Domain overview
Seeing and steering a fusion plasma
Original Fusenergy explanation, framed against public technical references. Educational, not engineering or investment advice.
You cannot control what you cannot measure, and a fusion plasma cannot be touched by any ordinary sensor. Diagnostics therefore infer temperature, density, current, and impurity content from light and fields the plasma emits: Thomson scattering measures electron temperature and density from laser light scattered off electrons, interferometry tracks line-integrated density, spectroscopy identifies impurities, and magnetic probes reconstruct the field and plasma shape in real time.
Those measurements feed real-time control loops that keep the plasma positioned, shaped, and stable on millisecond timescales. The highest-stakes control problem in a tokamak is the disruption — a sudden loss of confinement that dumps energy and current into the walls — so predicting and mitigating disruptions, increasingly with machine-learning models trained on operational data, is a major research front. Fault detection and robust actuators keep a shot from ending badly when a subsystem misbehaves.
Behind the control room sits a growing data infrastructure: high-rate pipelines, digital twins that model the machine alongside the real one, and ML operators that assist scenario planning. The ten topics — Thomson scattering, interferometry, spectroscopy, magnetic probes, real-time control, disruption prediction, fault detection, data pipelines, digital twins, and machine-learning operators — are where plasma physics becomes an operable, reproducible machine.
Non-contact measurement
Temperature, density, current, and impurities are inferred from scattered laser light, emitted spectra, and magnetic fields — no probe survives the core, so everything is remote sensing.
Disruption prediction
In tokamaks a disruption can damage the machine in milliseconds. Predicting and mitigating it — now often with ML on operational data — is a top control priority.
Real-time control and data
Shape, position, and stability are held by feedback loops on millisecond timescales, backed by high-rate data pipelines and digital twins for scenario development.