[CP2K-user] [CP2K:21287] Inconsistent potential energy between DFT-MD and SCF during NPT_F ensemble simulation

[Denan Li]

Dear CP2K developers and community,

I am running DFT-MD (NPT_F ensemble) simulations in CP2K (version 2024.3) 
and using a machine learning potential (Deep Potential, DP) to predict 
energies along the DFT-MD trajectory to validate my DP model. However, I 
observed inconsistencies between the potential energies calculated by 
DFT-MD, SCF, and DP, while no such discrepancies were found in NVE or NVT 
ensembles.

My key question:
In the NPT_F ensemble, some potential energies and forces differ 
significantly from those calculated by SCF. Is this an expected behavior, 
or could it be influenced by additional factors (e.g., barostat 
contributions)?

Any suggestions would be greatly appreciated. Thank you for your time and 
expertise!

- Denan LI

Below are my test results:
*1) *I have run a 10-ps long DFT-MD trajectory and compared the energies 
with those predicted by DP. The overall energy profiles show significant 
discrepancies (see Fig-1a). DP and DFT-MD energy trends in some regions align 
well, while they diverge completely in others (e.g., around frame 9600) 
(Fig-1b).
[image: fig1.png]

*2) *I selected two representative frames and performed OT calculation to 
obtain the SCF energy:
*a)* The last MD step (matching trends): DFT-MD and OT  energies agree 
closely.
*b) *The 9600th frame (divergent trends): DFT-MD and OT energies show a 
large dismatch.
*c)* Note: Both DFT-MD and OT calculations confirmed SCF convergence;
[image: fig2.png]

*3)* I further conducted ML-MD simulations in CP2K using the DeepMD-kit 
interface and compared the energies/forces between the ML-MD trajectories 
and DP predictions.
*a)* In NVE/NVT ensembles, forces/energies from ML-MD trajectory match will 
with DP re-evaulations:
[image: fig3.png]
*b)* In NPT_F ensemble, ML-MD energies/forces deviate from DP predictions, 
with errors showing periodic oscillations:
[image: fig4.png]

-- 
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 visit https://groups.google.com/d/msgid/cp2k/ce59ff3f-04cb-43ae-8ab9-66926ac364fen%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250316/164b4359/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fig4.png
Type: image/png
Size: 1199556 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250316/164b4359/attachment-0004.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fig1.png
Type: image/png
Size: 431991 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250316/164b4359/attachment-0005.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fig2.png
Type: image/png
Size: 36570 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250316/164b4359/attachment-0006.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fig3.png
Type: image/png
Size: 699848 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250316/164b4359/attachment-0007.png>