Machine Learning Integration
Developing novel deep learning approaches for creating accurate interatomic potentials that maintain quantum mechanical precision while enabling large-scale simulations.
Machine Learning Potentials for Atomistic Systems
My research focuses on the development and application of machine learning potentials augmented with enhanced sampling simulations for atomistic systems and condensed phase reactive systems. This interdisciplinary approach combines the accuracy of quantum mechanical methods with the computational efficiency required for large-scale molecular simulations.
Chemical Reaction Modelling in Explicit Solvent
We have developed a deep learning-based active learning strategy to create ab initio level accurate machine-learned (ML) potentials for solution-phase reactive systems. Using these ML potentials, we perform enhanced sampling simulations to efficiently sample reaction processes. Multiple transitions between reactant and product states enable us to calculate converged free energy surfaces for chemical reactions.
As a prototypical example, we investigated the Menshutkin reaction, a classic bimolecular nucleophilic substitution reaction (SN2) in aqueous medium. Our analyses revealed that water stabilizes the ionic product state through enhanced solvation, facilitating the reaction and making it more spontaneous. This approach significantly expands the scope of studying chemical reactions under realistic conditions, such as explicit solvents at finite temperatures, closely mimicking experimental conditions.
Research Highlights
Enhanced Sampling Methods
Implementing advanced enhanced sampling techniques like metadynamics to explore rare events and calculate free energy landscapes of complex chemical processes.
Reactive System Modeling
Studying chemical reaction mechanisms in realistic environments, including explicit solvent effects and finite temperature conditions that closely mimic experimental setups.
High-Performance Computing
Leveraging advanced computational resources and algorithms to bridge the gap between theoretical models and experimental observations at the molecular level.
Quantum Chemistry Integration
Combining DFT calculations, ab initio molecular dynamics, and quantum mechanical methods to provide accurate descriptions of electronic structure and chemical bonding.
Data-Driven Approaches
Implementing active learning strategies and efficient data sampling techniques to optimize training datasets for machine learning potential development with minimal computational cost.