289

Papers

9/9

Roadblocks Active

4

Connections

Gyrokinetic simulations of the ARC compact tokamak show that fusion-born alpha particles — moving far faster than the background plasma — substantially cut ion-scale turbulent heat and particle transport in the inner core (within half the plasma radius). The suppression works through fast-ion-driven waves that energise zonal flows, which in turn break up turbulent eddies. This matters because alpha heating is what will sustain a burning plasma; if alphas simultaneously reduce the turbulence that tries to cool it, reactor performance projections could be considerably more optimistic than models that ignore this effect.

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

ReadSaveConnections

Sintered boron pebble rods were inserted into the DIII-D lower divertor and exposed to heat fluxes up to 80 MW/m² — comparable to what a fusion power plant divertor must survive. The boron eroded substantially, injecting boron ions into the plasma (which actually helps wall conditioning), but roughly half the material was lost as fine dust that dispersed into the vacuum chamber rather than being recoverable. This is the first direct test of a granular, self-replenishing plasma-facing concept in a real tokamak divertor, revealing both the promise and the dust-management challenge that must be resolved before renewable liquid or granular divertor surfaces can be used in a reactor.

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

ReadSaveConnections

Tearing instabilities are the seed of magnetic islands that can grow into full plasma disruptions — the abrupt, machine-damaging collapses fusion reactors must avoid. This paper analytically extends classical tearing theory to include anomalous (turbulence-enhanced) resistivity, showing it introduces sharp spatial singularities and a hyperbolic growth-rate divergence during the early phase-slip stage. Understanding this mechanism is a prerequisite for predicting when and how fast magnetic islands form in high-performance plasmas, which is essential for any disruption prediction or avoidance system.

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

ReadSaveConnections

Global gyrokinetic simulations using the TRIMEG code show that when the full nonlinear response of the background thermal plasma is included, fast-ion (energetic particle) transport becomes significantly less 'stiff' — meaning it takes a much larger drive to push transport to high levels. Specifically, the scaling of energetic-particle heat flux with the pressure gradient weakens from a quartic to a roughly quadratic dependence. Because stiff transport is one reason reactor designs must operate well above threshold, this result suggests current models may be overestimating confinement losses for fast ions, with implications for neutral beam and alpha-particle confinement in ITER and beyond.

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

ReadSaveConnections

Liquid metal divertors (using lithium or tin flowing over the plasma-facing surface) are a leading alternative to solid tungsten for handling the enormous heat loads in a fusion reactor, but designing them requires accurately modelling how magnetic fields interact with the flowing liquid. This paper extends the open-source FreeMHD code to self-consistently evolve the magnetic field induced by flowing liquid metal currents — an effect that prior versions ignored — and validates the result against laboratory liquid-metal experiments. The improvement allows engineers to predict electromagnetic braking, surface stability, and heat transfer in liquid-metal divertor channels under conditions that closer to a real reactor.

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

ReadSaveConnections

Electron cyclotron heating (ECH) can stabilise dangerous tearing instabilities if microwave power is deposited precisely at the right location inside the plasma, but calculating this in real time is too slow for conventional ray-tracing codes. The ECHO algorithm replaces the ray-tracing code with a fast neural-network surrogate and uses a genetic algorithm to optimise mirror angles and power in real time on DIII-D, demonstrating robustness even when individual gyrotrons fail mid-discharge. Successfully automating ECH profile control is a concrete step toward the closed-loop plasma control systems that ITER and future plants will require to avoid disruptions.

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

ReadSaveConnections

High-harmonic fast waves (HHFW) used for plasma heating can decay into daughter waves near the scrape-off layer (the thin boundary region between hot core plasma and the wall), a process this paper analyses theoretically for an inhomogeneous plasma. The decay produces ion-cyclotron Bernstein waves and leads to parametric turbulence, with anomalous ion heating as a side effect. Understanding where RF power is actually deposited — versus lost to parasitic processes near the edge — is critical for designing efficient heating systems that do not inadvertently damage plasma-facing components through unexpected edge heating.

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

ReadSaveConnections

Turbulence and plasma instabilities in tokamaks involve stochastic, high-dimensional dynamics where the relevant cause-and-effect relationships are not obvious. This paper presents a method that learns which variables actually drive a system's behaviour by examining how the full probability distribution — not just the average — changes with inputs, then automatically prunes uninformative variables. For fusion, the approach could help build interpretable digital-twin surrogates that identify the handful of plasma parameters genuinely governing confinement or disruption onset, reducing the dimensionality challenge in real-time control.

██████████ 0.5 turbulence-modeling Preprint

ReadSaveConnections

Particle-in-cell (PIC) simulations of plasma behaviour are computationally expensive because they must solve for electromagnetic fields at every timestep across millions of particles. The LRC-FNO architecture presented here reduces this cost by learning a fast neural-network approximation of the field solver, achieving relative errors around 4–5% on scrape-off-layer plasma benchmarks and remaining stable over simulation times twice the training window. Faster, accurate field solvers could eventually make turbulence simulations of the plasma edge — a critical region for heat exhaust and erosion — more tractable for routine engineering design.

██████████ 0.5 turbulence-modeling Preprint

ReadSaveConnections

This theoretical paper recasts the equations governing collisionless plasma (including gyrokinetics, the standard framework for fusion turbulence simulations) into a Schrödinger-like form, then derives fluctuation theorems — statistical relations that constrain how entropy and energy fluctuate in non-equilibrium systems. The reformulation connects the plasma physics eigenstates (Case-Van Kampen modes) to the mathematical structure of quantum mechanics, potentially opening new tools from quantum information theory for analysing plasma transport. The practical payoff is still distant, but the framework could offer new analytical bounds on transport coefficients without running expensive simulations.

██████████ 0.4 turbulence-modeling Preprint

ReadSaveConnections