Conversation
…strain field is correct
…cing, internal still relative)
The two-parameter PoissonSurface basis was only ever used in test code, so no migration path is added — this just documents the change in the Unreleased changelog entry.
UnitCell.F_bulk's semi-infinite geometric lattice sum used the raw, untransformed l index instead of the refHKLTransform-converted index in its attenuation phase, in both the numba/C++ and plain-Python paths. This was only correct for the default case of a component using its own bulk cell as the reference (no reference_uc set); any explicit reference_uc whose out-of-plane axis differs from the bulk's -- including a plain scale difference, not just rotation or reindexing -- was affected. Present in all releases up to and including v1.5.0. Document the correct convention and the fix in ctr_structure_factors.rst, note it as a breaking change alongside the also-unreleased PoissonSurface and Wyckoff convention changes in the changelog, and regenerate release_notes.rst. BREAKING CHANGE: UnitCell.F_bulk now uses the refHKLTransform- converted l index in its lattice-sum denominator. Bulk CTR amplitudes computed with an explicit reference_uc whose out-of-plane axis differs from the bulk's will change.
Wyckoff parameter fitting's delta->absolute convention change (2fa9e43) was covered for newly-created parameters but not for loading a pre-fix, value_kind="delta" parameter dict through parametersFromDict/updateFromParameters. Add a round-trip regression test that simulates the legacy dict shape (no reference_value, stored value is the delta) and asserts it reproduces the same basis as the current absolute-value convention.
test_legacy_xtal_uses_corrected_interface_support's fresh reference
values (added previously) disagree with the pre-existing
CTRs_reference.dat baseline by ~2.6%, and it's unclear yet which is
correct. Rather than picking one, keep both paths visible:
- Restore the original pre-beee738 test verbatim (minus two
assertions on removed legacy internals unrelated to the physics
question) as test_legacy_xtal_reconstructs_reference_interface,
comparing against the unmodified CTRs_reference.dat. This fails
now, intentionally, so the discrepancy stays visible instead of
being silently dropped.
- Add test_simple_unitcell_bulk_structure_factor_matches_reference,
reusing the already-validated testdata/0V12_calculated.{xpr,dat}
fixtures (real fit-derived reference intensities, already used by
TestCTRcalculationNumPy/Numba/Cpp) for the un-layered unit-cell +
bulk case. This passes, which localizes the CTRs_reference.dat
discrepancy to the layered Film/EpitaxyInterface/SkellamProfile
path rather than the underlying unit-cell or semi-infinite bulk
calculation.
Determining which of CTRs_reference.dat or the current code is
correct is deferred until other PR feedback is addressed.
_readDispersion_cached's key includes a continuous energy value, so functools.cache could grow without limit over an energy scan in a long-running batch job, unlike the new byte-budgeted C++ FormFactorCache in this PR. Switch it to functools.lru_cache with a generous maxsize. _readWaasmaier_cached is left on plain @cache since its key space (element/ion names) is inherently small and bounded.
find_grid was a linear scan over all cached Q-grids; touch (called on every cache hit) was a linear scan over all cached species across every grid -- the bigger hotspot of the two. Add a hash-bucketed index for find_grid (bucketed by the low 64 bits of the Q-grid's hash, still verified by full 128-bit hash + memcmp within a bucket, so a bucket collision can only cost extra comparisons, never a wrong result) and store each CachedSpecies's own list iterator so touch can splice it directly. Verified with a 500-call stress test at a tiny cache budget (473 evictions) with no crashes and correct hit/miss behavior, plus the existing cpp-backend test suite.
Install the extendedfilesupport and symmetry extras via cibuildwheel's CIBW_TEST_EXTRAS (pip install <wheel>[extras]).
Correct distributed bulk replacement and add strain_coupling and offset controls for transitions between independent and fully strain-coupled lattices. Propagate the resulting geometry through films, stacked components, structure factors, density profiles, serialization, and fitting. Preserve v1.5.0 compatibility by loading legacy two-parameter text and HDF5 records with strain_coupling=1 and offset=0. Document the model with equations, interpretation guidance, RuO2/TiO2 density plots, and an executable CTR example notebook.
Preserve areal optical content under normal strain and prevent transitive merging of layers beyond the z tolerance.
Solve angle arrays with NumPy-vectorized wavefield propagation and preserve finite p-polarized behavior at exact critical angles.
Use ionic forward scattering factors for charged species, quote Meson setup arguments correctly, and preserve multidimensional angle shapes.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds optical wave field calculation for CTR modeling, built around a new interface-based optical (delta/beta) profile — a building block toward distorted-wave Born approximation (DWBA) support, not yet a full DWBA implementation. Also includes a chain of correctness fixes, Wyckoff-parameter handling improvements, and performance work needed to make it practical.
Physics / scientific behavior
CTRoptics.py), replacing/reducing profile numerical complexity vs. the prior approach. This lays groundwork for full DWBA but does not itself implement the distorted-wave scattering amplitude.CTRresolution.py).split_in_layers().Performance
Other
test_CTRcalc.py(+1204 lines), newtest_CTRoptical_profile.py,test_CTRresolution.py,test_scattering_factor_cache.py.ctr_resolution.rst, updates toctr_structure_factors.rst,wyckoff_fitting.rst, changelog/release notes.Notes for reviewers
2fa9e43changes Wyckoff parameter fitting convention (absolute vs. relative) — flagged as user-facing behavior change per repo conventionDWBAbranch name is a working name; this PR delivers the wave field calculation piece, not the full distorted-wave scattering amplitude.cf2a59b), so this should merge/rebase cleanly with 0 commits behind.