[CP2K-user] [CP2K:20321] Re: ADMM and stress tensor

[Arthur France-Lanord]

Hi, 

Yes, these parameters have been converged (except for EPS_SCHWARZ_FORCES, 
which I didn't know of). Also, thanks for pointing out the PBE and not PBE0 
pseudopotential (differences are minor though). 

I've computed the stress tensor through numerical differentiation and the 
results are consistent with the analytical result, so I'll assume these 
results are appropriate, and that large differences with the reference PBE 
stress tensor are due to differences in forces. 

best, 
Arthur

On Tuesday, June 11, 2024 at 8:37:57 PM UTC+2 Frederick Stein wrote:

> Dear Arthur,
> - Did you check the convergence of all relevant parameters (CUTOFF_RADIUS, 
> EPS_SCF (if possible), EPS_SCHWARZ, EPS_SCHWARZ_FORCES (in the HF section), 
> EPS_DEFAULT)?
> - EPS_PGF_ORB should be chosen less tight (or just stick to the default). 
> You should also employ a PBE0/hybrid-based pseudo potential. If there is 
> none in the data directory, you can checkout this repository: 
> https://github.com/juerghutter/GTH.
> - You may also check other auxiliary basis sets for ADMM in BASIS_ADMM, 
> but any other basis set smaller than your primary basis set will do the job 
> too.
> Best,
> Frederick
>
> Arthur France-Lanord schrieb am Dienstag, 11. Juni 2024 um 19:51:26 UTC+2:
>
>> Hi Frederick, 
>>
>> Thank you for your quick reply - here are the files. I'm using RUN_TYPE 
>> MD, but it's basically similar to ENERGY_FORCE - I prefer outputting 
>> quantities this way. The kinetic contribution to the stress tensor is zero, 
>> since velocities and temperature are also zero. 
>>
>> For comparison, here are the stress tensor components for the parent gga 
>> run used as an initial wavefunction: 
>>
>> #   Step   Time [fs]            xx [bar]            xy [bar]           
>>  xz [bar]            yx [bar]            yy [bar]            yz [bar]       
>>      zx [bar]            zy [bar]            zz [bar]
>>        0       0.000    19290.8730848902     -591.1275884811     
>> -502.6298831727     -591.1275884811    19948.5377135713    -2305.6236157682 
>>     -502.6298831727    -2305.6236157682    21069.0573734849
>>
>> best, 
>> Arthur
>>
>>
>> On Tuesday, June 11, 2024 at 7:12:20 PM UTC+2 Frederick Stein wrote:
>>
>> Hi Arthur,
>> can you provide your input and output files?
>> Best,
>> Frederick
>>
>> Arthur France-Lanord schrieb am Dienstag, 11. Juni 2024 um 18:34:58 UTC+2:
>>
>> Hi, 
>>
>> Using cp2k 9.1, I'm performing a PBE0 calculation using the ADMM 
>> approximation (ADMM_PURIFICATION_METHOD MO_DIAG) and OT. I'm seeing some 
>> weird numbers on the stress tensor components (STRESS_TENSOR ANALYTICAL), 
>> although when comparing to a PBE calculation, forces seem reasonable. 
>>
>> In this version, is there an analytical implementation of the stress 
>> tensor for ADMM? According to https://github.com/cp2k/cp2k/pull/1647 
>> there should be, but I'd rather check. 
>>
>> thanks, 
>> Arthur
>>
>>
>>

-- 
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/0e9564c1-9cae-49e4-a419-8e3a0713a210n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20240612/93a2a906/attachment.htm>