filament, also called a blob, is a coherent field-aligned structure of dense 📝plasma that detaches from the edge of a 📝tokamak and propagates radially outward through the 📝scrape-off layer, carrying much of the particle transport that crosses it.
The two names describe the same object from different angles: elongated along the magnetic field over metres, it is a filament; localised to a centimetre or so across the field, it appears in a cross-field image as a blob. The propagation mechanism is well understood. Curvature and grad-B drifts separate charge vertically inside the overdense structure, the resulting internal electric field crosses the 📝toroidal field, and the E-cross-B drift pushes the whole structure radially outward — a plasma analogue of a buoyant bubble. How fast it goes depends on how that internal charge is drained, which is what defines the velocity scaling regimes: sheath-connected, when the current closes through the 📝divertor sheath, versus resistive regimes, when parallel resistivity or neutral collisions break that connection.
Filaments matter because they make scrape-off-layer transport intermittent and convective rather than diffusive. They flatten the far-SOL density profile, deliver particles and heat directly to the 📝first wall and main-chamber 📝plasma-facing components instead of to the divertor targets, and drive main-chamber recycling and erosion. They are also a distinct phenomenon from an 📝edge localized mode: filaments are continuous turbulent output of the boundary, present in every discharge, while an ELM is a discrete 📝pedestal collapse that ejects far more energy at once. Filaments are one of the clearest concrete instances of 📝anomalous transport — transport that classical theory cannot account for and that is instead measured directly.
And measured is the operative word. Gas puff imaging on 📝Alcator C-Mod showed that in Ohmic lower-single-null discharges both average blob size and radial velocity grow with core density, shifting from sheath-connected toward resistive scaling as the empirical density limit is approached (Garcia and colleagues, 2026, DOI 10.1088/1361-6587/ae6bbb); the same paper's extension of that scaling to 📝SPARC is a prediction, not a C-Mod result. On TCV, the TCV-X21 validation case compared measured filaments against GBS turbulence simulations and found radial and poloidal velocities in agreement while the simulations overestimated filament size (Wang and colleagues, 2026, DOI 10.1088/1361-6587/ae6d71) — a useful reminder that the codes used to project boundary behaviour in future machines still get the amplitude of these structures wrong.