Gas bubbles and atomic diffusion mechanisms in CdTe solar cells

This thesis examines various atomistic mechanisms associated with the production of cadmium telluride (CdTe) thin film solar cells and gives a fundamental understanding of processes that are used in their production. Thin

film CdTe is the most important thin film photovoltaic technology with annual module production increasing to 8 GW. Currently, the champion conversion efficiency achieved from a CdTe solar cell in a research environment is 22.1%. Cells are produced by a deposition process followed by a CdCl2 treatment at 400 C which improves cell efficiency and cell structure. The thesis uses a combination of theoretical modelling tools, Molecular Dynamics, binary collision algorithms, Ab-initio calculations, Parallel trajectory splicing and Kinetic Monte Carlo for the investigation.

The viability of magnetron sputtering as an alternative to the closed space sublimation method of thin film deposition is examined since experiment had shown the formation of inert gas bubbles in the structures after the elevated temperature CdCl2 treatment in sputtered cells. Using Molecular Dynamics, the penetration threshold energies are determined for both Ar and Xe, typical working gases in a magnetron, with CdTe in both zinc-blende and wurtzite

phases. These calculations show that more Ar than Xe can penetrate into the growing film with most penetration across the (111) surface. The mechanisms and energy barriers for interstitial Ar and Xe diffusion in zinc-blende CdTe were determined. Energy barriers are reduced near existing clusters, increasing the probability of capture based cluster growth. Barriers in wurtzite are higher with non-Arrhenius behaviour observed. This provides an explanation for the

increase in the size of voids observed experimentally after the CdCl2 treatment removes these wurtzite structures through its stacking fault removal process. Blister exfoliation was also modelled, showing the formation of shallow craters with a raised rim. Using ab-initio methods, we find that a saturation of Cl at the grain boundaries promotes Cl into interstitial and substitutional sites which passivate electronic defects at the boundaries and leads to the efficiency increase. This saturation and subsequent Cl substitution is also shown to promote removal of

stacking faults which are prevalent in the as-deposited films via a cascading Te diffusion mechanism in the grain interior which is a side effect of the CdCl2 treatment.

We simulate CdTe growth as would occur in magnetron sputtering and examine the growth formation mechanisms using adaptive Kinetic Monte Carlo (aKMC). Chains of Cd and Te are formed, via low barrier adatom surface transitions, which develop into more rounded structures pinning them in place. In the case considered, a stacking fault layer grows and an Ar di-interstitial forms by diffusion below the surface. Recently proposed optimisation strategies for the aKMC saddle searching methods are tested and found to provide a 10% speed-up in our systems.

Since group-V doping could increase efficiency of the CdTe cell to 25 % through increased electron hole concentration, the viability of arsenic incorporation methods is examined. We find that As di-interstitials are stable structures within the CdTe lattice which have a > 1.8 eV ‘break-away’ barrier.

This suggests that substitutional As atoms will be difficult to form during a diffusion based doping process (as the As will be incorporated as As2) whereas ion implantation could minimise As2 formation since there would be single As

atoms in the beam. Using a binary collision algorithm, the density of vacancies was shown to be an order of magnitude higher than As concentration making As capture via a vacancy more likely than di-interstitial formation. A further aspect that could improve efficiency is if the films are grown under Cd-rich conditions where the formation of CdTe antisites could occur. These defects are shown to capture As interstitials and provide a site for AsTe substitution.

Finally some suggestions are made as to how deposition methods might be further investigated and how doping by group-V elements could be improved.