Skip to content

Fields, output, and restarts

Discover and read fields

After loading a mesh and setting magnetization:

print(simulation.get_all_field_names())
print(simulation.is_subfield_available("H_demag"))

nodal_field = simulation.get_subfield("H_demag")
average_field = simulation.get_subfield_average("H_demag")

Common fields include magnetization direction m, magnetization M, applied field H_ext, exchange field H_exch, demagnetization field H_demag, anisotropy field H_anis, total field H_total, energy densities, pinning, and the scalar-potential quantities phi and rho. Availability depends on the loaded materials and whether demagnetization is enabled.

Material-specific averages can be requested by material name:

py_average = simulation.get_subfield_average("m", "Py")

Probe a position

probe_subfield_siv takes positions in metres and returns ordinary SI values:

field = simulation.probe_subfield_siv("H_demag", [0.0, 0.0, 0.0])

It returns None outside the mesh. probe_subfield provides the same behavior and can accept an explicit unit argument.

NDT and HDF5 output

simulation.save_data()                  # averaged quantities only
simulation.save_data(fields=["m", "H_total"])
simulation.save_data(fields="all")      # every available spatial field

For a simulation named sample, Nmag writes:

  • sample_dat.ndt: tab-separated, SI-labelled averaged quantities, one row per save event;
  • sample_dat.h5: mesh metadata and requested nodal or cell fields.

Use HDF5 readers such as H5py for analysis. The current rewrite does not ship the legacy ncol executable. It does ship a modern single-snapshot nmagpp exporter:

nmagpp --vtk results/simulation.vtu \
  results/simulation_dat.h5 \
  --mesh assets/model.nmesh.h5 \
  --field m

Use --field NAME more than once or --all-fields to select additional point arrays. Modern spatial HDF5 output contains point coordinates and fields; an external mesh is required when it does not contain tetrahedral cell topology. The command exports one current snapshot and does not recreate the legacy time-series nmagpp format.

The same exporter is available as a normal Python import:

from nmag import export_vtk

export_vtk(
    "results/simulation_dat.h5",
    "results/simulation.vtu",
    mesh_path="assets/model.nmesh.h5",
    fields=("m", "H_demag"),
)

To write selected spatial fields to another compact HDF5 file:

simulation.save_spatial_fields("snapshot.h5", ["m", "H_demag"])

Output lifecycle

NmagConfig.output_policy controls files created by a new simulation:

  • "error" refuses to overwrite existing NDT/HDF5 output;
  • "replace" starts those output files again; and
  • "append" validates the old NDT schema before adding rows.

Append is an output choice, not a physical restart.

Restart checkpoints

checkpoint = simulation.save_restart_file("relaxed_state.h5")

# Create a compatible simulation and load the same mesh/materials first.
continued.load_restart_file(checkpoint)

Native checkpoints preserve magnetization, pinning, current density, external field, clock, and supported dynamics state. Mesh and material fingerprints must match. DOP853 is reconstructed from the physical state, so later adaptive step sizes need not reproduce an earlier run exactly.

load_m_from_h5file(path) transfers only magnetization and requires the same mesh, making it useful when other simulation settings should remain new.

These are native Python 3 checkpoint files. Legacy Nmag restart files are not supported.