Polarizable Force Field-Based Molecular Dynamics Simulations of Ionic Liquid Electrolyte Combined with a Neural Network-Assisted Analysis by GDyNets
발표자
()
초록
내용
A deep learning method called Graph Dynamical Networks (GDyNets) was suggested to learn atomic scale dynamics from MD simulations. GDyNets can classify local states around a target atom and obtain transition matrices. However, there are few studies using this method to analyze amorphous systems. In this work, an MD trajectory of LiTFSI/PYR14TFSI ionic liquid electrolyte (ILE) is trained by GDyNets. The trajectory is generated with the APPLE&P polarizable force field. Li+ ions are treated as target atoms and classified into 3 local states. State-wise structural properties are calculated to identify the states. The identified states are a Li+ ion coordinated by 2 or 3 TFSI- ions, or a cluster composed of multiple Li+ and TFSI- ions. Also, dynamical properties of the states or transitions are measured to reveal which state or transition dynamics is faster than others. Finally, a design rule for ILEs with faster Li+ ion conduction is suggested.