Source: Sergi Molins
Before any supercomputer is accepted at the National Energy Research Scientific Computing Center (NERSC), scientists are invited to put the system through its paces during an “early science” phase. The new-generation pore-scale flow and reactive transport model developed at the ESD-led DOE Energy Frontier Research Center (EFRC) for Nanoscale Control of Geologic CO2 was among those chosen to test the capabilities of NERSC’s new supercomputer Edison (Related story: “Early Edison Users Deliver Results”).
ESD scientists Sergi Molins, Carl Steefel, Jonathan Ajo-Franklin, and Li Yang are collaborating with Computational Research Division scientists David Trebotich and Terry Ligocki on the problem of subsurface geological carbon sequestration (Related story: “Modeling Pore-Scale Geological Carbon Sequestration”). The first-of its-kind simulation of pore-scale reactive transport processes associated with CO2 injection and sequestration uses a centimeter-long capillary tube filled with crushed calcite (calcium carbonate) as an experimental and computational analogue of a flow path through the subsurface. The capillary tube is infiltrated by a high-CO2 solution, which reacts with the calcite as it flows through the tube and dissolves it, raising the solution pH and releasing calcium into the solution. The high-resolution pore-scale simulations (which solve the Navier-Stokes equations with a resolution of 1 micron, together with the multicomponent reactive-transport equations for the chemical system) are able to capture the micron-scale hydrodynamic and diffusion boundary layers at interfaces between reactive calcite grains and the flowing solution (Figure 1).
While earlier simulations were successfully performed on Hopper, NERSC’s previous flagship system, the unprecedented spatial resolution and process fidelity of the model has very large memory requirements: the simulation generates datasets of one terabyte for just a single 100 ms time step. Edison’s high memory-per-node architecture means that more of each calculation (and the resulting temporary data) can be stored close to the processors working on it. As a result, simulations on Edison are running 2.5 times faster than on Hopper, reducing the time it takes to get a solution from months to just weeks.
This research was carried out as part of the DOE Energy Frontier Research Center (EFRC) for Nanoscale Control of Geologic CO2. Important additional support came in the form of SciDAC-e ARRA funding that builds on years of algorithm and software development at LBNL within the SciDAC (ASCR) program in DOE. The Center is headed by Associate Lab Director Don DePaolo and is a collaborative effort led by Lawrence Berkeley National Laboratory that includes three other national laboratories and four universities. The extreme computing results shown here have been recently highlighted by NERSC [2] and have been included in the FY15 DOE ASCR budget narrative submitted for Congressional approval [3].
Related Citations:
Steefel, C.I., S. Molins, and D. Trebotich (2013), Pore scale processes associated with subsurface CO2 injection and sequestration. Reviews in Mineralogy and Geochemistry, 77.
NERSC 2012 Annual Review. “New Model Will Help Predict Stability of CO2 Reservoirs.” Science at scale highlight. to appear. http://www.nersc.gov/news-publications/publications-reports/nersc-annual-reports/
FY15 Budget Request to Congress, “Massively-Parallel Simulations Verify Carbon Dioxide Sequestration Experiments”, to appear, http://science.energy.gov/ascr/about/ascr-budget/
Funding Source: BES, EFRC-NCGC