A GPU Accelerated Lennard-Jones System for Immersive Molecular Dynamics Simulations in Virtual Reality

Abstract

Interactive tools and immersive technologies make teaching more engaging and complex concepts easier to comprehend are designed to benefit training and education. Molecular Dynamics (MD) simulations numerically solve Newton’s equations of motion for a given set of particles (atoms or molecules). Improvements in computational power and advances in virtual reality (VR) technologies and immersive platforms may in principle allow the visualization of the dynamics of molecular systems allowing the observer to experience first-hand elusive physical concepts such as vapour-liquid transitions, nucleation, solidification, diffusion, etc. Typical MD implementations involve a relatively large number of particles N = O($$10^4$$) and the force models imply a pairwise calculation which scales, in case of a Lennard-Jones system, to the order of O($$N^2$$) leading to a very large number of integration steps. Hence, modelling such a computational system over CPU along with a GPU intensive virtual reality rendering often limits the system size and also leads to a lower graphical refresh rate. In the model presented in this paper, we have leveraged GPU for both data-parallel MD computation and VR rendering thereby building a robust, fast, accurate and immersive simulation medium. We have generated state-points with respect to the data of real substances such as CO$$:2$$. In this system the phases of matter viz. solid liquid and gas, and their emergent phase transition can be interactively experienced using an intuitive control panel.