Presets
Playback
Parameters
Camera
Layers
Quality
Model check
The science
What you see. A ferrofluid is a suspension of ~10 nm magnetite grains in oil, so it behaves as a liquid with a large magnetic permeability. Above a magnet, the field pulls fluid toward the strongest field and heaps it up. Past a threshold field the flat surface becomes unstable and breaks into a hexagonal lattice of spikes: the Rosensweig (normal-field) instability.
Threshold. For a linearly magnetisable fluid (relative permeability μr) with a normal induction B, surface waves of wavenumber k obey ρω² = ρgk + σk³ − k²(μr−1)²B² / (μ0μr(μr+1)) (Cowley & Rosensweig, 1967). The field term is negative, so a mode grows when B exceeds B(k)² = μ0μr(μr+1)/(μr−1)² · (ρg/k + σk). The lowest point of that curve is the critical field Bc² = 2μ0μr(μr+1)/(μr−1)² · √(ρgσ) at kc = √(ρg/σ), i.e. a wavelength λc = 2π√(σ/ρg) of about 1 cm for a kerosene-based fluid. Three plane waves at 60° superpose to hexagonal peaks with nearest-neighbour spacing 2λc/√3. Low gravity (Moon, near weightlessness) widens the pattern; high surface tension does too.
The magnet. A uniformly magnetised disc is equivalent to a current sheet K = Br/μ0 on its side wall. The field on the fluid plane comes from a numerical Biot–Savart sum over 14 ring currents, so it falls off radially and spikes only form inside the radius where B ≥ Bc. The heap under the magnet comes from the linear-fluid normal traction (μr−1)B²/(2μ0μr) balanced by ρgh.
Honest simplifications. Linear magnetisation (real fluids saturate, which raises the threshold at large fields); a deep, laterally uniform layer (the real layer is 5 mm deep and the field decays across it); the field is not re-computed from the deformed surface. The onset in real fluids is subcritical with hysteresis (spikes persist when the field is lowered below Bc); this model switches on and off at Bc. Spike height is schematic (grows like √(B²/Bc² − 1) and is capped), the hump is clipped to conserve the fluid volume, and the animation is a viscous relaxation, not a time-dependent solution of the Navier–Stokes equations. The presets use plausible round numbers, not a particular product.
Checks. (1) numerical scan of the neutral-stability curve against the closed-form Bc and kc; (2) the same length scale for water gives the tabulated 2.7 mm capillary length; (3) the ring-current field against the exact on-axis disc formula; (4) the lattice spacing of the pattern used for the mesh.