For most of his career, Middlebury College physicist Noah Graham has been focused on the physics of the very large and very small. He's recently been investigating oscillons, clumps of waves that are localized in space but oscillate in time and do not disperse. Because of the large amount of energy required to form oscillons, a natural place to study them is in the early universe.

Using compute clusters at Middlebury College, the California NanoSystems Institute High Performance Computing Facility at the University of California, Santa Barbara, and the Applied Mathematics Computational Lab at MIT, Graham proved oscillons exist in models of particle physics. That work was published in 2007 in Physical Review Letters and Physical Review D.

Now Graham's using supercomputing resources to scale up to more realistic situations to see if these oscillons form spontaneously from the hot, chaotic conditions of the early universe as it expands. He used NCSA's now-retired Tungsten to begin scaling up his simulations; currently he relies on NCSA's Abe and Pittsburgh Supercomputing Center's Big Ben.

"To see oscillons form in a simulation, you have to describe a large volume of universe compared to the fundamental scales in the system, meaning you need a large lattice, which then expands as the simulation proceeds," he says.

The importance of localized objects has been recognized for a long time, Graham says, but most of the attention has been focused on the case of static solutions, which are easier to analyze.

"This work, however, suggests that the more general case of oscillating solutions, while more difficult to calculate, might be of considerably broader applicability. Stable objects of this kind are particularly interesting in the early universe, because they are out of thermal equilibrium—at equilibrium they would disperse as the universe cools. Such a departure from thermal equilibrium is a necessary condition for baryogenesis, the process that formed the protons and neutrons we see today," he explains. "Finding the mechanism of baryogenesis remains a major unsolved problem in cosmology; while it's too early to say if oscillons played a role in this process, it's at least an intriguing possibility."

Large-scale computing is needed not only to run the simulations, but also to extract meaningful results from the correspondingly large output, says Graham. Due to the extremely vast quantity of the data, the team was unable to analyze relationships within the entire dataset using off-the-shelf visualization software. So they turned to NCSA's Advanced Applications Support Visualization Group. Visualization expert David Bock is using a custom renderer he developed to volume visualize the entire dataset.

"Using my custom renderer, we were able to analyze the entire oscillons dataset using a variety of volume visualization methods," says Bock. "Modifications were made to the simulation based on these representations, more data was generated, and new visualizations were produced. This analysis and discovery process has been iterated several times over the past year."

Frame from a visualization depicting a single oscillon dataset rotating about its axis. The data is rendered using a volume visualization method that maps color to the maximum data value along a ray cast through the volume at each pixel location. Dark, cold hues (purple, dark blue) represent lower values while bright, warm hues (yellow, bright white) represent higher values of data.

Frame from a visualization depicting a single oscillon dataset rotating about its axis. The data is rendered using a volume visualization method that maps color to the maximum data value along a ray cast through the volume at each pixel location. Dark, cold hues (purple, dark blue) represent lower values while bright, warm hues (yellow, bright white) represent higher values of data.