Developing machine learning methods for chemical problems.
Here we are mainly interested in modelling potential energy surfaces of nanoclusters,nanoalloys.
Studying small gas adsorptions on the surface of nanoalloys using ML based potentials.
Developing transferable ML based interatomic potentials and atomic/molecular based descriptors to study structure, dynamics, repnose properties of chemical systems
Using hardware technologies(FPGA,GPUs) to parallelize our programs.
Here we are interested in developing a heterogenous computing platform, a hardware-software co-design,
to implement parallel algorithms in an FPGA framework to reduce computation time. In case of a typical MD simulation, we developed a hetrogenous model
such that the computationally expensive force calculations are implemented on FPGA while the rest of the simulation is carried on a PC. We believe such a system will drastically
reduce the computational time and pave the way to parallelize the code efficiently.
IEEE Access 10,40338(2022)
3.
ML based Kinetic energy functionals for developing orbital free DFT(OFDFT).
Developing a Kinetic energy density functionals is a major challenge in OFDFT. We are trying to solve this problem with the help of ML methods
to directly map the electron density to Kinetic energy density functionals.