Surface and Bulk Passivation of Silicon by LF-PECVD Hydrogenated Silicon Nitride SiNx:H J-F. Lelièvrea), Y. Roziera), A. Bernaudeaua), O. Palaisb), A. Kaminskia), S. Quoizolaa), O. Nichiporuka), M. Bérenguer a), P. Girard a), J-C. Loretz c), C. Giral c), Y. Pellegrin c), M. Lemitia) a) Laboratoire LPM, INSA de Lyon, UMR-CNRS 5511, Villeurbanne France, email:
[email protected], b) Laboratoire TECSEN, UMR CNRS 6122, Marseille, France, email:
[email protected] c) SEMCO-Engineering SA, Montpellier, France, email :
[email protected] Abstract: This work is focused on the surface and bulk passivation properties of hydrogenated silicon nitride SiNx:H deposited on n-type and p-type silicon wafers by a new vertical-configured Low-Frequency (440kHz) PECVD reactor. The best surface passivation was obtained for the Si-richest SiN, with surface recombination velocities as low as 5cm/s. The evolution of passivation quality according to SiN stoechiometry was discussed. Interestingly, the best effective lifetime after annealing the samples was obtained for the near stoechiometric SiN. Bulk passivation was analyzed thanks to effective lifetime measurements performed on multicrystalline Si, showing an enhancement with increasing SiN deposition temperature. Key Words: Passivation, Silicon Nitride, PECVD. laboratory [4], giving Surface Recombination Velocity (SRV) and b of the studied samples. As these latter values were found for an injection level ( n) of 7.1014cm-3, the eff values reported in this work are given for the same n. The influence of Rapid Thermal Annealing (RTA-ADDAX R1000) was studied in order to simulate the firing of solar cell contacts. All the samples were annealed at around 800°C and lifetime measurements were performed again. In parallel, the electrical characteristics of SiN layers were obtained by dark capacitance-voltage (CV) measurements at different frequencies (1, 10, 100kHz and 1MHz using a HP4284A) performed on metal-insulator-semiconductor (MIS) structures: Al/SiN/Si/Al. For this purpose, single-sided SiN layers were deposited on 10 .cm p-type c-Si CZ wafers during the same runs as the FZ c- Si and mc-Si samples. Two CZ c-Si wafers were placed in the reactor for each run, one of which was RTA annealed before the deposition of aluminium by electron-beam evaporation on both sides of the structures. The MIS samples were then annealed at 400°C for 15 minutes under H2 in order to obtain ohmic contacts. The CV measurements, performed with applied voltages swept between the inversion and the accumulation modes and back to the inversion, allowed to determine Qf in the SiN layer thanks to both the computation of the flat-band capacitance CFB and the Maserjian and Vincent method [5]. The thickness, wavelength-dependant refractive index n(λ) and extinction coefficient k(λ) of the different SiN films were deduced by spectroscopic ellipsometry (JobinYvon UVISEL), using Tauc-Lorenz dispersion model [6]. SiN thickness was found to lie between 75 and 125nm and the deduced Qf were normalized for 100nm. Finally, the influence of the deposition temperature on the SiN passivation properties was investigated in the same way as previously described, but varying the SiN deposition temperature between 300 and 400°C while maintaining the gas flow ratio at R=7.7.
1 Introduction Hydrogenated Silicon Nitride SiNx:H films (abbreviated to “SiN” in this work) are largely used as antireflective coating as well as passivation layer for industrial crystalline (c-Si) and multicrystalline (mc-Si) silicon solar cells. The reduced recombination rates at SiN passivated silicon surfaces are attributed to (i) a reduction of the interface state density (Dit,) and (ii) a band bending at the silicon surface due to fixed positive SiN charges (Qf), resulting in a field-effect passivation in the form of an accumulation or an inversion layer near the Si/SiN interface for n-type and p-type Si, respectively [1,2]. On the other hand, Si bulk passivation is believed to be due to the release of hydrogen from the SiN layer after a high temperature step, which will neutralize defects in the silicon substrate for well chosen annealing temperature and time [1]. This work is focused on the surface and bulk passivation properties of SiN deposited on different silicon wafers by Low Frequency (440kHz) Plasma Enhanced Chemical Vapour Deposition (LF-PECVD).
2 Experimental SiN layers were deposited by direct LF-PECVD on both surfaces of 5 .cm p-type and n-type FZ c-Si wafers as well as on textured p-type mc-Si substrates. The reactor used was developed by SEMCO-Engineering and has the particularity to present a vertical configuration chamber. Different stoechiometries were obtained modifying the ammonia-tosilane R=NH3/SiH4 gas flow ratio while the temperature (T=370°C) and all the deposition parameters (pressure, plasma power, total gas flow) were maintained constant. Lifetime measurements were carried out using the PhotoConductance Decay (PCD) method developed by Sinton and Cuevas [3], in both the transient and quasi-steady state measurement modes. Under the assumption that all studied FZ c-Si samples have equally high values for the bulk lifetime ( b), the extracted values of the effective lifetime ( eff) can be used as a measure of the surface passivation quality of the different SiN layers. The results were confirmed independently by the microwave phase-shift (µW-PS) technique developed at TECSEN
3 Results and Discussion 3.1 Passivation and SiN Stoechiometry The as-deposited SiN refractive index n at 605nm, the total SiN charge density before (Qf) and after (Qf*) the high temperature step are displayed in Table I. Qf decreases, then
1
R=NH3/SiH4
10
7.7
6
3.5
3
2.5
2
n(605nm) Qf (×1012 cm-2) Qf*(×1012 cm-2)
1.98 8 3.3
2.02 6 4.6
2.08 4.6 3.1
2.47 2.9 1.4
2.56 3.2 1.4
2.69 -
2.98 3.2 1.5
[2], although values for n/ p may be discussed [8]. The high temperature step induced a strong decrease of the SRV for Sirich SiN/c-Si samples whereas almost no change was observed for the N-rich alloys (R=10, 6, 7.7). This is believed to be due to the release of hydrogen from the SiN layer, breaking first the weaker Si-H bonds [9]. The reduction of Qf (see Table I) can be explained by the unpassivated negatively charged dangling nitrogen bonds [2]. Interestingly, the best surface passivation after annealing was obtained for R=7.7, which corresponds to near stoechiometric SiN. The fired mc-Si samples showed a slight increase of eff for all the SiN stoechiometries (except a small decrease for R=2) indicating that hydrogen bulk passivation effectively occurs during annealing.
Table I As-deposited SiN refractive index at 605nm, total SiN charge density before (Qf) and after (Qf*) RTA according to different gas ratios R.
saturates, as the Si-content of the SiN layer increases. The Si dangling bond is the dominant deep defect in SiN films and as the Si-content is reduced, its predominant configuration is ⋅Si≡N3 (the so-called K center) [7,8]. The positively charged K+ defect contributes to the total Qf, which hence increases as the stoechiometric SiN is approached. On the contrary, eff increases when the Si-content of the SiN layer is increased for both the n- and p-type FZ c-Si samples as well as for the ptype mc-Si samples (See Fig.1 and 2). A high eff of 3.26ms was found for Si-richest SiN layer deposited on n-type c-Si, corresponding to a SRV of 5cm/s, which is extremely low for LF-PECVD. N-rich alloys present a higher defect density [7], due to their higher porosity [9]. These defects should be fieldeffect passivated by the higher Qf of these layers. However, as suggested by Schmidt and Aberle [8], Qf must be drastically reduced to ~1011cm-2 under illumination (PCD measurements case) in comparison with the values determined by dark CV measurements: illumination lifts the quasi-Fermi level n above the energy level of the K centers, leading to their neutralization. Hence, an increased number of K centers in Nrich SiN alloys is not sufficient to compensate the higher defect density of these layers. On the other hand, a high SiH4 gas flow during PECVD is believed to be responsible for the passivation of dangling bonds at the Si interface, following the model developed by Mäckel and Lüdemann [7].
3.2 Effect of the Deposition Temperature It was found that there is an optimum deposition temperature of 370°C for c-Si substrates, which is attributed to a more effective field-effect passivation, probably due to a higher Qf value for this temperature (see Fig.3). The highest temperature studied (400°C) offers the best passivation for mcSi. This suggests that hydrogen bulk passivation occurs during SiN deposition and is enhanced as the temperature is increased.
Figure 3
14 -3 eff for n=7.10 cm according to the SiN deposition temperature. The corresponding Qf (×1012cm-2) values are indicated near the curve.
4 Conclusion Figure 1 Injection-dependant effective lifetime
eff(
This study presents a good overview of surface and bulk passivation properties of hydrogenated silicon nitride SiNx:H films deposited on n-type and p-type c-Si as well as p-type mcSi thanks to a new vertical-configured LF-PECVD reactor.
n) according to different R.
5 References
Figure 2
eff
[1] F.Duerinckx and J.Szlufcik, Sol. Ener. Mat. & Sol. Cells, 72, 2002, 231. [2] S.Wolf, G.Agostinelli and G.Beaucarne, J. Appl. Phys., 97, 2005, 063303. [3] R.Sinton, A.Cuevas Appl. Phys. Lett., 69(17), 1996, 2510. [4] O.Palais and A.Arcari, J. Appl. Phys., 93(8), 2003, 4686. [5] E.H.Nicollian and J.R.Brews, MOS Physics and Technology, 1982 (Wiley, New-York). [6] G.E.Jellison and F.A.Modine, Appl. Phys. Lett., 69(3), 1996, 371, 2137. [7] H.Mäckel, R.Lüdemann, J. Appl. Phys., 92(5), 2002, 2602. [8] J.Schmidt, A.G.Aberle, J. Appl. Phys., 85(7), 1999, 3626. [9] J-F.Lelièvre et al., 20th European Photovoltaic Solar Energy Conference, Barcelona, 2005.
for n=7.1014cm-3 according to the SiN refractive index.
The better surface passivation obtained for n-type c-Si (see Fig.2) could be explained by a large capture cross-section ratio ( n/ p~100 at midgap) of minority to majority charge carriers, leading to more critical surface passivation for p-type silicon
2
