Building an Error Measure for a Kinetic Monte Carlo Model

Abhijit Chatterjee
Indian Institute of Technology Kanpur

The evolution of materials defects is often governed by atomic processes that occur at time scales beyond the reach of accurate atomistic simulation techniques, such as the molecular dynamics method. Therefore, obtaining an accurate picture about the role of defects in influencing the long-time material structure and properties introduces major challenges. The development of new materials modeling approaches that are efficient and can accurately
study these “rare-event” atomic processes is essential to overcome these challenges. The kinetic Monte Carlo (KMC) method is a promising materials modeling technique that can
reach long time scales by using a fixed catalog of atomic processes and advancing time by selecting processes from the catalog. However, the KMC models have their own limitations, which prevent their direct use for study of defect evolution. These include the lattice
approximation in KMC where the continuous atomic positions are mapped to a discrete lattice and assuming the types of processes in the fixed KMC catalog. Except for a handful of
simple materials, it is nearly impossible to guess the important process mechanisms for a given material system. As a result, there are a number of situations, e.g., multicomponent systems, defects in materials including point defects, dislocations, grain boundaries and
interfaces, nucleation and growth problems, where the fixed process catalog assumption and on-lattice representation can be inaccurate. In this talk, I shall describe recent developments in my group to overcome many of the challenges of standard KMC. In particular, I shall focus on our approach for building a KMC catalog on-the-fly for a given material system using parallel MD simulations. I will also discuss how an error measure can be developed for
a KMC catalog. Finally, I will describe other developments in our group, such as the localenvironment
KMC method, enables us to improve the computational efficiency of KMC.

Back to Workshop II: Atomistic and Mesoscale Modeling of Materials Defects