Integration of spherical harmonics descriptor with neural network
In order to decrease the computational cost for calculations of energy and forces for nanoparticles (>100 atoms), we have developed spherical harmonics based descriptor which is applied to neural network. This integrated technique reduces the complexity in molecular dynamics simulations for long time scales.
Neural network potential for metallic nanoclusters
We have modelled a global potential energy surface using neural network for sodium clusters (Na20 to Na40) and gold clusters (Au17 to Au58). We have applied these potentials to study the thermal stability, fluxionality, and probabilities along with many other thermodynamic properties.
Neural network potential for nano-alloys
To examine the effect of composition for gold nanoclusters, we have modelled a global potential energy surface for (AgAu)55 system. By applying this potential we have derived c-T Phase Diagram and Landau Free Energies to check the thermal stability and fluxionality throughout the composition range.
Force field model for nucleobase clusters
In order to study the non-covalent interactions in nucleobase clusters, we have used force fields such as AMOEBA and OPLSAA.
Global optimizations and structural evolution in molecules and clusters