I am going to do some research with geometry optimization in Abinit and then electronic properties analysis with Elk. For that, I must use the same functionals in both codes. Yes, I can simply use the same libxc and do not care, but I just want to know.
The default Elk functional is PW92 exchange-corellation. As far as I could see from the code, it follows the original PW92 paper [1]. Perdew and Wang introduced a correlation functional, while they apparently used Hartree-Fock-based exchange[2]. The question is, if I used Libxc LDA_X exchange [3] with Libxc PW92 correlation, would I get the same functionals and the same results, as with elk's xctype=3? The problem is that the exchange energy formula in [1] and in elk differs from the one in [2,3] by the factor of 2. Is that correct and/or what do I miss?
The question also extends to other built-in functionals in Elk. Is physics in Elk versions and Libxc versions the same?
I used elk-4.0.15, gonna use 4.3.6.
Thanks.
UPD: As Lars Nordström noticed above, the factor of 2 is units differece. So, Libxc's LDA_X + LDA_C_PW must be the same as Elk's xctype=3.
Hi,
I am going to do some research with geometry optimization in Abinit and then electronic properties analysis with Elk. For that, I must use the same functionals in both codes. Yes, I can simply use the same libxc and do not care, but I just want to know.
The default Elk functional is PW92 exchange-corellation. As far as I could see from the code, it follows the original PW92 paper [1]. Perdew and Wang introduced a correlation functional, while they apparently used Hartree-Fock-based exchange[2]. The question is, if I used Libxc LDA_X exchange [3] with Libxc PW92 correlation, would I get the same functionals and the same results, as with elk's xctype=3? The problem is that the exchange energy formula in [1] and in elk differs from the one in [2,3] by the factor of 2. Is that correct and/or what do I miss?
The question also extends to other built-in functionals in Elk. Is physics in Elk versions and Libxc versions the same?
I used elk-4.0.15, gonna use 4.3.6.
Thanks.
UPD: As Lars Nordström noticed above, the factor of 2 is units differece. So, Libxc's LDA_X + LDA_C_PW must be the same as Elk's xctype=3.
[1] link.aps.org/doi/10.1103/PhysRevB.45.13244
[2] https://journals.aps.org/prb/abstract/10.1103/PhysRevB.13.4274
[3] http://octopus-code.org/wiki/Lda_x
Last edit: Andrew Shyichuk 2017-05-16
It might be units, Hartree vs Rydbergs ...
Last edit: Lars Nordström 2017-05-16
Silly me, it is Rydberg in [2] and Hartree allover [1]. Thanks. General question remains, however.