AI Framework Controls Fusion Plasma in Milliseconds
Princeton Plasma Physics Laboratory and Princeton University researchers have demonstrated PACMAN, an integrated AI control system that monitors and adjusts tokamak plasma on timescales of about 20 milliseconds. In tests on the DIII-D facility, it predicted a tearing-mode instability roughly 200 milliseconds ahead and altered plasma conditions to prevent it, while also coordinating multiple heating systems and meeting density and rotation targets. The modular design lets multiple machine-learning models share data and run in a continuous loop far faster than human operators.
Tokamaks confine plasma hotter than the Sun’s core with magnetic fields; small disturbances can grow and quench the reaction within milliseconds. Conventional simulations are too slow for real-time use, and earlier AI controllers were often isolated. PACMAN packages live sensor data, runs predictive models, resolves conflicting commands, and enforces hardware safety limits before sending instructions, keeping human operators in charge of high-level goals.
Key uncertainties remain around generalization to larger or differently configured machines and long-term reliability under continuous operation. Performance gains must still be weighed against the risk of unforeseen model failures, even with hard safety overrides. Scaling from research shots lasting minutes to sustained power-plant conditions is unproven.
Sources: ScienceDaily, Princeton Plasma Physics Laboratory, Nuclear Fusion.
Realistic Model Advances Detection of Fractons
Helmholtz-Zentrum Berlin researchers report numerical evidence that fractons—nearly immobile quasiparticles—can appear in a more realistic two-dimensional quantum spin-1 model of a solid. Earlier predictions relied on highly abstract gauge theories; the new simulations show a gapless fracton quantum spin liquid phase can survive under quantum conditions that previously erased the particles or reduced them to classical objects. Limited mobility arises because a single fracton cannot move without coordinated help from others.
Fractons emerge at corners of magnetic domain walls in systems where electron spins fluctuate like a liquid even at absolute zero. Their restricted motion has been proposed as a route to more robust quantum information storage, since local errors would be harder to propagate. The improved modeling of spin interactions overcame earlier trade-offs between quantum effects and particle stability.
Experimental confirmation is still missing. Identifying or engineering materials that match the model’s assumptions is the next barrier; Rydberg-atom simulators are one candidate platform. Whether the predicted phase is stable enough for practical use, or whether detection signatures will be unambiguous, remains open.
Sources: ScienceDaily, Helmholtz-Zentrum Berlin, Nature Communications.
Ringdown Signals Offer Test for Black Hole “Hair”
A Nagoya University-led team has calculated how surrounding matter or deviations from pure general-relativity black holes would alter the gravitational-wave ringdown that follows a merger. Frequency and damping rate respond differently to extra “hair,” producing a distinctive mismatch that depends on the amount and pressure distribution of the hidden matter. For spinning black holes the effect further splits according to whether waves co-rotate or counter-rotate with the spin.
In the simplest Kerr black holes, ringdown depends only on mass and spin. Any additional structure—ordinary matter or new physics—should leave an imprint on the fading waveform. The researchers linked photon-orbit properties to ringdown modes, added small amounts of extra matter to standard models, and mapped the resulting shifts across several theoretical cases.
Current detectors may lack the sensitivity to resolve the predicted differences cleanly. Distinguishing genuine hair from other environmental effects or modified gravity will require higher-precision future observations and careful modeling of multiple hair types. The method supplies concrete templates rather than a detection.
Sources: Phys.org, Nagoya University, Journal of Cosmology and Astroparticle Physics / arXiv.