Skip to main content

Pure-Python crystallography, CIF I/O, and differentiable structure factors for X-ray diffraction.

Project description

midas-hkls

Crystallography toolkit for MIDAS: HKL list generation (sginfo-equivalent), CIF I/O, and differentiable structure factors in PyTorch for intensity-aware peak fitting in pf-HEDM, ff-HEDM, and powder diffraction.

What it provides

Always-on (numpy only)

  • SpaceGroup — load by number, Hermann-Mauguin symbol, or Hall symbol; expose symmetry operations, systematic absences, equivalent reflections, multiplicities, Laue class, centering. All 230 SGs.
  • Lattice — direct/reciprocal metric tensors, d-spacings, Bragg 2θ, with per-crystal-system convenience constructors.
  • generate_hkls() — enumerate Laue-unique allowed reflections within a d-spacing or 2θ cutoff, sorted by d-descending, with multiplicities.
  • Atom, Crystal — asymmetric-unit description with symmetry expansion to the full unit cell (special-position dedupe).
  • Cromer-Mann (IT92) form factors f(s) for 98 neutral elements.
  • CLI: midas-hkls gen|info|list (drop-in for GetHKLList).

Optional: [cif] (gemmi) or [cif-pure] (pycifrw)

  • read_cif(path) -> Crystal and write_cif(crystal, path) — full CIF1.1 with anisotropic ADPs (gemmi) or isotropic-only (pycifrw fallback).
  • Origin-choice and rhombohedral/hexagonal settings handled correctly via the resolved Hall symbol.

Optional: [torch] — differentiable structure factors

  • structure_factors(crystal_t, hkl, *, anomalous=False) returns complex F_hkl tensor, differentiable through:
    • atomic fractional coordinates,
    • occupancies,
    • isotropic B-factors and anisotropic U-tensors,
    • the six lattice parameters,
    • wavelength (when anomalous=True).
  • intensity_from_crystal(...) and powder_intensity(F, m, 2θ) for Lorentz-polarization-weighted powder I_hkl.
  • anomalous_correction(elements, wavelength_A) for f', f'' (Cromer-Liberman tables, 92 elements × 401 log-spaced energies, 100 eV–200 keV).
  • Symmetry expansion is exact integer arithmetic; the autograd graph is rebuilt each forward call so gradients flow through ASU handles to UC atoms.

Quick start

1. Generate HKL list

from midas_hkls import SpaceGroup, Lattice, generate_hkls

sg  = SpaceGroup.from_number(225)              # CeO₂ / Cu / Au / NaCl  (Fm-3m)
lat = Lattice.for_system("cubic", a=5.411)
refs = generate_hkls(sg, lat, wavelength_A=0.173, two_theta_max_deg=15.0)
for r in refs:
    print(r.ring_nr, (r.h, r.k, r.l), r.d_spacing, r.two_theta_deg, r.multiplicity)

2. Read a structure & compute differentiable F_hkl

import torch
from midas_hkls import read_cif, generate_hkls, structure_factors, intensity_from_crystal

xt = read_cif("ceo2.cif")
xt_t = xt.to_torch(requires_grad={"B_iso": True})         # mark B-factors trainable
refs = generate_hkls(xt.space_group, xt.lattice,
                     wavelength_A=0.173, two_theta_max_deg=20.0)

F, I = intensity_from_crystal(xt_t, refs, wavelength_A=0.173, polarization=0.5)

# Fit B-factors against an experimental I_obs (log-space residual)
opt = torch.optim.Adam([xt_t.B_iso_asu], lr=0.05)
for _ in range(300):
    opt.zero_grad()
    _, I = intensity_from_crystal(xt_t, refs, wavelength_A=0.173)
    loss = ((torch.log(I + 1e-3) - torch.log(I_obs + 1e-3)) ** 2).mean()
    loss.backward()
    opt.step()

3. Anomalous scattering (resonant f', f'')

from midas_hkls import structure_factors, anomalous_correction

# Add f', f'' from Cromer-Liberman tables at the experimental wavelength
F_anomalous = structure_factors(xt_t, hkls,
                                wavelength_A=1.5418, anomalous=True)

# Or get f', f'' directly per element
fp, fpp = anomalous_correction(["Fe", "O"], wavelength_A=1.5418)

CLI

midas-hkls gen --sg 225 --lat 5.411 5.411 5.411 90 90 90 --wavelength 0.173 \
               --two-theta-max 15.0 -o ceo2.csv
midas-hkls info --sg "Fm-3m" --ops
midas-hkls list

Install

pip install midas-hkls                       # base: numpy only
pip install "midas-hkls[cif]"                # + gemmi (CIF I/O)
pip install "midas-hkls[torch]"              # + torch (structure factors)
pip install "midas-hkls[all]"                # all of the above

Parity & validation

  • HKL generation: byte-for-byte parity vs. MIDAS's GetHKLList (sginfo) on CeO₂, LaB₆, Si, α-Fe, α-Ti, calcite, Pnma, P21/c.
  • Structure factors: |F| matches gemmi.StructureFactorCalculatorX to <0.01% on CeO₂, Si, α-Fe, LaB₆, calcite (after applying gemmi's change_occupancies_to_crystallographic to align conventions).
  • Anomalous f', f'' matches gemmi.cromer_liberman exactly on grid energies and within 0.05 between grid points.
  • torch.autograd.gradcheck verified on |F|² w.r.t. lattice parameters and atomic positions in float64.

Conventions

  • Lengths in Å; angles in degrees.
  • B-factor B = 8π² U (Ų); CIF U_ij stored in fractional basis.
  • Wavelengths in Å; energies in eV (E_eV = 12398.4 / λ_Å).
  • Symmetry operations stored as integer Seitz matrices over translation base STBF=12 — exact-arithmetic absence detection, no float fuzz.
  • Equivalent HKLs include Friedel pairs (centric structure factor under X-ray Laue symmetry).

Roadmap (post v0.2.0)

  • Wyckoff special-position constraints during refinement.
  • Aspherical / multipole atomic form factors.
  • Magnetic structure factors.
  • Expanded ion form factors (currently only neutral atoms).

Origin

The 530-entry Hall-symbol table is extracted verbatim from sginfo (© 1994-96 Ralf W. Grosse-Kunstleve, public domain). IT92 form factors and Cromer-Liberman anomalous tables are exported from gemmi at packaging time and ship as JSON.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

midas_hkls-0.3.0.tar.gz (367.5 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

midas_hkls-0.3.0-py3-none-any.whl (361.3 kB view details)

Uploaded Python 3

File details

Details for the file midas_hkls-0.3.0.tar.gz.

File metadata

  • Download URL: midas_hkls-0.3.0.tar.gz
  • Upload date:
  • Size: 367.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for midas_hkls-0.3.0.tar.gz
Algorithm Hash digest
SHA256 9e1f055ffa8ccb8ab9afed89ad1cdb2ede4600229a0897f1fa40d250486f7c24
MD5 224e812397c53181046ded7157c38c54
BLAKE2b-256 11ef93fa0bec4399ed6f8a95e9f9ac6780707399d5f4fdedc18ecc446f7a586b

See more details on using hashes here.

Provenance

The following attestation bundles were made for midas_hkls-0.3.0.tar.gz:

Publisher: python-packages.yml on marinerhemant/MIDAS

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file midas_hkls-0.3.0-py3-none-any.whl.

File metadata

  • Download URL: midas_hkls-0.3.0-py3-none-any.whl
  • Upload date:
  • Size: 361.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for midas_hkls-0.3.0-py3-none-any.whl
Algorithm Hash digest
SHA256 0bd181d65058dc2bb8ac82dfbe86b8ee65b7a856bae0c21908c3e0fc06cdab9b
MD5 869ec88170531837a1b84e16346f6e28
BLAKE2b-256 2eeec69acf11cf1e77b2bf776bafa66aacd99cbb9e2c127c2b661568695f38b4

See more details on using hashes here.

Provenance

The following attestation bundles were made for midas_hkls-0.3.0-py3-none-any.whl:

Publisher: python-packages.yml on marinerhemant/MIDAS

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page