276

Papers

8/8

Roadblocks Active

2

Connections

Gyrokinetic simulations show that when the parameter β_eff = q²βₑ exceeds a critical value, turbulence abruptly transitions from a low-transport state to a high-transport state — driven not by any linear instability crossing a threshold, but by magnetic (Maxwell) stress overtaking the Reynolds stress that normally drives self-regulating zonal flows. Because zonal flows are the plasma's own mechanism for suppressing turbulence, their suppression by magnetic fluctuations causes heat losses to spike. This identifies a concrete, calculable design limit that reactor plasma conditions must stay below to maintain good confinement.

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

ReadSaveConnections

Using the well-benchmarked GENE gyrokinetic code on MAST-U relevant parameters, this study demonstrates that externally imposed E×B flow shear — long treated as a turbulence-suppression tool — can instead sharply increase heat transport when the imposed shear is strong enough to destroy the plasma's self-generated zonal flows. The mechanism is a spatial incompatibility: imposed shear layers and self-organized zonal shear layers cannot coexist, and when the former wins, the usual regulatory feedback is lost. This overturns a common design assumption and suggests that momentum injection schemes for confinement improvement must be carefully matched to local zonal-flow dynamics.

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

ReadSaveConnections

GyroFlow is a latent generative model that estimates the steady-state statistical distributions of gyrokinetic plasma turbulence in 5-dimensional phase space without simulating the slow, expensive approach to saturation — exploiting the mathematical equivalence between time averages and ensemble averages in ergodic systems. It outperforms autoregressive and reduced-order baseline models. If validated on reactor-relevant geometries, this approach could reduce turbulence characterization time from hours to seconds, making exhaustive reactor parameter scans computationally feasible.

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

ReadSaveConnections

Gyrokinetic simulations of the STEP spherical tokamak show the same q²βₑ threshold found in companion work governs the transition to large electromagnetic heat fluxes in that device. A key finding is that large-major-radius tokamaks reach this damaging regime at lower electron beta than small spherical tokamaks — meaning that reactor-scale devices like ITER may actually benefit from MHD-regulated turbulence saturation that is harder to access in compact designs. This has direct implications for which plasma operating points are safe in devices currently under design.

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

ReadSaveConnections

Runaway electrons generated during plasma disruptions — one of the most dangerous threats to tokamak walls — are shown to transport through disordered magnetic field regions not via simple diffusion, as assumed by the widely used Rechester-Rosenbluth model, but via anomalous power-law dynamics caused by electrons getting temporarily trapped near magnetic islands. This means energetic electrons linger far longer than classical models predict before escaping to the wall. Mitigation strategies based on the diffusive approximation may be systematically underestimating wall damage risk.

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

ReadSaveConnections

Gauss Fusion presents a computational design for a stellarator plasma equilibrium scaled to deliver 3 GW of fusion power and 1 GW of electricity, starting from the proven Wendelstein 7-X geometry. The optimized configuration retains more than 85% of the alpha-particle self-heating power and reduces the neoclassical transport loss metric (effective ripple) below 0.01 — both necessary conditions for a commercially viable stellarator. This is a commercial roadmap document showing that W7-X heritage configurations can, in principle, be scaled to power-plant conditions using current optimization codes.

██████████ 0.9 q-engineering Preprint

ReadSaveConnections

A new neoclassical transport module built into the NLT gyrokinetic code is validated against the standard NEO code for multi-species plasmas, including carbon and tungsten impurities, across conditions relevant to the EAST tokamak. Accurate modeling of heavy-impurity transport matters because tungsten atoms sputtered from reactor walls can migrate into the plasma core, where they dilute the deuterium-tritium fuel and radiate energy, potentially extinguishing the plasma. The successful benchmark against NEO gives confidence that this tool can predict tungsten accumulation risks in future devices.

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

ReadSaveConnections

Applying directional wavelet analysis to drift-wave turbulence simulations reveals that zonal flows produce distinct statistical fingerprints: turbulent fluctuations are more anisotropic (elongated) perpendicular to the flow and more intermittent (spiky, bursty) along it, at specific spatial scales. These directional signatures are quantitatively different from what isotropic turbulence models would predict. This matters because such statistical patterns could serve as experimental validation targets — allowing measurements from plasma diagnostics to test whether turbulence models are correctly capturing zonal-flow physics.

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

ReadSaveConnections

TokaGrad is a JAX-based tokamak simulator where every physics step — plasma equilibrium, heat transport, L-to-H transition, pedestal formation — is mathematically differentiable, forming a single computational graph from actuator inputs to plasma outputs. This means gradient-based optimizers can directly tune heating waveforms and current profiles by back-propagating through the full simulation, rather than relying on expensive finite-difference approximations around a black-box code. Benchmarked against ASTRA and TRANSP on ITER-relevant scenarios, this architecture could substantially accelerate scenario optimization for current and future devices.

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

ReadSaveConnections

Helicon waves are a candidate method for driving off-axis plasma current in tokamaks — important for sustaining steady-state operation — but full-wave modeling is too slow for integrated plasma simulations. This paper delivers a compact reduced model that captures the dominant physics (parallel electron Landau damping) using a cold-plasma dispersion relation plus a single correction term, validated against established reference models over 1.6 million parameter samples from EAST, HL-3, DIII-D, and KSTAR. The model is fast enough to include in whole-discharge simulations, enabling integrated current-drive optimization for reactor designs.

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

ReadSaveConnections