[CP2K-user] [CP2K:20181] Different Total energy and ENERGY| Total FORCE_EVAL ( QS ) energy [a.u.]

[Jürg Hutter]

Hi

yes, your energy is not well converged and your forces will have a
considerable non-SCF error.
The EPS_SCF default works for most systems using OT. If you use diagonalization,
you need to check if it is tight enough for your needs.
In your case I would suggest to improve SCF convergence.

regards
JH

________________________________________
From: cp2k at googlegroups.com <cp2k at googlegroups.com> on behalf of Benjamin Shi <benjaminshi97 at gmail.com>
Sent: Monday, May 6, 2024 9:43 AM
To: cp2k
Subject: Re: [CP2K:20180] Different Total energy and ENERGY| Total FORCE_EVAL ( QS ) energy [a.u.]

Hi Jurg,

Thank you very much for the information. Just to confirm, the "Total FORCE_EVAL (QS) energy" quantity is the correct energy to use?

For this calculation, I have used an EPS_SCF of 5E-7, which is tighter than default. Are you suggesting that the discrepancy between "Total energy:" and "Total FORCE_EVAL (QS) energy" which I'm observing here means that my forces could be noisy or energy isn't well-converged? Previously, I have confirmed that the force is well-converged by modifying/increasing EPS_DEFAULT and it changes atomic forces by less than 0.01 eV/A.

Best wishes,
Benjamin

On Monday, May 6, 2024 at 5:34:01 PM UTC+10 Jürg Hutter wrote:
Hi

this is a problem of consistency of forces and energies. During the SCF
some energy terms (especially with diagonalization) are calculated from the
input density and some from the output density. Upon convergence
the energy is recalculated in a consistent manner and forces are wrt this
energy.
However, this shouldn't matter as long as the density is sufficiently converged.
In your case I would suggest to improve SCF convergence, as that will also
considerably reduce the noise (or error) on your forces (good for ML).

regards
JH

________________________________________
From: cp... at googlegroups.com <cp... at googlegroups.com> on behalf of Benjamin Shi <benjam... at gmail.com>
Sent: Sunday, May 5, 2024 11:56 AM
To: cp2k
Subject: [CP2K:20171] Different Total energy and ENERGY| Total FORCE_EVAL ( QS ) energy [a.u.]

Dear CP2K developers,

I am writing to enquire as to what the difference is between the energy given by "Total energy:" and "ENERGY| Total FORCE_EVAL ( QS ) energy [a.u.]:". As seen in attached output file, I'm finding a slight difference in the two energy values. Does this arise because I incorporate Fermi smearing? In the context of training machine learning potentials, which quantity would be the correct quantity to use that is consistent with the forces that are printed out?

Thank you for your help,
Benjamin

--
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... at googlegroups.com<mailto:cp2k+uns... at googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/34a1a6d5-ed13-4c3b-b6ed-574aea8d4228n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/34a1a6d5-ed13-4c3b-b6ed-574aea8d4228n%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
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+unsubscribe at googlegroups.com<mailto:cp2k+unsubscribe at googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/bf021d53-15b0-48fa-8ed6-acd9a95b66d8n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/bf021d53-15b0-48fa-8ed6-acd9a95b66d8n%40googlegroups.com?utm_medium=email&utm_source=footer>.

-- 
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+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB0759547AD146935DEE4787119F1C2%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.