Surface, Emitter and Bulk Recombination in Silicon and Development of Silicon Nitride Passivated Solar Cells

Abstract

Recombination within the bulk and at the surfaces of crystalline silicon has been investigated in this thesis. Special attention has been paid to the surface passivation achievable with plasma enhanced chemical vapour deposited (PECVD) silicon nitride (SiN) films due to their potential for widespread use in silicon solar cells. The passivation obtained with thermally grown silicon oxide (SiO2) layers has also been extensively investigated for comparison. Injection-level dependent lifetime measurements have been used throughout this thesis to quantify the different recombination rates in silicon. New techniques for interpreting the effective lifetime in terms of device characteristics have been introduced, based on the physical concept of a net photogeneration rate. The converse relationships for determining the effective lifetime from measurements of the open-circuit voltage (Voc) under arbitrary illumination have also been introduced, thus establishing the equivalency of the photoconductance and voltage techniques, both quasi-static and transient, by allowing similar possibilities for all of them. The rate of intrinsic recombination in silicon is of fundamental importance. It has been investigated as a function of injection level for both n-type and p-type silicon, for dopant densities up to ~5x10cm. Record high effective lifetimes, up to 32ms for high resistivity silicon, have been measured. Importantly, the wafers where commercially sourced and had undergone significant high temperature processing. A new, general parameterisation has been proposed for the rate of band-to-band Auger recombination in crystalline silicon, which accurately fits the experimental lifetime data for arbitrary injection level and arbitrary dopant density. The limiting efficiency of crystalline silicon solar cells has been re-evaluated using this new parameterisation, with the effects of photon recycling included. Surface recombination processes in silicon solar cells are becoming progressively more important as industry drives towards thinner substrates and higher cell efficiencies. The surface recombination properties of well-passivating SiN films on p-type and n-type silicon have been comprehensively studied, with Seff values as low as 1cm/s being unambiguously determined. The well-passivating SiN films optimised in this thesis are unique in that they are stoichiometric in composition, rather than being silicon rich, a property which is attributed to the use of dilute silane as a process gas. A simple physical model, based on recombination at the Si/SiN interface being determined by a high fixed charge density within the SiN film (even under illumination), has been proposed to explain the injection-level dependent Seff for a variety of differently doped wafers. The passivation obtained with the optimised SiN films has been compared to that obtained with high temperature thermal oxides (FGA and alnealed) and the limits imposed by surface recombination on the efficiency of SiN passivated solar cells investigated. It is shown that the optimised SiN films show little absorption of UV photons from the solar spectrum and can be easily patterned by photolithography and wet chemical etching. The recombination properties of n and p emitters passivated with optimised SiN films and thermal SiO2 have been extensively studied over a large range of emitter sheet resistances. Both planar and random pyramid textured surfaces were studied for n emitters, where the optimised SiN films were again found to be stoichiometric in composition. The optimised SiN films provided good passivation of the heavily doped n-Si/SiN interface, with the surface recombination velocity increasing from 1400cm/s to 25000cm/s as the surface concentration of electrically active phosphorus atoms increased from 7.5x10cm to 1.8x10cm. The optimised SiN films also provided reasonable passivation of industrial n emitters formed in a belt-line furnace. It was found that the surface recombination properties of SiN passivated p emitters was poor and was worst for sheet resistances of ~150Ω/ . The hypothesis that recombination at the Si/SiN interface is determined by a high fixed charge density within the SiN films was extended to explain this dependence on sheet resistance. The efficiency potential of SiN passivated np cells has been investigated, with a sheet resistance of 80-100Ω/ and a base resistivity of 1-2Ωcm found to be optimal. Open-circuit voltages of 670-680mV and efficiencies up to ~20% and ~23% appear possible for SiN passivated planar and textured cells respectively. The recombination properties measured for emitters passivated with SiO2, both n and p, were consistent with other studies and found to be superior to those obtained with SiN passivation. Stoichiometric SiN films were used to passivate the front and rear surfaces of various solar cell structures. Simplified PERC cells fabricated on 0.3Ωcm p-type silicon, with either a planar or random pyramid textured front surface, produced high Voc’s of 665-670mV and conversion efficiencies up to 19.7%, which are amongst the highest obtained for SiN passivated solar cells. Bifacial solar cells fabricated on planar, high resistivity n-type substrates (20Ωcm) demonstrated Voc’s up to 675mV, the highest ever reported for an all-SiN passivated cell, and excellent bifaciality factors. Planar PERC cells fabricated on gettered 0.2Ωcm multicrystalline silicon have also demonstrated very high Voc’s of 655-659mV and conversion efficiencies up to 17.3% using a single layer anti-reflection coating. Short-wavelength internal quantum efficiency measurements confirmed the excellent passivation achieved with the optimised stoichiometric SiN films on n emitters, while long-wavelength measurements show that there is a loss of short-circuit current at the rear surface of SiN passivated p-type cells. The latter loss is attributed to parasitic shunting, which arises from an inversion layer at the rear surface due to the high fixed charge (positive) density in the SiN layers. It has been demonstrated that that a simple way to reduce the impact of the parasitic shunt is to etch away some of the silicon from the rear contact dots. An alternative is to have locally diffused p regions under the rear contacts, and a novel method to form a rear structure consisting of a local Al-BSF with SiN passivation elsewhere, without using photolithography, has been demonstrated.