Different boxes and trajectories obtained from the NPT simulations using VCAIMAGES

[Yicheng]

Dear VASP developers,
Our group has encountered a problem related to the use of VCAIMAGES in VASP 6.4.2. I will attempt to articulate the problem clearly below, and appreciate your attention and patience.
We aim to investigate the free energy difference between Na-Cl and K-Cl ion pairs in aqueous solutions through thermodynamic coupling constant integration, following the detailed procedures outlined in your paper on the melting point of silicon (Dorner et al., 2018, Phys. Rev. Lett.). Our plan is to transition Na(+) gradually into K(+) using the keyword VCAIMAGES. The first few lines of our POSCAR file (consistent in both "01" and "02" folders) are presented as below:
NaCl-64H2O_Na-to-K_TI
1.0
12.43000000 0.00000000 0.00000000
0.00000000 12.43000000 0.00000000
0.00000000 0.00000000 12.43000000
H O Cl Ex
128 64 1 1
Direct
in which Na or K is denoted as “Ex”. The pseudopotentials for Na_pv and K_sv are employed in the POTCAR files in “01” and “02” folders respectively, in which all “Na” or “K” are replaced with “Ex”. The masses of "Ex" in these two POTCAR files are set to be the same (i.e., 22.990, the atom mass of Na).
With the settings mentioned above, we obtained exactly the same trajectories in “01” and “02” folders from NVT simulations. All input and output files are provided in the zip folder named “

NVT_100-steps.zip

” attached in this post. This validate the statement in VASP wiki: “since the energies and forces are averaged, the trajectories generated by the two simulations will be exactly identical.”
However, when it comes to NPT simulations, whether using the ICONST file to constrain the shape of the simulation cell or not, both boxes and trajectories differ in the "01" and "02" folders. (Please review the input and output files in “

NPT_10-steps_with-ICONST.zip

” and “

NPT_10-steps_without-ICONST.zip

” attached in this post.) Besides, although the ICONST file was used, the shape of the simulation cell was not exactly cubic.
In fact, if the NPT simulations are not performed for such a “strange” alchemical case, but for a relatively “ordinary” case to conduct the thermodynamic integration for one system between two exchange-correlation functionals (e.g., from PBE to revPBE), we also obtain different boxes and trajectories. (Please review the input and output files in “NPT_PBE-to-revPBE_with-ICONST.zip” attached in the reply of this post.)
Due to the different boxes and trajectories, we are concerned about the usability of the averaged energies obtained from NPT simulations for thermodynamic integrations. Could you please kindly clarify if there are any issues in our input files, or if the problem may be caused by VASP?
Thank you for your time and I am looking forward to your reply.
Sincerely,
Yicheng

[Yicheng]

Attached is the zip folder named “NPT_PBE-to-revPBE_with-ICONST.zip”.

[jonathan_lahnsteiner2]

Dear Yicheng,

Thank you for bringing this to our attention. Currently, the VCAIMAGES tag cannot be used in conjunction with the NPT ensemble. This is because the stress tensor is not averaged like the forces and energies are. We plan to address this issue in a future release.

All the best,

Jonathan

[Yicheng]

Dear Dr. Lahnsteiner,

Thank you very much for your time and response! Sincerely looking forward to the early release of the new version of VASP with such improvements.

Best wishes,
Yicheng

Dear Yicheng,

Thank you for bringing this to our attention. Currently, the VCAIMAGES tag cannot be used in conjunction with the NPT ensemble. This is because the stress tensor is not averaged like the forces and energies are. We plan to address this issue in a future release.

All the best,

Jonathan

[jonathan_lahnsteiner2]

Dear Yicheng,

I just wanted to let you know, that the new VASP was released including the fix to the bug you encountered.
I hope that your calculations will run without issues now.

All the best Jonathan

[Yicheng]

Dear Dr. Lahnsteiner,

Thank you very much for your time and effort in addressing this issue! I have been using the new version since last week, and all calculations are running smoothly.

Best wishes,
Yicheng

Dear Yicheng,

I just wanted to let you know, that the new VASP was released including the fix to the bug you encountered.
I hope that your calculations will run without issues now.

All the best Jonathan