Electronic structure simulations of polymer(s) on graphene through a Physics-informed neural network

Machine learning is a rapidly growing area that has recently gained ground in electronic structure simulations. Conventionally, density functional theory is the most prominent method of electronic structure method but requires one of the highest demands on academic high performance computing systems worldwide. Accelerating these with machine learning reduces the resources needed and allows for simulations of complex systems. Here we use a Physics-informed neural network to determine the electronic structure and calculate the adhesion ability and identify types of surface interaction of polymers angelica lactone (ALP) and polymethyl methacrylate (PMMA). ALP is a green (biodegradable) polymer that is capable of polymer assisted transfer of graphene, which is important for graphene device fabrication. The simulations show that the ALP binds to graphene more strongly in the presence of PMMA, which increases the Van der Waals interactions of graphene with angelica lactone by increasing the electrostatic forces between the polymers.