2020/12/03 by Gianluca Tirimbó, Tirimbó, Gianluca, Onur Çaylak +3
Materials Science · Physics and Astronomy · Biochemistry, Genetics and Molecular Biology · #Machine Learning in Materials Science #Advanced Chemical Physics Studies #Protein Structure and Dynamics
paper · pdf · doi:10.48550/arxiv.2012.01787
We present a Kernel Ridge Regression (KRR) based supervised learning method\ncombined with Genetic Algorithms (GAs) for the calculation of quasiparticle\nenergies within Many-Body Green's Functions Theory. These energies representing\nelectronic excitations of a material are solutions to a set of non-linear\nequations, containing the electron self-energy (SE) in the GW approximation.\nDue to the frequency-dependence of this SE, standard approaches are\ncomputationally expensive and may yield non-physical solutions, in particular\nfor larger systems. In our proposed model, we use KRR as a self-adaptive\nsurrogate model which reduces the number of explicit calculations of the SE.\nTransforming the standard fixed-point problem of finding quasiparticle energies\ninto a global optimization problem with a suitably defined fitness function,\napplication of the GA yields uniquely the physically relevant solution. We\ndemonstrate the applicability of our method for a set of molecules from the\nGW100 dataset, which are known to exhibit a particularly problematic\nstructure of the SE. Results of the KRR-GA model agree within less than 0.01 eV\nwith the reference standard implementation, while reducing the number of\nrequired SE evaluations roughly by a factor of ten.\n