Developing and maintaining fusion reactions — essentially recreating stellar conditions on Earth — is profoundly challenging, and Nathan Howard PhD ’12, a leading research scientist at the MIT Plasma Science and Fusion Center (PSFC), considers it one of the most captivating scientific puzzles of our era. “The science itself and the overarching potential of fusion as a renewable energy source are truly compelling. That inspired me to pursue graduate studies [at MIT] and engage with the PSFC,” he states.
Howard is a member of the Magnetic Fusion Experiments Integrated Modeling (MFE-IM) group at the PSFC. Together with MFE-IM group director Pablo Rodriguez-Fernandez, Howard and his team leverage simulations and machine learning to foresee how plasma will act in a fusion apparatus. MFE-IM and Howard’s investigations seek to predict the performance of specific technologies or setups prior to their testing in a real fusion setting, enabling more intelligent design decisions. To assure their precision, these models are constantly validated against data from prior experiments, ensuring their simulations are anchored in reality.
In a recent open-access article titled “Prediction of Performance and Turbulence in ITER Burning Plasmas via Nonlinear Gyrokinetic Profile Prediction,” published in the January edition of Nuclear Fusion, Howard elucidates how he employed high-resolution simulations of the turbulent swirls found in plasma to verify that the world’s largest experimental fusion facility, currently being constructed in Southern France, will function as anticipated upon activation. He also illustrates how an alternative operational setup could yield nearly identical energy output with reduced energy input, a finding that could significantly enhance the overall efficiency of fusion devices.
The most substantial and extraordinary project never built
Forty years ago, the United States, alongside six other countries, united to construct ITER (Latin for “the way”), a fusion facility designed to produce 500 megawatts of fusion power, alongside a plasma capable of generating ten times more energy than it receives from external heating. The plasma arrangement formulated to attain these objectives — the most ambitious of any fusion endeavor — is known as the ITER baseline scenario, and as fusion research and plasma physics have advanced, methodologies to realize this plasma have been honed using increasingly potent simulations similar to the modeling framework Howard utilized.
In his efforts to validate the baseline scenario, Howard employed CGYRO, a computational code created by Howard’s partners at General Atomics. CGYRO applies a sophisticated plasma physics model to a clearly defined set of fusion operational parameters. While time-consuming, CGYRO produces highly detailed simulations of plasma behavior at various locations within a fusion device.
The exhaustive CGYRO simulations were subsequently analyzed through the PORTALS framework, a suite of tools originally developed at MIT by Rodriguez-Fernandez. “PORTALS takes the high-fidelity [CGYRO] simulations and employs machine learning to construct a rapid model termed a ‘surrogate’ that can replicate the outcomes of the more complex runs, but at a much faster pace,” Rodriguez-Fernandez clarifies. “Only high-fidelity modeling tools like PORTALS afford us a preview into the plasma core before it even takes shape. This predictive-first strategy allows us to engineer more efficient plasmas in a device like ITER.”
Following the initial assessment, the surrogates’ precision was evaluated against the high-fidelity runs. If a surrogate did not align with CGYRO’s results, PORTALS was executed again to fine-tune the surrogate until it closely resembled CGYRO’s outcomes. “The advantageous aspect is, once a well-trained [surrogate] model is established, it can be employed to forecast conditions that differ, with substantially less necessity for the complete intricate runs.” Once fully trained, the surrogates were utilized to investigate how varying combinations of inputs might influence ITER’s expected performance and how it achieved the baseline scenario. Notably, the surrogate processes consumed a fraction of the time, and they could be leveraged alongside CGYRO to enhance its performance and deliver detailed results more expeditiously.
“Just dropped in to see what condition my condition was in”
Howard’s endeavors with CGYRO, PORTALS, and surrogates scrutinized a specific array of operational conditions forecasted to realize the baseline scenario. These conditions encompassed the magnetic field employed, the techniques used to regulate plasma shape, the external heating applied, and numerous other variables. Through 14 iterations of CGYRO, Howard confirmed that the current baseline scenario configuration could attain 10 times greater power output than input into the plasma. Howard remarks on the findings, “The modeling we executed is perhaps the highest fidelity achievable at this moment, and almost certainly the highest fidelity recorded.”
The 14 iterations of CGYRO used to validate the plasma performance entailed executing PORTALS to construct surrogate models for the input parameters and then linking the surrogates to CGYRO for more efficient operation. It took just three additional iterations of CGYRO to investigate an alternative scenario that suggested ITER could generate nearly the same quantity of energy with approximately half the input power. The surrogate-enhanced CGYRO model indicated that the temperature of the plasma core — and consequently the fusion reactions — was not significantly impacted by reduced power input; less power input equates to more efficient functioning. Howard’s discoveries also serve as a reminder that there may be additional avenues to enhance ITER’s efficacy; they have yet to be uncovered.
Howard reflects, “The ability to utilize the outcomes of this modeling to shape the design of experiments like ITER is exhilarating. For many years, I’ve asserted that this was our research goal, and now that we’ve actualized it — it’s a remarkable journey, and incredibly rewarding.”