283

Papers

9/9

Roadblocks Active

2

Connections

Sheared plasma rotation (E×B flow), conventionally regarded as a turbulence suppressant, can worsen heat transport when the imposed shear is comparable in strength to the self-generated 'zonal flows' that naturally regulate turbulence — because the imposed shear dismantles those zonal flows. Gyrokinetic simulations with the GENE code, validated by a reduced fluid model and applied to MAST-U geometry, show a non-monotonic transport response: moderate imposed rotation increases losses before stronger rotation eventually reduces them. This finding challenges the assumption that increasing plasma rotation always improves confinement, with direct implications for how rotation is used as a control tool in future reactors.

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

ReadSaveConnections

Local gyrokinetic simulations across two tokamak configurations (Cyclone Base Case and ST40) show that once the parameter q²β_e exceeds a critical value C_nl, electromagnetic Maxwell stress becomes large enough to suppress the self-generated zonal flows that normally cap turbulence, triggering a transition to a high-transport state. Crucially, this transition is driven by nonlinear dynamics and occurs before any linear stability boundary is crossed, so linear analysis cannot predict or warn against it. Identifying and respecting this threshold is a prerequisite for reliable confinement projections in high-β devices.

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

ReadSaveConnections

Using the GENE gyrokinetic code on the proposed STEP spherical tokamak reactor equilibrium, this paper shows that catastrophic turbulent heat fluxes appear at a plasma pressure (q²β_e) well below the linear MHD stability limit, meaning STEP could lose confinement without triggering any classical instability alarm. A key result for reactor design: larger-major-radius tokamaks reach this electromagnetic transport transition at lower β_e than smaller spherical tokamaks, making the operating margin tighter as machines scale up. This directly affects how STEP's operating point must be chosen to stay in the well-confined regime.

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

ReadSaveConnections

GyroFlow uses a generative machine-learning method (flow matching in latent space) to predict the saturated statistical state of 5D gyrokinetic turbulence directly from five plasma parameters, skipping the expensive transient simulation phase entirely. Trained on roughly 256 high-fidelity gyrokinetic simulations and evaluated against established fast models (QuaLiKiz, TGLF), GyroFlow is more accurate while running substantially faster. If the approach generalizes to a broader parameter range, it could enable turbulence-informed reactor scans that currently require supercomputing time, with code available at ml-jku/neural-gyrokinetics.

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

ReadSaveConnections

Laboratory tritium breeding tests use compact D-D neutron sources, but these produce a broad joint distribution of neutron energies and angles rather than the idealized single-energy isotropic beam typically assumed — and this mismatch matters enormously. Monte Carlo transport calculations (OpenMC) show that using realistic source spectra reduces the predicted tritium breeding ratio from 1.54 to 1.44, a 6.5% drop concentrated in the ⁶Li(n,t)⁴He channel, which is the primary breeding reaction. Since tritium self-sufficiency requires a breeding ratio above roughly 1.05, this source-fidelity correction could shift a blanket design from surplus to deficit and needs to be factored into ITER and DEMO blanket validation campaigns.

██████████ 0.9 tritium-breeding Preprint

ReadSaveConnections

During disruptions, high-energy runaway electrons travel through chaotically stochastic magnetic fields and can damage tokamak walls, but this paper shows their escape is not simple diffusion (as standard Rechester-Rosenbluth models assume) — instead, electrons get temporarily trapped near magnetic island boundaries, producing a power-law tail in the survival time distribution. A new semi-analytical model combining exponential and power-law decay reproduces this behavior in both a simplified symplectic map and a realistic JOREK simulation of a JET disruption. More accurate runaway transport predictions are needed to assess first-wall damage risk in ITER and to design effective mitigation strategies.

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

ReadSaveConnections

Experiments on the EAST tokamak show that coating the vessel wall with boron (via ICRF-assisted carborane boronization) triples the fraction of discharges that reach the desirable I-mode confinement regime compared to lithium conditioning (51% vs. 8%), and expands the accessible plasma density range from a Greenwald fraction of 0.35–0.54 to 0.26–0.77. I-mode is attractive for reactors because it sustains good energy confinement without triggering ELMs (the dangerous periodic edge instabilities associated with standard H-mode). Boron's lower sputtering yield appears to suppress impurity influx while allowing higher edge recycling, with implications for first-wall material choices in ITER and beyond.

██████████ 0.9 elm-control Preprint

ReadSaveConnections

TokaGrad is a JAX-based tokamak simulator in which every physics component — plasma equilibrium, transport, heating, L-H transition, and pedestal — is connected in a single automatic-differentiation computational graph. This means it can compute how any machine setting (heating waveform, coil current, fueling) affects the final plasma state without numerical finite-differencing, enabling gradient-based optimization of full discharge scenarios for the first time. Benchmarked against ASTRA and TRANSP on ITER-relevant L-to-H scenarios, the framework could substantially accelerate the search for optimal operating strategies that current trial-and-error approaches approach only slowly.

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

ReadSaveConnections

This paper presents the optimized magnetic equilibrium design for GIGA, a proposed 1-GW-electric stellarator power plant derived from the Wendelstein 7-X configuration and scaled to a 1500 m³ plasma volume producing 3 GW of fusion power. The optimized design achieves alpha particle confinement above 85% and neoclassical transport losses (effective ripple) below 1% — both necessary for net energy gain — while eliminating the disruption risk inherent to current-carrying tokamaks. As the first complete equilibrium basis for a GW-class stellarator, it provides a concrete reference point for comparing stellarator and tokamak reactor pathways.

██████████ 0.8 q-engineering Preprint

ReadSaveConnections

Tearing modes are magnetic reconnection instabilities that can grow into large magnetic islands and precipitate disruptions in tokamaks; this paper extends the classical Rutherford island-growth model to include sheared plasma flow. The key result is that a flow profile parameter r controls whether shear hastens or slows island saturation, and that the widely used constant-ψ approximation breaks down in the presence of Alfvén resonances — meaning existing disruption models based on that approximation may be quantitatively unreliable. The work provides updated theoretical groundwork for predictive disruption avoidance tools.

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

ReadSaveConnections