Next Generation Multi-Scale Quantum Simul. Software for Strongly Corr. Materials

This project is multi-funded between ERKJD44 and DPDP318. This project is part of the Scientific Discovery through Advanced Computing (SciDAC), "A Multi-Scale Many-Body Science Application for Strongly Correlated Materials," which proposes the development of a massively parallel multi-scale method for the study of strongly correlated materials. This method separates the problem into short length-scales treated explicitly with Quantum Monte Carlo (QMC) methods, intermediate length-scales treated diagrammatically using vertices obtained from the Monte Carlo, and long length-scales treated at the mean-field level. The codes we develop with this formalism will be used to gain a better understanding of materials where correlations over many length-scales are important, and to aid in materials design to improve and search for update correlated materials. Oak Ridge National Laboratory (ORNL) staff will be responsible for developing a parquet-based multi-scale many-body (MSMB) code. This task includes the optimization and parallelization of the measurements of four-point correlation functions needed for the multi-scale code, the optimization of memory requirements, and the optimization and parallelization of the linear algebra based parquet multi-scale code. Extending these codes to describe more complex systems will further require the implementation of longer-ranged interactions. This project addresses these challenges in an interdisciplinary manner, bringing together experts from computational many-body theory as well as from computer science and applied math into a coherent team.