Magnetic Garden (2D) - Dipole Chains, Field Lines and the Rosensweig Threshold

Top-down view of freely rotating disc magnets with exact dipole-dipole forces and field lines; a draggable probe magnet; and the hexagonal spike lattice of a ferrofluid dish.

Switch to 3D

Presets

Playback

Magnet garden

Ferrofluid dish (Rosensweig)

Layers

Model check

The science

Disc magnets as point dipoles

Each magnet is treated as a point dipole of moment m lying in the plane of the felt board. Two dipoles at separation r have energy U = (mu0/4 pi) [ m1.m2 - 3 (m1.r^)(m2.r^) ] / r^3. Head-to-tail (a chain) gives -2 E0 at contact, side-by-side parallel gives +E0, so magnets snap into chains. The unit is E0 = (mu0/4 pi) m^2 / d^3; for a 5 mm NdFeB disc (about 0.019 A m^2) it is about 0.28 mJ, so a contact pair holds roughly 0.56 mJ. The force and the torque on each magnet are the exact gradient of this energy; the page checks that numerically.

Field of a magnet falls as 1/r^3

On the axis a dipole has B = (mu0/4 pi) 2m / r^3. The graph compares this with an independent picture of two opposite magnetic poles a tiny distance apart (each giving a 1/r^2 field); the difference of the two 1/r^2 fields leaves 1/r^3, with log-log slope -3. The force between two magnets therefore falls as 1/r^4.

Dynamics

Magnets slide and turn on felt, so friction dominates inertia: the motion is overdamped, dx/dt = F / gamma, d phi/dt = tau / gamma_r, plus random kicks of size sqrt(2 kT dt) for temperature. A steep (sigma/r)^12 repulsion keeps the discs from overlapping (contact at about 1.01 d). At zero temperature the energy only goes down (a gradient flow) and the garden gets stuck in the nearest arrangement; a little temperature lets chains reconnect and close into rings, which have no free poles. Closed rings are the lowest-energy state for small clusters of dipoles in a plane. The probe is a vertical dipole hovering above the board: its field in the board plane is 3 M h (x,y) / (rho^2 + h^2)^(5/2) (pointing toward the point under it) and pulls every magnet along the field gradient.

Ferrofluid dish and the Rosensweig threshold

For a ferrofluid with linear magnetisation (relative permeability mu_r) under a vertical field B, the surface waves obey rho w^2 = rho g k + sigma k^3 - k^2 (B^2/mu0) (mu_r-1)^2 / (mu_r (mu_r+1)) (Cowley and Rosensweig 1967). The flat surface becomes unstable when the last term overcomes gravity and surface tension for some k. The most unstable wavenumber is k_c = sqrt(rho g / sigma) (wavelength about 10 mm) and the threshold is B_c^2 = 2 mu0 mu_r (mu_r+1)/(mu_r-1)^2 sqrt(rho g sigma). The dish shows the spikes of a hexagonal lattice with spacing 4 pi / (sqrt(3) k_c). The page scans the neutral curve numerically to confirm the closed form.

Honest limits

Real discs are not point dipoles at contact (their field is not exactly dipolar) and their moments could be slightly different; board friction is a simple viscous term; the 3D picture shows a flat board with in-plane magnets only. The spike height above threshold is a schematic sqrt(B^2/B_c^2 - 1) shape, not a solution of the nonlinear problem; real ferrofluids saturate, show hysteresis, and have much lower thresholds in practice. The fluid and the magnets do not interact in this model (separate experiments in one scene). Family-friendly science toy, no product claims.

Energy per magnet over time
Dipole field on the axis (log-log, slope -3)
Ferrofluid neutral curve and Rosensweig threshold
Cluster sizes