HOT FUSION: EXASCALE COMPUTING DRIVES EARTHLY EFFORTS TO PRODUCE THE ENERGY THAT POWERS THE SUN

Nuclear fusion processes similar to the sun’s have the potential to transform energy production on Earth. At least that’s the goal of California-based TAE Technologies, which plans to build the world’s first prototype hydrogen-boron fusion power plant, called Da Vinci, in the early 2030s. With Da Vinci, TAE’s seventh-generation test machine, the company seeks to fuel a renaissance in world energy production.

Although existing commercial nuclear plants make energy with fission, which breaks down large atoms into smaller ones, fusion combines smaller atoms to make heavier ones. Viable fusion energy could generate affordable, abundant carbon-free energy for everyone with no meltdown risk and far fewer environmental impacts and harmful byproducts than fission or fossil fuels. Some of the problems are that triggering fusion requires a lot of energy and an ongoing reaction is difficult to maintain.

In a journey toward making fusion an economically viable source of commercial energy, TAE has embraced the U.S. Department of Energy (DOE) Exascale Computing Project (ECP). As Toshiki Tajima, TAE chief science officer, says, “We want to create a whole-device model of a fusion reactor, and to do that we need the fast computation of the ECP framework.” As a member of the ECP Industry and Agency Council, Tajima can also interact with other experts from industry and government agencies on future objectives in high-performance computing. Work already underway at TAE promises a collection of important applications, from transforming energy production to modeling atmospheric processes.

Software for Simulations

Before constructing a prototype of a new reactor, TAE develops numerical simulations of the design. Simulating a fusion reactor requires connecting many different physics models into one framework. TAE took its workhorse 3D particle-in-cell code for monitoring global reactor stability and completely rewrote it to work with new exascale-capable ECP software products in order to jump start TAE’s use of the large-scale, fastest level of computers. One such product, WarpX, sits atop a software framework called AMReX, which provides the infrastructure for key ECP algorithms, including systems of partial differential equations and particle-mesh operations.

WarpX, which was designed to simulate laser particle accelerators, received the Association for Computing Machinery’s 2022 Gordon Bell Prize for outstanding achievement in high-performance computing. This code is simple enough to run a quick job on a fast desktop computer but versatile enough to run a far more complicated fusion computation on an exascale computer.

TAE already collaborates with the team at Lawrence Berkeley National Laboratory that led the international effort to develop WarpX. Sean Dettrick, TAE’s director of computational science, compares the WarpX and AMReX frameworks to a box of LEGO blocks. “You’re trying to assemble these new physics models with the same LEGO blocks,” he explains. TAE can use software tools like these to eventually model a complete reactor, and such gigantic simulations depend on software designed to run on exascale systems.

The Need for Exascale Speed

Although donut-shaped tokamak fusion reactors have dominated the field for decades, TAE’s devices have all used a cylindrical design based on its field reversed configuration (FRC), which keeps the plasma stable and confines the heat.

Time on DOE supercomputers — Summit at Oak Ridge National Laboratory, Theta at Argonne National Laboratory, and Perlmutter at the National Energy Research Scientific Computing Center, which is operated by the Lawrence Berkeley National Laboratory — gives a big boost to TAE’s 3D particle-in-cell code. With collaborators at the University of California, Irvine, TAE also accesses Oak Ridge’s exascale Frontier supercomputer via the DOE INCITE project. This world-leading supercomputer performs a quintillion — a million trillion — calculations per second. Although TAE has already begun designing Da Vinci, the company plans to use Frontier or whatever succeeds it as the most advanced computer available in its future work. As Tajima says, “We need this kind of large-scale, fastest-level of computers.”

To produce energy, a fusion reactor must maintain plasma — electrically charged gas — at temperatures as hot as the sun. With exascale computing, TAE can simulate the massive numbers of microscopic particles that interact with one another over the vastly different scales of time and space that operate in plasma physics.

The top magnetically confined fusion plasma of the type called Field Reversed Configuration (FRC) at TAE Technologies “C-2W” is visualized by a WarpX simulation. White curves are the magnetic field-lines and colors represent the plasma densities; the enclosing magnetic coils and the cylindrical vessel are also shown [The external (red) rings are magnets, surrounding the cylindrical vacuum vessel (gray) , which confine the cocoon-like plasma inside]. Fusion Plasma Image courtesy of TAE Technologies.

“Fusion is complicated and fast-moving,” says Tajima. “It’s not a human time scale.” The blink of an eye takes about 100 milliseconds. The plasma dynamics of a fusion-energy system occur in less than a millisecond.

Containing Jell-O with Rubber Bands

Creating and maintaining fusion depends on precise temperature and spatial control of the plasma, but that’s a complex task. As Edward Teller once put it: containing a plasma is akin to holding Jell-O together using rubber bands. To make a plasma more controllable, TAE thickens it with beams of high-energy particles and uses ECP to optimize that process.

“The plasma motions have to be controlled by our computers, but those have to be done very fast to simulate millisecond dynamics,” Tajima says.

Tajima compares the task to riding a bicycle, with the many continuous muscular, balancing and steering adjustments needed to stay on two wheels.

“If you don’t pedal, the bicycle is not going anywhere,” Tajima says. Pedaling also keeps the bicycle upright. Similarly, a fusion reactor must drive the plasma by injecting a particle beam at a certain energy, angle and timing to keep it stable.

A moving bicycle also needs steering, or it might veer to one side and topple. Feedback control serves as the plasma’s handlebars. TAE will incorporate full feedback control in Da Vinci’s design. Here, TAE researchers will use external magnetic fields to modify the plasma as it evolves.

“We inject a beam into the plasma, and then we have to monitor the beam direction or energy or plasma temperature or other properties,” Tajima explains. “Based on this property, we have to monitor which way the beam should now be energized or slightly different, just like our bicycle steering is adjusted.”

“For us to have a complete simulation of that system, we need to be able to represent those external actuators, as they’re called, which can change the state of the system,” Dettrick says. These external actuators include factors such as the magnetic field and injected beams of neutral atoms that interact with the plasma. Exascale technology makes simulating these factors possible.

Expanding the Applications

Although ECP-generated tools, such as the WarpX and AMReX frameworks, enable TAE’s designs of future fusion reactors, that’s only one possible application. As Dettrick  says, “It’s not just TAE people who are interested in this code,” which could be applied to questions in other fields.

As just one example, Dettrick points out that the FRC plasma model that TAE added to WarpX could be used for space plasma studies of the Earth’s ionosphere or magnetosphere.

As Dettrick concludes: “We’re proving the wisdom of ECP investments in the software stack and in the hardware by showing industry can take advantage of it, and it helps America and the world.”

More info on TAE Technologies.