[CP2K-user] [CP2K:13128] How to understand/tune parameters for GAPW calculations

[Krack Matthias (PSI)]

Dear Natalie

I will try to answer your questions:


1)      Yes, I think that’s right when using an all-electron potential.

2)      EPSFIT is the key input parameter to define the split of the basis set into a soft and a hard part. The default value of 1.0E-4 is too large for most systems. I would rather consider values of 1.0E-6 for EPSFIT to ensure that the soft basis includes sufficiently large exponents, e.g. if the largest exponent in the soft basis set for oxygen is only 4 this will give poor results.

3)      GAPW splits the basis set into a hard (frozen) and a soft part. This can be applied for all-electron calculations but also for hard pseudopotentials. There is no additional pseudopotential.

4)      In a first step, you can use the default values for these input parameters. More import is EPSRHO0. Consider smaller values than the default one for better accuracy. The electron counts printed (use PRINT_LEVEL medium) during and after the SCF procedure can serve as an accuracy check. In addition, the parameters defining the local atomic (Lebedev) grids (see &KIND section) might need larger values in some cases, e.g. for heavier elements.

5)      See above (EPSFIT)

HTH

Matthias

From: cp... at googlegroups.com <cp... at googlegroups.com> On Behalf Of Natalie Holzwarth
Sent: Dienstag, 14. April 2020 17:16
To: cp2k <cp... at googlegroups.com>
Subject: [CP2K:13128] How to understand/tune parameters for GAPW calculations

Dear cp2k developers,
      We are very much looking forward to possibly using CP2K  particularly using the GAPW scheme to extend our calculations which up to now have been performed using ABINIT and QUANTUM ESPRESSO using PAW.  We have tried a number of your tutorials and test inputs, but are have a hard time understanding how to properly pick the parameters.   Is there perhaps a writeup somewhere that describes the detailed implementation or perhaps a list of the main subroutines that a user should study?   Or perhaps someone might have time to answer some the following silly questions.
1.  These questions are based on the notion that we should use the all-electron bases since these would allow n^1 to be represented analogously to Blochl's PAW formulation.   Is this the correct way to think about it?
2.   According the paper by Lippert et al (1999), the soft wavefunctions and densities are chosen by eliminating the Gaussian basis primitives with too large an exponent.   Which of the input parameters determines the soft basis parameters?
3.   In addition to the soft density, Blochl's PAW scheme has a local pseudo potential.    I am not sure what is the analogous local pseudo potential for GAPW and which parameter controls it. I have assumed that we should use "ALL" potentials in the input files.   Is this correct?
4.   Some of the examples use the following parameters which seem to be important, but I am not sure how to choose them

MAX_RAD_LOCAL
ALPHA0_H
HARD_EXP_RADIUS

5.   Is there a provision/recommendation to use the frozen core approximation and if so, what parameters control that?

Thanks in advance for any suggestions you can provide and much gratitude for sharing and maintaining the cp2k software package.


Sincerely,   Natalie Holzwarth
--
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 cp... at googlegroups.com<mailto:cp... at googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/e59f80e3-1dda-4bb4-b969-93b783399ad9%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/e59f80e3-1dda-4bb4-b969-93b783399ad9%40googlegroups.com?utm_medium=email&utm_source=footer>.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200415/690f8446/attachment.htm>