INDEEDopt: a deep learning-based ReaxFF parameterization framework

Abstract Empirical interatomic potentials require optimization of force field parameters to tune interactions mimic ones obtained by quantum chemistry-based methods. The the is complex and requires development new techniques. Here, we propose an INitial-DEsign Enhanced Deep learning-based OPTimization (INDEEDopt) framework accelerate improve quality ReaxFF parameterization. procedure starts with a Latin Hypercube Design (LHD) algorithm that used explore parameter landscape extensively. LHD passes information about explored regions deep learning model, which finds minimum discrepancy eliminates unfeasible regions, constructs more comprehensive understanding physically meaningful space. We demonstrate here for parameterization nickel–chromium binary tungsten–sulfide–carbon–oxygen–hydrogen quinary field. show INDEEDopt produces improved accuracies in shorter time compared conventional method.