285

Papers

8/8

Roadblocks Active

1

Connections

Nonlinear gyrokinetic simulations of an ARC V3A reactor scenario (Q~20) show that fusion-born alpha particles significantly reduce turbulent heat and particle fluxes in the inner half of the plasma core. The mechanism is indirect: fast alphas excite modes that interact with zonal flows, which then suppress the Ion Temperature Gradient turbulence responsible for most heat loss. This matters because standard transport models use thermalized alpha distributions and therefore miss this suppression effect, potentially underestimating confinement in burning plasma conditions.

██████████ 0.9 turbulence-modeling Preprint

ReadSaveConnections

Sintered amorphous boron pebble rods were exposed to heat fluxes up to 80 MW/m² in the DIII-D tokamak divertor for the first time, testing a concept for a replenishable boronization source in reactor-relevant conditions. Significant dust emission was observed, with roughly half of the released boron escaping as fine particles into the plasma and vacuum chamber rather than being ionized and recycled. This is important because it provides the first direct data on erosion behavior and plasma contamination potential of a renewable boron format, which is being explored to replace batch boronization in future devices.

██████████ 0.8 divertor-thermal Preprint

ReadSaveConnections

This analytical study extends the classical theory of tearing mode instability — the magnetic reconnection process that seeds disruptions — to include anomalous resistivity that switches on above a current density threshold. The analysis finds that this threshold creates mathematical singularities that generate second-order corrections to mode stability, producing early-stage phase-slip layers not predicted by classical theory. Accurate tearing mode thresholds are critical because disruptions in reactor-scale tokamaks could cause irreversible first-wall damage, so more realistic stability criteria could improve disruption avoidance systems.

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

ReadSaveConnections

The open-source FreeMHD2 solver, built on OpenFOAM, now computes the induced magnetic field self-consistently in free-surface liquid metal flows rather than assuming it is negligible — the inductionless approximation that fails at high magnetic field strengths relevant to fusion divertors. A vector-potential formulation enforces zero magnetic divergence by construction, and the solver has been experimentally validated against free-surface height measurements from the LMX-U liquid metal facility. This provides a validated, publicly available tool for designing liquid metal divertor concepts (e.g., flowing lithium or tin surfaces) where inaccurate MHD modeling could lead to flawed thermal-hydraulic predictions.

██████████ 0.8 divertor-thermal Preprint

ReadSaveConnections

This work reframes tokamak magnetic equilibrium reconstruction as a machine learning transfer problem: a neural operator pretrained across multiple tokamak geometries can be fine-tuned to a new, unseen device using as few as 100 labeled examples, achieving mean relative errors below 4%. Single-geometry pretraining transfers poorly, showing that geometric diversity in training data is the key enabler. Faster and more adaptable equilibrium solvers matter because real-time equilibrium reconstruction underpins plasma control decisions, and building models from scratch for each new device is expensive.

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

ReadSaveConnections

A neural operator surrogate called LRC-FNO is introduced to replace the expensive field-solving step in particle-in-cell plasma simulations, using a two-level architecture that separately captures coarse field responses and fine-scale corrections. On a 2D scrape-off-layer fusion plasma test case it achieves relative errors below 5% and maintains physical consistency in closed-loop simulations out to nearly twice the training time horizon. Faster PIC surrogates could substantially accelerate turbulence and transport modeling in the plasma edge, a region where conventional simulations are computationally prohibitive at reactor scale.

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

ReadSaveConnections

This framework learns the full statistical distribution of a system's outputs conditional on its inputs — not just the average — allowing it to detect influential variables that only manifest through variability or tail behavior rather than mean shifts. Applied to stochastic dynamical systems including PDE-governed problems, it identifies parsimonious input sets that match the predictive accuracy of full models. For fusion, this approach could help construct digital twins of plasma systems that identify which sensor signals genuinely drive confinement changes without overfitting to large but noisy diagnostic streams.

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

ReadSaveConnections

The linear Vlasov-Poisson system governing collisionless plasma is reformulated as a Schrödinger-type equation, enabling the application of fluctuation theorems from nonequilibrium statistical mechanics to quantify how far a plasma state is from equilibrium. Explicit solutions are derived using Case-Van Kampen eigenvectors, and a stochastic relative entropy is defined as a rigorous measure of nonequilibrium dynamics. While theoretical, this framework could provide new mathematical tools for characterizing turbulent fluctuations and energy transfer in magnetically confined plasmas beyond standard quasilinear approximations.

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

ReadSaveConnections

This theoretical monograph distinguishes two regimes of chaos in coupled oscillatory systems: thin chaos, where instability remains locally confined with limited transport consequences, and thick chaos, where instability propagates to produce large-scale failure. A Structural Exit Theorem is proposed asserting that protective dynamical structure is lost through exactly four geometric mechanisms: separatrix splitting, resonance overlap, spectral-gap collapse, or mode activation. If the framework is formally validated, it could offer a diagnostic language for classifying which plasma perturbations are genuinely disruptive versus benign — though the work is a self-published non-peer-reviewed deposit with low methodological confidence and no numerical validation provided.

██████████ 0.5 plasma-disruption Peer-reviewed

Read

Stellarator magnetic configurations are optimized for alpha-particle confinement using a data-informed parameter space derived from existing stellarator designs via quantile transformation and PCA, combined with GPU-accelerated guiding-center tracing inside the optimization loop. The method finds configurations with good fast-ion confinement without requiring quasi-symmetry constraints, which conventionally dominate stellarator optimization. Retaining alphas long enough for them to slow down and heat the plasma is essential for a self-sustaining stellarator reactor, and this approach significantly reduces the computational cost of searching for viable configurations.

██████████ 0.4 q-engineering Preprint

ReadSaveConnections