Optimisation fails in version 6.6.1 (8d0f1dd) but succeeds in 6.5.0 (97a52f5) (conda installation)

[andrewtarzia]

Describe the bug The attached structure is successfully optimised in one version but hits some error (unclear what) and fails with convergence after 2 steps (same outcome on two machines, Ubuntu 22.04/Mac) in the latest version.

I assume there is something I have missed between the versions that solves this issue?

To Reproduce Steps to reproduce the behaviour:

1. happens with input: input_structure_1.txt

2. start xtb with: xtb input_structure_1.xyz --gfn 2 --opt --verbose --parallel 4

Please provide all input and output file such that we confirm your report.

Older version: output_650.txt

Latest version: output_661.txt

Expected behaviour Successful optimisation in both versions (it takes 415 steps for the working version).

[andrewtarzia]

Closing this because I uploaded a structure before checking if appropriate. I will reopen if possible.

[andrewtarzia]

Reopened, with files attached now that permission was granted.

[MtoLStoN]

Thank you for reporting this. I tried to reproduce your issue, but it seems I cannot optimize your structure with either xTB 6.5.0 nor xTB 6.6.1. However, looking at your structure this looks more like a problem with your input than like a bug. At least, it seems like you are trying to calculate some kind of fusion reaction between H109 and N301 (and others). Can you double check, that your attached input file is not corrupted?

[andrewtarzia]

Hey @MtoLStoN , thank you for the response.

Looking at the structure, it is definitely in a high energy state, which we probably could/should preoptimise before using xtb (this would be our normal approach). Due to the nature of our work, we often throw some pretty messy structures to xtb and it normally does a great job of figuring it all out.

Ultimately, I am not that surprised it failed, but I am surprised that there are different outcomes in different versions, which I thought was worth mentioning to you all. Note that I have just tested it on another machine (Pop OS, * xtb version 6.6.1 (8d0f1dd) compiled by 'conda@1efc2f54142f' on 2023-08-01) and it seems to work fine.

I am assuming this is not a simple issue to diagnose, and it is also not a barrier to any of our work (assuming we have no issues if the structure was better prepared), so feel free to close or mark this as nonurgent.