Drift in forces

[KWH]

Hello:
I am trying to converge both force and energy using SCAN+rVV10 and B88-vdW functionals. Even though I am using PREC=ACCURATE and EDIFFG=-0.02, The force drift in one direction with the SCAN functional is always large (about 0.3 eV/A) in one direction (Not always z). The B88 system does converge to low drift forces using the same convergence controls. I have attached all the relavant files for both the SCAN and B88 cases.

How big a problem is this? What can I do to fix it?

Thank you for your help

[alex]

Hi KWH,

your structure is partially fixed ... Release all atoms, forces will become smaller.

Best regards,

alex

[fabien_tran1]

Hi,

Without excluding what Alex mentioned, it seems that the drift is very sensitive to ENCUT when a MGGA functional (e.g., SCAN) is used. Using a larger ENCUT reduces the drift, however it is not clear why it is so and we were not aware of this problem. We will try to understand this and let you know.

[fabien_tran1]

By investigating further, I could see that these large values for the drift seem to be specific to the SCAN functional, which is known to lead to numerical problems. Such large values for the drift do not occur with the r2SCAN functional (METAGGA=R2SCAN). Therefore, I suggest to use R2SCAN instead of SCAN (and no need to use larger ENCUT), and related to this change I also suggest to use BPARAM=11.95, see Nonlocal_vdW-DF_functionals.

[KWH]

Thank you for your responses.

As you can see from the attached file, I still get rather large single axis forces with r2scan even when using ENCUT = 1.25x larger than the largest ENMAX. I am also using the recommended BPARAM.

Is this related to the fixed bottom carbon layer?

Best wishes and thanks

[fabien_tran1]

Ok I see. I was using another system for my investigations on the drift, for which r2SCAN was giving very small values. For the moment, the only suggestion that I have is to increase ENCUT (to 700 or 800) and to continue with r2SCAN. I will let you know in case of new information.

[KWH]

Thank you for your help.
Can you tell me how I should interpret the force drift values? What do they mean?

[fabien_tran1]

The sum over the atoms in the cell of the forces should in principle be zero. Because of numerical noise/inaccuracy this may not be the case, resulting in a non-zero sum of forces (the drift). Discussions are also provided here:
https://www.vasp.at/forum/viewtopic.php?t=16851
https://www.vasp.at/forum/viewtopic.php?t=18891
https://www.vasp.at/wiki/index.php/Ener ... d_FFT_mesh

[KWH]

Thank you! Those links were very informative