PPPL researchers have used Artificial Intelligence (AI) to create an innovative technique to improve the prediction of disruptions in fusion energy devices — a grand challenge in the effort to capture on Earth the fusion reactions that power the sun and stars.

PPPL physicist Michael Churchill based the new method on some of the same technology used in a commercial smart speaker and home assistant. The method enables the deep learning form of AI machine learning to predict disruptions directly from raw, high-resolution data from fusion experiments.

This capability provides a major advantage over many existing machine learning methods. The technique not only utilizes unmodified raw data, but also recognizes and remembers early developments in a lengthy sequence of events, which is difficult for current techniques to do.

Disruptions are a sudden loss of plasma control that halts fusion reactions and can seriously damage the inner walls of doughnut-shaped tokamak fusion facilities that house the reactions. Accurate predictions that lead to avoidance of disruptions will be essential for future devices.

Churchill tested the new method on data from a single diagnostic that measures the temperature of electrons in the DIII-D National Fusion Facility that General Atomics operates for the U.S. Department of Energy (DOE). The results showed that the technique accurately predicts early-on if a disruption is imminent. This information can be used to trigger mitigation systems that safely shutdown the discharge.

“These findings provide proof-of principle,” said Churchill, who published the results in a refereed paper for the 2019 Conference on Neural Information Processing Systems (NeurIPS) in Vancouver, Canada. “The exciting new capability,” he says, “is that deep learning directly applied to raw data from a single diagnostic can produce useful predictions.”

Scientists at PPPL have found a novel way to prevent pesky magnetic bubbles in plasma from interfering with fusion reactions – delivering a potential way to improve the performance of fusion energy devices. The method applies to managing radio frequency (RF) waves to stabilize the magnetic bubbles, which can expand and create disruptions that can limit the performance of ITER, the international facility under construction in France to demonstrate the feasibility of fusion power.

The new way modifies the standard technique for controlling these magnetic bubbles, or islands, by steadily depositing radio (RF) rays into the plasma. That technique proves inefficient when the width of an island is small compared with the characteristic size of the region over which the RF ray deposits its power, a condition called “low-damping.” In that case, much of the power leaks from the island.

The new model predicts that depositing the rays in pulses rather than steady state streams can overcome the leakage problem, said Suying Jin, a graduate student in the Princeton Program in Plasma Physics and lead author of a paper that describes the method in Physics of Plasmas. Pulsing can also increase stabilization in high-damping cases — those in which the width of the island is larger than the RF-deposit region — for the same average power, she noted.

For the novel process to work, “the pulsing must be done at a rate that is neither too fast nor too slow,” she said. “This sweet spot should be consistent with the rate that heat dissipates from the island through diffusion,” she said, and the predicted process lends itself to experimental verification.

The new model draws upon past work by Jin’s co-authors and advisors Allan Reiman, a Distinguished Research Fellow at PPPL, and Professor Nat Fisch, director of the Program in Plasma Physics at Princeton University and associate director for academic affairs at PPPL. Their research provides the framework for the study of RF-power deposition trends to stabilize magnetic islands.

A challenge to creating fusion energy on Earth is trapping the charged gas known as plasma that fuels fusion reactions within a strong magnetic field and keeping the plasma as hot and dense as possible for as long as possible. Now, scientists at PPPL have gained new insight into a common type of hiccup known as the sawtooth instability that cools the hot plasma in the center and interferes with the fusion reactions. These findings could help bring fusion energy closer to reality.

“Conventional models explain most instances of the sawtooth crashes, but there is a tenacious subset of observations that we have never been able to explain,” said PPPL physicist Christopher Smiet, lead author of a paper reporting the results in Nuclear Fusion. “Explaining those unusual occurrences would fill a gap in understanding the sawtooth phenomenon that has existed for almost 40 years.”

Researchers have known for decades that the temperature at the core of fusion plasma often rises slowly and can then suddenly drop — an unwanted occurrence since the cooler temperature reduces efficiency. The prevailing theory is that the crash occurs when a quantity called the safety factor, which measures the stability of the plasma, drops to a measurement close to 1. The safety factor relates to how much twist is in the magnetic field in the doughnut-shaped tokamak fusion facilities. However, some observations suggest that the temperature crash occurs when the safety factor drops to around 0.7. This is quite surprising and cannot be explained by the most widely accepted theories.

The new insight, coming not from plasma physics but from abstract mathematics, shows that when the safety factor takes specific values, one of which is close to 0.7, the magnetic field in the plasma core can change into a different configuration called alternating-hyperbolic. “In this topology, the plasma is lost in the core,” Smiet says. “The plasma is expelled from the center in opposite directions. This leads to a new way for the magnetic cage to partially crack, for the temperature in the core to suddenly fall, and for the process to repeat as the magnetic field and temperature slowly recover.”

The new insights suggest an exciting new research direction toward keeping more heat within the plasma and producing fusion reactions more efficiently. “If we can’t explain these outlier observations, then we don’t fully understand what’s going on in these machines,” Smiet said. “Countering the sawtooth instability can lead to producing hotter, more twisty plasmas and bring us closer to fusion.”

A major roadblock to producing safe, clean and abundant fusion energy on Earth is the lack of detailed understanding of how the hot, charged plasma gas that fuels fusion reactions behaves at the edge of doughnut-shaped fusion facilities called tokamaks. Recent breakthroughs by researchers at PPPL have advanced understanding of the behavior of the highly complex plasma edge in tokamaks on the road to capturing the fusion energy that powers the sun and stars.

Understanding this edge region will be particularly important for operating ITER, the international fusion experiment under construction in France to demonstrate the practicality of fusion energy.

Among the first-of-a-kind findings has been the discovery that accounting for the turbulent fluctuations in the magnetic fields that confine the plasma that fuels fusion reactions can significantly reduce the turbulent particle flux near the plasma edge. Computer simulations show that the net particle flux can go down by as much as 30 percent, despite the fact that the average magnitude of turbulent particle density fluctuation goes up by 60 percent — indicating that even though the turbulent density fluctuations are more virulent, they are moving particles out of the device less effectively.

Researchers have developed a specialized code called “Gkeyll” — pronounced like “Jekyll” in Robert Louis Stevenson’s “The Strange Case of Dr. Jekyll and Mr. Hyde” — that makes these simulations feasible. The mathematical code, a form of modeling called “gyrokinetics,” simulates the orbiting of plasma particles around the magnetic field lines at the edge of a fusion plasma.

“Our recent paper summarizes the Gkeyll group's efforts in the area of gyrokinetic simulation,” said PPPL physicist Ammar Hakim, lead author of a Physics of Plasmas paper that provides an overview of the group’s achievements. The research, coauthored by scientists from six institutions, adapts a state-of-the-art algorithm to the gyrokinetic system to develop the “key numerical breakthroughs needed to provide accurate simulations,” Hakim said.

Noah Mandell, a graduate student in the Princeton University Program in Plasma Physics, built on the team’s work to develop the first gyrokinetic code able to handle magnetic fluctuations in what is called the plasma scrape-off layer (SOL) at the edge of tokamak plasmas. The British Journal of Plasma Physics has highlighted his report as a featured article.

Mandell’s findings are best described as “proof-of-concept” with regard to the magnetic fluctuations, he said. “We know there are more physical effects that need to be added to the code for detailed comparisons with experiments, but already the simulations are showing interesting properties near the plasma edge.”

For a video of first computer simulations of kinetic plasma turbulence near the edge of fusion devices click here.

Carefully manipulating the outer skin of plasma can create cascades of effects that help create the stability needed to sustain fusion reactions, scientists have found. The findings, led by PPPL and Princeton University physicist Dylan Brennan, could provide insight into the physics required to stabilize plasma in doughnut-shaped fusion facilities called tokamaks. These include ITER, the multinational facility being built in France to demonstrate the practicality of fusion power.

Researchers used computing power unavailable even a few years ago to reveal that controlling the electric current at the edge of the plasma in tokamaks can affect areas of the plasma closer to the core and is key to maintaining the stability of fusion plasmas. “We have shown that all the layers matter when you try to stabilize the plasma,” said Brennan, lead author of a paper reporting the results. “It turns out that the plasma edge strongly couples to the inner regions.”

This result, reached in collaboration with physicists at General Fusion, a Canadian company exploring the development of fusion energy, grew from analysis of what happens when the plasma is compressed — a technique sometimes used to heat plasma to the super-high temperatures needed to sustain fusion reactions. The researchers found that the current that courses through the outer layer of compressed plasma can either prevent instabilities from interfering with the fusion process or amplify them.

The findings show that to determine the stability of the system one must assume that all the layers matter, Brennan said, since the interactions between layers can either stabilize or destabilize the plasma. “The conditions at one local point in the plasma are not sufficient for you to characterize the total situation,” he said. “You have to consider the whole system.”

Brennan furthermore believes that scientists analyzing the behavior of fusion machines in the future should assume that all the layers of the plasma matter — even those that do not use compression to heat the plasma — to increase the efficiency of the devices. “This exercise of showing a physics result for a complicated mechanism of coupling between surfaces suggests that we can now do this kind of work routinely on any kind of device,” said Brennan. “That’s exciting news.”

Bringing the power of the sun to Earth requires sound theory, good engineering, and a little finesse. The process entails trapping charged, ultra-hot plasma so its particles can fuse and release enormous amounts of energy. The most widely used facilities for this process are doughnut-shaped tokamaks that hold plasma in place with strong magnets that are precisely shaped and positioned. But inevitable errors in the shaping or placement of these magnets can lead to poor confinement and loss of plasma, shutting down fusion reactions.

These irregularities can produce what are called 3D error fields that trigger a disruption in the plasma, causing it to escape its magnetic confinement. “The question is how large an error field can ITER tolerate without disrupting,” said PPPL physicist Nikolas Logan, referring to the international tokamak under construction in France to demonstrate the practicality of fusion energy. “We want to prevent disruptions in ITER because they could both interfere with fusion reactions and damage the walls of the tokamak,” said Logan, lead author of a paper reporting the findings.

Since ITER is under construction, the researchers used two codes to model the effects of error fields on plasmas in tokamaks in South Korea, China, the United Kingdom, and other countries. The codes strengthened the errors until the plasmas disrupted. “By combining these codes, we were able to simulate a wide range of conditions that could occur in a variety of devices, including ITER,” said PPPL physicist Qiming Hu, a paper co-author.

The findings “are promising for ITER and other future devices,” said Logan. “We need to be careful, but this helps physicists and engineers know just how careful we need to be when it comes to avoiding these 3D fields before putting lots of power into ITER.”

Bringing to Earth the virtually unlimited fusion energy that powers the sun and stars requires modeling the hot, charged plasma gas that fuels fusion reactions. PPPL scientists have now discovered a new theoretical method for designing computer algorithms to model the fuel and speed the development of a safe and clean energy source for humanity.

“Our research rigorously demonstrates how to preserve fundamental mathematical properties of plasmas in simulation algorithms,” said Alexander Glasser, a graduate student in the Princeton Program in Plasma Physics at PPPL and lead author of a paper in the Journal of Plasma Physics that lays out the findings. “We provide a blueprint for the construction of algorithms that more faithfully describe the real world.”

The new method identifies algorithms that preserve the underlying mathematical structure of plasmas. Glasser and physicist Hong Qin, his coauthor and advisor, demonstrate how algorithms can preserve the “gauge structure” of plasmas— their symmetries and conservation laws—in simulations.

The research derives from a theorem developed in 1915 by German mathematician Emmy Noether. According to her theorem, the “gauge symmetry” of electromagnetic systems gives rise to their conservation of electric charge.

While Noether’s theorem applies to the differential equations that describe electromagnetic systems, Glasser and Qin show that its key mathematical features apply to properly constructed plasma simulation algorithms as well. Such algorithms, which they call “gauge-compatible splitting methods,” conserve electric charge with the highest precision possible on computers.

The proposed methods now require implementation in a simulation code that models a fusion plasma. “The design proposed in this paper is guided by the philosophy that space-time is discrete and not continuous, and all the laws of physics can be established on discrete space-time,” Qin said. “It turns out that the techniques enabled by this philosophy can lead us to more accurate and reliable simulation algorithms for fusion energy devices.”

An international team led by PPPL has upgraded a key computer code for calculating forces that act on magnetically confined plasma in fusion energy experiments. The revised code, called the free-boundary stepped-pressure equilibrium code (SPEC), becomes part of a suite of three computational tools that will allow scientists to further improve the design of breakfast-cruller-shaped fusion facilities known as stellarators and help bring efficient fusion reactors closer to reality.

When incorporated with a stellarator-optimization code and a coil-design code, SPEC will help find a magnetic configuration that improves the performance of the twisty stellarator design. The suite of complementary codes together determine the optimal location for the plasma and determine the shape that the external electromagnets must have to hold the plasma in the proper position.

“We want to optimize both the plasma position and the magnetic coils to balance the force that makes the plasma expand while holding it in place,” said physicist Stuart Hudson, interim head of the Theory Department at PPPL and lead author of the paper reporting the results. “That way we can create a stable plasma whose particles are more likely to fuse. The upgraded SPEC code enables us to know where the plasma will be for a given set of magnetic coils.”

Stability is crucial for fusion plasma. A plasma that bounces around inside a stellarator can escape, cool off, and tamp down the fusion reactions — in effect quenching the fusion fire. “That’s one of the problems with plasmas,” Hudson said. “They move all over the place.”

The new version helps solve the problem by allowing researchers to calculate the boundary of the plasma without knowing its position beforehand. Used in coordination with a coil-design code called FOCUS and an optimization code called STELLOPT — both of which were also developed at PPPL — SPEC lets physicists simultaneously ensure that the plasma will have the best fusion performance and the magnets will not be too complicated to build. “There’s no point optimizing the shape of the plasma and then later finding out that the magnets would be incredibly difficult to construct,” Hudson said.

Hudson and collaborators from Germany, Australia, and Switzerland used pencil and paper to determine the solution steps and powerful PPPL computers to verify the results. “We demonstrated that the code works,” he said. “Now it can be used to study current experiments and design new ones.”

Scientists seeking to reproduce fusion on Earth must deal with sawtooth instabilities — up-and-down swings in the central pressure and temperature of the plasma that fuels fusion reactions similar to the blades of a saw. If these swings are large enough, they can lead to the sudden collapse of the entire discharge of the plasma. Such swings were first observed in 1974 and have so far eluded a widely accepted theory that explains experimental observations.

Researchers at PPPL now propose a new theory to explain the swings that occur in doughnut-shaped tokamaks, or fusion facilities. The theory, created through high-fidelity computer simulations, appears consistent with observations made during tokamak experiments, the researchers said. Understanding the process could prove vital to next-generation fusion facilities such as ITER, the international experiment under construction in France to demonstrate the practicality of fusion power.

The new findings demonstrate that when the pressure in the core of the plasma reaches a certain point, other instabilities can be excited that produce the sudden pressure and temperature drops. These instabilities create stochastic — or jumbled — magnetic fields in the core of the plasma causing the collapse, said physicist Stephen Jardin, lead author of a paper describing the process and highlighted in a featured American Institute of Physics publication called “SciLight.”

This advanced model provides a new way to understand sawtooth phenomena. Looking ahead, the scientists want to explore the applicability of the model to tasks such as describing the evolution of “monster sawteeth,” or very large oscillations, and using high powered Radio Frequency antennas to control sawtooth swings. “We want to develop a simulation model of a whole tokamak plasma,” Jardin said, “and this new theory of the sawteeth is an important part of the effort.”

Electric current is everywhere, from powering homes to controlling the plasma that fuels fusion reactions to perhaps giving rise to vast cosmic magnetic fields. PPPL scientists have recently found that the current can form in previously unknown ways, providing researchers with new astrophysical insights and possible new ways to harvest on Earth the fusion energy that drives the sun and stars.

“It’s very important to understand which processes produce electrical currents in plasma and which phenomena could interfere with them,” said Ian Ochs, graduate student in Princeton University’s Program in Plasma Physics and lead author of a paper selected as a featured article in Physics of Plasmas. “Currents are the primary tool we use to control plasma in magnetic fusion research,” he said.

Currents develop in the plasma in doughnut-shaped tokamaks in the presence of electromagnetic waves. These waves, similar to those that radios and microwave cookers emit, cause some of the already-moving plasma electrons to “ride the wave like surfers on a surfboard,” Ochs said. “And the resulting motion of electric charge, if uncompensated, produces a current in the plasma.”

Ochs found that researchers could surprisingly create these currents when a low-frequency form of the waves, called “ion acoustic waves,” arises. “There are magnetic fields throughout the universe on different scales, including the size of galaxies, and we don’t really know how they got there,” Ochs said. “The mechanism we discovered could have helped seed cosmic magnetic fields, and any new mechanisms that can produce magnetic fields are interesting to the astrophysics community.”

His results from pencil-and-paper calculations consist of mathematical expressions that give scientists the ability to calculate how these currents develop and grow. “The formulation of these expressions was not straightforward,” Ochs said. “We had to condense the findings so they would be sufficiently clear and use simple expressions to capture the key physics.”

The results deepen understanding of a basic physical phenomenon. “What especially excites me,” said co-author Nat Fisch, professor and associate chair of the Department of Astrophysical Sciences and director of the Program in Plasma Physics, “is that the mathematical formalism that Ochs has built, together with the physical intuitions and insights that he has acquired, now put him in a position either to challenge or to put on a firm foundation even more curious behavior in the interactions of waves with resonant particles in plasma.”