[CP2K-user] Question about usage of libxc

[Yike Huang]

Dear Frederick,

Thank you for your answer and explanation with so many details, they help 
me a lot. 
Indeed after I posting this question, I tried to run a toy calculation 
without defining any functional in XC_FUNCTIONAL section and in output file 
I found exchange-correlation contribution was 0, but I was still not sure 
until I got your answer. 
Thank you again.

Best regards, 
Huang Yike

在 2020年8月14日星期五 UTC+8下午8:46:27，Frederick Stein写道：
>
> Dear Huang,
>
> There are some examples with correlation functionals in the test suite 
> (tests/QS/regtest-libxc/H2O-hybrid-b3lyp_libxc.inp). LibXC correlation 
> functionals are treated like exchange functionals. Technically, there isn't 
> any difference between calculating a LibXC exchange or correlation or 
> kinetic energy functional of the same kind.
>
> RPBE affects only the enhancement factor of the exchange part. The LDA 
> exchange energy density is known analytically. For the correlation part, 
> you have to manually add the PBE correlation functional. The LibXC 
> implementation and our implementation, both exploit the PW92 LDA functional.
>
> The input section looks like the following
>
> &XC_FUNCTIONAL
>   &LIBXC
>     FUNCTIONAL XC_GGA_X_RPBE
>   &END LIBXC
>   &LIBXC
>     FUNCTIONAL XC_GGA_C_PBE
>   &END LIBXC
> &END XC_FUNCTIONAL
>
> or with the cp2k implementation of PBE
>
> &XC_FUNCTIONAL
>   &LIBXC
>     FUNCTIONAL XC_GGA_X_RPBE
>   &END LIBXC
>   &PBE
> # We don't need PBE exchange here because we are using RPBE exchange
>     SCALE_X 0.0
>   &END PBE
> &END XC_FUNCTIONAL
>
> I hope I could help you.
>
> Best,
>
> Frederick
> yk... at gmail.com schrieb am Freitag, 14. August 2020 um 06:15:02 UTC+2:
>
>> Greetings, cp2k users.
>>
>> I want to use rpbe functional in my calculation, after reading manual I 
>> find I should use libxc cuz rpbe is not implemented directly in cp2k. 
>> However, when I am reading input files in test/, I find there are examples 
>> in which only exchange term is defined, so I wonder how will cp2k calculate 
>> the correlation term? Will it use original LDA?
>> Answers will be appreciated, thanks in advance. 
>>
>> Huang Yike 
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200814/775767d8/attachment.htm>