Presets
Playback
Charge separation
Breakdown
Dielectric breakdown leader
Structures on the ground
Return stroke and observer
Layers
Model check
The science
The cloud map
The left panel shows a vertical cross-section through a tripole thundercloud: a main negative layer near the reversal temperature level (about -15 C), a positive layer above and a small positive pocket near 0 C. Charge builds up by ice-graupel collisions, dQ/dt = Isep - Q/τ, with Isep ≈ 1 A (w/15 m/s)(LWC/1.5 g m-3). Each layer is a uniformly charged disc over a conducting ground (image charges). On the axis the field is exact, Ez = (σ/2ε0)[sgn(dz) - dz/√(dz²+R²)]; off the axis the discs are sampled by rings of point charges. The colour shows β|E|/Ek with Ek ≈ 2.6 MV/m · e-z/8.4 km the conventional breakdown field and β an enhancement factor for hydrometeors (measured storm fields are about ten times below Ek). A flash starts when the ratio reaches 1 below the main negative layer.
The leader as a dielectric breakdown model
In the right panel the stepped leader grows on a grid, following Niemeyer, Pietronero and Wiesmann (1984). The conducting channel is held at potential 0, the ground and grounded structures at 1, and the potential between them obeys ∇²φ = 0, solved by red-black successive over-relaxation (SOR, ω = 1.85) with a warm start after each step. A new channel cell is added next to the existing channel with probability pi ∝ φiη / Σ φjη, i.e. according to the local field. Small η gives bushy, fractal channels, large η a nearly straight one. Because grounded masts and rods are equipotentials with sharp tips, the field concentrates there and the leader is drawn towards them: this is why a lightning rod works, and the strike statistics (run 40 flashes) show how the rod changes where the flashes land.
Return stroke, energy and thunder
The leader charge is QL = λ L (L = lattice path length); the peak current is Ip = 10.6 QL0.7 (kA, C; Cooray). The return stroke runs up the channel at about vRS ≈ c/3 with the Heidler current i(t) = (I0/η) xn/(1+xn) e-t/τ2, x = t/τ1 (first stroke 1.8 / 95 us, n = 10). The energy is estimated as W ≈ ½ QL Δφ with Δφ from the potential of the disc model at the channel origin. The flash-to-bang delay is d / cs, cs = 331.3 √(1 + T/273.15) m/s.
Honest simplifications
A uniformly charged thin disc makes the field jump by σ/ε0 across its plane, which is why the on-axis curve has steps at the layer altitudes; a real layer is a few hundred metres thick and smoother. The model is two-dimensional: the leader lives on a square lattice (4-neighbour steps overestimate the straight-line length), one cell is one step of 60-90 m rather than the 50 m of real steps, and the SOR relaxation is only partly converged between steps in the interest of speed (the solver itself is verified against exact solutions in the checks). The cloud is a set of discs, Isep is a single number, and the strike statistics depend on the lattice size, η and the start scatter. Timings are slowed down for display; physical durations are in the readouts. No lightning-safety advice is intended: go indoors when thunder is audible.
Model checks: SOR against the exact linear potential between plates; Laplacian residual, maximum principle and tip enhancement for a grounded needle; ring-sampled vs exact disc field and ∫E dz = Δφ; Heidler peak and charge by two quadratures.