[CP2K-user] Nuclear energy contributions in GAPW scheme

[pavan kumar behara]

I think I missed the cancelling term in periodic calculations, the part of 
nuclear potential that cancels out the long-range part of pseudopotential. 

On Tuesday, July 2, 2019 at 1:18:38 PM UTC-4, pavan kumar behara wrote:
>
> Hello CP2K developers,
>
> I am trying to separate the electronic and nuclear energy contributions in 
> the GAPW scheme. I could do so successfully for a molecule with 
> non-periodic calculation by passing nuclear-only density to the pw_poisson 
> and hartree_1c energy calculation routines. By adding nuclear_pw, 
> nuclear_1c with the self-energy term I get exact same value as analytical 
> calculation (q1*q2 / r ). 
>
> I did this by passing rho0_s_gs, with only parts relevant to the core part 
> and ignoring the hard and soft coefficients that pertain to electron 
> density, to pw_poisson_solve() to get nuclear_pw. This includes ignoring 
> the Qlm_h, Qlm_s under calculate_rho0_atom() routine. So, only Qlm_z goes 
> into Qlm_tot, thus rho0 contains only the multipole of core charges.
>
> And, by considering only the rhoz relevant terms in hartree_local_methods 
> to get nuclear_1c. For this under hartree_local_methods, I considered only 
> vrrad_z and vrrad_0 terms in Vh1_h, Vh1_s respectively. And, ecoul_1_z, 
> ecoul_1_0 energy terms calculated using those potentials. 
>
> When it comes to periodic calculations doing the same results in negative 
> nuclear energies for small cell sizes. If I do a very big box (of 100A) by 
> placing the molecule at the center, I do see that this energy value is 
> positive and it is converging towards the non-periodic value but still not 
> so close. 
>
> My question is whether I am not accounting for the ewald summation effects 
> correctly? Or my assumption or way of separating the nuclear and electronic 
> densities is not correct? 
>
> I also saw a very old discussion about parameter tuning (link 
> <https://groups.google.com/d/msg/cp2k/RUFQScjSDn0/lWdtYskTm9IJ>) and 
> there's a mention about periodic interactions not being correct for small 
> cells and system type. I am testing the periodic case on a hematite 
> supercell (3x3x2) structure (cell size of [14.7A, 14.7A, 26.5A]), with DZVP 
> basis and GTH pseudopotential, and facing difficulties. I tried tuning the 
> input parameters (alpha0_hard, max_local_rad, eps_fit) but still end up 
> with negative nuclear energy. I am a bit lost on where to look, any 
> pointers are highly appreciated. 
>
> I am sorry for such a long post. Thank you very much for your time.
>
> Best regards,
> Pavan.
>
> *Edits*: Made changes to the question, I went through Blochl's paper on 
> PAW method and understood the electrostatic term contributions and 
> compensation density well. Also, included more details about the parts I am 
> modifying.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190703/dc52e868/attachment.htm>