National Center for Computational Sciences

Many achievements in technology—from magnets to high-temperature superconductors to the magnetoresistive materials used in modern disk drives—owe their usefulness to strong electronic correlations. Current methods such as quantum Monte Carlo (QMC) combined with dynamical mean field theory have made considerable progress in explaining correlated electronic and magnetic systems—for example, illuminating the origins of antiferromagnetism and superconductivity in these systems. Nevertheless, such techniques are fundamentally limited in their ability to simulate these enormously complex systems and have necessarily employed approximations that introduce systematic biases.

Over the last several years, Mark Jarrell of the University of Cincinnati and his group have developed the new algorithms that are required for the next generation of codes. Using codes such as Jarrell and his team will use the Dynamic Cluster Approximation (DCA), Multiscale Many-Body (MSMB) and the Maximum Entrophy Method (MEM) codes to systematically incorporate local corrections, the team will use 15 million hours on ORNL’s petascale Jaguar computer to calculate the complete set of length and energy scales for the correlated electronic and magnetic systems in these materials at a quantitative, material-specific level. The Dynamical Mean Field Approximation (DMFA) and its cluster extensions, including DCA, are at the heart of this approach, which This approach maps the lattice where electronic interactions occur onto a cluster embedded in a calculated medium. The correlations within the cluster are treated explicitly, while those at longer length scales are treated in a mean field approximation. The embedded cluster problem is solved with a massively parallel QMC simulation. The broad impact of this work is to tackle, at an unprecedented scale and accuracy, the physical process governing the emergence of new phenomena in systems where electron correlation plays a central role.