Unexpected Optimized Structure from GFN2-xTB

[murilucass]

Dear all,

I am attempting to optimize a structure using xTB with the settings --chrg -1 --gfn2 --alpb DMF --uhf 0, but the resulting geometry appears unexpected. Specifically, the optimization yields a five-membered ring in which one of the carbons is part of a phenyl group.

I would like to ask whether this behavior could be related to a bug, or if there are alternative approaches or settings you would recommend to avoid obtaining such a distorted structure. The input file, optimized geometry, and log file are attached.

Thank you very much for your time. I look forward to your reply.

Kind regards,
Murilo

xtbopt.xyz.txt
eqeq-input.xyz.txt
outXTB.txt

[chselz]

Hi Murilo thanks for your interest in xTB and sorry for the late follow up.
I tested this with the latest bleeding edge f811460 as some rework regarding the optimizer has been done since the latest release and i cannot confirm your findings. There is now strange rearrangement. See attached folder with my results. Also in the structures that you provided there is no structure with a five membered ring as you described so i am not quite sure what the problem is present in your calculations.

Also i think the correct flag is -P or --parallel and not --T as this i think the keyword in CREST for parallel execution

optimization.zip

[murilucass]

Hi Christian! Thank you for your support. I highlighted the five-membered ring in your output because this structure doesn't match with what we obtain in nmr experiments, which I have attached below too. What adjustments do you think we could make to obtain this structure? Thank you again for your help! Em seg., 30 de mar. de 2026 às 09:43, Christian Selzer < ***@***.***> escreveu:

…

*chselz* left a comment (grimme-lab/xtb#1395) <#1395 (comment)> Hi Murilo thanks for your interest in xTB and sorry for the late follow up. I tested this with the latest bleeding edge f811460 as some rework regarding the optimizer has been done since the latest release and i cannot confirm your findings. There is now strange rearrangement. See attached folder with my results. Also in the structures that you provided there is no structure with a five membered ring as you described so i am not quite sure what the problem is present in your calculations. Also i think the correct flag is -P or --parallel and not --T as this i think the keyword in CREST for parallel execution optimization.zip <https://github.com/user-attachments/files/26350213/optimization.zip> — Reply to this email directly, view it on GitHub <#1395?email_source=notifications&email_token=CAITF2A7VV3UYSCSOJYYBZL4TJTYXA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMJVGQ3TMNBWGY4KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-4154764668>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/CAITF2CH2XINK6A5PKAGID34TJTYXAVCNFSM6AAAAACWYLF76KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCNJUG43DINRWHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[chselz]

Hi I'm afraid i didn't receive your attachment thus i cannot see what you marked in the structure. I am assuming that you are seeing a bond between atoms 17 and 19. At least when i open the structure with molden no bond is shown to me so this may be due to your structure visualizer. But please not that these "bonds" are not quantum mechanically founded but displayed as a eye guiding help and if i am right based on van der waals radii. So just because a visualization tool like molden displays a bond does not mean there is one.

Furthermore if you have a look into the Wiberg bond order there is only a bond order of 0.471 maybe due to the negative oxygen interacting with the positive carbon atom of the ring.

Hope that helps ☺️