# OPTIM + ORCA: refine a transition state and follow its path (N6 chain -> 3 N2)

BFGSTS refines the TS (hybrid eigenvector-following, no Hessian), then PATH pushes off along the
imaginary mode and minimises both sides. Start (POINTS): a wB97X-D3/cc-pVDZ TS for the N6 chain -> 3 N2 step.
Run: `sbatch job.sh` (~4 h on 2 cores). Output: `OPTIM.out`, `path.info`, `EofS`, `path.xyz`.
Expected: TS -328.21668 Eh; one side -> chain (-328.24921 Eh, barrier 20.4 kcal/mol); other side -> 3 N2
(-328.50949 Eh, 163.3 kcal/mol below the chain); one maximum in EofS. Most of the run is the 3 N2 drifting
apart slowly until the BFGSMIN convergence test is met.

ORCA and MPI: job.sh swaps the Open MPI 5.0.5 of module orca/6.1.1 for 4.1.7 (5.0.5 now and then segfaults inside
ORCA; the "reloaded with a version change" message in slurm-*.out is expected) and turns core binding off.
The ORCA line of odata runs `./orca_retry.sh`, which runs ORCA and reruns a failed call (up to 3 times); failures are
logged in `orca_failures/failures.log`, and job.sh prints a NOTE if that file exists. A hung call is not caught.

`expected_output/`: OPTIM.out, path.info and EofS from a run of this input (Pat, 4 Nov 2026, Open MPI 5.0.5; the
ORCA settings and results are otherwise the same).