[calvinp0]

The defect

ORCA's third argument in a %geom Scan constraint is a number of points, not a step size. ARC treated it as a step size, so every ORCA rotor scan it generated swept a few degrees instead of a full rotation.

For a torsion at 92.8° with the default rotor_scan_resolution = 8.0, ARC asked for:

%geom Scan
D 0 1 2 3 =  92.8, 84.8, 8.0
end

which ORCA reads as 8 points spanning 8 degrees, backwards.

Two fixtures in the repository settle the semantics independently. arc/testing/rotor_scans/orca/cc.txt:209 echoes D 2 1 0 3 = -180.0, 180.0, 45.0 and the run contains 45 RELAXED SURFACE SCAN STEP banners with a 45-row surface table stepping by 360/44 = 8.18°. A step size of 45 would have given 9 rows. dft.txt:211 echoes a third argument of 40.0 and produces 40 steps, not 9.

A second defect sat on the same lines: a single-torsion scan closed the block with one end while the multi-torsion path wrote two, leaving %geom unterminated. Both fixtures show two. Every ARC-generated ORCA 1D rotor input was therefore syntactically invalid as well as mis-ranged.

The change

%geom Scan
D 0 1 2 3 = 92.8, 452.8, 46
end
end

46 points spanning 360° forwards, step exactly 8.0.

The endpoint deliberately duplicates the start. arc/job/trsh.py:1729 invalidates a scan on abs(energies[-1] - energies[0]) > inconsistency_az, a hysteresis test that is only meaningful when the last point is the same dihedral as the first; trsh.py:1621-1631 builds int(360 / scan_res) + 1 conformers on the same convention, and determine_rotor_symmetry carries a wrap-around branch that exists only for a duplicated endpoint. The cc.txt fixture shows how tight a real closure is: −180° and +180° differ by 0.0022 kJ/mol.

The point count is int(360 / scan_res) + 1, so the step is exactly scan_res and the count is always an integer. No guard is added for a resolution that does not divide 360, because arc/main.py:1334 and arc/job/adapters/common.py:709 already reject one, the latter inside _initialize_adapter, which covers directly-constructed adapters.

The sweep is now positive, matching the Gaussian, xtb, TeraChem and Q-Chem adapters. The start value is not folded into ±180: calculate_dihedral_angle returns 0–360 and the three sibling adapters pass dihedral + 360.0 unfolded in production today.

Tests

arc/job/adapters/orca_test.py never asserted the generated %geom Scan block, which is why this shipped. 12 tests now pin it — the rendered input verbatim, the point count, start, end and implied step, the resolution table (8→46, 4→91, 2→181, 1→361), integer rendering, the multi-torsion case, block closure, the TS case, and a start above 180°. All 12 fail against the previous line.

A cross-ESS test renders a real Gaussian scan input and asserts ORCA's point count equals Gaussian's step count plus one, so the two programs stay comparable.

arc/job/adapters/orca_test.py + arc/parser/: 188 passed.

Note for reviewers

The companion parser-side change, which converts ORCA scan energies from absolute Hartree to the relative kJ/mol every consumer expects, is in #917 (arc/parser/adapters/orca.py). Both are needed for an ORCA rotor scan to produce a usable barrier.

🤖 Generated with Claude Code

https://claude.ai/code/session_01MndwJm4BEs4DN3vJQsEwD3

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 66.17%. Comparing base (6703600) to head (ec0a877).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1052      +/-   ##
==========================================
- Coverage   66.18%   66.17%   -0.02%     
==========================================
  Files         122      122              
  Lines       41824    41825       +1     
  Branches    10751    10751              
==========================================
- Hits        27680    27676       -4     
- Misses      11084    11093       +9     
+ Partials     3060     3056       -4     

Flag	Coverage Δ
functionaltests	66.17% <ø> (-0.02%)	⬇️
unittests	66.17% <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.