Python API Reference

Field Classes#

All field classes follow the same interface:

• compute(magnetization) – returns the effective field as an (n, 3) array
• energy(magnetization) – returns the energy as a scalar (Joules)

Magnetization m is always an (n, 3) array of unit vectors.

DemagFieldFK#

Demagnetizing field via the Fredkin-Koehler scalar potential method. This is the primary solver and supports multiple BEM backends.

DemagFieldFK(coordinates, elements, Ms, *,
           scale=1.0, tolerance=1e-8,
           solver=SolverType.AMG_CG,
           h2_cache="", h2params=None,
           fmm_params=None, backend=None,
           tree_type=None)

Additional methods:

• energy_weak_form(m) – energy via element-level FEM weak form, more stable on meshes with sharp potential transitions

Convenience subclasses select a specific backend:

• DemagFieldFKH2Lib – CPU H2-matrix BEM
• DemagFieldFKFMM – GPU FMM (same as default)
• DemagFieldFKDense – dense BEM, $O(n^2)$, exact reference
• DemagFieldFKGalerkin – Galerkin BEM, $O(n^2)$, smoothed

DemagFieldShell#

Shell-transform demagnetizing field solver. Replaces the BEM boundary integral with a mapped shell region. Fully GPU – no CPU round-trip.

DemagFieldShell(coordinates, elements, element_zones, Ms, *,
                scale=1.0, tolerance=1e-8,
                solver=SolverType.AMG_CG,
                transform=ShellTransform.PARALLELEPIPED)

DemagFieldVolFMM#

Volume FMM demagnetizing field solver. Bypasses the Fredkin-Koehler decomposition entirely – computes the demagnetizing field directly via volumetric FMM. Fully GPU.

DemagFieldVolFMM(coordinates, elements, Ms, *,
                 scale=1.0, fmm_vol_params=None)

ExchangeField#

Exchange field via the FEM stiffness matrix with GPU SpMV.

ExchangeField(coordinates, elements, A, *, scale=1.0)

AnisotropyField#

Uniaxial magnetocrystalline anisotropy field.

AnisotropyField(coordinates, elements, Ku, axis, *, scale=1.0)

ZeemanField#

External (Zeeman) field.

ZeemanField(coordinates, elements, Ms, field, *, scale=1.0)

compute() returns $\mu_0 M_s \mathbf{H}_\text{ext}$.

Additional methods:

• set_field(field) – update the external field without rebuilding

Composable Effective Fields#

EffectiveField#

Low-level compositor. You construct individual field objects and add them by name.

eff = EffectiveField(coordinates, elements, scale=1.0)
eff.add("demag", DemagFieldFK(coords, elems, Ms, scale=scale))
eff.add("exchange", ExchangeField(coords, elems, A, scale=scale))

EffectiveFieldFK#

High-level compositor for BEM-based models. Owns the mesh and $M_s$, and provides add_*() / set_*() methods.

eff = EffectiveFieldFK(coordinates, elements, Ms, scale=1.0)

Also has compute(), energy(), compute_component(), energy_component(), component_names, node_volume, num_nodes – same as EffectiveField.

EffectiveFieldShell#

Same interface as EffectiveFieldFK, but for shell-transform meshes. Exchange, anisotropy, and Zeeman operate on body-only elements.

eff = EffectiveFieldShell(coordinates, elements, element_zones, Ms, scale=1.0)

The element_zones array tags each tetrahedron: >0 = magnetic body, 0 = air, -1 = shell.

EffectiveFieldVolFMM#

Same interface as EffectiveFieldFK, but uses DemagFieldVolFMM (direct volumetric FMM, no BEM or FK decomposition) for the demagnetizing field.

eff = EffectiveFieldVolFMM(coordinates, elements, Ms, scale=1.0)

All other methods (add_exchange, add_anisotropy, add_zeeman, set_*, recompute, compute, energy, etc.) are the same as EffectiveFieldFK.

Enums#

SolverType#

GPU linear solver presets (AMGCL with CUDA backend).

BemBackend#

TreeType#

FMM tree construction strategy (only affects FMM_GPU backend).

String shortcuts: "octree", "adaptive", "kdtree".

ShellTransform#

Parameter Structs#

ParamsFKH2Lib#

H2-matrix compression tuning (only affects H2LIB backend).

ParamsFKFMM#

GPU FMM accuracy and performance.

ParamsVolFMM#

Volume FMM parameters.