The Scientific Grand Challenges for National Security workshop, held in Washington in fall 2009, centered on two questions:
- What challenges could scientists working in national security arenas tackle with exascale computers?
- What’s necessary to ensure the new machines can address these challenges?
The gathering also let scientists provide input for high-performance computing development and gave them a chance to learn about Department of Energy plans for the new machines.
Panels in six technical areas suggested a range of priority research directions in which exascale computing could play a major role.
Multiphysics simulation problems: Plotting the route to total nuclear disarmament, ensuring U.S. energy security, increasing human interaction with sophisticated technology, protecting data-gathering networks and improving data capture from them, and enhancing industrial competitiveness.
Nuclear physics: Computing and describing the physics behind fission and fusion reactions, including those in astrophysical phenomena like supernovae, and simulating integrated nuclear systems such as reactors.
Materials science: Simulating and understanding, often at atomic scale, the chemical, nuclear and mechanical interactions and stresses that can degrade materials.
Chemistry: Developing faster and more accurate predictive chemistry models, including quantum chemistry, molecular dynamics and multiscale simulations.
Science of nonproliferation: Establishing models and methods to better design sensor networks and to gather and analyze data from them, including programs to sift masses of data for unusual behavior.
Uncertainty quantification and error analysis: Enhancing ways to evaluate the potential error and uncertainty in models and their predictions, especially those incorporating a high number of uncertain parameters and large datasets.
Participants also recommended research on themes that were common to all six areas, including improving algorithms and computer codes, developing models spanning large time and space scales, addressing systems that couple multiple disciplines (such as climate and energy) and others.