/ fp-disk ← micromagnetism.com
University of Vienna · magnetic particle imaging & hyperthermia

Dynamic loop of a thin magnetic disk in an AC field

A thin disk (diameter D, thickness t) is treated as one spin: its shape anisotropy confines the moment to the disk plane, a weak uniaxial anisotropy Ku selects the in-plane easy axis x, and the alternating field H(t) = H0 sin(2πf t) is applied along x. The page solves the Fokker–Planck equation for the in-plane angle φ and shows the dynamic loop mx(H), the dissipated energy per cycle, and the quantities relevant for magnetic particle imaging (point-spread-function width) and hyperthermia (power density).

Everything on this page is computed live in your browser (Web Worker, no server, no lookup tables) from the one-spin model of a thin disk below; the Python reference solver gives identical numbers (area/(4B0) = 0.0214 for the default case).

How the model works: energy, the unknown W(φ,t), the reduced equation, symbols, numerics, validityfor a physicist who has not read the paper

What is computed

A magnetic disk of diameter D and thickness t ≪ D, small enough to be uniformly magnetised, is treated as a single macrospin μ = MsV, V = πD²t/4. The disk is immobilised and does not interact with neighbours. Thermal agitation makes the moment direction fluctuate; the applet solves the Fokker–Planck equation for the probability density of the in-plane angle, live in your browser, and integrates the loop of the ensemble average mx(t) = ⟨cos φ⟩ over one period.

The energy

The shape anisotropy of a thin cylinder is an easy plane of strength K⊥ = μ0Ms²(Nz − Nx)/2, with Nx = (τ/π)(ln(4/τ) − 1/2), Nz = 1 − 2Nx, τ = t/D. It is so strong (μ0HK,⊥ = 2K⊥/Ms is of order 1 T) that the moment stays in the plane, m = (cosφ, sinφ, ε) with a small tilt ε. In the plane the energy density is

f(φ) = −Ku cos²φ − μ0Ms H(t) cosφ,   H(t) = H0 sin(2πf t) along x

For Ku > 0 there are two wells, φ = 0 and φ = π, separated by the barrier KuV (reduced: σu = KuV/kBT) at zero field; the field tilts the landscape. For Ku = 0 the disk is a free planar rotor whose equilibrium magnetisation is I1(ξ)/I0(ξ), ξ = μ0MsVH/kBT (dashed in the loop plot, next to the 3-D Langevin function L(ξ) of an isotropic moment).

The unknown: W(φ, t)

W(φ, t) is the probability density of the in-plane angle: W dφ is the probability that the moment points within [φ, φ+dφ] of the easy axis at time t; it is periodic in φ and normalised, ∫02π W dφ = 1. W is not the magnetisation; the magnetisation along the field is its first harmonic mx(t) = ∫ cosφ W dφ. The heat map below shows W itself.

The reduced equation

Mobility in three lines. LLG in the angles, with γ' = γ/(1+α²):

ε' = −(γ'/Ms) [ 2αK⊥ε + f'(φ) ],    φ' = −(γ'/Ms) [ α f'(φ) − 2K⊥ε ]

The tilt relaxes fast, ε' ≈ 0 ⇒ 2K⊥ε = −f'/α; inserting gives φ' = −(γ'/Ms)(α + 1/α) f' = −(γ/(αMs)) f'(φ): the in-plane mobility is γ/(αMs), a factor (1+α²)/α² (101 for α = 0.1) larger than for an isotropic macrospin — the precession about the strong easy-plane field is converted into in-plane motion. With thermal noise this becomes the Fokker–Planck (Smoluchowski) equation

∂W/∂t = ∂/∂φ [ (γ/(αMs)) ( W ∂f/∂φ + (kBT/V) ∂W/∂φ ) ]

Symbols

SymbolMeaningUnit · how it enters
D, t, τ = t/Ddisk diameter, thickness, aspect rationm · volume and demagnetising factors
V = πD²t/4disk volumem³ · thermal energy scale kBT/V
Mssaturation magnetisationA/m · Zeeman energy, K⊥, mobility, energy per cycle
Kuin-plane uniaxial anisotropy (easy axis x)J/m³ · −Ku cos²φ
Nx, Nzthin-cylinder demagnetising factors– · Nx = (τ/π)(ln(4/τ) − 1/2), Nz = 1 − 2Nx
K⊥, μ0HK,⊥easy-plane (shape) anisotropy and its field 2K⊥/MsJ/m³, T · keeps the moment in the plane, validity limit
α, γGilbert damping, gyromagnetic ratio 1.760859×1011–, rad s−1T−1 · mobility γ/(αMs)
T, kBTtemperature, thermal energyK, J · diffusion term
H(t), H0, B0 = μ0H0, fAC field along x, amplitude, frequencyA/m, mT, Hz · drive
φin-plane angle of the moment from xrad · the single variable
W(φ, t)probability density of φ1/rad · the solved quantity
mx = ∫ cosφ W dφreduced magnetisation along the field– · loop mx(μ0H)
A = −∮ mx d(μ0H)loop areaT · area/(4B0) = fraction of the ideal square loop
E = MsA, P = fEenergy per cycle and power densityJ/m³, W/m³ · hyperthermia
σu = KuV/kBTreduced in-plane barrier– · blocking of the in-plane switching
kBT/m, 4.16 kBT/mthermal field of the moment m = MsV; Langevin PSF FWHM estimatemT · MPI resolution scale
Hc, Mr|H| at mx = 0; mx at H = 0mT; units of Ms
α²B⊥/4validity limit of the overdamped reduction, B⊥ = μ0HK,⊥mT · compare with B0

Numerics — what “live” means

φ ∈ [0, 2π) is divided into N periodic finite-volume cells (default 256). Fluxes use the Scharfetter–Gummel scheme, so the discrete equilibrium is exactly the Boltzmann distribution. Time stepping is implicit Euler (L-stable, robust for the stiff high-field part) with S steps per period (default 4000); each step is a periodic (cyclic) tridiagonal solve, done with the Thomas algorithm plus a Sherman–Morrison correction. The periodic steady state comes from the propagator over half a period and the symmetry (φ, H) → (φ + π, −H) (a shift by N/2 cells): W* = R Mhalf W*, one dense N×N solve with one row replaced by the normalisation. From W* one period is propagated to record mx(t) and W(φ, t). Cost: about 0.6–1 s per loop in a Web Worker (shown in the status line).

Validity

Eliminating the tilt assumes that it relaxes faster than the in-plane motion. This holds strictly for in-plane fields B ≪ α²B⊥/4 (B⊥ = 2K⊥/Ms); for a 1-nm, 200-nm permalloy-like disk with α = 0.1 that is about 2.4 mT, i.e. the default 20 mT is outside the strict range and the card is flagged amber. Stochastic-LLG checks are running; at 300 kHz / 20 mT / Ku = 500 J/m³ the reduction agrees with micromagnetics within 5 %.

Steady-state loop mx(μ0H)

dynamic loopplanar rotor I₁/I₀ (equil., Ku=0)3-D Langevin L(ξ)

Distribution W(φ, t)

The solved quantity: probability density of the in-plane angle over one field period; the loop above is its first harmonic ⟨cosφ⟩. Colour = log10 W; vertical axis φ (0 = along +x in the middle, ±180° = along −x at the edges); thin line = h(t)/h0 (+1 top, −1 bottom).

Frequency scan

fixed B0, 40 log-spaced points 1 kHz–100 MHz, evaluated from high to low frequency
area/(4B0) versus f at the current disk, at fixed B0 or along H0 = C/f; about 0.5–1 s per point (more where the steps are raised). Click a point to adopt its frequency.