[CP2K-user] [CP2K:11256] DF2 van der Waals (vdW) and XC functionals

[Patrick Gono]

Dear Ronald,

it is true that documentation can be sparse. I suggest following the
articles cited in the manual for more information.

I note some related problems with the documentation. I find examples like
> for LIBXC:
>
> FUNCTIONAL XC_MGGA_X_M06_L XC_MGGA_C_M06_L
>
> but the docs at
>
> say
>
>    - FUNCTIONAL
>    <https://manual.cp2k.org/trunk/CP2K_INPUT/FORCE_EVAL/DFT/XC/XC_FUNCTIONAL/LIBXC.html#list_FUNCTIONAL>
>    - This keyword cannot be repeated and it expects precisely one word.
>
> Yet there are "two words"
>
> and including VWN afterwards also seems like more than one functional, as
> I said above.
>

In older versions (prior to 6.1), the FUNCTIONAL keyword accepted a list of
words. In the newest version (6.1), this keyword only accepts a single
word. Note, however, that the section &LIBXC can be repeated. So, instead
of:

&LIBXC
  FUNCTIONAL XC_MGGA_X_M06_L XC_MGGA_C_M06_L
&END LIBXC

you can do:

&LIBXC
  FUNCTIONAL XC_MGGA_X_M06_L
&END LIBXC
&LIBXC
  FUNCTIONAL XC_MGGA_C_M06_L
&END LIBXC

You can also mix different correlation or exchange functionals, like so:

&XC_FUNCTIONAL
  &LIBXC
    FUNCTIONAL  XC_GGA_X_RPW86
  &END LIBXC
  &LIBXC
    FUNCTIONAL  GGA_C_PBE
  &END LIBX
&END XC_FUNCTIONAL
&vdW_POTENTIAL
  DISPERSION_FUNCTIONAL NON_LOCAL
  &NON_LOCAL
    TYPE RVV10
    PARAMETERS  9.3   0.0093
    KERNEL_FILE_NAME  ./rVV10_kernel_table.dat
    CUTOFF      400
  &END NON_LOCAL
&END vdW_POTENTIAL



where I used RPW86 exchange and PBE correlation. The reason behind that is
that I am trying to use the (modified) rVV10 nonlocal van der Waals
functional. In the original paper:
https://journals.aps.org/prb/abstract/10.1103/PhysRevB.87.041108 the
authors define the total exchange-correlation as:
XC_rVV10 = X_RPW86 + C_PW_LDA + C_non-local_vdW

However, in later work it was found that the non-local vdW correction from
the rVV10 works quite well with other combinations of correlation and
exchange functionals. Hence, I am now using
XC_rVV10 = X_RPW86 + C_PBE + C_non-local_vdW
with a specific choice of the b parameter in the vdW correction.

What I am trying to say is that you can mix and match the non-local vdW
correction with different exchanges and correlations. The specific choice
will always depend on the system you are trying to study. There are many
papers testing and comparing various combinations, such as this one
https://pubs.acs.org/doi/10.1021/ct4003527 where the VV10 corrections is
added to various exchanges and correlations.

In your input file you use the Lee-Murray-Kong-Lundqvist-Langreth (LMKLL)
nonlocal van der Waals density functional (via the keyword TYPE LMKLL). If
you look up the relevant publication, you will find what correlation and
exchange functional it is supposed to be used with. You may also find
benchmark papers that use this non-local correction on top of
other-than-originally-intended exchange and correlation functionals. This
allows one to choose the best combination for the system at hand (large
system vs small; periodic vs isolated; metals vs semiconductors vs small
molecules; etc.).

Hope this helps.
Yours sincerely,
Patrick Gono

On Tue, 12 Feb 2019 at 15:02, Ronald Cohen <reco... at gmail.com> wrote:

> I am quite confused about the (lack of) documentation for the XC
> functionals and non-local vdW. I find examples such as:
>
> &XC
>
> &XC_FUNCTIONAL NO_SHORTCUT
>
> &LIBXC T
>
> FUNCTIONAL XC_GGA_X_RPW86
>
> &END LIBXC
>
> &VWN T
>
> &END VWN
>
> &END XC_FUNCTIONAL &VDW_POTENTIAL
>
> POTENTIAL_TYPE NON_LOCAL &NON_LOCAL
>
> TYPE LMKLL
>
> VERBOSE_OUTPUT T KERNEL_FILE_NAME ./vdW_kernel_table.dat CUTOFF 160
>
> &END NON_LOCAL &END VDW_POTENTIAL
>
> &END XC
>
>
> (from
> http://archer.ac.uk/training/course-material/2014/08/CP2K/Slides/NSCCS_2014_Sanliang_Ling.pdf
>  )
>
>
> and similar even in the regtests.
>
> This seems to call both XC_GGA_X_RPW86 and VWN (lda) .
>
> What does it mean to call two different xc potentials in cp2k, and why
> would one do it?
>
> Tracing this in the code is not so easy either.
>
>
> Ins eom docs it seems to suggest that LDA only should be used for
> correlation. Is that because GGA is included
>
> in the vDW function in cp2k? Also there is the question which
> pseudopotentials to use.
>
>
> It seems to me that  XC_GGA_X_RPW86 is correct without the VMN, and this
> would correapond to what is done for DF2 in
>
> quantum espresso it seems.
>
>
> I note some related problems with the documentation. I find examples like
> for LIBXC:
>
> FUNCTIONAL XC_MGGA_X_M06_L XC_MGGA_C_M06_L
>
> but the docs at
>
> say
>
>    - FUNCTIONAL
>    <https://manual.cp2k.org/trunk/CP2K_INPUT/FORCE_EVAL/DFT/XC/XC_FUNCTIONAL/LIBXC.html#list_FUNCTIONAL>
>    - This keyword cannot be repeated and it expects precisely one word.
>
> Yet there are "two words"
>
> and including VWN afterwards also seems like more than one functional, as
> I said above.
>
>
> Thank you!
>
>
> Ronald Cohen
>
>
>
>
> --
> 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... at googlegroups.com.
> To post to this group, send email to cp... at googlegroups.com.
> Visit this group at https://groups.google.com/group/cp2k.
> For more options, visit https://groups.google.com/d/optout.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190212/502e4779/attachment.htm>