Research Projects

Molecular dynamics (MD) simulations play a vital role in understanding complex molecular and condensed-phase phenomena. However, traditional approaches face limitations in balancing computational scalability with predictive accuracy. Ab initio MD, while providing high accuracy through quantum mechanical (QM) calculations, is restricted to small systems and short timescales due to its intensive computational requirements. On the other hand, classical force fields (FFs), although computationally efficient, struggle to model reactive events and intricate chemical interactions accurately, particularly in diverse and multicomponent environments. To overcome these limitations, this project proposes the development of a novel Hessian-Trained Machine Learning Force Field (ML-FF) that integrates the precision of quantum chemistry with the efficiency and scalability of machine learning. Unlike conventional ML force fields trained solely on energies and forces, the proposed approach incorporates the Hessian matrix—second-order derivatives of the potential energy surface—to capture curvature information and improve accuracy, especially for non-equilibrium geometries along intrinsic reaction coordinates (IRC). The methodology involves three key steps: (1) constructing a comprehensive database of chemical reactions, including energy, force, and Hessian data; (2) training ML models capable of reproducing quantum-level accuracy; and (3) applying the trained ML-FF to simulate complex chemical reactions, including those in polymerization, crystallization, and catalysis, such as the oxygen-evolving complex (OEC) in photosystem II (PSII). The incorporation of Hessians enhances the model's generalization capacity and transition state prediction capabilities. This project is expected to transform computational chemistry by enabling scalable, accurate simulations of reactive systems that are otherwise inaccessible using existing methods. In addition to its scientific goals, the proposal emphasizes education and capacity building by establishing a world-class simulation and training program in atomistic modeling, contributing to the advancement of computational research and training in India.