In inertial confinement fusion, a capsule of fuel begins at temperatures near zero and pressures close to vacuum. When lasers compress that fuel to trigger fusion, the material heats up to millions of degrees and reaches pressures similar to the core of the sun. That process happens within a miniscule amount of space and time.
To understand this process, scientists need to know about the large-scale conditions, like temperature and pressure, throughout the target chamber. But they also want detailed information about the material - and the atoms - contained within. Until now, computer models have struggled to bridge that gap across the wide range of conditions encountered in such experiments.
In a recent study published in Physical Review E, researchers at Lawrence Livermore National Laboratory (LLNL) and the University of California, Davis created a new framework that couples tiny, atom-scale simulations to code that describes the macroscopic world, all within the same simulation.
"We're talking about atoms on the order of nanometers versus, on the other hand, large flow fields on the order of meters," said Tim Linke, a UC Davis Ph.D. candidate conducting his research in-residence at LLNL. "The connection between those two lies in the material."
To create this tie, the team combined a hydrodynamics code from LLNL with a molecular dynamics code from Sandia National Laboratories. The former describes big-picture conditions and how they evolve over time at specific locations. That information feeds into the latter, which calculates how individual atoms in the material respond to those conditions. The work provides a crucial advantage: both simulations run concurrently. The atomistic simulations are conducted 'on-the-fly' alongside the large-scale code.
Running the framework requires massive computational power. After encountering insurmountable bottlenecks on other systems, the authors specifically tailored their code to work with the accelerated processing unit architecture of LLNL's Tuolumne supercomputer, which mimics that of its exascale sister system El Capitan.
The method could have an entire suite of applications, from studying fusion to planetary science to astrophysical phenomena like asteroid impacts.
"This paper is very general, and we kept it that way on purpose because we could go into phase transitions or defects or chemical reactions, or really all sorts of things that require this microscopic insight to be bridged into the macroscopic world," said Linke.
The new approach can be particularly useful for systems that evolve towards a different chemical composition during the simulation.
"That is the case of wetted foam targets used at the National Ignition Facility. The targets start as a microscopically mixed system consisting of a foam and deuterium but eventually, as the system is compressed and heats up, the chemical bonds will break," said LLNL scientist and author Sebastien Hamel. "The system becomes more atomically mixed, more homogeneous."
The code provides a path to better understand the material properties of such scenarios.
"This opens the door to new applications where nonequilibrium material behavior, including phase transitions (e.g., liquid to solid) and chemical reactions, can be concurrently simulated with the hydrodynamics using atomistic insights, which is the next stage of this research," said LLNL scientist and author Dane Sterbentz.
Other co-authors include LLNL's Philip Myint, Lorin Benedict, Kevin Korner and Jon Belof; and UC Davis' Jean-Pierre Delplanque. This work is a result of Jon Belof's Presidential Early Career Award.
