1. Simulating Intermolecular Collisional Energy Transfer and Chemical Reaction in Bath Model

​A simulation bath model is developed recently [J. Chem. Phys. 140, 194103 (2014)] to study collisional energy transfer phenomena. In this model, a 3D box implemented with periodic boundary conditions can be taken with a thousand or more solvent molecules. Any solvent bath density from liquid to gas can be considered. Collisional energy transfer as well as chemical reactions can be studied in this model. In the group, we have been performing many such projects. 

 

2. Unimolecular Dissociation Reaction

​Unimolecular dissociation is studied by chemical dynamics simulations of randomly excited clusters of aromatic molecules. Benzene dimer, benzene-hexafluorobenzene, benzene-hexachlorobenzene, etc. are studied at a temperature range of 700-2000 K. This temperature range is relevant for combustion chemistry. These studies provide detailed knowledge of their bindings and dynamical aspects. Moreover, the dissociation of the complex of aromatic molecule and metal ions were also studied. The dissociation rate constants are compared with canonical and microcanonical theories. The role of anharmonicity can also be extracted for these dissociations.  

   ​

3. Electronic Non-adiabatic Dynamics in Gas and Condensed Phase

​It is already well known by now that a strong or even a moderate coupling between the electronic states break down the Born-Oppenheimer (BO) approximation and the inclusion of non-adiabatic coupling terms into the Schrödinger equation becomes important. Being a pure quantum phenomenon, this beyond Born-Oppenheimer field become very popular among the quantum physicists and theoretical physical chemists. Some smart theoretical developments have been made over the last few decades to this field. Non-adiabatic quantum dynamics has also become art for reproducing experimental result, especially photoabsorption spectra. MCTDH, TDDVR are among the packages which are used rigorously for this purpose. In spite of its wide applicability, a fully quantum mechanical dynamics is still impossible even with a huge computation facility of current time for systems of many degrees of freedom. Therefore the classical mechanical molecular dynamics will be used as the workhorse for many coming years as well. It is of interest, therefore, to find out ways to incorporate quantum mechanical effects into molecular dynamics simulations. In 1990, John Tully showed us how to tackle quantum non-adiabaticity within a classical trajectory through his famous "fewest switch" algorithm. In this project a very efficient software for doing non-adiabatic surface hopping molecular dynamics in underway.

​

4. Development of Two-Body Potential Energy Function

​The potential energy surface (PES) of a molecular system holds many of its chemical properties. Dynamics, which is nothing but the spatial evolution of nuclei around the PES with time. The chemical dynamics simulations performed nowadays involve the integration of the classical equations of motion (Newtonian or Langevin), calculating the forces on atoms at each step either directly by electronic structure calculations, called direct dynamics or on-the-fly dynamics, and from analytical PES. Even for small-size systems, the use of an analytical surface may be a convenient choice because the direct dynamics may become unstable for some cases. GAfit is a genetic algorithm-based software which can parameterize PES in an easier way. It needs the atom-atom distances for a particular orientation and the energy value. For PES scans as a function of a distance for different orientations, the software can fit them simultaneously and provide generalized atom-atom potential in form of Buckingham, modified Buckingham, Lennard-Jones, or any other functional form. This is indeed a very powerful tool to parameterize PES for dynamics of larger system as well as system that is unstable in direct dynamics.

​

5. Machine Learning Approaches in Reaction Dynamics 

​At present, machine learning (ML) tools are being used in many fields of research, including physical, chemical, and biological sciences. In our group, we are trying to apply machine-learning approaches to predict dynamical behaviour of molecules without being simulated. In addition, we are also trying to predict non-bonded interactions of molecular species using artificial neural network (ANN) models. So far, we are able to predict the dissociation behaviour of aromatic complexes at different excitations using ML techniques. The interactions between solute solvent molecules are also predicted using the same technique. Currently, we are trying to build a dynamics software to perform reaction dynamics using ML model(s).