Issues with stress tensor calculations.

[Dina Kussainova]

Hi,

I am doing single-point calculations of Li2CO3 with PBE functional, DZVP basis sets, and GTH pseudopotentials. Please see attached input file.  However, I am getting different values of stress tensors when I am choosing analytical or numerical methods. When I am doing NPT simulations with the analytical method for stress tensor calculations, looks like volume is also overestimated. Do you know what can be a reason for this? Any help would be appreciated. 

Below are snapshots of the output files.

Thanks,

Dina

[Jürg Hutter]

Hi

unfortunately, without the complete input it is not possible to reproduce your findings.
From your input I would suggest:

- use a recent version of CP2K (version 9.1 or the recently released 2022.1)
- Don't use LONG_RANGE_CORRECTION in D3
- for debugging purposes it might also be good to switch off CALCULATE_C9_TERM
- Use EPS_SCF 1.E-7 or lower
- Don't use &XC_GRID
XC_DERIV SPLINE2
XC_SMOOTH_RHO NN50
&END XC_GRID
- Increase REL_CUTOFF to 60 or higher
- Use CELL_REF with 20% increased cell parameters to stabilize the finite difference calculation.

best regards

Juerg Hutter

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Dina Kussainova <di...@princeton.edu>
Sent: Friday, July 15, 2022 10:59 PM
To: cp2k
Subject: [CP2K:17324] Issues with stress tensor calculations.

[analytical.png]

[numerical.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+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/61ebfe8b-437f-42ad-affb-8d3b64a344b1n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/61ebfe8b-437f-42ad-affb-8d3b64a344b1n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Dina Kussainova]

Dear Professor Juerg Hutter,

Thank you for your help. Turning off LONG_RANGE_CORRECTION helped to get similar stress tensors for both analytical and numerical methods. However, I have also noticed that XC_GRID parameters significantly affect stress tensor calculations as well as forces/energy calculations. Please see below snapshots of some test calculations with different parameters. Considering this, what is the best way to identify the correct parameters for my calculations? 

Thanks,

Dina

[Jürg Hutter]

Hi

as I wrote in my first answer, I strongly advice against using smoothing methods for the density or
the gradients. As far as I know, the stress dependence on the smoothing function is not implemented.
If the grid convergence is a problem, I would use other methods, e.g. NLCC or GAPW.

regards

Juerg Hutter

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Dina Kussainova <di...@princeton.edu>

Sent: Tuesday, August 2, 2022 9:34 PM
To: cp2k
Subject: Re: [CP2K:17429] Issues with stress tensor calculations.

Dear Professor Juerg Hutter,

Thank you for your help. Turning off LONG_RANGE_CORRECTION helped to get similar stress tensors for both analytical and numerical methods. However, I have also noticed that XC_GRID parameters significantly affect stress tensor calculations as well as forces/energy calculations. Please see below snapshots of some test calculations with different parameters. Considering this, what is the best way to identify the correct parameters for my calculations?

Thanks,
Dina

[orig.png]
[test2.png]

[test8.png]

[test11.png]

[test12.png]

[test13.png]

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/fddf3eab-9bb6-4b56-b127-9c019eb0bee3n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/fddf3eab-9bb6-4b56-b127-9c019eb0bee3n%40googlegroups.com?utm_medium=email&utm_source=footer>.