Novel asynchronous computational framework for efficient fragment-based electronic etructure and dynamics for gas-phase and condensed phase problems

We present a weighted graph based molecular fragmentation approach for electronic structure, ab intiio molecular dynamics and for computing molecular potential energy surfaces. Our approach utilizes the local nature of chemical transformations to rigorously, and adaptively, partition any given chemical problem into a family of computational tasks that present varying degrees of complexity. Following this, we exploit the well-known MapReduce formalism in BigData, and the idea of asynchronous, stochastic computation in concurrency theory, simultaneously, to develop a seamless algorithm for correlated electronic structure, and high-dimensional potential surfaces. The computational implementation seemlessly integrates with a large range of available gas-phase and condensed phase electronic structure codes as is apparent from the attached graphic.