Learning why fusion plasmas
sometimes act unpredictably
a numerical simulation tool called NIMROD,
a group of plasma physics researchers is making strides toward understanding
why the magnetic “thermos” that insulates plasma in toroidal
fusion experiments sometimes springs a leak.
Fusion research seeks to generate energy by heating fuels like deuterium
and tritium to a very high temperature. The intense heat causes atoms
to ionize, form a plasma, and ultimately fuse to release energy. These
electrically charged particles are guided by a magnetic field. If a
reactor based on the concept of magnetic confinement can sustain the
plasma at sufficiently high pressure, it will produce energy in abundance.
However, the plasma doesn’t always cooperate, says Associate Professor
Sovinec. “As we add more particles and more energy, the plasma
can organize itself into collective motions that distort the magnetic
field and poke holes in it,” he says.
Sovinec recently became head of NIMROD, a computer code development
project sponsored by the U.S. Department of Energy (DOE) Office of Fusion
Energy Sciences. In this collaborative effort, researchers from across
the country are studying how to model this set of distorting motions,
commonly known as magnetohydrodynamic, or MHD, instabilities.
Within the last two years, the group has focused on an instability called
an edge-localized mode, or ELM, which releases some of the plasma’s
stored internal energy in short bursts. ELMs manifest near the boundary
between closed and open magnetic-field regions in tokamak-style experiments.
surfaces and magnetic field line trajectories in a numerical simulation
of the nonlinear MHD evolution of shot 86144 in the DIII-D tokamak
at General Atomics.
(View larger image)
Inside the boundary, where
the field lines are closed, there are large net plasma flows—but
the profile drops quickly near the open-field region. This sudden change
in flow leads to a transport barrier, which, says Sovinec, is good news
for fusion. “Ultimately, the inside is hotter and you’re
closer to getting a reactor-grade plasma,” he says. But the associated
pressure and charge-current gradients also tend to drive MHD instabilities,
In the past few years, researchers have significantly increased our
understanding of the linear properties of ELMs, says Sovinec. However,
NIMROD researchers are among the first to conduct global nonlinear calculations,
which yield a more comprehensive description of ELM events—in
this case, ELMs in a discharge of the DIII-D tokamak at General Atomics
in San Diego.
Capitalizing on recent advances in NIMROD’s modeling capabilities,
the team used a two-fluid model that enables the plasma electron dynamics
to evolve separately from the ion behavior. “It makes the simulations
more difficult because not only are there new effects to account for,
but it also extends the range of temporal scales you need to follow,”
The expanded model, which includes electron and ion behavior, reproduces
an important stabilization mechanism in the linear spectrum. “What
we find, then, is nonlinear coupling among the modes that are still
unstable,” he says. “It’s similar to waves beating
together and generating higher and lower harmonics. And those harmonics
group the wiggles from these ELMs into a helical band that goes around
the torus. So instead of having ripples that cover the whole outboard
side, we’re seeing perturbations that group together in just one
The findings are significant to fusion
researchers because certain classes of ELM events would damage components
of the planned ITER (International Thermonuclear Experimental Reactor)
experiment, the first magnetic-confinement configuration that will produce
a self-sustaining plasma. Improved understanding of ELMs will lead to
techniques for controlling them and, as a result, avoiding reactor damage.
In addition, the results point to a time when
the NIMROD team can incorporate ELM models into more comprehensive tokamak
simulations. “We would like to be able to do the modeling more
efficiently,” says Sovinec. “In doing this exercise, we’ve
really stretched the algorithm into places where we haven’t been
The recent calculation required
15 12-hour segments on the IBM SP5 computer at the National Energy Research
Scientific Computing Center, using 344 processors rated at 7.6 gigaflops
Currently comprising researchers from
UW-Madison, Utah State University, the University of Colorado at Boulder,
the University of Washington, the University of Oklahoma at Tulsa, and
X Corporation, the NIMROD team has existed for nearly a decade. In the
future, says Sovinec, its challenges not only will include developing
more accurate plasma models, but also developing new strategies for
moving into the petaflop generation of parallel computing, which will
enable them to run calculations on tens of thousands of processors.
Sovinec says the NIMROD project is a good
vehicle for training future computational physicists. At UW-Madison,
he advises or co-advises six PhD-level students and two postdoctoral
researchers who use the NIMROD code for everything from conducting simulations
of on-campus plasma experiments to applying it to astrophysical situations.
“They see how all the different aspects, ranging from analytical
plasma physics to the mathematics of numerical analysis, to getting
an efficient implementation of the algorithm, to running numerical experiments,
come together when modeling a particular situation,” he says.