Help: cp2k vibration analysis stopped in the middle, how to renew the calculation?

[Zac Smith]

Hello  cp2k users,  

I used a virtual machine to do vibration analysis with cp2k, but the task was killed when the hard disk was full, and the output file only contains out, molden, cube, and the eig file of size 0. Can the calculation be continued in this case?

[Krack Matthias]

It is not possible to restart a vibrational analysis run. Such a functionality is not implemented.

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/9d914e4f-bcb1-42ae-9b02-6c14f6fec8a5n%40googlegroups.com.

[Zac Smith]

Understood, thank you for your reply.

[Eric Patterson]

On a closely-related note…

Since normal vibrational analysis is not restartable, but mode-selective vibrational analysis is… Is there any drawback to using mode-selective and choosing all atoms, and then restarting as necessary until all modes have converged? Given the queue time limit imposed by my HPC resources, this seems to me the only way to get the job done. Are there any pitfalls one should bear in mind?

Thanks,

Eric

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZRAP278MB082784950ADD367A217A8C43F4D99%40ZRAP278MB0827.CHEP278.PROD.OUTLOOK.COM.

[Leili Rassouli]

Hi Eric,

I've faced a similar issue and was considering the approach you suggested. Have you had a chance to try it, and does it work well in your case? Your insights would be greatly appreciated.

Thank you for your assistance.
Lili Rassouli

[Mikhail Povarnitsyn]

Dear Leili,

There is one solution to suggest. We have modified the "vibrational_analysis.F" file a little bit to specify a range of atoms to be shifted. By dividing the full range into several parts, it is feasible to calculate all forces by parts and then gather them to reconstruct the full Hessian and diagonalise it afterwards. If you like we could send you two source files to recompile the cp2k.popt. 

Which version of cp2k do you use?   

Best regards,

Mikhail  

[Eric Patterson]

Hello Lili,

I did do a little bit of testing of the mode-selective method vs the normal method. The results are comparable if I choose all atoms and restart until all modes are converged (or at least nearly so — I didn’t care vey much about low energy modes).

However, I’ve also managed to compile a working MPI version on my HPC, so I am able to run on enough nodes to get through a default vibrational analysis before the queue kills the job. 

Best,

Eric

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/0d3984ac-a0a6-4040-b500-6a5d67f5c556n%40googlegroups.com.

[Leili Rassouli]

Dear Mikhail and Eric,

Thank you for your responses. My job is completed after 10 days, but to enhance the computational efficiency, I am considering the use of a mode-selective method or the dividing the full range into several parts to potentially expedite the calculation process. Mikhail, I would be appreciative if you could provide me access to the source files, given that I am currently utilizing CP2K v2022.1.

I'm currently facing an issue with negative frequencies. While using the ground state wavefunction in the ground state geometry, frequencies are positive and within the expected range. However, switching to the first excited wavefunction yields a significant number of negative frequencies.

Enclosed is my input file for 221 supercells of Fe2O3 with 120 atoms. Your insights and recommendations on this matter would be greatly valued.

I've explored various parameters, including:

Adjusting the CUTOFF value of TDDFPT from 300 to 600 and 700
Refining the optimized geometry
Increasing the TDDFT convergence from 1.0e-5 to 1.0e-6 and 1.0e-7
Despite these efforts, a notable proportion of frequencies persist as negative values. 

I express my gratitude in advance for your time and assistance.

Best regards,
Lili

[Eric Patterson]

Dear Lili,

I suggest you make your TDDFT inquiry a new thread since it is a separate issue. I have no experience with TDDFT in CP2K, so I’m afraid I’m of no use in this case..

Best,

Eric

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/3e2997be-cac2-48d0-aeca-78080fa420f9n%40googlegroups.com.
<Fe2O3_vib.inp>

[Leili Rassouli]

Thank you, Eric. I made a new inquiry.

[Mikhail Povarnitsyn]

Dear Leili,

Sorry for the late reply. Two files for you version 2022.1 are attached. You have to make substitution into src/motion and recompile the code: 

make ARCH=local VERSION="psmp".

To specify the range of atoms to shift from equilibrium positions, just add the keywords START_ATOM_ID and END_ATOM_ID:

&VIBRATIONAL_ANALYSIS
DX 0.005
START_ATOM_ID  1
END_ATOM_ID      100
.....................

.....................

For instance, you can split the task of 600 atoms into 6 parts 1-100,101-200, etc and then gather the force derivatives into the dynamical matrix.

If you need further assistance, please let me know.

Best regards,

Mikhail