Modelling and Optimisation of Single Junction Strain Balanced Quantum Well Solar Cells

In an attempt to find the optimum number of wells for maximum conversion efficiency a pair of otherwise identical strain balanced samples, one containing 50 wells and the other 65 wells have been characterised. The 65 well sample is found to possess a lower predicted efficiency than the 50 well sample, suggesting that the optimum well number lies between these values. Devices grown using tertiary butyl arsine (TBAs) are found to possess comparable conversion efficiencies to the control cells grown using arsine and slightly superior dark IV characteristics, indicating that TBAs may be substituted for arsine without loss of device efficiency and may even be beneficial to cell performance. Several fundamental refinements to the existing quantum efficiency model of are explored. Firstly, expressions for the strained band gaps are derived. A value for the conduction band offset is . determined using the difference in energy between the heavy and light hole exciton peaks in low temperature photo current scans and found to be 0.55??0.03. The magnitude of the el-hhl exciton binding energy is also estimated from these scans and found to be in excellent agreement with the value obtained from a simple, parameterized expression for the exciton binding energy. Finally, an absolute calculation for the absorption coefficient is incorporated into the quantum efficiency model and values for the heavy and light hole in-planes masses are obtained. The model is found to underestimate the level of absorption in the intrinsic region by an amount consistent with estimates of the magnitude of the reflection from the back surface. The conversion efficiency of a sample predicted using SOL is compared to an independently obtained value. Good agreement is observed between the two results (25.3% and 25.7% for 317 suns AM1.5D). Additionally, an optimum structure for illumination by the AM1.5D spectrum was found to be a 120A well ofIno.lGaAs.Imperial Users onl