The dangling-bond defect in amorphous silicon: Statistical random versus kinetically driven defect geometries

Abstract Amorphous and micro-crystalline silicon (a-Si:H, μ c-Si) are key materials for resource-saving thin-film solar cells. However, the efficiency of such devices is severely limited by light-induced Si dangling-bond defects, which can be detected by electron paramagnetic resonance (EPR). We report density-functional theory calculations on a set of random dangling bonds created in supercell models of a-Si:H and compare calculated hyperfine and g-tensor distributions to the ones obtained from a recent multi-frequency EPR spectral analysis. Our results show that the g-tensor does not exhibit axial symmetry as has been previously assumed, but is clearly rhombic. The hyperfine coupling to the undercoordinated Si atom, on the other hand, is almost perfectly axial. This apparent discrepancy in the symmetry properties is shown to be a consequence of the underlying coupling mechanisms and how these are influenced by structural disorder. However, the hyperfine distribution calculated from our random models underestimates the experimentally observed 30% red-shift when going from c-Si to a-Si:H. We suggest that only a subset of possible dangling-bond configurations is observed in experiment. We discuss plausible mechanisms that would give rise to such a selection, and new experiments to test these hypotheses.