different results from same settings

Question

different results from same settings

Closed this issue 2 months ago · 4 comments

Hi everyone,

Is it normal to get three different results, i.e., three different numbers of conformers, when I run a CREST calculation (version 3.0.1) three times with the same settings and starting structures? I noticed in the output that the MTD simulations produce the same number of conformers, but the number of conformers differs after optimizations in all three cases.

Thank you in advance.

Answer 1 · 2024-07-24T17:27:09.000Z

Yes, the search is not deterministic due to the MD/MTD parts. Especially the number of high-lying conformers can differ. Only the lowest energy conformer(s) should be the same in the sampling limit.

Answer 2 · 2024-07-24T17:41:27.000Z

But why is the number of conformers the same after the MTD simulations in all three cases, yet varies only after the optimizations?

Answer 3 · 2024-07-24T17:58:42.000Z

Because the MTD/MD runtime and snapshot dump frequency is the same. That means, while the same number of candidates are generated, since the MD/MTD can non-deterministically move to different regions of the PES, the candidates are not the same in different runs.

Answer 4 · 2024-07-24T18:05:09.000Z

Ah ok, thank you very much for your time and effort.