A dangerous instability in which the plasma column moves rapidly upward or downward from its equilibrium position and strikes the first wall — one of the most severe mechanical loading scenarios for a tokamak's vacuum vessel and internal components.
In a tokamak with an elongated (vertically stretched) plasma cross-section, the plasma is inherently unstable to vertical displacements. Active feedback control systems using external coils continuously correct small perturbations to keep the plasma centred. A vertical displacement event occurs when this feedback control is lost — typically during or immediately following a disruption — and the plasma column drifts rapidly toward the upper or lower wall of the vacuum vessel.[1]
The consequences are twofold and compounding: the moving plasma scrapes against the wall, depositing its remaining thermal and magnetic energy onto an asymmetric, localised area; and the decaying plasma current drives large "halo currents" through the vacuum vessel structure, creating electromagnetic forces that can reach hundreds of meganewtons in reactor-scale devices.
Tokamak plasmas are elongated (typical elongation κ = 1.6–1.9) because vertically stretched cross-sections improve energy confinement and stability limits, allowing higher plasma pressure and current for a given magnetic field. However, an elongated plasma sitting in a purely external poloidal field is like a ball balanced on a saddle point: any vertical displacement reduces the restoring force and increases the driving force, making the equilibrium vertically unstable.[2]
The instability growth rate is set by the ratio of the plasma elongation to the time constant of the nearest conducting structures (passive stabilisation plates or the vacuum vessel wall). For ITER, the passive stabilisation system slows the vertical instability to a growth rate of roughly 5–10 ms, giving the active control system enough time to respond under normal conditions. During a disruption, however, the plasma current and shape change too rapidly for feedback to track.
A typical VDE proceeds in stages:
1. Loss of vertical control. A disruption or control-system failure allows the plasma to begin drifting vertically. The drift is initially slow (tens of milliseconds) as eddy currents in passive stabilisers resist the motion.
2. Wall contact. The plasma edge touches the first wall or a limiter surface. Current continues to flow in the plasma, but the cross-section is now partially bounded by a material surface rather than a magnetic separatrix.
3. Current quench with wall contact. The plasma current decays while the plasma is pressed against the wall. During this phase, a fraction of the plasma current transfers into the scrape-off layer and flows through the wall as a halo current. The combination of halo current and toroidal magnetic field produces large vertical forces on the vessel; toroidal asymmetry in the halo current adds lateral (sideways) forces and torques.[1]
VDEs are classified by whether the thermal quench occurs before or after significant vertical displacement:
Cold VDE: The thermal quench happens first (as in a normal disruption), cooling the plasma before it moves far. The subsequent vertical drift and current quench produce halo currents but relatively modest thermal loading on the contact surface.
Hot VDE: The plasma retains much of its thermal energy as it drifts into the wall, depositing a large thermal load on a small, asymmetric area before the current quench. Hot VDEs are less common but more damaging to plasma-facing components. Some analyses suggest they are the design-limiting thermal transient for the ITER first wall.[3]
The primary defence against VDEs is disruption avoidance and early mitigation: if a disruption can be predicted and mitigated (e.g., by shattered pellet injection) before vertical control is lost, the VDE is prevented entirely. Additional measures include robust passive stabilisation structures with low resistivity (to slow the instability), fast-response vertical stability coils, and structural design margins that assume a certain number of full-load VDEs over the device lifetime.[2]