The new Princeton University supercomputer, Traverse, enhances PPPL research to develop the science to bring the fusion that powers the sun and stars to Earth. Traverse houses the same processing architecture as the leadership class supercomputers at Oak Ridge National Laboratory and Lawrence Livermore National Laboratory.
The powerful supercomputer will enable PPPL scientists to adapt their codes to the leadership class computing cluster and prepare for exascale machines that will process a billion billion (1018 ) calculations per second. “At that scale we will be able to simulate and optimize fusion reactors, speeding the deployment of fusion energy in the global battle against climate change,” Cowley said at the ribbon-cutting ceremony. “We are very grateful to the University for this marvelous facility.” Click here to read the full release on the University website.
Artificial intelligence (AI), a branch of computer science that is transforming scientific inquiry and industry, could speed the development of safe, clean and virtually limitless fusion energy for generating electricity. A major step in this direction is under way at PPPL and Princeton University, where a team of scientists working with a Harvard graduate student is for the first time applying deep learning — a powerful new version of the machine learning form of AI — to forecast sudden disruptions that can halt fusion reactions and damage the doughnut-shaped tokamaks that house the reactions.
Unlike traditional software, which carries out prescribed instructions, deep learning learns from its mistakes. Accomplishing this seeming magic are neural networks, layers of interconnected nodes — mathematical algorithms — that are weighted by the program to shape the desired output. For any given input the nodes seek to produce a specified output, such as correct identification of a face or accurate forecasts of a disruption. Training kicks in when a node fails to achieve this task: the weights automatically adjust themselves for fresh data until the correct output is obtained.
"Artificial intelligence is the most intriguing area of scientific growth right now, and to marry it to fusion science is very exciting,” said Bill Tang, a principal research physicist at PPPL and leader of the project. “We’ve accelerated the ability to predict with high accuracy the most dangerous challenge to clean fusion energy.” Concurred Harvard graduate student Julian Kates-Harbeck, chief author of the software and lead author of a paper describing it in Nature: “The ability of deep learning to learn from complex data makes it an ideal candidate for the task of disruption prediction.”
Crucial to demonstrating this ability has been access to huge databases provided by the DIII-D National Fusion Facility that General Atomics operates for the DOE in California and the Joint European Torus (JET) in the United Kingdom, the largest fusion facility in the world. These vast databases have enabled reliable predictions of disruptions on tokamaks other than those on which the system was trained — in this case from the smaller DIII-D to the larger JET. The achievement bodes well for the prediction of disruptions on ITER, a far larger and more powerful tokamak that will have to apply capabilities learned on today’s fusion facilities.
To capture and control on Earth the fusion reactions that drive the sun and stars, researchers must first turn room-temperature gas into the hot, charged plasma that fuels the reactions. Scientists at PPPL have conducted an analysis that confirms the effectiveness of a novel, non-standard way for starting up plasma in future compact fusion facilities.
The innovative technique, known as “transient coaxial helical injection (CHI),” eliminates the central magnet, or solenoid, that launches the plasma inside tokamaks, the most widely used fusion facilities. Such elimination could facilitate constant fusion reactions and free-up valuable space in the center of compact spherical tokamaks, whose cored-apple shape has less room inside than conventional doughnut-shaped tokamaks that are more common.
The freed-up space could be used to strengthen the magnetic field that confines the plasma and thereby improve its performance. Elimination of the solenoid could also simplify the design of compact tokamaks.
Solenoids run down the center of a tokamak and induce current in the uncharged gas that researchers inject into the facility. The current also creates a magnetic field that combines with the field produced by magnets that surround the tokamak to bottle up and control the plasma.
By contrast, the CHI process creates the crucial current with electrodes placed near the bottom or top of the tokamak, eliminating the space-eating solenoid. “What we primarily focused on was the beginning stage of forming the plasma,” said PPPL physicist Kenneth Hammond, lead author of a paper describing the study in Physics of Plasmas. “This helped paint a fuller picture of how CHI discharges work.”
The transient CHI technique has valuable potential, said Tom Brown, a principal engineer at PPPL. “If successful, CHI could provide space for interior components that could enhance the performance of spherical devices,” Brown said. However, he added, “further engineering details need to be developed at the experimental level that also can work within a higher-level demonstration device and also in an eventual fusion power plant.”
A key requirement for future facilities that aim to capture and control fusion on Earth is accurate predictions of the pressure of the plasma that fuels fusion reactions inside doughnut-shaped tokamaks that house the reactions. Central to these predictions is forecasting the pressure that the scrape-off layer, the thin strip of gas at the edge of the plasma, exerts on the divertor — the device that exhausts waste heat from fusion reactions.
PPPL researchers have developed fresh insight into the physics governing the balance of pressure in the scrape-off layer. “Previous simple assumptions about the balance of pressure are incomplete,” said PPPL physicist Michael Churchill, lead author of a Nuclear Fusion paper that describes the new findings. “The codes that simulate the scrape-off layer have often thrown away important aspects of the physics, and the field is starting to recognize this.”
Churchill and PPPL colleagues determined the key factors behind the pressure balance by running the state-of-the-art XGCa computer code on the Cori and Edison supercomputers at the National Energy Research Scientific Computing Center. The code treats plasma at a detailed kinetic — or particle motion— level rather than as a fluid.
Among key features found was the impact of the bulk drift of ion particles in the plasma, an impact that previous codes have largely ignored. Also seen to be important in the pressure balance were the kinetic effects due to ions having different temperatures depending on their direction. The new findings could improve understanding of the scrape-off layer pressure at the divertor, Churchill said, and could lead to accurate forecasts for the international ITER experiment under construction in France and other next-generation tokamaks.
PPPL physicists have discovered key information about how electrically charged gas known as “plasma” flows inside the edge of doughnut-shaped tokamak fusion devices. The findings mark an encouraging sign for the development of machines to produce safe and clean fusion energy for generating electricity without creating long-term hazardous waste.
The result partially corroborates past PPPL findings that the width of the heat exhaust produced by fusion reactions could be six times wider, and therefore less narrow, concentrated, and damaging, than had been thought. “These findings are good news for ITER,” said PPPL physicist C.S. Chang, lead author of a description of the research in Physics of Plasmas, referring to the international fusion experiment under construction in France. “The findings show that the heat exhaust in ITER will have a smaller chance of harming the machine.”
The superhot plasma within tokamaks, which can reach hundreds of millions of degrees, is confined by magnetic fields that keep the plasma from the walls of the machines. However, particles and heat can escape from the boundary between the magnetically confined and unconfined plasmas. At this boundary, the field lines cross at the so-called X-point, the spot where the waste heat and particles escape confinement and strike a target called the “divertor plate.”
The new findings reveal that a hill-like bump of electric charge occurs at the X-point. This electrical hill makes the plasma circulate around it, like cars maneuvering around a construction site.
The researchers produced these findings with XGC, an advanced computer code developed with external collaborators at PPPL that models the plasma as a collection of individual particles rather than as a single fluid. The model, which showed that the connection between the upstream plasma located above the X-point and the downstream plasma below the X-point formed in a way not predicted by simpler codes, could lead to more accurate predictions about the exhaust and make future large-scale facilities less vulnerable to internal damage.
Scientists who use magnetic fields to bottle up and control on Earth the fusion reactions that power the sun and stars must correct any errors in the shape of the fields that contain the reactions. Such errors can have a damaging impact on the stability and confinement of the hot, charged plasma gas that fuels the reactions.
Researchers led by scientists at PPPL have found clear evidence of error fields in the initial 10-week run of the National Spherical Torus Experiment—Upgrade (NSTX-U), the flagship fusion facility at the laboratory. The exhaustive detection method they used could provide lessons for error correction in future fusion devices such as ITER, the international experiment under construction in France to demonstrate the practicality of controlled fusion energy.
One major find stood out: a slight misalignment of the magnetic coils that run down the center of the tokamak and produce the fields that wrap horizontally — or “toroidally” — around the interior of the vessel. “We looked for the source of the error with the biggest impact on the plasma,” said physicist Nate Ferraro, first author of the research that reported the search and discovery in Nuclear Fusion. “What we found was a small misalignment of the center-stack coils with the casing that encloses them.”
The slight misalignment generated errors that resonated in the behavior of the plasma. Among the issues was an effect that kept the edge of the plasma from rotating, and increased localized heating on plasma-facing components inside the tokamak.
Discovery of the misalignment followed shut-down of the tokamak for ongoing repairs in the wake of a coil failure. The misalignment findings are now being used “to drive new engineering tolerance requirements for NSTX-U as it is rebuilt,” the researchers said. “Every tokamak is concerned about error fields,” noted Ferraro. “What we are trying to do is optimize the NSTX-U.”
Machine learning (ML), a form of artificial intelligence that recognizes faces, understands language and navigates self-driving cars, can help bring to Earth the clean fusion energy that lights the sun and stars. Researchers at PPPL are using ML to create a model for rapid control of plasma — the state of matter composed of free electrons and atomic nuclei, or ions — that fuels fusion reactions.
Researchers led by PPPL physicist Dan Boyer have trained neural networks — the core of ML software — on data produced in the first operational campaign of the National Spherical Torus Experiment-Upgrade (NSTX-U), the flagship fusion facility at PPPL. The trained model, reported in a paper in Nuclear Fusion, accurately reproduces predictions of the behavior of the energetic particles produced by powerful neutral beam injection (NBI) that is used to fuel NSTX-U plasmas and heat them to million-degree, fusion-relevant temperatures.
These predictions are normally generated by a computer code called NUBEAM that requires complex calculations to be made hundreds of times per second to analyze the behavior of the plasma during an experiment. But each calculation can take several minutes to run, making the results available to physicists only after an experiment that typically lasts a few seconds is completed.
The new ML software reduces the time needed to accurately predict the behavior of energetic particles to under 150 microseconds — enabling the calculations to be done online during the experiment.
The technique combines ML predictions with the limited measurements of plasma conditions available in real-time. The combined results will help the real-time plasma control system make more informed decisions about how to adjust beam injection to optimize performance and maintain stability of the plasma — a critical quality for fusion reactions.
The rapid evaluations will also help operators make better-informed adjustments between experiments that are executed every 15-20 minutes during operations. Said Boyer: “Accelerated modeling capabilities could show operators how to adjust NBI settings to improve the next experiment.”
The swirls created by milk poured into coffee or the shudders that can jolt airplanes in flight are examples of turbulence, the chaotic movement of matter found throughout nature. Turbulence also occurs within doughnut-shaped tokamaks that house the plasma that fuels fusion reactions. PPPL scientists have now discovered that turbulence may play an increased role in affecting the self-driven, or bootstrap, current in plasma that is necessary for the reactions.
For fusion to occur in tokamaks, the plasma must be confined by a cage of magnetic fields that shape and control the hot, charged matter. The bootstrap current plays an important role in creating the cage and disturbing the current could affect the confinement of the plasma and the production of fusion reactions.
“It turns out that turbulence may significantly affect the bootstrap current,” said PPPL physicist Weixing Wang, lead author of a paper reporting the results of simulations in Nuclear Fusion. “This is a potential effect that needs to be taken fully into account.”
The simulations show that turbulence can weaken the bootstrap current in relatively high temperature plasma and strengthen the current in plasmas with lower temperatures. Weakening means that more of the overall plasma current would have to be created by external means like radio waves rather than by the self-generated bootstrap current that naturally occurs in tokamaks.
Findings from this theoretical study can be tested and validated by experiments on scenarios achievable in tokamaks including the DIII-D National Fusion Facility operated by General Atomics. More massive supercomputer simulations will be required to answer further questions, such as how much self-driven current may be expected in future burning plasma experiments.
Whether zipping through a star or a fusion device on Earth, the electrically charged particles that make up the fourth state of matter, better known as plasma, are bound to magnetic field lines like beads on a string. Unfortunately for plasma physicists who study this phenomenon, the magnetic field lines often twist and knot like pretzels.
Now, findings by an international team led by PPPL scientists show that the twisted magnetic fields can evolve in only so many ways, with the plasma inside following a general rule. As long as there is high pressure on the outside of the plasma pressing inward, the plasma will spontaneously take on a doughnut, or torus, shape and balloon out in a horizontal direction. Constraining this ballooning is the average amount of twisting in the plasma, a quality known as “helicity.”
“The helicity prevents the configuration from blowing apart and forces it to evolve into this self-organized, twisted structure,” says Christopher Smiet, a physicist at PPPL and lead author of a paper reporting the results in the Journal of Plasma Physics. “By studying the magnetic field in this more general framework, we can learn new things about the self-organizing processes within tokamaks and the instabilities that interfere with them,” he said.
The findings can provide insight into the behavior of masses of plasma emitted from the sun that can expand and collide with the Earth’s magnetic field. In mild form, the collisions cause the northern lights. If powerful enough, these collisions can disrupt the operations of satellites and interfere with cell phones, global positioning systems, and radio and television signals. “It’s fascinating what you can learn when you study how knots unravel,” Smiet said.