286

Papers

9/9

Roadblocks Active

2

Connections

Using a chain of coupled plasma transport codes, this study shows that injecting argon gas into a compact high-field tokamak like ARC can cool the divertor target below 2 eV — the threshold needed to avoid melting tungsten surfaces — while sustaining near-gigawatt fusion power in H-mode. Unlike neon, argon does not accumulate excessively in the plasma core, so it suppresses edge heat loads without degrading fusion performance. This is one of the most complete quantitative demonstrations that a single impurity-seeding strategy can simultaneously satisfy core performance and exhaust requirements for a next-step device.

██████████ 0.9 divertor-thermal Preprint

ReadSaveConnections

This paper addresses the problem that disruption-prediction models trained on one tokamak fail on another because diagnostic layouts and plasma conditions differ. By generating synthetic diagnostic signals from MHD simulations and combining them with Fourier domain adaptation, the authors transfer a model trained on China's EAST tokamak to J-TEXT with zero experimental data from the target device, improving early-warning rates from 50% to 57% on nearly 1,600 discharges. The approach matters because every new fusion device currently requires years of experimental data before reliable disruption prediction is possible — this could compress that timeline.

██████████ 0.9 plasma-disruption Preprint

ReadSaveConnections

Runaway electrons — beams of relativistic electrons produced during plasma disruptions — can punch holes in reactor walls if not controlled. This paper trains physics-informed neural networks on the equations governing runaway electron kinetics, enabling predictions of current, energy, and energy distribution across arbitrary plasma conditions orders of magnitude faster than conventional solvers. Speed matters here because real-time disruption mitigation systems need to predict and respond to runaway electron dynamics on millisecond timescales that traditional simulation cannot match.

██████████ 0.9 plasma-disruption Preprint

ReadSaveConnections

This theoretical paper clarifies how toroidal magnetic flux — a quantity rarely used as a primary coordinate in tokamak analysis — simplifies the description of magnetic equilibria and the physics of the separatrix, the boundary between magnetically confined and unconfined plasma. The author argues that using toroidal flux as the radial coordinate reduces spurious sensitivity to the central current profile and gives a cleaner interpretation of loop voltage and flux slippage. While purely analytical, this type of foundational clarification directly affects how equilibrium codes and q-profile analyses are constructed, with downstream implications for disruption and current-profile control.

██████████ 0.8 q-engineering Preprint

ReadSaveConnections

Electron cyclotron heating (ECH) is a key tool for controlling the internal current profile of a tokamak, which in turn determines stability and confinement. This paper demonstrates a real-time optimizer (ECHO) that uses a fast neural-network proxy for a ray-tracing code to steer microwave beams to their target deposition location, and shows it keeps working correctly even when individual gyrotron launchers fail mid-shot. Experimental validation on DIII-D confirms that the algorithm matches the accuracy of offline ray tracing, which is significant because robust ECH control is central to maintaining the q-profiles that suppress disruptions and ELMs.

██████████ 0.8 long-confinement Preprint

ReadSaveConnections

Doppler backscattering (DBS) measures plasma flow velocity and density turbulence by bouncing microwaves off density fluctuations inside the plasma. This paper designs a DBS system for Pegasus-III, a tight-aspect-ratio (spherical) tokamak where measuring turbulence from the outer core to the plasma edge is especially challenging due to the unusual magnetic geometry. Beam-tracing simulations show that a Ka-band system with toroidal steering can simultaneously measure ion-scale turbulence and infer the local magnetic pitch angle — information critical for validating turbulence models that underpin confinement predictions.

██████████ 0.8 turbulence-modeling Preprint

ReadSaveConnections

Knowing the exact shape of the plasma boundary in real time is essential for magnetic control, but conventional equilibrium solvers are too slow for 10 kHz feedback loops. This paper trains a physics-informed neural network on 10,000 TCV discharges to reproduce the magnetic flux map in under 100 microseconds — fast enough for active plasma shape control at reactor-relevant rates. The system was validated in closed-loop operation on TCV, and the fast inference also enables real-time adjustment of the divertor strike-point position to manage heat loads, directly connecting to the divertor-thermal roadblock.

██████████ 0.7 plasma-wall Preprint

ReadSaveConnections

VMEC, the standard stellarator equilibrium code, assumes the plasma is organized into perfect nested magnetic surfaces — an assumption that breaks down at high plasma pressure. This study compares VMEC against HINT, a code that allows magnetic surfaces to break up, across 33 configurations of the Large Helical Device (LHD) at pressures up to 5%. The codes agree at low pressure, confirming VMEC's validity there, but diverge above a configuration-dependent threshold where stochastic field regions erode plasma volume — a regime that matters for high-performance stellarator operation and for understanding confinement limits.

██████████ 0.7 long-confinement Preprint

ReadSaveConnections

The Grad-Shafranov equation describes the magnetic equilibrium of a tokamak, and solving it quickly across many configurations is needed for design studies and control applications. VEQPy introduces a compact parametric solver that uses harmonic shape functions and polynomial radial profiles, handling six different types of physics inputs through a unified framework and solving in about 1.6 milliseconds on a D-shaped case. Fast and flexible equilibrium solvers like this are enabling infrastructure for integrated scenario modeling and machine learning-based control, where thousands of equilibrium evaluations per second may be required.

██████████ 0.6 turbulence-modeling Preprint

ReadSaveConnections

Energetic particle-induced geodesic acoustic modes (EGAMs) are plasma oscillations driven by fast ions — the fusion-born alpha particles in a burning plasma — and can cause these particles to be redistributed or lost before they heat the bulk plasma. Using gyrokinetic simulations with the ORB5 code, this paper shows that EGAMs develop nonlinear amplitude oscillations through the same physical mechanism as simpler beam-plasma instabilities, providing a cleaner theoretical handle on predicting their saturation behavior. Understanding saturation is necessary for estimating how much fast-ion transport EGAMs drive and whether they threaten confinement in a reactor.

██████████ 0.6 turbulence-modeling Preprint

ReadSaveConnections