Development of metal oxide solar cells through numerical modeling.

Photovoltaic (PV) devices become increasingly important due to the foreseeable energy crisis, limitation in natural fossil fuel resources and associated green-house effect caused by carbon consumption. At present, silicon-based solar cells dominate the photovoltaic market owing to the well-established microelectronics industry which provides high quality Si-materials and reliable fabrication processes. However ever increased demand for photovoltaic devices with better energy conversion efficiency at low cost drives researchers round the world to search for cheaper materials, low-cost processing, and thinner or more efficient device structures. Therefore, new materials and structures are desired to improve the performance/price ratio to make it more competitive to traditional energy. Metal Oxide (MO) semiconductors are one group of the new low cost materials with great potential for PV application due to their abundance and wide selections of properties. However, the development of MO solar cells is very limited so far mostly due to the poor materials and poor understanding of the materials and devices. This research conducts a systematic numerical investigation on MO thin film solar cells. Various MO semiconductors are used to explore different structures and combinations for solar cells; and the effects of material properties and structures are optimised for the best performances. For the ideal cases, it is found that a TiO2/CuO hetero-junction solar cell shows a conversion efficiency of ~16% with the CuO film thickness only 1.5μm. When a back surface field layer, such as Cu2O, is added at the back of this device, the open circuit voltage (VOC) can be improved by 70% without sacrificing short circuit current, resulting in a conversion efficiency of ~28%, increased by ~70% as compared to the two-layered structure. This is close to the theoretical maximum efficiency of silicon single junction solar cell, which requires a 200~400μm thickness film. Modelling also shows the alternative Schottky barrier type MO semiconductor solar cells can perform well. For an ideal Metal/CuO Schottky barrier solar cell, the conversion efficiency could be as high as ~17%, better than the TiO2/CuO hetero-junction solar cell. The effects of defects and interface states are then considered for more realistic cases as there exists vast amount of defects mostly due to oxygen/metal vacancies/interstitials in the films, and vast amount of interface states due to the large lattice mismatch of the two materials used. All defects and interface states in the solar cell layers, hetero-junction interfaces and metal/semiconductor contacts are found detrimental to the cells. For example, if the defect concentration in the CuO layer in TiO2/CuO structure is compatible to the acceptor concentration of 1x1016cm-3, the cell efficiency would be reduced dramatically to 7%. With defect concentration even as low as 1x1013cm-3, the significant VOC improvements in the TiO2/CuO/Cu2O would be reduced to an ignorable value. For interface states, they capture and recombine both electrons and hols passing through the hetero-junction interface, leading to deteriorated performance. The simulation shows that the interface states have a detrimental effect on the performance if its density is higher than 1012cm-2. However it was found that by increasing the difference of doping concentration in p-n junctions, the interface state effect minimized significantly. Furthermore, it is found the optical reflection at hetero-junction interface may induce a serious conversion efficiency loss, if the n-type semiconductors and p-type semiconductors have very different refractive indices. For some MO devices such as TiO2/CuO and ZnO/Cu2O, the reflection rate is around 5%, while for other material systems such as ZnO/Si, or ITO/Ge, the interfacial optical reflection may reach 10~30%, resulting in an efficiency loss by ~10%. It is also found that the interfacial reflection should be calculated through experimental data of refractive index at each photon frequency, rather than the dielectric constant. Otherwise, huge error may be introduced to the simulation results