Skip to content

ACTINV 1.7.2

Latest

Choose a tag to compare

@Connoravila1 Connoravila1 released this 02 Nov 21:41
· 70 commits to master since this release

ACTINV v1.7.2 — release notes

ACTINV 1.7.2 makes actinv data fetch work again for first-time installs and ships the OpenMC R2S adapter inside
the Python package. It also:

  • adds an uncertainty band for OpenMC shutdown-dose tallies;
  • adds mixture self-shielding;
  • makes rate pruning faster;
  • fixes the potential-scattering estimate of composition-dilution self-shielding.

The bundled TENDL-2025 data, the data catalog (1.2.0) and default results are unchanged. Everything else in the
1.7.1 release notes and 1.7.0 release notes applies, including
their known limitations.

Upgrade the Python package with:

python -m pip install --upgrade actinv

Rust CLI users can install this exact version with:

cargo install --locked --force actinv-cli --version 1.7.2

Data download

The two decay-data archives (ENDF/B-VIII.0 and JEFF-3.3) come from the IAEA Nuclear Data Section. Its site now
answers automated downloads with HTTP 403, so actinv data fetch 1.7.1 and earlier stop on them for anyone
installing the data for the first time.

When the IAEA URL fails, 1.7.2 takes ACTINV's byte-identical copy from the
decay-sources-v1 release. It holds the copy to
the catalog's size and SHA-256, so it installs exactly the files the catalog names. With an older release, install
the two files by hand as described in the
install guide. Existing data folders need nothing.

OpenMC

from actinv.openmc import ActinvR2SManager now comes with pip install actinv. It replaces copying the contrib
script, which is kept as a shim. The adapter needs OpenMC 0.15.3 or newer. It is tested on 0.15.3 and 0.16.0, and
weekly on OpenMC's development branch. Importing actinv alone never imports OpenMC.

  • Defaults: the library, decay data and solver come from the installed data and package unless you pass them.
  • Step 1 tallies only the flux on the library's groups and needs no depletion chain.
  • Reloading: load_results(path) reopens a finished run without solving again.
  • MPI: under MPI, actinv mesh runs once, on rank 0, and a failure there stops every rank.
  • Uncertainty band: uncertainty={} gives every photon tally a first-order band in
    results["dose_uncertainty"]. The band covers activation cross sections (MF=33), decay constants and flux-tally
    noise, plus the photon tally's own statistics. It is a lower bound on the data uncertainty: nuclides outside the
    bands and channels without covariance data add nothing. The band has not yet been checked against sampling.

See Use ACTINV with OpenMC.

Also in this release

  • Mixture self-shielding (self_shielding.mixture, opt-in): computes resonance self-shielding factors at run
    time from the material's own pointwise cross sections, so overlapping resonances of different nuclides shield each
    other. It has not yet been qualified against measurements.
  • Faster rate pruning on large reaction networks. Results are unchanged. ledger.rate_pruning may list more
    dropped nuclides, with looser but still valid bounds on nuclides that are dropped anyway.
  • Python: the extension module is now actinv._actinv, re-exported by actinv. import actinv and the
    actinv command are unchanged.
  • Reproducible tables: build-shielding no longer records the input directory, so identical inputs give a
    byte-identical table.
  • Source package: the Python source distribution lists only the workspace crates it ships, so building it from
    source no longer fails in tools that read the top-level Cargo.toml.

Results to re-check

  • Self-shielding with dilution: "composition".
    • The defect: the analytic potential-scattering estimate, used for nuclides the table does not cover, was 100
      times too small. The ENDF channel radius was read as femtometres.
    • Who is affected: materials whose main scatterers are light nuclides missing from the table, such as concrete
      and water. They were over-shielded.
    • Affected releases: 1.1.0 to 1.7.1.
    • What to do: rerun calculations that used this option.
    • Not affected: default runs, which do not self-shield.