Kurt Brorsen

My research interests lie in theoretical and computational chemistry, specifically the development of new techniques in electronic structure theory (EST) and their application to important problems in the chemical and related sciences. As in most computational sciences, recent computer advances have allowed the ability to calculate the answers to problems once thought impossible. These advances have also led to the development of new methods that have been designed with high performance computing in mind.

One of these methods is the Fragment Molecular Orbital (FMO) method. FMO breaks a large system into essentially independent smaller fragments and performs EST calculations on each fragment. The impact of the remainder of the molecular system is accounted for with high accuracy by a Coulomb bath. The division of a large molecular system into many smaller, independent fragments provide a means for FMO to take advantage of a massively parallel computing environment by employing multi-level parallelism: each fragment on a separate node and fine-grained parallelism with nodes. This makes the study of complex systems like proteins and polymers with accurate EST methods feasible. Because of the complexity of accurate EST methods, extending the FMO method to the petascale and beyond will not be trivial, and novel new approaches in both EST algorithm design and middleware computational developments will be needed to overcome the new bottlenecks that will be encountered. For example, we will likely have to rethink how each of the myriad tasks encountered in an EST calculation is coded. Computational science approaches like adaptive programming and computational quality of service will become very important.