Research Accomplishments 2007
We highlight here recent success stories of LANS researchers
in theoretical and algorithmic results, software development, and applications.
We also have two-pagers for our
2006 accomplishments.
The PETSc Sieve library
has been used to reengineer the popular PyLith code for tectonic and seismic simulation, enabling parallelism and simulation in varying dimensions.
The new solver FilMINT reduces previous "optimal" solutions by 4% in a case study of the design of a load-bearing thermal insulation system.
We have developed a computationally efficient method for optimizing electrostatic interactions between biological molecules such as proteins. These optimizations offer valuable insights into the means by which molecules recognize one another, binding their targets tightly and yet not binding other molecules. Drug design processes can also be guided by biomolecule electrostatic optimization.
We have been studying
the multiscale nature of the self-organization process exhibited by
active biological systems
such as biopolymer networks.
These systems operate far from equilibrium and actively dissipate energy to shape and maintain large-scale coherent structures.
We have developed a large-scale computational code, NEKCEM, for wake field and wake potential calculations in 3D accelerating cavity structures. NEKCEM employs a high-order numerical scheme that enables accurate and efficient computations with high performance in parallel on body-conforming meshes. This work is critical for accelerator design.
Petascale computing platforms, featuring P=10^4-10^5 processors are soon to be available to the Nation's science programs. Current developments in the transport code, Nek5000, are focusing on P > 10^6 to prepare for the next generation of computers and beyond. Several recent applications of Nek5000 illustrate the scientific potential of state-of-the-art solvers at the petascale.
|
