From the Inner Ear to the Stars in NPACI's Strategic Applications Collaborations

ew NPACI Strategic Applications Collaborations (SAC) have been announced by Jay Boisseau, associate director for scientific computing at SDSC and coordinator of the SAC program. Scientific targets range from the inner ear and the nervous system to the stars. The target computers include the new IBM SP teraflops system being installed at SDSC and a new Sun HPC 10000 server with 64 processors and 64 GB of shared memory. The projects have been selected with arrival of these systems in mind.

The SAC program seeks to enhance the effectiveness of computational science and engineering research conducted by NPACI users. The results of the first projects have been encouraging (see enVision articles on AMBER in the October-December 1998 issue, and on PARTREE and SCF in April-June 1999), and Boisseau expects that the program will achieve similar successes with the new group. The goal of these collaborations is to develop a synergy between the researchers and NPACI staff that accelerates the researchers' efforts by using NPACI resources--human and computational--most effectively. The idea, according to Boisseau, is to enable new science on relatively short timescales and to discover and develop general solutions that benefit not only these projects, but also specific disciplinary communities and users of high-performance computing generally.

SACs pair academic researchers who have scientific expertise with NPACI staff possessing computational expertise. The teams pursue the goals of capability computing by moving important codes to more modern computational architectures. The seven new collaborations and the collaborating principal investigators are listed below.

NONSTANDARD STELLAR ATMOSPHERES

THEORETICAL STUDY OF LEPTON ANOMALOUS MAGNETIC MOMENTS

PARALLEL CLIMATE MODEL

CONSTRUCTING A MODEL OF THE COCHLEA

QUANTUM CHROMODYNAMIC

GENESIS SIMULATOR

PHYSICS OF HIGH-DENSITY MAGNETIC RECORDING

Figure 1. The Givelberg Cochlear Model

A test run of a preliminary model of the cochlea. A wave can be seen starting to propagate down the long and narrow basilar membrane.

NONSTANDARD STELLAR ATMOSPHERES

Peter H. Hauschildt, University of Georgia
Edward A. Baron, University of Oklahoma

Hauschildt, Baron, and collaborator François Allard of the University of Lyon in France have spent many years developing their general stellar atmosphere code PHOENIX, which is capable of modeling stars ranging from cool brown dwarfs to supernovae. Spectra calculated from first principles within PHOENIX can be compared with stellar spectra obtained observationally. PHOENIX includes the effects of deviations from local thermodynamic equilibrium (known as NLTE effects), in contrast to "standard" stellar atmosphere models, which results in greater realism. The code has been parallelized for large platforms, and additional optimization will help broaden the utility of the code for myriad astrophysical investigations. Working with Hauschildt and Baron are Boisseau and Stuart Johnson of SDSC's Scientific Computing Group.

THEORETICAL STUDY OF LEPTON ANOMALOUS MAGNETIC MOMENTS

Toichiro Kinoshita, Cornell University

The electron anomalous magnetic moment is a fundamental constant of physics that has been studied with increasing precision over the past 50 years. Current computational results match the latest experimental results in precision, but the calculation and experimental verification of this constant to a higher precision will enable fundamental tests of quantum mechanics.

"At a small additional cost," Kinoshita says, "we will be able also to get a new value for the muon anomaly, which will provide the most rigorous test of the standard model of the electroweak interaction." The algorithm is an adaptive-iterative Monte Carlo routine, used to evaluate hundreds of Feynman integrals. Working on this fundamental theoretical problem with Kinoshita will be applied mathematician Bob Leary of the Computational Science Research Group, computational scientist Bob Sinkovits of the Scientific Computing Group, and computational scientist Sharon Brunett of the Center for Advanced Computing Research (CACR) at Caltech, an NPACI resource partner site.

PARALLEL CLIMATE MODEL

Tim Barnett, Scripps Institution of Oceanography (SIO)
Warren M. Washington, National Center for Atmospheric Research (NCAR)

The elements of this large climate model were developed by many investigators over a number of years. The Parallel Climate Model (PCM) is a joint effort, sponsored by DOE, among investigators at Los Alamos National Laboratory (LANL), the Naval Postgraduate School, the U.S. Army Corps of Engineers Cold Regions Research and Engineering Laboratory, and NCAR. The Climate Modeling Group at NCAR maintains the atmospheric portion of the model as part of the Community Climate Model (CCM). The Climate Change Research group maintains the PCM, which includes an overall flux coupling code that links the CCM atmosphere and land surface models to an ocean model (POP) from LANL, a model of the polar caps and sea ice originally developed by Yuxia Zhang of the Naval Postgraduate School, and a river transport model developed by a group at UT Austin.

Other investigators, including Barnett of SIO, participate in testing and refining elements of the PCM. Giri Chukkapali and Dong Ju Choi of the Scientific Computing Group are working with them to determine the best choices for moving this group of codes and the flux coupler into a next-generation configuration that can estimate climate parameters over tens of decades or longer as external conditions are varied.

CONSTRUCTING A MODEL OF THE COCHLEA

Edward Givelberg, University of Michigan

Edward Givelberg recently joined the mathematics faculty of the University of Michigan after completing his doctorate under the direction of Charles Peskin at New York University. He has used the theory of elasticity and differential geometry to derive a new shell theory that describes the elastic properties of the basilar membrane of the cochlea. The cochlea is the part of the inner ear responsible for hearing. Within the cochlea the basilar membrane is a complex structure, described in the model as an elastic shell (Figure 1).

Givelberg's code extends the general immersed-boundary method of Peskin's heart code to the case of an elastic shell. He will be assisted in optimizing his code for the new teraflops platform by Scientific Computing Group computational scientist Richard Charles, who is also continuing his SAC collaboration with Peskin's group. Working with them also will be physicist Julian Bunn of CACR and computational scientists Abhijit Bose and Randy Crawford of Michigan's Center for Parallel Computing.

QUANTUM CHROMODYNAMICS

Robert L. Sugar, UC Santa Barbara

Sugar is a leader of a group called the MIMD Lattice Computational Collaboration (MILC), which is a DOE Grand Challenge Application group and one of the world's premier theoretical particle physics groups. It also includes Claude Bernard of Washington University, Tom DeGrand of the University of Colorado, Carleton DeTar of the University of Utah, Steven Gottlieb of Indiana University, Urs M. Heller of Florida State University, James Hetrick of the University of the Pacific, Kari Rummukainen of Nordita, and Douglas Toussaint of the University of Alabama. Their MILC code is used to predict and calculate the properties and interactions of fundamental particles as tests of both experimental and theoretical results in particle physics. Sugar will work with Dominic Holland of the Scientific Computing Group and CACR's Brunett to capitalize on the incoming teraflops platform.

GENESIS SIMULATOR

James M. Bower, California Institute of Technology

Bower and colleague David Beeman of the University of Colorado are the main developers of GENESIS, a widely used environment for neural simulation that can model systems ranging from a single neuron to large networks, delivering outputs that match the results of many standard laboratory experimental techniques. "Optimization of GENESIS routines would benefit a wide community," said SDSC's Chukkapalli. He and Holland will work on the project with Bower.

PHYSICS OF HIGH-DENSITY MAGNETIC RECORDING

H. Neal Bertram, Center for Magnetic Recording Research, UC San Diego

A recent New York Times article proclaimed that IBM had announced a new record in magnetic storage density, packing 20 billion bits in a square inch. Advances like these are due in part to new understandings of the fundamental physics of magnetic recording that come from the experimental and theoretical program of the UC San Diego Center for Magnetic Recording Research. Neal Bertram has used a group of extremely sophisticated codes on SDSC and other supercomputers over the years, and the SAC project with Bertram and his group will be led by Larry Carter, professor of computer science at UC San Diego, working with Stuart Johnson and Amitava Majumdar of the Scientific Computing Group. "We're targeting a new, large-memory architecture and taking advantage of other optimization opportunities," Carter said. --MM *