SI2-SSE: Quantum Monte Carlo Software for a Broad Electronic Structure Research Community via Minimal Explicit Dependency (MED) programming

Cornell University, Ithaca NY

Investigators

Abstract

This project, jointly supported by the Advanced Cyberinfrastructure (ACI) Division, Division of Materials Research (DMR), and Division of Chemistry (CHE) at NSF, will provide a robust, usable software infrastructure for methods for materials analysis, which, when run on high-performance computers, have the potential to revolutionize and accelerate the discovery of materials for a variety of engineering applications ranging from better catalysts, to solar cells to thermoelectrics used for refrigeration. The big hurdle to doing this is that these methods for solving the fundamental equation of quantum mechanics, known as the Schroedinger equation, are either not sufficiently accurate or are too computationally expensive. The purpose of this proposal is to develop software for a class of stochastic techniques, called quantum Monte Carlo methods, that will allow scientists and engineers to solve the equation more accurately and thereby predict materials properties with greater accuracy. The software uses a new programming paradigm to make the use as well as the extension of the software by developers easier. If researchers find this programming paradigm useful, it may propagate to a much larger research community and may eventually become an integral part of a new programming language. Thus this project will contribute to the basic sciences and potentially contribute to the field of computer science as well. The goal of this proposal is to develop a software package consisting of a variety of quantum Monte Carlo (QMC) analysis tools. The software will have the capability to compute static and dynamic response properties, thereby greatly enhancing the range of applicability of QMC methods. Past research on optimizing many-body wave functions will aid in this goal since some of the formalism is the same. Both real space QMC methods, namely variational Monte Carlo (VMC) and diffusion Monte Carlo (DMC), and, determinant space QMC methods, namely semistochastic QMC (SQMC), which is an extension of the full configuration interaction QMC (FCIQMC) method, will be included, since each of these methods is the method of choice for some set of applications. The software uses the Minimal Explicit Dependency (MED) programming paradigm that greatly facilitates the development of complex programs. In particular, it lowers the barrier for researchers, who are not familiar with the inner workings of the program to contribute new functionality, thereby contributing to the long term survival of the software.