Reconstructing photoluminescence spectra at liquid nitrogen temperature from heavily boron-doped regions of crystalline silicon solar cells

When measured at low temperature (79 K), the photoluminescence (PL) spectra from silicon wafers containing a diffused heavily doped layer exhibit a second peak due to band gap narrowing in the diffused region. This work aims to decompose this peak into components arising from the various doping concentrations within the diffused layer. Whilst the peak position of silicon band-to-band PL spectra changes significantly with the doping concentration in silicon, the shape of the spectra also varies strongly with doping concentrations due to the broadening effects of band-filling and band-tail states. By measuring PL spectra on a range of uniformly heavily doped wafers, we show that these changes in spectral position and shape can be accurately modelled for doping concentrations above 1 x 10(19) cm(-3) using simple parameterisations, with minimal impact of variations in excitation intensity or injection level. This allows the PL spectra for a range of arbitrary doping concentrations to be reconstructed. We then show that the PL spectra from a thermally boron-diffused wafer, in which the boron concentration changes with depth, can be reconstructed based on a superposition of PL spectra arising from the layers of different doping beneath the surface. Furthermore, the scaling factor for each layer can be accurately estimated based on the doping profile and the fraction of incident light absorbed in the layer.