Use the frequency log's geometry for the TS normal mode displacement …

[calvinp0]

Base of a 3-PR stack — review this one first. Reading order: #967 → #968 → #970. All three change the head of analyze_ts_normal_mode_displacement(); they are split by cause, not by file.

What breaks

• The TS normal mode displacement (NMD) check accepted or rejected a transition state using a geometry that was not in the same coordinate frame as the normal modes it was compared against.
• The geometry came from the optimization job's final_xyz; the modes came from the frequency job's log. Gaussian reports modes in that job's own orientation, which is not generally the orientation of the opt geometry.
• Measured over 996 real benchmark frequency jobs: 82% agree to within 1°, but 17.6% differ by ~180°. These are exact rigid rotations (fit RMSD ~1e-16), not numerical noise.
• Adding a mode vector expressed in one basis to a coordinate expressed in another corrupts the cross term of every displaced bond length, so the rotation-invariance of a bond length does not rescue the result.

What the fix does

• get_ts_xyz_in_normal_mode_frame() takes the TS geometry from the frequency job's output file — the same file the displacements are parsed from — so geometry and modes share a frame by construction.
• For Gaussian that resolves to the last Standard orientation: block, falling back to Input orientation:. Both are right: the fallback only applies when Gaussian suppressed the standard orientation (e.g. under nosymm, which ARC's troubleshooting can add), and in that case the frequency analysis is done in the input orientation too.
• Falls back to the TS species geometry when the freq log yields no geometry, or when the parsed geometry's element symbol sequence differs from the species'. Both fallbacks log a warning that the frame may not match.
• The fallback is one-directional. When the species has no geometry — ARCSpecies.get_xyz() returns None for a TS with no final_xyz/initial_xyz/conformer and no TS guess carrying an initial_xyz — the parsed geometry is used rather than discarded. Returning None there would hand the caller nothing to measure: it immediately does len(ts_xyz['symbols']).
• is_ts_atom_order_consistent_with_reactants() makes the analysis return None when the TS atom order differs from the concatenated reactant order. Nothing checked this before: the forming/breaking bond indices are indices into the reactant order, so on a mismatch the check silently scored the wrong atoms and still returned a confident verdict.

Why this shape

• A frame mismatch or an atom-order mismatch is a missing precondition, not evidence against the TS — hence None (analysis not performed) rather than False (TS rejected).
• ARC already pairs a geometry with modes from the same log elsewhere: arc/job/trsh.py::trsh_negative_freq parses both from one log_file before perturbing along the mode. The NMD check was the outlier, not the precedent.

How it was verified

• Independently confirmed by the Eckart condition. Modes expressed in their own frame satisfy Σ mₐ(rₐ − R_com) × dₐ = 0, and that quantity is frame-dependent, so it is a direct test of whether the pairing is consistent: 73% of affected jobs exceed 3σ when evaluated in the opt frame, versus 0 of 111 in the freq frame.
• 9 new tests: frames agree, a deliberately rotated geometry, an element symbol mismatch, a log with no geometry, a log that fails to parse, a species with no geometry at all, the atom-order guard (unit and end-to-end), and a scheduler-level test that an unknown (None) NMD verdict does not trigger a TS switch.
• arc/checks/ passes in full — 60 tests.

Open note for the reviewer

• The line just above the new guard, if n_ts != n_expected: return False, is the same class of missing precondition but still returns False. It predates this PR and is left alone here; worth deciding whether it should become None too.
• analyze_ts_normal_mode_displacement() still does not tolerate a None geometry, which after this change can only arise when the freq log and the species both yield nothing. That len(ts_xyz['symbols']) on a possibly-None value is unchanged from main and is left as-is rather than widened here.

What I searched for

• Whether ARC already had a helper that resolves a geometry into the frame of a job's parsed modes — searched by behaviour (geometry / frame / orientation reconciliation) rather than by an intended name. trsh_negative_freq does it inline for a different purpose; there was no reusable helper to import.
• Confirmed the geometry parse goes through the existing arc/parser entry point rather than a new reader, and that the symbol-sequence comparison had no existing equivalent.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 64.41%. Comparing base (d033cbb) to head (b5e1935).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #967      +/-   ##
==========================================
- Coverage   64.43%   64.41%   -0.02%     
==========================================
  Files         119      119              
  Lines       39601    39617      +16     
  Branches    10269    10270       +1     
==========================================
+ Hits        25515    25520       +5     
- Misses      11109    11112       +3     
- Partials     2977     2985       +8     

Flag	Coverage Δ
functionaltests	64.41% <ø> (-0.02%)	⬇️
unittests	64.41% <ø> (-0.02%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.