Applications of Nickelate perovskites for neuromorphic computing from electronic structure and Machine Learning

Metal to insulator (MIT) transition present in strongly correlated materials (SCM) is crucial in designing materials for neuromorphic applications, as they show great sensibility to small environment changes. We performed studies on controlling MIT in SCM’s through means of DMFT calculations. Our in-house DMFTwDFT framework provided DMFT total energies of various strongly correlated bulk SmNiO₃ and GdNiO₃ configurations mimicking a Boltzmann distribution under a Debye model at a finite temperature that was used to train an atomic interaction potential based on artificial neural networks (ANN). Results indicate very good correlation between the DMFT energies, and the total energies predicted by the machine learning model. Next, we introduce oxygen vacancies in order to tune the MIT and incorporate that for training the machine learning model. We also study the oxygen vacancy diffusion energy barrier through means of Nudged Elastic Band analysis and DMFT to elucidate the effect of vacancy migration on MIT. Elastic constants are calculated from DMFT energy calculations and are fed to the NN potential. The relation between the deformation and the energy obtained from DMFT allow us to describe the role of vacancies to create distortions which impact the electronic properties.