I would like to apply for participation in the upcoming ACTS tools Workshop at Lawrence Berkeley National Laboratory. Currently I am involved in a computational project at Argonne National Laboratory under the supervision of Dr. Hans Kaper and Dr. Gary Leaf, as well as in on-going research in High Performance Scientific Computing at Old Dominion University under the supervision of Dr. David Keyes and Dr. Constance Schober.

While at Argonne I work on numerical simulation of dynamics of nano-dispersed spring magnets in presence of external fields, addressing the computationally expensive far-field evaluation procedure. The far-field (stray field or the de-magnetization field) comes from the non-local part of the energy functional and essentially couples the problem, making its evaluation the most expensive part of the total force calculation. Since the force field drives the dynamics of the system, the field-evaluation has to take place at each timestep, becoming the bottleneck of the simulation; thus, it requires efficient algorithms and significant computational resources. At present the 2D codes have been developed and are mostly sequential and in need of parallelization and further optimization. However, it is 3-dimensional models that are of ultimate interest and 3D codes are to be developed in the nearest future as soon as lower-dimensional problems are well-understood. The codes run on Argonne's 80-node IBM SP2 and networks of workstations running Linux, Solaris and IRIX. In software development we have successfully used PETSc - a parallel PDE solver library developed at Argonne. In the future we would like to be able to use other tools to facilitate the development of efficient parallel codes. Of particular interest to us are PDE libraries, parallelization and visualization tools.

In my work in High-Performance Scientific Computing at Old Dominion University I am involved in research and development of geometric PDE integrators for time-dependent non-linear problems. Such integrators are designed to preserve geometric features that PDEs might possess, such as conservation laws, Hamiltonian and symplectic structures, various symmetries.

Although still mainly a research topic, these numerical schemes have been gaining popularity and are already being used in various applied problems with success. We have designed a family of non-canonical symplectic integrators for the Non-linear Schroedinger equation and tested their performance against other comparable solvers. In the future I would like to investigate a particular class of PDE integrators, the so called multi-symplectic schemes, for Hamiltonian non-linear wave equations in 2 and 3 spatial dimensions. I am planning to design and implement a family of these schemes to access their accuracy, stability, run-time performance and other properties in comparison with more standard methods. For this I would like to be able to use existing code development environments, packages and visualization tools, and to devote most of the time to more specific research.

It is my experience that robust and versatile software tools are extremely important in any computational study. Thus, I feel that learning about such new tools would benefit my research in many ways.