High-performance computing is the key to UND scientists’ explorations in nanoscience
|Juana Moreno utilizes high-speed computer siimulations to test hypotheses and study particles on the atomic and molecular levels.
Juana Moreno, assistant professor of physics, and Brian Moritz, postdoctoral research associate, are explorers in the realm of nanoscience, the study of structures and interactions that occur on the scale of billionths of a meter.
A human hair is about 10,000 nanometers in diameter. Moreno and Moritz study particles on the atomic and molecular scale, requiring the use of high-speed computer simulations to verify their hypotheses by building a bridge between more traditional theoretical and experimental scientific research methods.
The establishment of the Computational Research Center (CRC) last year at UND has been a boon to these researchers, as it has been to others in a growing number of fields (see Page 14 for the profile of another CRC user, Chemistry Professor Mark Hoffmann).
“High-performance computing clusters enable theorists to extend their studies of physical, chemical, and biological systems,” Moreno explains. “Multi-scale, multi-phenomena modeling is enhancing our understanding of such complex systems as strongly correlated materials, biomolecules, or atmospheric phenomena. A high-performance computing cluster that can efficiently run simulation codes in parallel is essential for this type of research.”
Moreno, Moritz, and their collaborators at Oak Ridge National Laboratory and the University of Cincinnati are searching for optimal materials from which to construct what are known as “spintronic” devices. These devices, she believes, may allow not only the transfer of electronic charges but also the transport of “spin,” a quantum property of the electrons. Right now the researchers’ focus is primarily theoretical, involving mathematical explorations at the tiniest scale.
But at the end of the trail lie some important practical applications.
Take computers, for example. A theoretical breakthrough in the physics of signal transfers like those that occur between a magnetic storage unit (hard disk) and the central processing unit (CPU) would have enormous impact. Moreno and her team hope to create a new kind of material for devices carrying what are known as “spin-polarized” electrical currents, thus allowing the integration of both storage and processing capabilities. This will lead to much faster, smaller, less expensive and more capable machines.
Much of their current work uses a technique known as “Monte Carlo Simulation.” It was so named in 1946 by its inventor, the mathematician Stanislaw Ulam, who had noticed that roulette wheels exhibited the kind of random behavior observed in nature at nano levels.
The technique solves a mathematical problem by generating suitable random numbers and observing that fraction of the numbers obeying some property or properties, thus obtaining solutions to problems that are too complicated to solve analytically. Moreno and Moritz use Monte Carlo simulations within the Dynamical Cluster Approximation (DCA) technique, one of the most powerful approaches for the study of electronic models.
“Dr. Moreno is a very promising young scientist whose research is already receiving national acclaim,” observed Peter Alfonso, UND’s vice president for research. “For example, she received one of the prestigious Ralph E. Powers Junior Faculty Enhancement Awards for her work in spintronics. Virtually all of the other award winners reside in the nation’s major research institutions.”
Moreno notes that UND’s computing facilities are equal to what she’s found elsewhere: “My group also has access to the Galaxy cluster at the Center for Functional Nanomaterials at the Brookhaven and Oak Ridge National Laboratories. The UND cluster compares favorably with these other resources available to our group.”
That’s music to the ears of Alfonso and to the partners who are in the process of expanding the UND facility.
Computational Research Center
The Computational Research Center (CRC), established last year to give University of North Dakota faculty and their research partners a new instrument for scientific computing, is moving to the next level through hardware upgrades and the addition of key support personnel.
The Center uses a cluster computing approach to give UND a capability once only available with a supercomputer and at an astronomical price, Alfonso said.
He noted that the Center, which reports to his division, is a partnership including UND’s Office of Information Technology Systems and Services (ITSS), the North Dakota Experimental Program to Stimulate Competitive Research (ND EPSCoR), and, on a contract basis, Meridian Environmental Technology of Grand Forks.
The original system featured a 48-node Linux Beowulf Cluster with dual Opteron 1.8-GHz 64-bit processors. Recent enhancements have included the addition of 16 computer nodes, increasing the system’s capacity to 64 dual-processor nodes, said ITTS Director Dorette Kerian.
Start-up and server administration costs for the initiative were funded by EPSCoR (Experimental Program to Stimulate Competitive Research), with ITSS providing secure physical space in Upson Hall, help desk services, cluster system backup, and system software. The CRC Users Committee, chaired by Richard Schultz, associate professor of electrical engineering, is now searching for a High-Performance Computing Specialist to assist researchers in migrating their custom simulation software programs to the multiprocessor cluster architecture.
Meridian, a spinoff company started by UND faculty member and Regional Weather Information Center Director Leon Osborne, administers the system. The current agreement includes a handoff of server administration duties to ITSS prior to July 2006.