Argonne Town Hall Energy

1. Argonne Town HallEnergy May 31, 2007

2. There have been two previous Town Hall meetings • Berkeley • Nanoscience • Oak Ridge • Nuclear Energy • fission

3. Berkeley next steps • Focused on materials • This subject is on web-site • Carrier transport simulation for solar cell • Numerical design of catalysts • Numerical design of hydrogen storage • Biofuel • Nuclear Energy • Fission • NIF • Hydrocarbon fuels efficiency (combustion) • Efficiency at low emmisions • Liquid fuels

4. Berkeley focused on the science • We need to also articulate to the payoffs. • Conservation of existing technologies • increases in efficiency - hydrocarbon fuels • Much lower emissions – hydrocarbons fuels • Solid state lighting – a factor of 2 • New systems and materials • Cheap and efficient solar cell (organic system, nano) • hydrogen storage with the right binding energy • Catalysis • Hydrogen, coal

5. Oak Ridge next steps • Focused entirely on nuclear energy • Fuel performance • Reactor design • Separations • Integration • Discussion centered around what challenges and issues are in the way of having exa-scale applications ready by 2015 (stretch goal) • Not only technically complex, but complex in that new and innovative ways for engaging industry and regulatory agencies must be devised • In some areas legacy codes simply do not exist (simulation has never been applied in these areas, e.g., waste forms) – will this make it “easier”? • Requirements definition still evolving • Driven by two basic needs: design/analysis, safety/licensing • Depends upon user turn-around time required • The case for exascale is easy to make • Not yet posted on web • 4 presentations yet to be translated into prose

6. ITER • Goal is to develop a world-leading integrated modeling capability of major benefit to harvesting science from ITER ($10B international project to be built in France in 10-12 years) • Key for US strategy beyond ITER to DEMO phase of fusion development • V & V a key component of this proposed “FSP” – Fusion Simulation Project • Associated workshop report available soon (July ’07 time-frame) • Some of key codes (e.g., multi-scale kinetic transport dynamics) presently demonstrate very good scaling on leadership class platforms (CRAY XT-3 Jaguar @ ORNL, BGL @ IBM Watson, Earth Simulator in Japan, etc.) – excellent potential for petascale & eventually exascale • Other key codes (e.g., MHD dynamics) need to significantly improve scalability on leadership class platforms – where the “science scales with the number of processors used • System integration will demand modern frameworks where modules with improved predictive physics capability can be readily incorporated • Connections to international facilities over the next decade important from a validation perspective

7. Some additional comments • Need exa-flop computing at all scales • 1,000’s of cpu’s to millions of cpu’s • Time to solution from minutes to weeks • Number of users is measured in multiple thousands not 10’s or 100’s • Theory and algorithm development will match or exceed the factor of 1,000,000 achieved by the transition from teraflops to exa-flops • Need a systems, economic overview