Science - Written by on October 31, 2011

Computation Proves Predictive Capabilities of Nuclei through Fluorine-14 Simulation

Tags: , ,

Scientists use Oak Ridge and Argonne supercomputers to gain insight into nuclear behavior

The Vary team made predictions about the behavior of fluoride–14 and published its results in Physical Review C. The predictions (ab initio bars) nearly matched an experiment done 6 months later at Texas A&M’s Cyclotron Institute (experiment bars). Image courtesy Texas A&M University Cyclotron Institute

As part of its quest to understand fluorine-14, a team led by Iowa State University physicist James Vary used Oak Ridge Leadership Computing Facility (OLCF) and Argonne Leadership Computing Facility (ALCF) resources to predict the behavior of this relatively unknown isotope. It published its predictions in Physical Review C in February 2010. Six months later, a group of researchers at Texas A&M University’s Cyclotron Institute performed an experiment producing fluorine-14, and the results nearly mirrored those of Vary’s group.

“Simulations have come to the stage of development where they are so precise that they can actually predict with some accuracy experimental results that have not yet been obtained,” Vary said. The team used the OLCF’s Cray XT5 Jaguar and the ALCF’s IBM BlueGene/P Intrepid supercomputers to run the Many Fermion Dynamics-nuclear (referred to as MFDn) code for an ab initio simulation of fluorine-14, meaning simulations focused on careful treatment of the neutron–neutron, neutron–proton, and proton–proton interactions responsible for predicted states of the element. The largest simulations ran on about 30,000 of the Jaguar’s 224,256 processing cores, and the team ran many simulations before submitting its predictions. “We held our breath,” Vary said in reference to waiting for experimental data to come from the Cyclotron Institute.

The research fits well with the Department of Energy’s (DOE’s) mission to explore extreme forms of nuclear matter, an area in which humans have limited understanding despite the fact that nuclei make up 99 percent of visible matter in the universe. By successfully charting the behavior of fluorine-14, the team’s research lays the groundwork for observing the strong interactions of subatomic particles in other unstable isotopes, moving toward a clear and comprehensive image of this type of nuclear matter and its properties. These comprehensive definitions take science closer to experimentally defining more unstable isotopes at DOE’s Facility for Rare Isotope Beams.

Vary’s principal scientific collaborators are Steven Pieper of Argonne National Laboratory, Pieter Maris of Iowa State University, Joseph Carlson and Petr Navratil of Lawrence Livermore National Laboratory, Andrey Shirokov of Moscow State University, Hai Ah Nam and David Dean of Oak Ridge National Laboratory (ORNL), and Witold Nazarewicz of University of Tennessee–Knoxville.

The team is supported by ALCF and OLCF resources through the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program as well as National Energy Research Scientific Computing Center resources. It is funded through DOE’s Scientific Discovery through Advanced Computing program and its office of Advanced Scientific Computing Research, as well as the National Science Foundation. The team is also part of a larger project aiming to define properties of all isotopes, Universal Nuclear Energy Density Functional (or UNEDF, for short), in which researcher Steven C. Pieper is working under the same INCITE allocation using 32,768 cores on the IBM BlueGene system at Argonne National Laboratory to make calculations of states of the carbon-12 nucleus.

Intricate interactions

In addition to predicting the mass and low-lying, excited-state behavior of the fluorine-14 nucleus, Vary’s team also correctly predicted that these unstable isotopes would stabilize by decaying proton emission, or “proton dripping,” which results in a lighter nucleus. “We can think of it as a wet sponge,” Vary said. “Fluorine-14 has protons in it, but it can’t hold them, so it must let one out very quickly. It doesn’t come out with a lot of energy, but enough to be observed.”

Many unstable nuclei appear for only microseconds and, in many cases, only in situations that would be very difficult to observe experimentally. Nuclear power reactors, for instance, create many short-lived materials during a nuclear reaction that are capable of influencing energy generation in reactors. “You can produce something that only lives for a second or two, but during that time it can absorb additional neutrons in the reactors,” Vary said.

Vary pointed to three main issues that make the project computationally difficult. The quantum many-particle problem means that the simulation must account not only for large numbers of particles interacting with one another, but also for the quantum mechanics—in which subatomic particles exhibit both particle-like and wave-like characteristics—that must be calculated with high accuracy.

The second issue comes from the extremely strong interactions among the neutrons and protons. Vary said many of the conventional tools physicists use to study particles are not applicable to their simulations of unstable nuclei. “We want to make sure we don’t lose some of the inherent predictive power that comes with treating [strong interactions] carefully,” Vary said. In treating both the quantum and strong interactions with such precision, a third issue arises—the massive calculations that must be performed to generate meaningful data for comparison with experiments.

ORNL’s David Dean was principal investigator on the team’s first 3-year INCITE project, which ran from 2008 to 2010. The team secured another 3-year allocation beginning in 2011 to work toward the same goals, and Vary, this project’s principal investigator, said he and his partners are definitely looking forward to continued collaboration with leadership computing facilities. “We’re very excited about the convergence of the technical capabilities that these new facilities represent with our own theoretical and algorithmic capabilities, meaning that we can productively use these new tools,” Vary said.—by Eric Gedenk