282

Papers

10/10

Roadblocks Active

3

Connections

A neural network adapted from a medical survival-analysis framework (DeepHit) uses 50 milliseconds of radar-like plasma edge measurements to predict when the first ELM — a violent heat burst that damages divertor components — will occur, with 100ms warning on the DIII-D tokamak. That 100ms lead time is critical because it exceeds the response latency of automated mitigation actuators such as pellet injectors or magnetic perturbation coils. This is a proof-of-concept on a limited dataset, so the statistical robustness is not yet established, but the architecture and diagnostic pairing are directly compatible with real-time control systems.

██████████ 0.9 elm-control Preprint

ReadSaveConnections

FIREFLY is a fast computational package that estimates how much heat and exhaust gas will hit the divertor — the reactor's exhaust system — for a given magnetic geometry, running far faster than full-physics codes by using a simplified heat transport model paired with the established EIRENE neutral-particle tracker. It was validated on the Wendelstein 7-X stellarator and enables engineers to scan thousands of divertor geometry variants (slot width, target angle, baffle position) to find configurations that keep peak heat flux below material limits. This kind of rapid design-space exploration is essential for DEMO-class devices where divertor survival is a hard engineering constraint.

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

ReadSaveConnections

A new pseudo orbit-averaging algorithm delivers a 30,000x speed-up for gyrokinetic simulations — particle-scale plasma models — making it feasible to compute steady-state plasma equilibria in high-field magnetic mirror devices like the WHAM experiment (17 T HTS mirror) without prohibitive compute cost. The calculated ion confinement time and electrostatic potential agree with analytic theory, validating the approach. This unlocks quantitative turbulence and transport studies for a class of alternative fusion concepts that has seen renewed commercial interest precisely because HTS magnets now make high mirror ratios practical.

██████████ 0.9 hts-magnets Preprint

ReadSaveConnections

When electrons are accelerated to relativistic runaway speeds during a plasma disruption, injecting a circularly polarized radio wave (R-wave) can trap them at a resonant velocity and reverse their acceleration — a 'firewall' effect shown via analytical theory and PIC simulations at tokamak-relevant parameters. Runaway electrons are dangerous because they can carry megaampere currents and punch through plasma-facing components when a disruption occurs. This wave-injection approach could complement existing mitigation techniques like massive gas injection by actively arresting electron acceleration rather than only diluting the plasma.

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

ReadSaveConnections

In stellarators, the inherently irregular 3D magnetic geometry causes turbulent ion-temperature-gradient (ITG) modes to become spatially trapped — a phenomenon mathematically identical to Anderson localization in disordered quantum systems — which naturally caps how far ion temperature can rise. The authors map the ITG eigenvalue equation onto the Aubry-André-Harper model and find a localization threshold for W7-X parameters at η*_i ≈ 1.54, using equilibrium data from the DESC code. If confirmed by gyrokinetic simulation, this predicts that stellarators may self-regulate transport without active stabilization, a significant potential advantage over tokamaks.

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

ReadSaveConnections

When external magnetic perturbations are applied to a tokamak to suppress ELMs, ions near resonant magnetic surfaces can partially shield out that perturbation — and this paper shows analytically that the shielding becomes significant precisely in the high-temperature, high-density regimes relevant to ITER and DEMO operation. New closed-form expressions for the shielding magnitude are derived and verified against the SLAYER numerical code, extending existing theory to include ion parallel flow. Quantifying this effect accurately matters because over- or under-estimating shielding directly affects how much perturbation coil current is needed to achieve ELM suppression.

██████████ 0.8 plasma-disruption Preprint

ReadSaveConnections

Measurements from the IMPED mirror device show that large-scale Rayleigh-Taylor instabilities (caused by plasma buoyancy in curved magnetic fields) transfer energy nonlinearly to lower-frequency drift waves through three-wave coupling, using two established spectral analysis methods (Ritz and Kim) validated first on synthetic data. This matters because understanding which turbulent channels carry free energy determines the dominant transport loss mechanisms, which feed directly into the reduced models used to predict confinement time in reactor designs. The paper also provides useful practical guidance on when each analysis method is statistically valid based on data kurtosis and stationarity.

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

ReadSaveConnections

Standard MHD equilibrium theory predicts singular (infinite) plasma currents at resonant magnetic surfaces in 3D stellarators, which is physically impossible and undermines the theoretical foundation for stability calculations. This paper proposes a statistical equilibrium framework based on ergodic magnetic fluctuations that produces smooth solutions, resolving this decades-old inconsistency. If validated, this framework would provide a more rigorous basis for computing stellarator equilibria, with knock-on effects for stability analysis and transport modeling in devices like W7-X and HSX.

██████████ 0.7 turbulence-modeling Preprint

ReadSaveConnections

Simulations show that a stellarator built from simple tilted circular coils — far easier to manufacture than the complex helical windings used in W7-X — can achieve nested magnetic flux surfaces and low neoclassical transport (the dominant loss mechanism in stellarators), verified using the DESC equilibrium solver and field-line tracing. Stellarators offer disruption-free operation but have historically been expensive to build due to coil complexity. A practical simplified coil geometry could substantially reduce cost and accelerate the experimental programme for this reactor concept.

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

ReadSaveConnections

Particle-in-cell simulations on a 1600³ grid show that 3D effects delay the onset of magnetic reconnection — the process that releases stored magnetic energy during disruptions — compared to 2D models, with a strong guide field delaying it further by suppressing oblique instability modes. However, once fast reconnection begins, all cases converge to the same normalized rate of 0.08–0.10 regardless of dimensionality or guide field strength. This convergence result suggests that 2D models may correctly capture the peak reconnection rate even if they mispredict timing, which is directly relevant to how disruption onset models are calibrated.

██████████ 0.7 plasma-disruption Preprint

ReadSaveConnections