A postdoctoral fellow at Lawrence Livermore National Laboratory is leading exhaustive computer modeling efforts to study a new approach to controlled fusion.
“It is critical to simulate these (fusion) systems as best we can before doing experiments,” says Frederico Fiuza, who is in Livermore’s Fusion Energy Sciences Program. “The complexity is so large and building and using these machines is so expensive that you need to know as best as you can what to expect.”
Fiuza spent much of his recently completed doctoral research embellishing OSIRIS, a modeling code to tackle uncertainties surrounding what’s called the fast-ignition approach to inertial-confinement fusion. He loaded OSIRIS on all 1,572,864 cores of Sequoia, the lab’s IBM Blue Gene/Q, achieving what Livermore says is the biggest simulation ever with that kind of program.
Fiuza has received 19.5 million processor hours for follow-up work on Mira, a smaller Blue Gene at Argonne National Laboratory. Fiuza’s award is through the Department of Energy Advanced Scientific Computing Research Leadership Computing Challenge program.
“I’m sure the time we have with Mira is going to be extremely useful and productive, but it will also give rise to new ideas we will want to test,” says Fiuza, eager to apply for even more computing hours.
“When we try to achieve fusion in a controlled way in the laboratory, we are trying to meet the conditions that exist in the sun,” Fiuza says, “which is in the state of a plasma” – charged particles interacting with magnetic fields.
The plasma physicist is driven by the aims of the National Ignition Facility (NIF): to channel what lights up stars into a controllable device delivering essentially carbon-free energy. “That is the beautiful thing about fusion energy,” he says. “In theory it allows us to have a clean and safe energy source that is very abundant.”
Because fusion fuels are largely derived from hydrogen isotopes found in water, each gallon of the ocean could produce the energy of 300 gallons of gasoline. Similarly, 50 cups of water have energy equivalent to two tons of coal. Such gains follow from Albert Einstein’s most famous formula: The helium atoms produced in a fusion reaction are slightly less massive than the hydrogen isotopes feeding it. Since Einstein’s equation posits that mass equals energy, that slight excess is converted to heat.
Controlled fusion technology generally takes two approaches. The first aims to confine the fuel, heated to temperatures like the sun’s, with a powerful magnetic field, usually in a donut-shaped chamber called a tokamak. The second takes advantage of inertia, using (in NIF’s case) a laser-driven implosion to squeeze the fuel, typically deuterium and tritium, to high densities and temperatures, igniting star-like energy conversions.
Inertial confinement fusion experiments are underway at NIF, Livermore’s $3.5 billion facility. Its central-hot-spot approach relies on simultaneously compressing and igniting a round fuel capsule until it implodes, much like how a diesel engine works.
The facility’s 192 intense laser beams focus on a pencil eraser-sized gold cylinder holding a BB-sized fuel pellet. The lasers make the cylinder emit X-rays that implode the pellet to a dense state capable of igniting a thermonuclear burn.
Fiuza and his colleagues are simulating another possible path. Their fast-ignition technique would separate fuel compression from ignition, he says. “Instead of compressing fuel all the way, we suggest compressing it to an even level with more relaxed constraints. Then shine another single, very intense, short-pulse laser into the fuel.” The additional ultra-fast laser pulse would accelerate high-energy electrons, transferring energy to the fuel’s center and heating it enough to trigger a fusion burn, even with the milder initial compression.
The catch: “We don’t yet have the short-pulse laser that would allow us to reach ignition conditions. This is a very complex process. So we need to understand very well how such intense lasers would interact with the fuel and generate these electrons.”
Fiuza’s simulations will employ the hybrid OSIRIS version he developed during doctoral work split between his home school, Lisbon’s Instituto Superior Tecnico (IST), and a guest stint at UCLA.
OSIRIS, developed jointly by UCLA and IST researchers for more than a decade, is based on a particle-in-cell (PIC) algorithm. PIC codes model the behavior of plasmas by computing on a grid the motions of individual charged particles under the effect of electromagnetic fields.
OSIRIS’ ability to flexibly describe complexities like laser-plasma interactions is crucial to modeling fast ignition. But a standard PIC code can’t simulate everything that happens where laser beams meet fuel pellets, Fiuza says. For example, fuel capsule sizes “are on the order of a millimeter, but to fully understand everything going on you have to also capture the motions of charged particles that populate the plasma, which can be down to the scale of an atom,” he says. “So we have many orders of magnitude difference between the largest and the smallest scales we need to capture to accurately model everything that is going on in our system.”
To address this multiscale problem, his Livermore colleagues proposed a hybrid algorithm Fiuza implemented in OSIRIS that modifies calculations in a fuel pellet’s environs.
“Our goal with this project is for the first time to capture all of this in a self-consistent way, and if possible in 3-D,” he says.