Photoluminescence characterisation of silicon bricks

Abstract

Quality electrical characterisation of photovoltaic silicon bricks after crystallization is of high importance as it can not only increase the productivity of the directional solidification but also permits selective cutting and specialised processing. In an industrial environment, the characterisation method can be considered ideal if it allows the non-destructive measurement (i) with high spatial resolution, (ii) at inline speed, (iii) with high bulk lifetime sensitivity, and (iv) with reliable accuracy. For this purpose, this thesis explores the use of spectrally dependent photoluminescence (PL) measurements in both imaging and full spectrum applications. The bulk lifetime signature of the PL spectrum can be accessed most simply by measuring a ratio of two differently filtered PL images and then translated to bulk lifetimes using a transfer function modelled in this thesis. The experimental apparatus is based on the equipment commonly used for PL measurements of wafers only adapted for a longer excitation wavelength. However, for quantitative brick measurements, the spectral dependence of the apparatus is needed to be characterised precisely. This thesis presents PL intensity ratio detected bulk lifetime images and discusses their accuracy. The technique is compared to existing brick measurement techniques with particular focus on the limitations. Light spreading in the silicon CCD was identified as a major cause of image blur and its strong effect on the quantitative analysis was effectively reduced by a proposed deconvolution procedure, whereby the light spreading is measured, mathematically described with a point spread function and then employed in an inverse Fourier transform. In addition, the image contrast in the vicinity of a grain boundary was electrically modelled in two dimensions and its effect on the PL intensity ratio analysis quantified. Additionally, this thesis experimentally verifies the analytical description of the PL emission of silicon bricks and employs various regression approaches to determine bulk lifetimes in an automated way using full spectral information. Overall, it is shown that quantitative bulk lifetimes can be determined via spectral photoluminescence analysis with a resolution of 160 µm and captured within seconds. At this point, good accuracy has been achieved and is only limited by the design and calibration of the experimental setup.