Acceptor defects in polycrystalline Ge layers evaluated using linear regression analysis

Polycrystalline Ge thin films have recently attracted renewed attention as a material for various electronic and optical devices. However, the difficulty in the Fermi level control of polycrystalline Ge films owing to their high density of defect-induced acceptors has limited their application in the aforementioned devices. Here, we experimentally estimated the origin of acceptor defects by significantly modulating the crystallinity and electrical properties of polycrystalline Ge layers and investigating their correlation. Our proposed linear regression analysis method, which is based on deriving the acceptor levels and their densities from the temperature dependence of the hole concentration, revealed the presence of two different acceptor levels. A systematic analysis of the effects of grain size and post annealing on the hole concentration suggests that deep acceptor levels (53–103 meV) could be attributed to dangling bonds located at grain boundaries, whereas shallow acceptor levels (< 15 meV) could be attributed to vacancies in grains. Thus, this study proposed a machine learning-based simulation method that can be widely applied in the analysis of physical properties, and can provide insights into the understanding and control of acceptor defects in polycrystalline Ge thin films.