q734738781/neb-analysis

neb-analysis

Overview

Use this skill after a NEB run has been collected and the task is to extract or interpret the barrier. It focuses on common post-analysis checks, missing endpoint-energy pitfalls, and how to read suspicious profiles. For setup, use neb-prepare. For the run protocol itself, use neb-calculation.

Quick Start

1. Run analyze_vasp_neb_results on the collected NEB result root.
2. If image energies are incomplete, stop and diagnose the missing images instead of inferring a barrier.
3. Keep the barrier convention explicit: forward, reverse, or both.
4. If the profile is wave-like rather than single-saddle, look for hidden intermediates rather than forcing one long pathway interpretation.

Allowed tools

• analyze_vasp_neb_results

Workflow

1. Require a complete energy profile

• The workflow is not complete until analyze_vasp_neb_results exports the barrier summary, CSV profile, and plot.
• If image energies are missing or partial, report that explicitly instead of inventing a barrier.
• Do not collapse a partial profile into a single barrier number.

2. Handle endpoint OUTCAR correctly

• VASP NEB endpoint images do not generate fresh endpoint energies for barrier analysis.
• If analyze_vasp_neb_results reports missing OUTCAR energy for image 00 or the last image, copy the original endpoint relax OUTCAR files into those endpoint image directories and rerun the analysis.
• Treat that copy as a normal post-processing repair step, not as a scientific change to the pathway.

3. Keep the energy reference explicit

• State whether the reported barrier is forward, reverse, or both.
• Unless the task says otherwise, take the forward barrier relative to the initial endpoint and the reverse barrier relative to the final endpoint.
• If endpoint energies come from copied relax OUTCAR files while the interior images come from the NEB run itself, say that plainly.

4. Interpret suspicious profiles as modeling clues

• If the profile shows wave-like repeated rises and falls rather than one localized saddle region, suspect hidden metastable intermediates.
• Use those local minima as hints for splitting the path into shorter primitive hops.
• If the original setup needed too many images or had a periodic displacement above about 6 Å, connect that observation to the possibility that the path was overextended from the start.

5. Keep convergence claims narrow

• A collected run is not automatically a trustworthy barrier.
• If the climbing-image stage was skipped or unclear, say so.
• If the profile is incomplete, oscillatory, or based on a questionable endpoint contract, report those QC limitations directly.

Method-critical defaults

• Never report a final barrier from an incomplete image-energy profile.
• Endpoint validation and endpoint-energy provenance are part of the barrier contract.
• Wave-like multi-bump profiles are often a modeling warning, not merely an ugly plot.
• Keep the barrier convention explicit in the final answer.

Output Contract

Return:

• result root
• summary JSON, CSV, and plot path
• stated barrier convention
• any QC flags or reasons the barrier should be treated as provisional

References

• Use neb-prepare if the pathway itself needs to be rebuilt.
• Use neb-calculation if the next step is a rerun or a refinement stage rather than interpretation.