Page 1 of 1

CSVR thermostat

Posted: Fri Mar 14, 2025 12:49 pm
by paulanatalia_abufager

Hello,
Should the SMASS tag be set for the CSVR thermostat?
thanks,
Paula


Re: CSVR thermostat

Posted: Fri Mar 14, 2025 5:43 pm
by martin.schlipf

SMASS is only used for Nose thermostat so its setting does not affect the result for MD_ALGO = 5. It is a good practice to verify this for yourself. You should see that the results are independent of the SMASS selected.


Re: CSVR thermostat

Posted: Sun Mar 16, 2025 3:06 pm
by paulanatalia_abufager

Hello, Calculations without explicitly setting the SMASS tag and those with SMASS=0 yield different results, even though the SMASS tag is not supposed to affect the outcome. Do you know why this happens? Thanks, Paula,


Re: CSVR thermostat

Posted: Mon Mar 17, 2025 12:31 pm
by martin.schlipf

Thanks for testing it for yourself. Perhaps you forgot to restart from scratch, e.g., because a WAVECAR file was in the directory. I tested setting SMASS to different values and I always get exactly the same trajectory.


Re: CSVR thermostat

Posted: Mon Mar 17, 2025 3:08 pm
by paulanatalia_abufager

Both calculations started from scratch.


Re: CSVR thermostat

Posted: Tue Mar 18, 2025 7:28 am
by martin.schlipf

Can you post the OUTCAR files of both calculations, please?


Re: CSVR thermostat

Posted: Tue Mar 18, 2025 9:44 am
by paulanatalia_abufager

Please find the files at the following link. The files are too large to upload.

https://drive.google.com/drive/folders/ ... sp=sharing


Re: CSVR thermostat

Posted: Tue Mar 18, 2025 10:32 pm
by martin.schlipf

You did not set the same random seed for both calculations. Without that even running the same INCAR file twice should result in different calculations. I think, if you set this tag in your INCAR file, you will see that your calculations agree.

Note that you typically should not set this tag in the INCAR file to avoid repeating the same MD run twice. This tag is really intended for testing purposes. If you want to test things you can read the RANDOM_SEED from the OUTCAR file and put it as input to a second calculation. Then the calculation should reproduce exactly the same trajectory at least for some time. Over many iterations small numerical fluctuations may result in diverging calculations anyhow.

Finally, I noticed that you ran big expensive calculations to test this. I would recommend that you read how to generate small reproducible examples. It is very important that you gradually move from small and simple systems where you understand the behavior to more complex ones. Vice versa, when you encounter something that does not behave like you think it should, you try reducing the system size and the amount of INCAR tags so that you can really understand what is going on. If you learn this skill of identifying errors not only will you be able to overcome more problems on your own, you will also receive better and faster support from us because with a minimal example there are much less possible causes that we need to examine.