University of Wisconsin-Madison
Brian Cornille always had a knack for math, but an undergraduate research experience at the University of Wisconsin-Madison shaped his computational interests. When a chemistry course matched Cornille with a computational research project, postdoctoral researcher Scott Gruenbaum showed him the ropes: how to navigate the Linux language and write and compile code.
“That had a major impact on me, having someone who mentored me that strongly and invested time in me,” says Cornille, a Department of Energy Computational Science Graduate Fellowship (DOE CSGF) recipient.
Although he majored in both chemistry and engineering physics, Cornille soon moved toward plasma physics research. Fusion energy had intrigued him since middle school, and he looked for related research projects on the UW campus.
Cornille soon landed in Carl Sovinec’s lab, where researchers develop computational simulations to approximate plasma’s complex behavior in fusion experiments, a difficult process to monitor. “It's much easier to diagnose a simulation than an experiment,” Cornille says. “With the simulation you have a whole picture of what’s going on all the time.” He began by building a parallel version of DEBS, a legacy magnetohydrodynamics code, and remained in Sovinec’s group for his Ph.D.
Cornille has dug deeper into numerical methods for his graduate research. He began by exploring whether an alternate numerical method – first-order system least squares – could help scientists better understand how magnetic fields evolve in fusion plasma experiments. He and his colleagues simulate this energetic mixture of positively charged ions and negatively charged electrons as a conducting fluid at time scales from nanoseconds to tens or hundreds of milliseconds.
Next Cornille will explore methods to simulate vertical displacement events, potentially catastrophic disruptions in the hot fluid that can damage fusion devices. It’s especially important to find ways to incorporate more realistic features into the simulations. For example, many plasma physics simulations approximate the vacuum vessel that encloses the hot fluid as an ideal conductor. But to accurately simulate how plasma and the vessel interact in vertical displacement events, researchers must model the wall as a resistive layer instead.
The DOE CSGF has let Cornille focus on his core mathematical and computational interests: writing software for understanding the numerical challenges in these physics problems. “I have the freedom to focus on the nitty-gritty of the software and just the numerics side of things,” he says. Without that funding, he adds, “I'd be focused more on the physics than the software component.”
During his 2017 Lawrence Livermore National Laboratory practicum, Cornille worked with Dan White on BLAST, a hydrodynamics code that boosts simulation accuracy and facilitates work on massively parallel computers. BLAST uses high-order finite element methods, mathematical tools that provide better approximations for solving differential equations. Cornille added a component in BLAST to address the evolving magnetic field so the code could be used to simulate plasmas.
The DOE labs support for researchers developing computational codes and their frameworks impressed Cornille. “There are all sorts of layers to their code bases, and they care about having very high-quality code, which you can't always afford to do at a university.”
Cornille aims to finish his Ph.D. within the next two years, when both he and his wife, a computational nuclear engineer, plan to seek academic positions. (They welcomed a daughter into their family in 2019.) But positions at one of the DOE national labs could also be the right fit.
Image caption: This simulation of a plasma disruption within a tokamak fusion reactor shows how heat is deposited on the outside. The rope-like features are magnetic field lines within the plasma, and colors show plasma temperatures. (Orange is hotter, blue is cooler.) The simulation was produced with the NIMROD code Brian Cornille and his colleagues continually improve. Credit: Scott E. Kruger, Tech-X.