269

Papers

9/9

Roadblocks Active

0

Connections

Using self-consistent 3D fluid simulations of a diverted tokamak, this paper shows that drift-wave turbulence spontaneously generates the sheared flow responsible for the L-to-H confinement improvement — no external trigger assumed. The same simulation framework reproduces three experimentally observed features simultaneously: the power threshold, a density at which the threshold is minimized, and the well-known but poorly understood asymmetry between favorable and unfavorable magnetic field directions. Understanding why less heating power is needed in one configuration directly informs how ITER and future reactors choose their operating conditions.

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

ReadSaveConnections

Rather than running one expensive plasma simulation at a time, this study ran 256 full-physics gyrokinetic simulations concurrently and without human intervention, scanning how triangularity and elongation of the plasma cross-section affect heat exhaust across four power levels. A key finding is that the mechanism controlling edge heat at low power — neoclassical trapped ions responding to magnetic geometry — is completely different from the turbulent mechanism that dominates at high power, meaning design rules derived in one regime can be misleading in another. The full 50 TB dataset is publicly released, making this a reusable resource for the broader community.

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

ReadSaveConnections

ELMs — periodic bursts of energy that strike the reactor wall — are a central threat to divertor materials in future fusion devices. This DIII-D experiment shows that deliberately wobbling the plasma vertically at 20 Hz forces ELMs to fire four times more frequently, shrinking the stored energy released per burst from roughly 10% down to below 1%. Smaller, more frequent ELMs deposit far less peak heat on plasma-facing surfaces, demonstrating a simple, coil-based pacing technique that could protect ITER's divertor without requiring pellets or external magnetic perturbations.

██████████ 0.9 elm-control Preprint

ReadSaveConnections

Liquid metal walls are a candidate solution for managing the extreme heat loads in a fusion reactor, but pure tin would sputter heavy impurities into the plasma. This study uses reactive molecular-dynamics simulations to show that adding small amounts of aluminum or lithium to liquid tin causes those lighter elements to preferentially migrate to the surface when oxygen or hydrogen is present — effectively giving the surface low-Z sputtering characteristics while tin provides thermophysical stability. Sn-Al and Sn-Li alloys with tuned compositions emerge as viable plasma-facing material candidates, narrowing the experimental search space.

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

ReadSaveConnections

This paper re-implements a standard plasma turbulence simulation code inside TensorFlow so that automatic differentiation — the same technology behind neural-network training — can compute how any output (e.g., heat flux) changes with any input (e.g., plasma density gradient). The core challenge is that turbulence is inherently noisy, so gradients are approximate, but the authors demonstrate these approximate gradients are still useful for optimization and for coupling gyrokinetic physics into AI workflows. This matters because tuning a tokamak's plasma profiles to minimize turbulent losses currently requires many expensive trial-and-error simulations; differentiable physics could reduce that to a single pass.

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

ReadSaveConnections

Conducting shells surrounding a tokamak plasma act as passive stabilizers that slow down disruptions by resisting rapid changes in magnetic flux. This paper derives an analytical formula for how fast magnetic fields penetrate a segmented conducting shell — one broken into toroidal sections like an armadillo's armor — finding that segmentation forces currents into a standing-wave pattern and that the effective resistivity grows quadratically with the number of gaps. This result directly informs how passive stabilizer structures in future devices must be designed to retain adequate disruption-braking performance while accommodating the mechanical and diagnostic penetrations that make segmentation unavoidable.

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

ReadSaveConnections

Controlling plasma instabilities in real time is hard because sensors measure only coarse averages (density, temperature at a few points) while the physics occurs at the microscopic particle level. This paper provides a theoretical framework for training a controller on sparse sensor data by imitating a hypothetical perfect controller that has full information, and derives mathematical guarantees on how good the resulting policy can be given the information gap. The work is currently restricted to a simplified 1D plasma model (Vlasov-Poisson), but the stability guarantees it provides could guide the design of practical controllers for disruption avoidance in real tokamaks.

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

ReadSaveConnections

SPARC, currently under construction, is designed to produce 140 MW of fusion power at an energy gain Q≈11. This study predicts the gamma-ray signals that SPARC's diagnostics will detect, using a chain of established plasma and radiation transport codes to simulate how DT fusion reactions, radio-frequency heating, and background neutrons all contribute to the measured spectrum. The value is practical: by knowing what gamma-ray signatures to expect before the machine operates, engineers can design detectors and shielding that will actually be able to measure fusion power, fast-ion distributions, and heating performance from day one of DT operations.

██████████ 0.7 q-engineering Preprint

ReadSaveConnections

This computational design study proposes adding 288 small programmable coils to a tokamak-like device, which the authors calculate would allow access to over 1.66 million distinct magnetic configurations — spanning the full spectrum from tokamak to stellarator geometries — using a single fixed hardware installation. Representative configurations in the scan show low neoclassical particle losses and good confinement of fusion-born alpha particles. The concept is entirely unbuilt, but if realized it would function as a high-throughput laboratory for testing which magnetic geometries perform best, potentially shortening the stellarator optimization pathway from decades to years.

██████████ 0.6 long-confinement Preprint

ReadSaveConnections

In a magnetized plasma, zonal flows (bands of sheared velocity) suppress turbulent transport, while streamers (radially elongated structures) enhance it — the balance between them strongly influences confinement. This laboratory experiment in a linear plasma column maps the transition between these regimes by varying neutral gas pressure, identifying the specific nonlinear coupling mechanism by which neighboring drift waves merge through a mediator mode to flip the system from flow-dominated to streamer-dominated behavior. Knowing the collision-frequency threshold at which this flip occurs provides a direct experimental handle for predicting and potentially avoiding degraded-confinement states in tokamak edges.

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

ReadSaveConnections