After discovering that gold is catalytically active at nanoscale, it is always important to understand such peculiar activity of gold at atomic level. This leads to knowing the nanoparticle's most stable configuration, dynamical and thermodynamic properties. Knowing the structure and properties of such nanoparticles will also allow biomedical engineers to identify appropriate binding sites for drugs used to treat cancer and other diseases. The findings could also optimize the use of gold nanoparticles in catalyzing the oxidation process that transforms dangerous carbon monoxide emissions into the less noxious carbon dioxide.
In Computational Chemistry, machine learning (ML) is being increasingly used in the past decades to save the computational cost and overcome the bottleneck of evaluating atomic forces required for geometry optimizations and molecular dynamics. Development of algorithms and softwares, involving ML techniques, will help us in extending quantum mechanics and molecular dynamics to nanoscale.
With above goals in mind, we are trying to develop ML methods, such as artificial neural networks (ANN) to construct interatomic potentials which are many orders of magnitude faster and accurate in evaluating atomic forces. We are trying to use these methods to evaluate total energies and atomic forces of medium sized gold clusters, nanoalloy clusters and thio protected gold clusters. Using a better input functions in ANN to decribe the atomic environments we could easily extend this to nanoparticles of sizes greater than 1.5nm.
We mainly focus on following:
1. Developing ANN methods and appying it to bare gold and coated gold nanoparticles.
2. Modelling of atomic enviroments
3. Developing sampling algorithms
4. Appying above methods to study gold nanoparticles, nanoalloys, coated gold nanoparticles.
5. Using hardware technologies to parallelize our programs that will help to extend the above methods to experimentally relevant sizes.
Dr. Satya S. Bulusu
Chemistry (Theoretical and Computational Chemistry)