Mismatch between i-pi/cp2k and cp2k results

[Sr V]

Hello all,

I am trying to optimise a TS. I was able to do it with cp2k and i-pi/cp2k at the MNDO level(semi-empirical). But the results between them are not matching. Especially the imaginary frequency mode has huge difference in value. i-pi/cp2k gives 1700cm-1 whereas cp2k gives 1200cm-1.

Can someone shed some light on it? I tried the same calculations for a different system (H+CH4) using same inputs at MNDO, there i-pi/cp2k and cp2k values match.

I have attached both i-pi/cp2k and cp2k inputs and relevant outputs.

[Mariana Rossi]

Hi there, I think I answered only to you personally last time. Sorry for that. Could you please share your geometry inputs as well?

[Sriram Vignesh AK]

Yeah sure, I am attaching coordinates of reactant and TS (MNDO optimised).

[Mariana Rossi]

Hi there, a last question: could you please check that geometries that are being passed into CP2K make sense? Can you look at the CP2K output file for these geometies, when CP2K is talking to i-PI? My question is because it seems you are starting i-PI with a pdb file, which is not the usual modus operando that most of us use, so there could have been a bug that creeped in. Can you build xyz files from the same geometries you input in cp2k?

[Sr V]

Yeah, I have checked the input geometries given to cp2k (they are correct). And I tried another optimisation using xyz file also as input(I am attaching input and outputs), but it didn't help either.

PS:I have attached a low print version of cp2k output because of file size. But i have checked the passing of coordinates from i-pi to cp2k is proper(even in pdb).

[yairl...@gmail.com]

Hi,

Thanks for reaching out.
The ratio of the two frequencies you report is suspiciously sqrt(2). What happens with the other frequencies? Have you checked the Hessians? Could you share the two sets of {optimized geometry, hessian files, frequencies}  you get, so we can have a closer look?

Also I notices that the TS.pdb and rctnt_E.pdb files have the atomic species in different order. i-PI and the client code must be initialized using geoemtry files that have the  same order of atomic species. Could you double check this?  This is also applies for the postprocessing script provided with i-PI, the geometry and hessian file should be consistent.
Let us know if this solves the issue.
Best,
Yair

[Sr V]

The other frequencies don't have this sqrt(2) factor as I could see. I am sharing the optimised geom and freq in both cp2k and i-pi/cp2k. But the hessian in cp2k isn't in readable format.  Hence I haven't attached that.

As far as the order is concerned, the reactant (rctnt_E.pdb) calculation is independent of TS (TS.pdb) right? But I am passing the same input only to both i-pi and cp2k in start. The post processing script (Instanton_postproc.py) takes the i-pi RESTART file to do postprocessing right.

Thanks

[yairl...@gmail.com]

Hi,
Thanks for sharing those files. Could you also share the RESTART file?
Best,
Yair

PS: one can read the cp2k hessian file in this way --> https://groups.google.com/g/cp2k/c/jevGAzeMo1c

[Sr V]

Sure, I am sharing the RESTART file and the cp2k hessian as well.

Thanks