around the mean value.

Acknowledgements This work has been supported in part by Progetto Finalizzato MADESS II (CNR).

References [1] Adams AC, Alexander FB, Capio CD, Smith TE. J Electrochem Soc 1981;128:1545. [2] Dominguez C, Rodriguez JA, Munoz FJ, Zine N. Thin Solid Films 1999;346:202. [3] Furukawa K, Liu Y, Gao D, Nakashima H, Uchino K, Muraoka K. Appl Surf Sci 1997;121/122:228. [4] Pai PG, Chao SS, Takagi Y, Lucovsky G. J Vac Sci Technol A 1986;4:689. [5] Temkin RJ. J Non-Cryst Solids 1975;17:215. [6] Kucirkova A, Navratil K, Zemek J. Thin Solid Films 1998;323:53. [7] Iacona F, Lombardo S, Campisano SU. J Vac Sci Technol B 1996;14:2693. [8] Philipp HR. J Non-Cryst Solids 1972;8–10:627. [9] Sassella A, Borghesi A, Corni F, Monelli A, Ottaviani G, Tonini R, Pivac B, Bacchetta M, Zanotti L. J Vac Sci Technol A 1997;15:377. [10] Sen PN, Thorpe MF. Phys Rev B 1977;15:4030.