Summary
Before the first R02 geometry search started, its authorized passport fixed the scope of the run and linked defined observations to actions. Ethanol later moved to a different pose while remaining inside the search region, so the movement was recorded and the same trajectory continued as required. ORCA ended the search after 95 cycles with four of five printed conditions met. The structure therefore remains a candidate requiring review, not a confirmed minimum.
Why fix the rules before launch
A geometry search can produce an inconvenient outcome: a molecule can move, the optimizer can reach its limit, or software can report convergence without satisfying every printed condition. If the response is chosen only after seeing that outcome, it becomes harder to separate an observation from a preference. For this search, the passport was fixed before launch so the run could be reported against the same scope and rules that authorized it.
What was fixed before launch
The authorization covered one geometry search with a limit of 100 optimization cycles. It allowed no automatic restart and no automatic additional quantum-chemistry job. The passport assigned these actions to the recorded outcomes:
| Recorded observation | Action fixed before launch |
|---|---|
| Native optimization converges; all five printed conditions and the required endpoint checks pass | If Cartesian max|g| ≤ 1 × 10⁻⁴ and RMS(g) ≤ 3 × 10⁻⁵ hartree/bohr, save the structure as a converged candidate |
| Native optimization converges, but the complete acceptance route above is not satisfied | Save the structure as a candidate requiring review |
| The cycle limit, an SCF or optimizer error, non-numerical data, or a crash is reached | Save the reason and the last valid calculated geometry, then stop |
| A checked geometry leaves the defined search region | Stop and report that the search left its defined region |
| The molecule migrates or changes pose while remaining inside the search region | Record the change, continue the same trajectory, and describe the result by the structure actually found |
The authorized run ended with evaluation of the search result. A Hessian or vibrational-frequency calculation was outside its scope and could not start automatically.
What happened
The ethanol molecule moved sideways by more than 1 Å and changed orientation, but remained inside the defined search region. The movement was recorded, and the same trajectory continued without a restart. ORCA ended the search after 95 optimization cycles.
- Four of five, not five of five. ORCA reported native convergence with four of its five printed conditions met. The force criteria were met by margins of about 6.8× and 20×; the maximum-step condition was not met.
- No automatic follow-on calculation. The run stopped after the search result was saved. No Hessian or frequency calculation followed automatically.
How the rules were applied
The change in ethanol pose did not trigger a stop: the passport said to record a change inside the search region and continue the same trajectory. The reported 4-of-5 ending did not satisfy the passport's complete 5-of-5 route to a converged candidate. The structure was therefore retained as a candidate requiring review.
Frequency analysis may later test the curvature around this candidate, but it is a separate diagnostic workflow with its own protocol and authorization. Until that work is completed and reviewed, the candidate is not described as a local or global minimum.
Why this matters
The pre-run rules did not guarantee a convenient outcome. They required the project to keep an unexpected trajectory rather than replace it with a preferred starting point, to report 4 of 5 conditions rather than “all conditions,” and to stop without adding an unapproved diagnostic. The useful result at this stage is a traceable candidate and a clear next question, not a stronger molecular claim.
A reporting pitfall we found
ORCA prints two different quantities under the name “MAX gradient.” One is a Cartesian force; the other is the internal-coordinate value used by the convergence test. In this search, they differed by roughly a factor of four. Early progress reports quoted the wrong one. R02 reports now identify which quantity they use.
What this note does not show
This note describes procedure, not an accepted molecular result. It does not claim that the candidate structure is a minimum, report a binding energy, or establish that the pilot method is accurate enough for the final study. It provides no evidence about formaldehyde, water, a functionalized target material, or cartridge performance. Independent external review has not yet been carried out.
Change history
v1.0 ·
First publication.