Units, materials, and anisotropy¶

SI quantities¶

nmag.SI attaches physical dimensions to a value:

import nmag

length = nmag.SI(5e-9, "m")
field = nmag.SI(8e3, "A/m")
exchange = nmag.SI(13e-12, "J/m")

Use quantity.in_units_of(nmag.SI(1, "unit")) when an external library needs an ordinary number. Nmag checks compatible dimensions at API boundaries; do not strip units from material constants merely to silence an error.

Magnetic materials¶

permalloy = nmag.MagMaterial(
    name="Py",
    Ms=nmag.SI(1e6, "A/m"),
    exchange_coupling=nmag.SI(13e-12, "J/m"),
    llg_damping=0.02,
)

The most commonly used parameters are:

Parameter	Meaning	Typical units
`Ms`	Saturation magnetization	A/m
`exchange_coupling`	Exchange constant	J/m
`llg_damping`	Gilbert damping	dimensionless
`llg_gamma_G`	Gyromagnetic ratio	m/(A s)
`llg_polarisation`	Current polarization for Zhang-Li torque	dimensionless
`llg_xi`	Nonadiabatic Zhang-Li coefficient	dimensionless

scale_volume_charges is an expert control that scales only the interior volume-charge source of the demagnetization calculation. The physical default is 1.0; surface charges are never scaled by this setting.

Uniaxial anisotropy¶

easy_axis = nmag.uniaxial_anisotropy(
    axis=[0.0, 0.0, 1.0],
    K1=nmag.SI(1e5, "J/m^3"),
    K2=nmag.SI(2e4, "J/m^3"),
)

material = nmag.MagMaterial(
    name="uniaxial",
    Ms=nmag.SI(8e5, "A/m"),
    exchange_coupling=nmag.SI(10e-12, "J/m"),
    anisotropy=easy_axis,
)

The axis is normalized by the constructor. K1 and K2 may be SI energy densities or plain values interpreted as J/m³.

Cubic and custom anisotropy¶

Use nmag.cubic_anisotropy(axis1, axis2, K1, K2, K3) for crystalline cubic terms. The two supplied axes must define an orthogonal orientation.

A custom polynomial energy can be supplied as a callable with an explicit order:

def energy(m):
    return nmag.SI(1e5 * m[2] ** 2, "J/m^3")

material = nmag.MagMaterial(
    name="custom",
    anisotropy=energy,
    anisotropy_order=2,
)

Predefined models use vectorized analytic derivatives. Custom callables use a validated finite-difference derivative, so they are intended for smaller or exploratory models. The experimental Diffsol backend does not currently support anisotropy.