Spring 2008
In This Issue:
Materials Modeling and Simulation
The Penn State Institute for Computational Science (ICS) is a small organization with a very large vision for advancing research across disciplines at Penn State. Padma Raghavan is the director of the new institute, which was created in July 2007, to take advantage of the opportunities arising out of the phenomenal growth in computational systems and discovery, and the emerging ability to translate the knowledge of individual disciplines into algorithms that can be run on a computer.
“The ICS is about using computing as the driving force to discovery,” says Raghavan, professor of computer science and engineering and a true believer in what may be the next revolution in scientific research. (Another true believer is the National Science Foundation, which is making a big push for what they call “multidisciplinary cyber-enabled discovery and innovation” through its Office of Cyberinfrastructure.)
As a tool for scientists, the computer and its attendant infrastructure of software, connectivity, and expertise, collectively known as the cyberinfrastructure, has the potential to be as transformative as the microscope and telescope in an earlier age or the atomic force microscope in the 1980s, opening up areas of knowledge previously beyond our ability to retrieve.
Recent examples are not hard to come by. In quantum cosmology, for instance, computing for the first time allowed physicists, led by Penn State Professor Abhay Ashtekar, to simulate the evolution of the universe prior to the Big Bang; in bioinformatics, the ability to analyze and compare large volumes of genetic data created by the Human Genome Project will someday allow individualized therapies for disease. To analyze the masses of data, Penn State biologist Webb Miller helped develop a widely used sequence comparison engine, called BLAST, that is a high speed program for looking at data to find similar sequences of DNA; in the field of materials science, superalloys that can retain their strength at high temperatures are being modeled by Penn State Professor Long-Qing Chen for NASA’s Ultra-Efficient Engine Technology program. Computer modeling is also helping Penn State scientists push the boundaries of miniaturization of electronic devices into the molecular and atomic scale, so-called molecular electronics.
A Contribution to Materials Research
Raghavan has made her own contribution to materials computation, along with Penn State materials scientists Long-Qing Chen and Zi-Kui Liu, and former research associate Keita Teranishi, now at Cray, Inc. Together they invented a method for storing the vast amounts of data collected in imaging the microstructure of materials, for which they have filed a provisional patent.
Understanding the microstructure of new materials and composites is important for determining how materials will perform under real world conditions. But digitizing the microstructure data is computer memory and storage intensive.
Using a software technique that compresses and regenerates the data through incremental changes to the resolution of images, along with a sparse skeletal technique that captures only the boundaries of different phases and grains, their method reduces the data to 1 percent of its original size, far smaller than existing compression software. Additionally, their method reduces the distortions common to compression by using a simulation program that predicts the realistic behavior of microstructures as the images are regenerated. Their software is so efficient that it can run on current workstations and PCs, compressing, decompressing, and regenerating images within a matter of seconds.
Leapfrogging the Competition
Penn State did not jump into the revolution in cyberscience from the beginning, and now there is catching up to do Raghavan says. This should not involve competing head-to-head with such established computational powerhouses as the University of Illinois-Urbana Champaign, or UC Berkeley, says Raghavan, but by using Penn State’s enormous research reach to carve out new areas of excellence. “We need to be known as Computational X – nano, infectious diseases, personalized therapies, or whatever. We may be able to leapfrog those universities with a 20-year lead, if we can head in new strategic areas, maybe energy and the environment, or health and human development,” she explains.
Computational discovery is inherently multidisciplinary, and the NSF, which sponsors the broadest range of basic science research, has made it clear that they are interested in sponsoring interdisciplinary teams through their Office of Cyberinfrastructure, which is structured to cross the seven discipline based directorates. But working across the traditional borders of departments and disciplines requires practice. “We need bigger teams and resources to provide seed grants to those groups to do the initial steps to be competitive in center-style proposals,” Raghavan states firmly. “You cannot walk down the hall and try to respond when these nationally competitive multidisciplinary solicitations come in. You only get the depth of ideas through existing teams, from working together. You need to find problems at the intersection of many disciplines that you can also take back to advance your own discipline.”
While Penn State decides whether to make a big commitment to computing and data analysis infrastructure, which Raghavan believes is inevitable for any university that wants to keep up in the era of big science, the Institute for Computational Science forges ahead in building support across departments and institutes. So far, sixty faculty members have signed up to be part of ICS, and a recent workshop had eighty participants, helping build the community of computational researchers. Penn State’s other multidisciplinary institutes and thirty faculty have partnered with ICS to go after a major equipment grant for a terascale computer node with a petabyte of storage. “We are not competing with the other institutes,” says Raghavan. “There are synergies across disciplines that we leverage to embark on an exciting, collaborative voyage of discovery.”
Contact:
Dr. Padma Raghavan, director of the Institute for Computational Science, is a professor in the Dept. of Computer Science and Engineering in the College of Engineering, http://www.research.psu.edu/ics/.
