Scientific Algorithms and Applications

On leading-edge computational science projects, NERSC staff scientists and mathematicians often collaborate directly with clients to develop methods, algorithms, and codes that address specific research problems. Recent collaborations have included research in nanoscience and materials science, cosmology, and climate modeling.

With the increasing interest in nanoscience, there is an urgent need for electronic structure calculations of ever-larger systems. NERSC staff scientist Lin-Wang Wang collaborates with Alex Zunger’s researcher group at the National Renewable Energy Laboratory (NREL) to address this issue. Recently Lin-Wang developed a charge patching method that allows the construction of the ab initio charge density of a nanosystem without doing a full self-consistent calculation (Phys. Rev. Lett. 88, 256402 [2002]). Based on the charge density, the single-particle Schrödinger equation can then be constructed. This procedure enables researchers to do ab initio quality electronic structure calculations for tens of thousands of atoms—a big advance over conventional methods, which can only calculate a few hundred atoms with the same accuracy. NREL researchers have used this method to calculate isoelectronic impurity levels for a series of semiconductors, and to calculate N pairs in gallium arsenide.

Andrew Canning continues to work with Malcolm Stocks and collaborators at Oak Ridge National Laboratory and elsewhere on computational methods for determining the magnetic structure of materials. Recently he collaborated on a study of the magnetic structure of y-ferromanganese (y-FeMn) alloy, a rare example of an fcc antiferromagnet, which has become a prototype for pinning layer studies in magnetoelectronic devices (J. Appl. Phys. 91, 7355 [2002]). The calculations were based on the constrained local moment model and used first-principles spin dynamics to obtain the ground state orientational configuration. Andrew collaborated on testing and optimizing the code for NERSC computers as well as running the code.

NERSC staff scientist Peter Nugent works with Greg Aldering and collaborators on the Nearby Supernova Factory (SNfactory), an international experiment designed to lay the foundation for the next generation of cosmology experiments that will measure the expansion history of the Universe using Type Ia supernovae. Peter is helping this group develop methods to quickly analyze about 50 GB of photometric and spectroscopic data each day in order to identify and calibrate discoveries and then plan and coordinate follow-up observations. These methods involve making statistical comparisons of the newly discovered events with large libraries of astrophysical objects, including stars, galaxies, novae, and supernovae. This work is currently being benchmarked and refined, with an expected implementation date of July 2003 to coincide with first-light for SNIFS, the integral-field-unit spectrograph on the University of Hawaii 2.2-meter telescope.

Two other examples of collaborative support by NERSC staff are described in detail elsewhere in this report: Chris Ding’s development of the MPH library for integrating multiple model components, and Julian Borrill's work on a new generation of cosmic microwave background data analysis tools.