MODELLING AND CHARACTERIZATION OF BBr3 BORON DIFFUSION PROCESS FOR N-TYPE SI WAFER SOLAR CELLS LI Mengjie 1, 2, a, HOEX Bram 3, MA Fa-Jun 3, DEVAPPA SHETTY Kishan 1, ABERLE Armin G. 1, 2, SAMUDRA Ganesh S. 2 1 Solar Energy Research Institute of Singapore (SERIS), National University of Singapore (NUS), Singapore 117574 2 Department of Electrical & Computer Engineering, NUS, Singapore 117576 3 School of Photovoltaic and Renewable Energy Engineering (SPREE), UNSW Australia, Sydney NSW 252, Australia a Corresponding author: LI Mengjie, Tel: + 65 828453; E-mail address: limengjie@u.nus.edu ABSTRACT: The non-uniformity of the BSG (borosilicate glass) layer formed during BBr3 tube diffusion is one of the major concerns for solar cell fabrication, as the BSG layer serves as the dopant source and thus determines the amount of boron dopants that are available. It is beneficial to have a deep understanding on the correlation between the properties of the BSG layer and the resulting doping profile. Process simulations on tube diffusion are of great relevance for the optimisation of industrial n-type solar cells. However, the simulation models and parameters were mostly calibrated for CMOS (Complementary Metal Oxide Semiconductor) device fabrication. In this work, we present an investigation on the influence of BSG non-uniformity on the resulting doping profiles. We demonstrate with both experiments and by simulations that the doping profiles are not strongly affected by the non-uniformity of the BSG layer once the BSG thickness is beyond a threshold. We present a set of model and parameters that is suitable for accurately predictive simulation for solar cell fabrication. In addition, the sensitivity of the key modelling parameters is discussed and reported. Keywords: Boron diffusion, c-si, Simulation, Solar cell 1 INTRODUCTION Solar cells on n-type silicon wafer substrates are promising alternatives compared to their p-type counterparts due to the potential of higher energy conversion efficiencies [1]. Their bulk minority carrier lifetime is less affected by common metallic impurities, such as Fe, due to a strong asymmetry in the capture cross sections for electrons and holes [1, 2]. Additionally, their energy conversion efficiencies do not suffer from light induced degradation related to B-O complexes [3]. BBr3 tube diffusion is a widely used doping technique in industry to introduce boron dopants into a Si substrate for the formation of boron emitter or FSF/BSF (front surface field/back surface field) in Si wafer solar cells [4]. The resulting boron dopant profile relies on the formation of BSG (borosilicate glass) layer as it serves as the dopant source [5]. The non-uniformity of the BSG (borosilicate glass) layer has always been the major concern in tube diffusions. The so-called BSG layer is known as the mixture of elemental boron, B2O3 and SiO2. In the diffusion process, N2-BBr3 vapour is introduced into the tube together with O2. In the tube, BBr3 reacts with O2 and deposits B2O3 on Si surface. Elemental boron is released from the reaction between B2O3 and Si. At a normal deposition temperature (>8 C), B2O3 presents in liquid phase, as it melts at above 5 C. Thus, the uniformity of the resulting BSG layer is hard to control, as it is determined by the reaction between B2O3 and Si. In this paper, we will address the influence of the BSG layer thickness on resulting doping profiles. From results obtained on Rsheet (sheet resistance) and simulated doping profiles, it will be demonstrated that the doping profile is strongly influenced by the boron concentration incorporated in the BSG layer and the boron diffusivity in BSG layer instead of the non-uniformity of the BSG layer. Process simulation offers accurate prediction of the resulting doping profiles and provides valuable guidance in process optimisation [6]. The models and parameters used in the diffusion simulation were mostly calibrated for CMOS (Complementary Metal Oxide Semiconductor) device fabrication, as tube diffusion has long been used in that field. However, the process conditions are quite different than that used for solar cell fabrication. Thus, the validity of the models and parameters has to be confirmed before directly transferring the knowledge from CMOS related simulation to solar cell related simulation. In this paper, we present a set of model and parameter that is suitable for solar cell simulations. In particular, we demonstrate the investigation on the sensitivity of the key modelling parameters, including boron diffusivity in Si and segregation coefficient of SiO2-Si interface. 2 EXPERIMENT N-type pseudosquare Czochralski (Cz) Si <1> wafers (156 mm 156 mm) were used as substrates. The substrates were saw damage etched with a 2% KOH solution and subsequently cleaned using a standard RCA cleaning process. Then, a pre-deposition step was carried out using a mixture of N2-BBr3 and O2 in an industrial tube diffusion furnace (Tempress, TS8113). The temperature (Tmax) of deposition was varied from 8 C to 1 C. The gas flow was kept the same. After deposition, the samples were cooled down in N2. A simple plateau temperature profile was used for both the pre-deposition step and the step (see Fig.1). The thickness and uniformity of resulted BSG layer were analysed using spectroscopic ellipsometry (Semilab, SE- 2). Prior to, the BSG layer was removed in 5% HF until the wafer turned hydrophobic. After that, a step was carried out at 1 C or 15 C for 3 min in an O2 or N2 ambient. All the ramp-up steps were carried out in a N2 ambient. The resulted active boron profiles were measured using electrochemical capacitance-voltage (ECV, WEP, CVP21) profiling. The Rsheet uniformity on a large area wafer was determined by
four point probe measurements (CMT-SR2N-PV). S E B conc. in Si S SExp (2) kbt B conc. in SiO2 To calibrate the modelling parameters, the doping profile before was imported into Sentaurus as the starting profile. The values of boron diffusivity and segregation coefficient were obtained by reproducing the experimental doping profiles after the steps [1]. By systematically varying each of the parameter (Di, DE, S and SE), the sensitivities of the parameters were investigated. Figure 1: Time-temperature profile used for the BBr3 boron diffusion process. 3 SIMULATION The modelling of boron diffusion process was done with Sentaurus TCAD [7]. The BSG layer was modelled as a thin layer of boron doped SiO2. A constant dopant concentration was assumed in the BSG layer. The properties of this newly defined layer was assumed to be comparable with thermally grown SiO2. To find out the influence of BSG non-uniformity on the resulting doping profiles, the thickness of the BSG layer defined in the simulation was systematically varied from sub-nanometer to 5 nm. As boron diffuses via interstitialcy mechanism exclusively [8], the vacancy mechanisms were turned off in the simulations. Both neutral state interstitial I and doubly positively charged interstitial I ++ were considered to contribute to the boron diffusion process. The dopant transport was modelled with Pair model. Pairing reactions between substitutional boron atoms and charged interstitials were considered while computing the particle flux. The boron distributions were obtained by numerically solving the differential equations for substitutional boron and boron-interstitial pairs. In the description of boron diffusion process, the key parameter is boron diffusivity [D, Eq. (1)], which determines the rate of dopant transport. It is expressed with an Arrhenius equation: D D i DE Exp k BT where Di is the pre-factor (i= represents the pre-factor of boron diffusivity of I channel, i=1 represents the prefactor of I ++ channel), DE is the activation energy, kb is the Boltzmann constant and T is the temperature in Kelvin. Upon thermal oxidation of Si, boron dopants will redistribute at the interface of SiO2-Si due to the segregation effect [9], which leads to a depleted region close to the Si-SiO2 interface. The default Segregation boundary condition was used in the description of boron segregation at SiO2-Si interface, where the total dopant fluxes at the interface were assumed to be balanced. The segregation coefficient [S, Eq. (2)] is the key parameter defining the boundary condition, which is also described by an Arrhenius equation: (1) 4 RESULTS 4.1 BSG layer thickness The BSG distribution on Si substrate shows a position dependent non-uniformity (Fig. 2) which is assumed due to the non-uniformity of the B2O3 deposited on Si. Despite the significant non-uniformity of the BSG layer, however, the Rsheet of the resulted boron emitters is quite uniform as shown in Fig. 3. Thus, the amount of boron diffused into the wafer seems not to be limited by the thickness of the BSG layer. In Fig. 4, measured active boron profile is shown together with the simulated profiles, where the thickness of the BSG layer was varied from.2 nm to 5 nm. Variation on the thickness of the BSG layer does not change the resulting doping profile until the BSG layer is less than 5 nm thick. Especially when the BSG thickness is in the range of sub-nanometer, the available boron dose is not sufficient. For a BSG layer thicker than 5 nm, further increasing the thickness of the BSG layer, the resulting active dopant profile is not significantly affected. As shown in the inset of Fig. 4, the boron incorporated in BSG layer is consumed within only a few nanometres from the BSG-Si interface. The boron concentration remains unchanged in the remainder of the BSG layer. This is explained by the extremely low diffusivity of boron in thermal SiO2 [11]. Thus, it was concluded that non-uniformity of BSG layer does not have a strong impact on the resulted doping profile. Once the minimum thickness of BSG layer exceeds a threshold (~ 5 nm), the resulting doping profile will not be affected by the thickness of the BSG layer. 4.2 Boron diffusivity and segregation coefficient In a step, the shape of the resulting doping profile is determined by boron diffusion in Si and boron segregation at the SiO2-Si interface. To be specific, in an inert step, the doping profile is determined by boron diffusion exclusively, while in an oxidizing drivein step, the doping profile is ruled by both diffusion and segregation mechanism. In the process simulation, the key parameters that have been used to describe the diffusion and segregation mechanism are boron diffusivity and segregation coefficient, respectively. As shown in Fig.5, variation on the boron diffusivity mainly affects the junction depth, while variation on the segregation coefficient leads to significant changes in the doping concentration in the near surface region. Fig. 5 (a) and (c) shows the variation on the activation energies of boron diffusivity and segregation coefficient, respectively. Fig. 5 (b) and (d) shows how the doping profiles will change with different pre-factors of boron diffusivity and segregation coefficient. The simulated
doping profiles are more sensitive to the changes on the activation energies, especially on the activation energy of boron diffusivity, as an increase or a decrease of only.1 ev leads to a change in junction depth of ~.4 μm. In [nm] contrast, a higher tolerance on the variation of pre-factors is demonstrated for both boron diffusivity and segregation coefficient. [nm] 4 3 2 1-1 -2-3 -4-4 -3-2 -1 1 2 3 4 95 o C, N 2 Max:3.5 Min:17.43 Ave:22.37 StDev:3.75 StDev/Ave:16.77 Figure 2: BSG thickness on Si substrate after a 95 C / 1 C pre-deposition step. 35. 31. 27. 23. 19. 15. 4 3 2 1-1 -2-3 -4-4 -3-2 -1 1 2 3 4 3. 27. 24. 21. 18. 15. 1 o C, N 2 Max:23.84 Min:15.97 Ave:19. StDev:2.37 StDev/Ave:12.49 6 4 2-2 -4-6 -6-4 -2 2 4 6 R sheet [ohms/sq.] 25. 23. 21. 19. 17. 15. 1 o C, N 2 Max:23.36 Min:18. Ave:21.9 StDev:.98 StDev/Ave:4.47 Figure 3: Rsheet of the boron emitter after a 95 C / 1 C pre-deposition step. 6 4 2-2 -4-6 -6-4 -2 2 4 6 R sheet [ohms/sq.] 55. 52. 49. 46. 43. 4. 95 o C, N 2 Max: 52.72 Min: 44.75 Ave: 48.22 StDev: 1.98 StDev/Ave: 4.11 1 22 1 2 1 14 1 o C, N 2 =.2 nm =.5 nm = 5 nm = 1 nm = 5 nm conc. [cm -3 ] 1 23 1 22 1 21 1 2 2 4 6 8 1 12 depth [nm] BSG Si -4-2 2 4 depth [nm] Figure 4: Simulated (lines) and measured (symbols) active boron depth profiles after pre-deposition step. The boron profiles in negative X-values present the simulated boron concentration in the BSG layer. The thickness of simulated BSG layer was varied from.2 nm to 5 nm. Inset: boron profile at the interface of BSG-Si.
1 2 Variation of 15 o C, N 2 = 4.1 ev = 4.2 ev = 4.3 ev (a)..5 1. 1.5 2. 2.5 3. 1 2 Variation of 1 o C, O 2 =.5 ev =.8 ev = 1.2 ev (c)..5 1. 1.5 2. 2.5 3. 1 2 Variation of D 1 (b)..5 1. 1.5 2. 2.5 3. 15 o C, N 2 D 1 = 5. D 1 = 492.75 D 1 = 1. 1 2 Variation of S 1 o C, O 2 S = 1. S = 97.99 S = 1. (d)..5 1. 1.5 2. 2.5 3. Figure 5: Simulated (lines) and measured (symbols) active boron depth profiles after the step. The sensitivity of the key modelling parameters: (a) activation energy of boron diffusivity (DE), (b) pre-fator of boron diffusivity (D1), (c) activation energy of boron segregation coefficient (SE) and (d) pre-factor (S) of boron segregation coefficient, were investigated by varying the parameter values. 1 2 1 15 8 o C, ECV Simulation 95 o C, ECV Simulation 1 o C, ECV Simulation..2.4.6.8 1. 1.2 Figure 6: Simulated (lines) and measured (symbols) active boron depth profiles with a BBr3 pre-deposition at different temperatures. Table I: Boron diffusivity and segregation coefficient calibrated in this work. Parameter Charge status Activation energy [ev] Pre-factor I ++ 4.2 (in N2) 492.75 cm 2 /s (in N2) 3.85 (in O2) 24.15 cm Diffusivity (D) 2 /s (in O2) I 4.2 (in N2) 5.1 cm 2 /s (in N2) 3.85 (in O2) 4.42 cm 2 /s (in O2) Segregation coefficient (S) -.82 97.99
4.3 Predictive simulation of boron diffusion After having re-calibrated the values for DE, Di, SE, S, the process simulation was tested for different BBr3 pre-deposition temperatures (see Fig. 6). The experimental active boron doping profiles are shown together with the simulated profiles. The doping source, i.e. the BSG layer was simulated with the boron doped SiO2. The modelling parameters were calibrated in previous step. Good agreement with the experimental doping profiles confirms the validity of the simulation model and the parameters that have been adopted. An increase of deposition temperature shows a strong influence on the boron dopant profile. This is explained by the temperature dependent boron diffusivity, as a higher diffusivity is experienced by the dopants at a higher temperature. The boron surface concentration, on the contrary, is not affected by the temperature as it is mainly determined by the boron concentration incorporated in BSG layer. 5 CONCLUSION In summary, we have investigated the influence of BSG non-uniformity on doping profiles. We demonstrate both experimentally and by simulation that the resulting doping profile is not strongly affected by the nonuniformity of the BSG layer once the BSG thickness exceeds 5 nm. We presented detailed study on the sensitivities of the modelling parameters, including the boron diffusivity and the segregation coefficient. The doping profiles are more sensitive to changes in the activation energies in both diffusivity and segregation coefficient. In contrast, the pre-factors show a limited impact on the doping profiles. [1] J. E. Cotter, J. H. Guo, P. J. Cousins, M. D. Abbott, F. W. Chen, and K. C. Fisher, "P-Type Versus n- Type Silicon Wafers: Prospects for High- Efficiency Commercial Silicon Solar Cells," Electron Devices, IEEE Transactions on, vol. 53, pp. 1893-191, 26. [2] D. Macdonald and L. J. Geerligs, "Recombination activity of interstitial iron and other transition metal point defects in p- and n-type crystalline silicon," Applied Physics Letters, vol. 85, pp. 461-463, 24. [3] S. W. Glunz, S. Rein, J. Y. Lee, and W. Warta, "Minority carrier lifetime degradation in borondoped Czochralski silicon," Journal of Applied Physics, vol. 9, pp. 2397-244, 21. [4] Y. Schiele, S. Fahr, S. Joos, G. Hahn, and B. Terheiden, "Study on boron emitter formation by BBr3 diffusion for n-type Si solar cell applications," in 28th European Photovoltaic Solar Energy Conference and Exhibition (EU PVSEC 213), 3 Sept.-4 Oct. 213, Munich, Germany, 213, pp. 1242-7. [5] J. Benick, "High efficiency n-type solar cells with a front side boron emitter," PhD Thesis, 21. [6] J. Schön, A. Abdollahinia, R. Müller, J. Benick, M. Hermle, W. Warta, et al., "Predictive Simulation of Doping Processes for Silicon Solar Cells," Energy Procedia, vol. 38, pp. 312-32, // 213. [7] Synopsys, "Synopsys TCAD," Available : http://www.synopsys.com, Release I-213.12 213. [8] S. Mirabella, D. De Salvador, E. Napolitani, E. Bruno, and F. Priolo, "Mechanisms of boron diffusion in silicon and germanium," Journal of Applied Physics, vol. 113, p. 3111, 213. [9] T. K. a. Y. Nishi, "Redistribution of diffusied boron in silicon by termal oxidation " Japanese Joutnal of Apoplied Physics, vol. 3, 1964. [1] M. LI, B. Hoex, F.-J. Ma, K. Devappa Shetty, A. S. Aberle, and G. S., "Numerical simulation of boron profiles in n-type Si wafer solar cells using tube diffusion with BBr3 ambient," To be submitted, 215. [11] T. Aoyama, H. Tashiro, and K. Suzuki, "Diffusion of Boron, Phosphorus, Arsenic, and Antimony in Thermally Grown Silicon Dioxide," Journal of The Electrochemical Society, vol. 146, pp. 1879-1883, May 1, 1999 1999. 6 ACKNOWLEDGEMENTS The authors would like to thank their colleagues from the Silicon Materials and Cells Cluster of the Solar Energy Research Institute of Singapore (SERIS) for their assistance in sample processing and measurements. SERIS is sponsored by the National University of Singapore and Singapore s National Research Foundation through the Singapore Economic Development Board. This research is supported by the National Research Foundation, Prime Minister s Office, Singapore under its Energy Innovation Programme Office (EIPO grant NRF212EWT-EIRP1-23). 7 REFERENCES