For more than half a century, scientists and engineers have pursued the promise of nuclear fusion power with a plan that seems impossible: employing powerful magnetic fields to contain plasmas 10 times hotter than the sun. It’s a task, as Edward Teller was quoted as saying, that’s like “trying to hold a blob of jelly with rubber bands.”
Nonetheless, today a mesh of magnetic field lines is capable of containing much of a superheated plasma. Too many high-energy particles still escape, however, frustrating efforts to maintain the desired fusion reactions and ravaging fusion machines’ linings. To solve these problems, researchers are developing simulations to run at Department of Energy national laboratories on some of the world’s most powerful computers.
A nuclear fusion power plant would have tremendous advantages over today’s fossil-fuel energy. Fusion works by merging small atoms into larger ones, converting wisps of mass into enormous amounts of energy. As a result, fusion-generated power would produce no carbon dioxide.
Existing nuclear installations use fission – splitting large atoms into smaller ones and generating long-lasting radioactive byproducts. Fusion would produce safer, shorter-lived wastes than fission. Fusion plants also fail safely, avoiding devastating accidents like those at Fukushima and Chernobyl. Fusion stops on its own the moment the device loses the strict conditions that allow atoms to merge.
But such a power source will become available only after scientists solve plasma containment problems. The solutions will be vital to the success of ITER, a major nuclear fusion experiment jointly run by 35 nations, including China, the European Union, India, Japan, Korea, Russia and the United States. Estimated to cost $17.8 billion, ITER is a
major investment in the future for all involved.
The intergovernmental collaboration is building the torus, or doughnut-shaped, ITER tokamak reactor in Cadarache, France, about 50 kilometers northeast of Marseille. Housed in a six-story building, ITER will be twice as large and 16 times heavier than any previous fusion reactor. It will use massive superconducting magnets to weave a mesh of magnetic fields more than 19 meters in diameter and 11 meters high.
Inside that torus, radio waves will superheat a mix of hydrogen isotopes – deuterium and tritium – until the gases reach temperatures between 100 million and 150 million degrees Celsius. As the isotopes heat up, electrons will break free from their atomic nuclei and the gas mixture will become a plasma, a free-flowing form of matter comprised of positively charged nuclei, or ions, and negatively charged electrons.
Finally, the nuclei will move fast enough for some to overcome their mutually repelling positive charges. They will collide, fuse and release some of the energy that had bound their original protons and neutrons.
Scheduled to start operating in 2020, ITER is designed to produce 10 times more energy than it uses. If it achieves that goal, the experiment will set the stage for DEMO, the first nuclear fusion reactor to generate energy for practical use.
A major obstacle to ITER’s success lies in the plasma itself. “A strong electrical current flows in the plasma. This current itself produces magnetic fields,” says Stephen C. Jardin, a professor of plasma physics at Princeton University who directs the Center for Extended Magnetohydrodynamic Modeling, a part of the Scientific Discovery through Advanced Computing (SciDAC) program.
One of ITER’s key goals is to contain a burning plasma – one that becomes self- sustaining after external heat sources trigger the first fusion reactions.
Thus, the plasma flow itself generates some of the jelly-restraining rubber bands, forming an important component of magnetic confinement. “However, if a disturbance gets set up, it could grow and eventually destroy the magnetic configuration and cause the current to stop,” Jardin continues. “We call this a major disruption. Our goal is to better understand what causes the major disruption to occur, and to thereby take steps to prevent it.”
Jardin is principal investigator of a project to simulate the internal dynamics of the plasma in ITER, in collaboration with researchers at General Atomics and Rensselaer Polytechnic Institute. Using the Hopper Cray XE6 and Edison Cray XC30 supercomputers at Lawrence Berkeley National Laboratory’s NERSC, the National Energy Research Scientific Computing Center, the group will work to replicate the causes of major disruptions and discover how to quell them.
Jardin likens a major disruption in a tokamak to a severe hurricane. “Because the disruption occurs interior to the tokamak, we have more preventive options at our disposal than does the National Weather Service in trying to prevent a hurricane.”
But the magnetohydrodynamic equations that describe the roiling of hot plasma are far more complex than those that describe the weather, Jardin notes, because they must solve the interactions with electric and magnetic fields. “These equations contain a large range of space scales and time scales that must be included in the calculation,” thus requiring a close-knit grid of data points distributed through the simulated plasma, and lots of time cycles. Computers calculate the physical changes at each point. Taken together, they form a complete picture of the plasma, much as millions of pixels comprise a digital image.
The team has written algorithms that can model plasma flow in especially large time steps. Collaborators at Rensselaer’s Scientific Computation Research Center, Jardin says, “form meshes for us that vary greatly in size and resolve only the regions in space that need high spatial accuracy.”
The group’s simulations qualitatively agree with major fusion experiments, including the National Spherical Torus Experiment at the Princeton Plasma Physics Laboratory (PPPL), DIII-D at General Atomics and C-Mod at the Massachusetts Institute of Technology. “We are in the process of making more detailed comparisons to make this more quantitative and thus define error bars for our calculations before we apply these to ITER,” Jardin says.
Meanwhile, a research team at the University of California, Irvine, is focusing on one especially troubling cause of instabilities in tokamak plasmas: the very particles the system is designed to produce.
These energetic particles, produced by fusion reactions and heat from outside the tokamak, can excite magnetohydrodynamic instabilities that let the particles escape, possibly ending the reaction and damaging the device, says Zhihong Lin, leader of the project and professor of physics and astronomy at UC Irvine. Collaborators include researchers at General Atomics, Oak Ridge National Laboratory, and computer chip maker NVIDIA.
The team is using one of the world’s fastest supercomputers, Titan at the Oak Ridge Leadership Computing Facility, to simulate the shear Alfvén wave, the most prominent type of burning-plasma instability in torus-shaped containment systems. “The goal of our project is to predict and to optimize the confinement properties of energetic particles in magnetically confined fusion experiments such as ITER, the crucial next step in the quest for fusion energy,” Lin says.
The group uses an established simulation program, the Gyrokinetic Toroidal Code (GTC), which simplifies the model by omitting tight spiraling motions charged particles undergo when they flow through a magnetic field.
“This simulation model has been very successful in simulating microturbulence – excited by thermal plasmas – in fusion plasmas,” Lin says.
The group has stretched GTC to new uses, modeling energetic particle turbulence and the transport of energetic particles out of the plasma in three state-of-the-art fusion simulation codes, Lin says. The researchers “have utilized the new capabilities to discover new physics of energetic particle confinement.”
Whereas Jardin’s and Lin’s teams search for clues to stabilize the burning plasma in tokamak reactors, Oak Ridge Staff Scientist David L. Green focuses on the most efficient means of heating the plasma to critical temperatures.
One of ITER’s key goals is to contain a burning plasma – one that becomes self- sustaining after external heat sources trigger the first fusion reactions. Only then could it generate 10 times more power than is required to run it.
“When it comes to ITER, efficiency is key,” Green says. The most cost-effective plasma-heating method known is called ion-cyclotron RF heating (ICH), he says. The term ion cyclotron refers to the tight spiraling motions of positively charged deuterium and tritium
ions – the same motions that are a dispensable nuisance for Lin’s simulations. These cyclotron orbits, or Larmor orbits, are vital to heating the plasma.
“However, ICH suffers from parasitic power losses that are not well understood,” says Green, Oak Ridge’s primary investigator for the SciDAC Center for Simulation of Wave Plasma Interactions.
With collaborators at MIT, PPPL, Tech-X Corp., and Lodestar Research Corp., Green’s group is running three-dimensional simulations to reveal the reasons for ICH’s mysterious power losses.
Green says Titan, a Cray XK7, “is enabling us to do the first simulation of radio frequency heating of ions that we think will include enough of the physics to elucidate where and how the efficiency losses are occurring.”
One key to capturing the rich physics in the simulation is the ability to portray complex problems. Before high-performance computing platforms such as Titan became available, the challenge was to make a problem simple enough for analytical work. Researchers adapted these approaches to large-scale computing as the high-performance computing
environment emerged.
“This approach, we’ve discovered, is not optimal for really using these leadership-class computers to their fullest,” Green says. “Instead, we have started attacking problems knowing that we don’t need to make the simplifications we used to, and also knowing that certain types of algorithms work better than others on the new computing architectures.”
The result has been better algorithms that are tailored to HPC platforms.
He describes the move forward. “We have to start by verifying our codes to give the answers we expect, and validating those answers with experimental results to ensure the physics models we are using are correct, he says. “Our team has been working on this aspect for some time now, and the component codes are ready to go.”
Now, with the ability to model the emission of radio waves from the antenna into the plasma and the plasma’s response to those waves, the group is poised for the next stage. “This project involves putting these pieces together for the first time to hopefully elucidate how the interaction between the launching antenna structure, and the plasma response, results in the unexplained loss of RF power we are seeing in experiments,” Green says.
The team soon will simulate past fusion experiments that exhibited the unexplained ICH power losses, finally with an ability to track all aspects of the experiments.
Says Jardin, “ITER is a monumental step toward the realization of fusion power. Success in this endeavor will be a turning point in the quest for a sustainable, carbon-free, and safe source of energy for the planet.”