HDF5 and Gmsh

treams_rs.io needs the [io] extra, which installs h5py. It mirrors treams.io.

T-matrices in HDF5 files

io.save_hdf5 writes spherical T-matrices to an open HDF5 group in the tmat.h5 layout used by treams, with units, positions, the embedding material and optional metadata about the scatterers and the computation. io.load_hdf5 reads them back as a TMatrix, or as an array of them for a parameter sweep:

import h5py
import numpy as np
import treams_rs as tr
from treams_rs import io

sphere = tr.sphere_tmatrix(k0=2 * np.pi / 700, lmax=2, radius=75, material=4)
with h5py.File("sphere.h5", "w", driver="core", backing_store=False) as file:
    io.save_hdf5(file, sphere, name="sphere", lunit="nm")
    loaded = io.load_hdf5(file, lunit="nm")
np.testing.assert_allclose(loaded.array, sphere.array)
np.testing.assert_allclose(loaded.k0, sphere.k0, rtol=1e-15)

lunit names the length unit of the positions and of 1 / k0, here nm. A loaded T-matrix supports scatter and cross_sections like any other; evaluate fields on the scattered wave. load_hdf5 also reads the older treams names of the embedding chirality and particle indices.

Meshes

io.mesh_spheres(radii, positions, model) adds spheres, their volume and surface groups and mesh sizes to a Gmsh model. You keep control of Gmsh: your code initializes it, generates and writes the mesh and finalizes it.