[CP2K-user] [CP2K:18203] Non-physical pressure behavior for non-local vdW functionals

[Vyacheslav Bryantsev]

Dear CP2K Developers,

I would like to follow up on unphysical stress tensor from AIMD using the 
van der Waals density functionals (e.g., optB88-vdW, optB86b-vdW).
The input file we use (attached in the previous emails) works perfectly 
fine for PBE-D3, SCAN, and SCAN-D3, which implies some problems with stress 
tensor computations in CP2K using the van der Waals density functionals.

Thank you,
Slava
Vyacheslav Bryantsev

On Thursday, December 1, 2022 at 6:38:27 PM UTC-5 ldgib... at gmail.com wrote:

> Hi Prof. Hutter,
>
> Thank you for your suggestion. We ran that DEBUG job with the input you 
> provided for optB86b-vdW. In the output, it compares the stress tensor 
> computed numerically and analytically, which showed that the two methods 
> gave the same result (within error). I have attached the output file from 
> that DEBUG job for reference. I have also included an xyz file for the 
> system we are looking at for any testing.
>
> We also discovered that we were running the SCAN-RVV10 calculations 
> without actually including the non-local vdW corrections from RVV10. So in 
> the figure I previously attached, the plot labeled "SCAN-RVV10" was 
> actually showing the pressures from SCAN without any dispersion 
> corrections. Therefore, we only see this non-physical behavior when using 
> non-local vdW functionals and showed that it is not attributed to the 
> method of computing the stress tensor (i.e., numerical vs analytical) based 
> on the DEBUG job we ran.
>
> We appreciate your help in identifying the source of this problem. If you 
> need any other information from us for testing, we are happy to provide it.
>
> Thank you,
> Luke
>
> On Wednesday, November 30, 2022 at 7:48:43 AM UTC-5 jgh wrote:
>
>> Hi 
>>
>> difficult to say what goes wrong here. No atomic coordinates for further 
>> testing available. 
>> But I would suggest to make the following test: 
>> Debug the stress tensor using finite differences and avoid 
>> STRESS_TENSOR DIAGONAL_ANALYTICAL 
>> use 
>> STRESS_TENSOR ANALYTICAL 
>>
>> Use the DEBUG run_type: 
>>
>> RUN_TYPE DEBUG 
>>
>> with 
>>
>> &DEBUG 
>> DEBUG_FORCES .FALSE. 
>> DEBUG_STRESS_TENSOR .TRUE. 
>> DEBUG_DIPOLE .FALSE. 
>> DEBUG_POLARIZABILITY .FALSE. 
>> DE 0.001 
>> STOP_ON_MISMATCH .FALSE. 
>> EPS_NO_ERROR_CHECK 5.e-5 
>> &END 
>>
>> regards 
>> JH 
>>
>> ________________________________________ 
>> From: cp... at googlegroups.com <cp... at googlegroups.com> on behalf of Luke 
>> Gibson <ldgib... at gmail.com> 
>> Sent: Monday, November 28, 2022 4:19 PM 
>> To: cp2k 
>> Subject: [CP2K:18105] Non-physical pressure behavior for non-local vdW 
>> functionals 
>>
>> Dear developers and users, 
>>
>> We are looking at the equilibrium density for an AlCl3 molten salt melt 
>> by measuring pressure values over 10 ps in NVT at a variety of volumes. We 
>> are comparing the pressures for a variety of functionals with various vdW 
>> corrections: PBE-D3, optB86b-vdW, optB88-vdW, SCAN-D3, and SCAN-RVV10. The 
>> data are shown in the attached plot with the mean pressure (± 1 standard 
>> deviation) for each volume (the plot for PBE-D3-smooth can be ignored). 
>>
>> It seems optB86b-vdW, optB88-vdW, and SCAN-RVV10 are reporting unexpected 
>> behavior in the stress tensors. Both optB86b and optB88 seem to be the most 
>> "wrong" with negative pressures across the board and an unusual increase in 
>> the pressure as the system's volume is expanded (all systems have the same 
>> number of particles). SCAN-RVV10 at least follows a believable trend and 
>> isn't too far off, so it may just not be appropriate for this system. 
>>
>> We are wondering if the stress tensor calculation is properly implemented 
>> for these non-local vdW functionals, since both PBE-D3 and SCAN-D3 are 
>> producing believable trends. 
>>
>> For reference, I have included the input files for the 3 problematic 
>> functionals I mentioned for the volumes corresponding to V_exp in the 
>> attached plots. The versions used for each of those runs are listed below: 
>>
>> * optB86b-vdW: CP2K-8.2 
>> * optB88-vdW: CP2K-6.1 
>> * SCAN-RVV10: CP2K-2022.2 
>>
>> Any help or insight is appreciated! 
>>
>> Thank you, 
>>
>> Luke Gibson 
>>
>> -- 
>> 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/afc955e0-daad-40f4-8bce-87b08f87cb8dn%40googlegroups.com
>> <
>> https://groups.google.com/d/msgid/cp2k/afc955e0-daad-40f4-8bce-87b08f87cb8dn%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/18cd4706-b0f6-4ee9-9a95-2e18a51d00b0n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20221214/bb7180ec/attachment.htm>